Etch Endpoint Detection Using Spectral Carpet Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current etch endpoint detection techniques in semiconductor manufacturing, such as time-based and optical end-pointing, suffer from inaccuracies due to variations in etch chamber conditions and wafer processing variations, especially with shrinking feature sizes.
Innovation Solution
The method generates three-dimensional intensity surface profiles, referred to as 'carpets,' from time-series spectra data during etch processing, which are used to create a virtual carpet for real-time prediction of etch depth, enabling precise endpoint detection by correlating virtual frame numbers with measured etch depths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical end-pointing is used to monitor spectral emissions, then endpoint detection capability is provided, but measurement precision deteriorates due to spectral condition variations at different time points
Solution Approach 1:
The patent transitions from analyzing single-point spectral data to analyzing time-series spectral data represented as three-dimensional carpets (wavelength × time × intensity). This dimensional expansion allows the system to capture spectral evolution over time, providing robust endpoint detection that accounts for spectral condition variations throughout the etch process rather than relying on inconsistent single-point measurements.
Solution Approach 2:
The system performs preliminary training by processing reference carpets obtained from etch processes with known endpoints. During this training phase, the system learns to correlate carpet features with endpoint conditions, building a knowledge base that enables accurate endpoint prediction in production without requiring real-time spectral condition consistency.
2Manufacturing precision
If time-based end-pointing is used to control etch duration, then process simplicity is maintained, but manufacturing precision deteriorates due to chamber and wafer variations
Solution Approach 1:
The patent replaces time-based mechanical control with optical measurement-based control. Instead of relying on pre-calibrated time estimates that ignore chamber and wafer variations, the system uses spectral emission analysis to directly measure etch progress and determine endpoint, achieving precision adaptive to actual process conditions.
Solution Approach 2:
The system creates virtual carpets that replicate the spectral evolution patterns of reference etch processes. By comparing production carpets against these virtual models, the system can predict endpoint and compensate for chamber and wafer variations without requiring complex real-time adjustments to the etch process itself.
3Productivity
If feature sizes are reduced to increase device density, then productivity is improved, but measurement precision deteriorates due to diminished spectral signal changes at endpoint
Solution Approach 1:
The system continuously monitors spectral emissions throughout the entire etch process, building a time-series carpet that captures the gradual evolution of spectral characteristics. This continuous measurement approach allows the system to detect subtle endpoint signals from small features by analyzing the cumulative spectral trajectory rather than relying on abrupt single-point changes.
Solution Approach 2:
By expanding from single-point spectral analysis to three-dimensional carpet analysis (wavelength × time × intensity), the system gains additional temporal and spectral dimensions to work with. This allows detection of endpoint through patterns and trends across multiple wavelengths and time points, compensating for the diminished signal strength from smaller features.
Data Source
AI summary
Methods and systems for using a time-series of spectra to identify endpoint of an etch process. One method includes accessing a virtual carpet that is generated from a time-series of spectra for an etch process. A polynomial with coefficients represents the virtual carpet. The method includes processing a fabrication etch process on a fabrication wafer and generating a carpet defined from a time-series of spectra while processing the fabrication etch process. While the processing the fabrication etch process and generating the carpet, comparing portions of the carpet and the virtual carpet to identify an endpoint metric of the fabrication etch process.