Photolithography Pattern Modeling With DNN Weight Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As semiconductor devices become more integrated, the proximity and narrow widths of mask patterns on photomasks lead to light interference and diffraction, resulting in distorted layouts during photolithography, which existing technologies struggle to accurately predict and correct.
Innovation Solution
A pattern modeling system and method using a deep neural network (DNN) with a weight filter to generate and train image data, determining specific areas for filter application, predicting pattern images with high accuracy by minimizing loss functions, and performing optical proximity correction (OPC) and process proximity correction (PPC).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the degree of integration of semiconductor devices is increased, then the productivity and device density are improved, but light interference and diffraction occur causing manufacturing precision to deteriorate
Solution Approach 1:
The system performs preliminary actions by generating multiple candidate correction patterns before the actual photolithography process. The DNN model predicts optimal OPC patterns in advance, allowing distortion compensation to be built into the mask design before manufacturing, thus preventing rather than correcting the interference issues during actual production
Solution Approach 2:
The system creates virtual copies of the layout patterns through simulated measurement data and uses these copies for training the DNN model. By copying and analyzing multiple pattern variations in the virtual domain, the system learns to predict and correct distortions without affecting the actual physical manufacturing process
2Measurement precision
If a weight filter is applied to specific areas of measurement data, then the prediction accuracy for those areas is improved, but the device complexity increases
Solution Approach 1:
The system applies local quality by using a weight filter that assigns different weights to different regions of the measurement data. Critical areas such as pattern edges and high-density regions receive higher weights, while less critical areas receive lower weights. This allows the DNN model to focus its learning capacity on the most important features, improving prediction accuracy for critical dimensions without requiring the entire system to be overly complex
Solution Approach 2:
The system changes parameters by dynamically adjusting the weight filter parameters based on the specific characteristics of the measurement data. The weight values are not fixed but are optimized during the training process to achieve the best prediction accuracy. This parameter optimization allows the system to adapt to different pattern types and manufacturing conditions without requiring a completely different system architecture
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enhances the prediction of measurement data and corrects distortions in semiconductor layouts, improving the accuracy of pattern formation and reducing errors in photolithography processes.
Implementation Method 1
generating first image data based on a sample pattern that is learned by a deep neural network (DNN), the DNN comprising a plurality of layers
Implementation Method 2
determining an area of the second image data to which a weight filter is to be applied, training the DNN by applying the weight filter to the determined area of the second image data
Implementation Method 3
The photomask may include a transparent area and an opaque area. The transparent area may be formed by etching a metal layer on the photomask and light may be passed therethrough. On the other hand, the opaque area may not pass light therethrough
Implementation Method 4
Due to this proximity, interference and diffraction of light may occur, and accordingly, a distorted layout different from a desired layout may be printed on the wafer
Implementation Method 5
Due to this proximity, interference and diffraction of light may occur, and accordingly, a distorted layout different from a desired layout may be printed on the wafer
Data Source
AI summary
A pattern modeling method of predicting image data includes generating first image data based on a sample pattern that is learned by a deep neural network (DNN), generating second image data by measuring the first image data, determining an area of the second image data to which a weight filter is to be applied, training the DNN by applying the weight filter to the determined area of the second image data, and predicting at least one pattern image based on a result of the training of the DNN.