Statistical Measurement Model for Semiconductor Metrology
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current optical metrology methods face challenges in accurately measuring complex three-dimensional semiconductor structures and multi-parameter correlations, leading to increased measurement time, computational complexity, and reduced throughput due to the need for extensive modeling and simulation, especially as semiconductor devices approach nanometer-scale dimensions.
Innovation Solution
A measurement model is created directly from measured training data, allowing for the direct calculation of process and structure parameter values from raw spectra, reducing the need for complex geometric models and simulations, and enabling faster measurement times by streamlining the modeling process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional optical metrology methods are used to measure complex three-dimensional semiconductor structures, then measurement coverage is improved, but measurement time increases significantly and throughput decreases
Solution Approach 1:
The patent pre-calculates and stores a library of synthetic spectra corresponding to various known structure parameters before actual measurement. This preliminary preparation eliminates the need for complex real-time calculations during measurement, enabling rapid comparison of measured spectra against the pre-built library to determine structure parameters, thus maintaining measurement coverage while dramatically reducing measurement time and increasing throughput
Solution Approach 2:
The patent measures optical spectra at a limited set of specific wavelengths rather than across the entire spectral range. By selecting only the most informative wavelengths for characterization, the method achieves sufficient measurement coverage for determining structure parameters while significantly reducing the data processing burden and measurement time, thereby improving throughput without sacrificing essential measurement capability
2Measurement precision
If extensive modeling and simulation are performed to characterize nanoscale structures, then measurement accuracy is improved, but computational time increases significantly
Solution Approach 1:
The patent performs all complex modeling and simulation work in advance to build a comprehensive library of synthetic spectra for various structure parameters. This preliminary computational effort captures the full complexity of light-matter interactions at nanoscale, ensuring measurement accuracy. During actual measurement, the system simply compares experimental data against this pre-computed library, reducing computational time to minimal levels while preserving the accuracy benefits of extensive modeling
Solution Approach 2:
The patent creates a digital copy of the complex physical measurement process by generating synthetic spectra that replicate what would be measured for known structure parameters. These synthetic spectra serve as virtual references that encode all the complex optical physics, allowing the system to determine structure parameters through simple pattern matching rather than repeated complex simulations, thus maintaining accuracy while eliminating computational time penalties
3Reliability
If a comprehensive library of synthetic spectra is generated for accurate parameter decoupling, then measurement reliability is improved, but the size of the library and computation time increase
Solution Approach 1:
The patent segments the comprehensive parameter space into discrete, manageable categories for structure parameters (such as critical dimension, film thickness, sidewall angle). For each parameter category, a dedicated subset of synthetic spectra is generated. This segmentation allows the system to maintain reliable parameter decoupling by having specialized spectral references for each parameter type, while keeping the overall library size and computational complexity manageable through organized modularity
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
This approach improves predictive results, reduces computation and user time, and enables efficient measurement of complex structures in less than an hour, with measurement times reduced to less than ten milliseconds per site, compared to traditional methods which can take days or weeks.
Implementation Method 1
optical metrology based techniques including scatterometry and reflectometry implementations
Implementation Method 2
spectra measurements are collected from a first plurality of measurement sites
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Methods and systems for creating a measurement model based on measured training data are presented. The trained measurement model is used to calculate process parameter values, structure parameter values, or both, directly from measured data collected from other wafers. The measurement models receive measurement data directly as input and provide process parameter values, structure parameter values, or both, as output. The measurement model enables the direct measurement of process parameters. Measurement data from multiple targets is collected for model building, training, and measurement. In some examples, the use of measurement data associated with multiple targets eliminates, or significantly reduces, the effect of under layers in the measurement result, and enables more accurate measurements. Measurement data collected for model building, training, and measurement, may be derived from measurements performed by a combination of multiple, different measurement techniques.