Semiconductor Spectral Difference Metrology for Nanosheet Process Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing metrology techniques face challenges in accurately measuring the thickness and material properties of nanoscale semiconductor structures like nanosheet structures due to complex measurement models and lack of sensitivity, which complicates process control in advanced fabrication processes.
Innovation Solution
A spectral difference based measurement model is employed to determine changes in structural parameters by analyzing differences in spectral measurements before and after process steps, using methods such as determining intensity, harmonic signals, or Mueller Matrix elements at discrete wavelengths, and applying machine learning models for parameter estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional spectroscopic ellipsometry is used to measure nanosheet structures, then measurement capability is provided, but measurement sensitivity is insufficient and model complexity increases
Solution Approach 1:
The patent segments the measurement process into two distinct stages: (1) collecting spectral data at multiple process steps, and (2) computing differences between spectra from different process steps. This segmentation transforms the complex problem of directly measuring subtle changes in nanosheet structures into a simpler difference computation problem, thereby improving measurement sensitivity while reducing model complexity.
Solution Approach 2:
The patent extracts the critical information by computing the difference between spectra obtained at different process steps. By taking out and analyzing only the differential changes rather than the full spectral data, the method isolates the relevant measurement signals from the complex background, improving sensitivity to structural changes while simplifying the analysis model.
2Reliability
If critical process steps are densely sampled for improved process control, then process control is improved, but measurement throughput decreases
Solution Approach 1:
The patent performs preliminary spectral measurements at multiple process steps and stores the data for later difference computation. By preparing the spectral data in advance during the fabrication process and only computing differences when needed, the system enables dense sampling without compromising throughput, as the actual analysis can be performed efficiently on pre-collected data.
Solution Approach 2:
The patent creates spectral difference copies by computing the difference between spectra from different process steps. This copying approach allows the system to analyze multiple process steps simultaneously through difference computations, maintaining high measurement throughput while achieving dense sampling for improved process control.
3Measurement precision
If complex high dimensional models are used to break correlations between measurement parameters and structural parameters, then measurement accuracy is improved, but computational efficiency decreases
Solution Approach 1:
The patent extracts only the differential information by computing spectral differences between process steps. This extraction eliminates the need for complex high-dimensional models to break correlations, as the difference computation inherently isolates the relevant structural changes. The result is improved measurement accuracy achieved through simple difference arithmetic rather than computationally intensive model inversion.
Solution Approach 2:
The patent replaces the complex mechanical system of traditional spectroscopic ellipsometry analysis (which requires complex physics-based models and iterative fitting) with a simpler computational approach based on spectral difference computation. This substitution maintains measurement accuracy while dramatically improving computational efficiency and reducing analysis time.
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 enhances measurement sensitivity and reduces model complexity, enabling accurate in-line monitoring and prediction of structural changes, thereby improving process control and performance in high-throughput semiconductor manufacturing.
Implementation Method 1
SE systems illuminate a structure under measurement with polarized light. The interaction between the illumination light and the structure under measurement changes the polarization of light reflected from the structure under measurement, and the change in polarization is sensitive to film thickness and material properties.
Implementation Method 2
The interaction between the illumination light and the structure under measurement changes the polarization of light reflected from the structure under measurement
Implementation Method 3
A spectral difference based measurement model determines changes in values of one or more parameters of interest based on differences in spectra measured before and after one or more process steps
Data Source
AI summary
Methods and systems for measuring values of one or more parameters of interest, including changes in values of one or more parameters of interest, based on measured spectral differences are presented herein. A trained spectral difference based measurement model determines changes in the values of one or more parameters of interest based on a measure of differences in spectra measured before and after one or more process steps. In some examples, a measure of spectral difference is determined based on a difference in measured intensity, a difference in harmonic signal values, or a difference in value of one or more Mueller Matrix elements. A measure of spectral difference may be expressed as a set of difference values, a scalar value, or coefficients of a functional fit to difference values. A measure of spectral difference may be determined based on a weighting of spectral differences according to wavelength.