ML-Based OPC Mask Pattern Generation for Low-k1 Lithography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lithographic projection technologies face challenges in accurately reproducing patterns with dimensions smaller than the classical resolution limit, particularly in low-k1 lithography, due to difficulties in manufacturing curvilinear masks and the complexity of optical proximity correction (OPC) processes, which are time-consuming and require numerous iterations to achieve high accuracy and process windows.
Innovation Solution
A method involving a trained machine learning model is employed to generate a mask pattern by modifying an input pattern based on differences between a target and simulated pattern, using a convolutional neural network (CNN) to speed up the optical proximity correction (OPC) process and address boundary merge issues in inverse OPC, while maintaining high accuracy and simplicity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional optical proximity correction (OPC) processes are used to generate mask patterns for low-k1 lithography, then manufacturing precision of patterns smaller than the classical resolution limit is improved, but the complexity of the correction process and time required increase significantly
Solution Approach 1:
The patent replaces conventional iterative OPC algorithms with a trained neural network model that directly generates corrected mask patterns. The neural network learns the complex optical proximity effects during training and applies corrections in a single forward pass, substituting the iterative computational mechanical process with a trained intelligent system that achieves comparable or superior accuracy without the iterative complexity
Solution Approach 2:
The patent performs preliminary training of the neural network model using a training set of patterns and their corresponding corrected versions. This preliminary action pre-loads the correction knowledge into the model, allowing it to perform rapid corrections during actual mask pattern generation without needing to execute complex iterative OPC algorithms at runtime
2Manufacturing precision
If iterative OPC methods are employed to achieve high accuracy in pattern reproduction, then manufacturing precision is improved, but the runtime and iterations required increase
Solution Approach 1:
The patent substitutes iterative OPC algorithms with a trained neural network that performs corrections in a single forward pass. The neural network is trained offline on a comprehensive training set, and during actual use, it directly generates corrected mask patterns without iterative loops, dramatically reducing runtime while maintaining high accuracy
Solution Approach 2:
The patent creates a trained neural network model that copies the correction behavior of complex iterative OPC algorithms. The network learns from a training set containing input patterns and their corresponding correctly corrected patterns, then reproduces this correction behavior rapidly without needing to re-execute the full iterative process
3Manufacturing precision
If curvilinear mask patterns are manufactured for advanced lithography, then manufacturing precision of sub-resolution features is improved, but the ease of manufacture decreases
Solution Approach 1:
The patent uses a neural network to generate the complex curvilinear mask patterns directly from simplified input patterns. The network learns to create the necessary curvilinear corrections during training, automatically generating manufacturable curvilinear patterns without requiring complex manual design or iterative adjustment processes
Solution Approach 2:
The patent transforms simple rectangular or linear input patterns into complex curvilinear corrected patterns through the neural network's parameter transformations. The network adjusts geometric parameters such as curve radii, line widths, and spacing to compensate for optical proximity effects, automatically generating the curvilinear features needed for high-precision sub-resolution printing
Data Source
AI summary
A method for determining a mask pattern and a method for training a machine learning model. The method for determining a mask pattern includes obtaining, via executing a model using a target pattern to be printed on a substrate as an input pattern, a post optical proximity correction (post-OPC) pattern; determining, based on the post-OPC pattern, a simulated pattern that will be printed on the substrate; and determining the mask pattern based on a difference between the simulated pattern and the target pattern. The determining of the mask pattern includes modifying, based on the difference, the input pattern inputted to the model such that the difference is reduced; and executing, using the modified input pattern, the model to generate a modified post-OPC pattern from which the mask pattern can be derived.