Dynamic Sampling Scheme Generation Model for Lithography Metrology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sampling scheme generators in lithographic processes are limited in customizing sampling locations for various patterns of measurement data, often requiring more measurements than necessary for effective process monitoring and control, leading to inefficiencies and increased costs.
Innovation Solution
A dynamic sampling scheme generation method using a trained model to infer preferred sampling schemes based on measurement data and context, with a Generative Adversarial Network (GAN) to determine whether further measurements are required, balancing the outputs of a generative and discriminative model to optimize sampling locations and reduce unnecessary measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a pre-determined discrete set of sampling schemes is used, then the sampling scheme generator is simple to operate, but it cannot truly customize sampling schemes for a wide variety of different patterns of measurement data
Solution Approach 1:
The patent transforms the static, pre-determined sampling scheme selection into a dynamic system that generates customized sampling schemes in real-time. The generative model continuously adapts to different patterns of measurement data, producing optimized sampling locations dynamically rather than selecting from fixed discrete sets. This resolves the contradiction by making the system adaptable to various data patterns while managing complexity through automated model-based generation.
Solution Approach 2:
The patent changes the parameters of the sampling scheme generation process by using a trained generative model that adjusts sampling locations based on input measurement data patterns. Instead of fixed discrete sets, the system varies sampling parameters dynamically according to the specific characteristics of the measurement data, enabling true customization while the model handles the complexity of parameter optimization.
2Measurement precision
If more sampling locations are measured to ensure adequate process monitoring, then measurement precision is improved, but metrology time increases and floor space requirements increase
Solution Approach 1:
The patent extracts only the essential sampling locations needed for adequate process monitoring by using a generative model to identify and select the minimum necessary set of sampling points. Instead of measuring all possible locations or using fixed discrete sets, the system extracts the specific subset of sampling locations that provide sufficient process monitoring quality, thereby reducing metrology time while maintaining measurement precision.
Solution Approach 2:
The patent applies partial action by measuring only the necessary portion of sampling locations rather than all possible locations. The generative model determines the optimal subset of sampling points that provide adequate process monitoring, avoiding excessive measurements. This resolves the contradiction by performing just enough measurement to achieve required precision without wasting metrology time.
3Adaptability or versatility
If a dynamic sampling scheme generator is used to select from pre-determined sets, then adaptability to different patterns is improved, but the system still requires more sampling locations than necessary
Solution Approach 1:
The patent implements self-service by using a generative model that autonomously generates optimized sampling schemes without requiring manual selection from pre-determined discrete sets. The model automatically adapts to different measurement data patterns and produces customized sampling locations, eliminating the need for operators to choose from fixed options. This resolves the contradiction by providing true adaptability while optimizing the number of sampling locations to improve measurement efficiency.
Data Source
AI summary
A method to infer a current sampling scheme for one or more current substrates is provided, the method including: obtaining a first model trained to infer an optimal sampling scheme based on inputting context and/or pre-exposure data associated with one or more previous substrates, wherein the first model is trained in dependency of an outcome of a second model configured to discriminate between the inferred optimal sampling scheme and a pre-determined optimal sampling scheme; and using the obtained first model to infer the current sampling scheme based on inputting context and/or pre-exposure data associated with the one or more current substrate.