Machine Learning Assist Feature Placement for Lithography
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lithographic processes face challenges in accurately transferring design layouts to substrates due to proximity effects and resolution limitations, particularly in achieving precise feature sizes and densities, which can result in defects such as line end pull back and corner rounding, impacting the manufacturing of integrated circuits.
Innovation Solution
The method involves using a machine learning model to determine assist features based on design layout characteristics, such as geometrical and statistical parameters, and incorporating these features into the design layout to enhance printability and process tolerance, employing techniques like optical proximity correction and independent assist features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional lithographic patterning is used, then manufacturing process is simple, but manufacturing precision deteriorates due to proximity effects and resolution limitations
Solution Approach 1:
The method applies optical proximity correction and determines assist features before the actual lithographic patterning process. By pre-calculating and pre-modifying the design layout to compensate for anticipated proximity effects and resolution limitations, the system ensures accurate feature reproduction without requiring complex real-time adjustments during manufacturing.
Solution Approach 2:
The invention introduces assist features as intermediary elements that mediate between the original design layout and the final patterned substrate. These assist features (such as scattering bars or serifs) are added to the design layout to modify the optical field distribution, thereby improving the printability of critical features without directly altering the functional design.
2Manufacturing precision
If assist features are added to design layout, then manufacturing precision improves, but device complexity increases
Solution Approach 1:
The system applies optical proximity correction and assist features selectively to specific regions of the design layout where proximity effects are most problematic. Rather than uniformly modifying the entire layout, the method identifies critical areas (such as dense regions or features with specific geometric characteristics) and applies corrections only where needed, thereby minimizing overall layout complexity while maintaining manufacturing precision.
Solution Approach 2:
The invention modifies parameters of the design layout such as feature dimensions, spacing, and the addition of assist features based on calculated proximity effects. By dynamically adjusting these parameters according to the local layout context and lithographic process conditions, the system optimizes pattern transfer fidelity while managing design complexity through parameter optimization rather than structural overhaul.
3Manufacturing precision
If machine learning model is used to determine assist features, then manufacturing precision improves, but computational requirements increase
Solution Approach 1:
The machine learning model for assist feature determination is trained in advance using comprehensive training data that captures various layout patterns and their optimal assist feature configurations. This preliminary training phase allows the model to learn complex relationships between design characteristics and optimal assist features, enabling accurate predictions during actual manufacturing with reduced real-time computational requirements.
Solution Approach 2:
The system uses the trained machine learning model to copy successful assist feature patterns from training examples to new design layouts. Rather than performing complex real-time optimization calculations, the model retrieves and adapts pre-learned solutions that have been proven effective for similar layout configurations, significantly reducing computational power requirements while maintaining high manufacturing precision.
Data Source
AI summary
A method including: obtaining a portion of a design layout; determining characteristics of assist features based on the portion or characteristics of the portion; and training a machine learning model using training data including a sample whose feature vector includes the characteristics of the portion and whose label includes the characteristics of the assist features. The machine learning model may be used to determine characteristics of assist features of any portion of a design layout, even if that portion is not part of the training data.