Semiconductor Image Simulation With ML-Based Missing Property Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in semiconductor examination is the limited availability of complete chip design specifications due to Intellectual Property concerns, leading to inaccuracies in simulating how different materials interact with light, resulting in less reliable synthetic images, which can compromise defect detection and yield.
Innovation Solution
A two-step approach involving a first ML model to estimate missing design specifications and a second ML model to generate accurate synthetic images, even when complete chip design information is not accessible, using abstract chip design data and actual scan images from various imaging configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complete chip design specifications are used for image simulation, then simulation accuracy is improved, but Intellectual Property protection is compromised
Solution Approach 1:
The patent extracts only the necessary geometric and structural information from chip design specifications while deliberately excluding sensitive material composition data and proprietary manufacturing parameters. This allows image simulation to proceed with sufficient accuracy for defect detection without exposing confidential Intellectual Property.
Solution Approach 2:
The patent introduces an intermediary processing layer that transforms complete design specifications into a sanitized representation suitable for simulation. This intermediary format preserves the geometric features needed for accurate image generation while filtering out sensitive material and process information, thus mediating between IP protection requirements and simulation accuracy needs.
2Loss of information
If incomplete design specifications are used for image simulation, then Intellectual Property is protected, but simulation accuracy deteriorates
Solution Approach 1:
The patent changes the parameter representation from complete material and geometric specifications to a simplified geometric-only format. By transforming the input parameters to include only essential structural information rather than detailed material properties, the system maintains IP protection while achieving sufficient simulation accuracy for defect detection purposes.
Solution Approach 2:
The patent applies partial action by using only the subset of design specification parameters that are necessary for image simulation accuracy, rather than requiring complete specifications. This selective approach includes geometric features, dimensions, and patterns while excluding sensitive material composition and manufacturing process data, thus achieving adequate simulation without full information exposure.
3Productivity
If traditional image simulation methods are used, then processing speed is maintained, but reliability of defect detection deteriorates due to inaccuracies
Solution Approach 1:
The patent replaces traditional physics-based image simulation mechanics with a machine learning-based approach. The ML model is trained on actual examination images and learns to generate synthetic images that accurately represent real-world variations in material properties and manufacturing imperfections, thereby improving defect detection reliability while maintaining processing efficiency through the trained model's rapid inference capability.
Solution Approach 2:
The patent performs preliminary training of the machine learning model using a dataset of actual examination images before deployment. This preliminary action allows the model to learn the complex relationships between geometric specifications and actual image appearances, including variations in material interactions with light. Once trained, the model can rapidly generate reliable synthetic images without requiring complete design specifications, thus improving both reliability and maintaining processing speed.
Data Source
AI summary
There is provided a system and method of image simulation for a semiconductor specimen. The method includes obtaining design data, and a plurality of actual images of the specimen acquired by an examination tool under a plurality of imaging configurations; processing the design data, the plurality of imaging configurations and the plurality of actual images by a first machine learning (ML) model, to obtain a set of estimated values for a set of physical properties characterizing the specimen, the physical properties being expected to result in varied image responses upon the specimen being imaged under different imaging configurations of the examination tool; and processing, by a second ML model, the design data, the set of estimated values, and a group of new imaging configurations of the examination tool, to obtain a group of synthetic images of the specimen simulating actual images acquired under the group of new imaging configurations.