Photomask Modeling With Image-Based ANN for Faster SEM Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current photomask manufacturing process modeling is time-consuming and expensive, requiring extensive effort from multiple engineers and is difficult to scale, as it involves complex mathematical and statistical methods that are limited by existing software tools, and often struggles with unforeseen effects.
Innovation Solution
An image-based Artificial Neural Networks (ANN) method is used to model photomask manufacturing, where a training phase builds an image-based ANN model by iteratively comparing predicted and actual SEM images, allowing for efficient prediction and correction of circuit shapes on photomasks, reducing the need for extensive measurement derivation and mathematical parameter updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mathematical and statistical methods are used to build photomask manufacturing models, then model accuracy can be achieved, but the process becomes extremely time-consuming and requires extensive effort from multiple engineers
Solution Approach 1:
The patent replaces traditional mathematical and statistical modeling methods with a neural network-based system. The neural network automatically learns patterns from SEM images and layout data, substituting the manual mathematical modeling process with an automated learning system that requires minimal human intervention while maintaining or improving model accuracy.
Solution Approach 2:
The neural network system is self-training, automatically learning from provided data without requiring extensive manual parameter tuning or expert intervention. The system derives its own modeling parameters from training data, making the process self-service oriented and dramatically reducing the time and expertise required compared to traditional methods.
2Ease of manufacture
If traditional modeling methods are used, then existing software tools can be leveraged, but the system cannot handle unforeseen effects and requires extensive manual updates
Solution Approach 1:
The neural network dynamically adjusts its internal parameters (weights and biases) based on training data, allowing it to adapt to various manufacturing effects automatically. This parameter learning capability enables the system to handle unforeseen effects by learning their patterns from data, rather than requiring pre-programmed mathematical models for each effect.
Solution Approach 2:
The neural network provides a universal modeling framework that can handle multiple types of photomask manufacturing effects simultaneously. Rather than requiring separate specialized tools for different effects, the single neural network system can model various effects by learning from diverse training data, improving versatility while maintaining ease of use.
3Measurement precision
If SEM images are used for measurements, then accurate manufacturing data can be obtained, but the process becomes complex and costly requiring experienced engineers
Solution Approach 1:
The system performs preliminary processing of SEM images during the training phase, automatically extracting relevant features and patterns before the actual modeling occurs. This preliminary action automates what would otherwise require manual measurement and analysis by experienced engineers, reducing both complexity and cost while maintaining measurement accuracy.
Solution Approach 2:
The neural network acts as an intermediary between raw SEM images and the final manufacturing model. Instead of requiring engineers to manually interpret SEM images and derive measurements, the neural network automatically processes the images and extracts meaningful data, simplifying the measurement process while maintaining accuracy.
4Ease of manufacture
If conventional modeling approaches are used, then established methods can be followed, but scaling the process to new circuits becomes difficult and time-consuming
Solution Approach 1:
The neural network model is dynamic and can be retrained or fine-tuned for new circuits by simply providing new training data. Unlike static mathematical models that require complete redevelopment for new circuits, the neural network adapts dynamically to new manufacturing conditions and circuit types, dramatically improving scaling efficiency while maintaining ease of use through its flexible architecture.
Data Source
AI summary
An image-based Artificial Neural Networks (ANN) is used for photomask modeling, which can self-construct an internal representation of the photomask manufacturing process, therefore allowing the modeling process to become unfettered by the limitations of existing mathematical/statistical tools, thus greatly reduces/eliminates the effort needed from tedious and costly model-builders. The ANN model requires mask layout data converted into image pixel form. In ANN training phase a first circuit image and its existing SEM image are modeled via multiple layers of convolution and rectification to pick out the salient features of transformed image. In ANN testing phase, a second circuit image and its existing SEM image are compared and verified to have a difference smaller than the predetermined requirement. The satisfactory second circuit image is converted back from pixel form to circuit layout data for photomask writing.