Optical CD Metrology Using ML Spectral Inference for 3D Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical critical dimension (OCD) metrology techniques face limitations in measuring small and complex three-dimensional semiconductor structures, as traditional optical systems are constrained by diffraction effects, and scanning electron microscopes struggle with shrinking device sizes and embedded features, necessitating the adoption of non-image-based OCD methods for accurate characterization.
Innovation Solution
A machine learning-based system that utilizes a computer network to process and analyze optical spectra data, employing feature extraction, regression analysis, and predictor variables to generate analytical models for OCD measurements, enabling sub-Angstrom sensitivity and fast, in-line characterization of complex features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional optical systems are used for OCD metrology, then the measurement process is straightforward, but the system is constrained by diffraction effects and cannot accurately measure small and complex three-dimensional structures
Solution Approach 1:
The patent replaces traditional image-based optical measurement systems with a non-image-based OCD metrology system that uses spectral analysis. Instead of relying on direct optical imaging constrained by diffraction limits, the system measures optical spectra (intensity, phase, polarization) as a function of wavelength and uses machine learning algorithms to extract critical dimension information, thereby overcoming the diffraction barrier and enabling accurate measurement of sub-10nm features.
Solution Approach 2:
The patent transforms the measurement approach by changing from spatial domain imaging to spectral domain analysis. The system measures optical properties (intensity, phase, polarization) across multiple wavelengths rather than capturing spatial images, and uses machine learning to map spectral parameters to critical dimension values, enabling precise measurement of features below the diffraction limit.
2Measurement precision
If scanning electron microscopes are used to measure shrinking device sizes, then image-based measurement is maintained, but the system struggles with embedded features and shrinking device sizes
Solution Approach 1:
The patent replaces electron beam-based imaging with optical spectral measurement. The non-image-based OCD system uses broadband light sources and spectral detectors to measure optical responses from semiconductor structures, enabling measurement of embedded features that are inaccessible to electron microscopes while maintaining precision for shrinking device dimensions.
Solution Approach 2:
The patent creates a universal measurement system that can handle both surface and embedded features using the same non-image-based spectral analysis approach. The machine learning model is trained on diverse structures including embedded features, allowing the system to universally measure various feature types without requiring different measurement techniques.
3Measurement precision
If non-image-based OCD methods are adopted for accurate characterization, then measurement accuracy for small features is improved, but substantial computing power and extended analysis time are required
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on extensive simulated spectral data that covers a wide range of possible semiconductor structures and critical dimension values. This pre-trained model can then rapidly predict OCD values from actual measurements without requiring extensive real-time computation, significantly reducing analysis time while maintaining high accuracy.
Solution Approach 2:
The patent uses computational models to create synthetic spectral data that copies the optical response of actual semiconductor structures. By training on these simulated spectra, the machine learning system learns to map spectral features to critical dimensions, enabling fast and accurate measurements without requiring complex real-time simulations during actual metrology operations.
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
The system provides accurate, fast, and non-destructive characterization of critical dimensions in complex three-dimensional semiconductor structures, overcoming the limitations of traditional methods by leveraging machine learning algorithms for enhanced sensitivity and precision.
Implementation Method 1
Optical critical dimension (OCD) metrology represents an inspection process that utilizes the reflection/refraction of wavelengths of light
Implementation Method 2
Optical critical dimension (OCD) metrology represents an inspection process that utilizes the reflection/refraction of wavelengths of light
Data Source
AI summary
A machine learning system and method for optical critical dimension measurement. From a training set of spectra and references, features are extracted and subjected to regression analysis to generate predictor variables. Using feature functions, inverse feature functions, a machine-learning predictor component and masks, a machine-learning optical critical dimension explainer is generated. A wafer is analyzed by metrology tools and the machine-learning predictor component calculates a critical dimension inference from measured spectra. Theoretical spectra are then generated by the predictor component based upon a modification of the critical dimension inference. The measured spectra are compared to the theoretical spectra and the fit of the measured spectra to the theoretical spectra is evaluated for acceptance. The results of the comparison and analysis is output in human readable form.