Spectroscopic Metrology Spot Size Reduction via Deconvolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As semiconductor and similar industries face shrinking geometries, optical metrology devices struggle to produce measurement spots that are spatially confined to target areas, making it difficult to achieve accurate measurements, and redesigning the optical system to reduce spot size is costly and time-consuming.
Innovation Solution
The effective spot size of a spectroscopic metrology device is reduced through deconvolution of measurement spectra combined with a training spectra set, using sparse sampling and deconvolution kernel weights to produce an estimated spectrum with a smaller spot size, allowing for more precise characterization of measurement targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the optical system is redesigned to reduce the spot size, then the measurement precision is improved, but the device complexity and cost increase
Solution Approach 1:
The patent creates a virtual copy of the measurement process by acquiring spectra at multiple grid positions around the target and using deconvolution algorithms to reconstruct what the spectrum would look like if the spot size were smaller. This computational approach replicates the effect of a smaller optical spot without physically redesigning the optical system.
Solution Approach 2:
The patent replaces the mechanical/optical approach of physically reducing spot size with a computational/mathematical approach. Instead of modifying optical components to achieve a smaller spot, the system uses deconvolution of spectral data to achieve the equivalent measurement precision that would result from a smaller spot size.
2Measurement precision
If the optical system is redesigned to reduce the spot size, then the measurement precision is improved, but the time and cost of implementation increase
Solution Approach 1:
The patent performs preliminary measurements by acquiring spectra at multiple grid positions surrounding the target area before performing the actual measurement. This preparatory data collection enables the deconvolution process to later extract high-precision measurements without requiring time-consuming optical system redesign.
Solution Approach 2:
The computational deconvolution process creates a virtual measurement scenario that replicates the conditions of a smaller spot size, eliminating the need for time-consuming physical optical system redesign and validation.
3Measurement precision
If the measurement spot is reduced to fit within the target area, then the measurement accuracy is improved, but the light intensity and signal strength decrease
Solution Approach 1:
The patent combines spectral information from multiple measurement positions (the full grid scan) through deconvolution to produce an estimated spectrum that represents what would be measured with a smaller spot. This merging of multiple weaker signals recovers the measurement accuracy of a small spot while utilizing the full light collection capability of the larger actual spot.
Solution Approach 2:
The deconvolution process creates a computational copy of the ideal small-spot measurement by processing the actual large-spot measurements, thereby achieving small-spot accuracy without sacrificing the light intensity advantages of the larger physical spot.
Data Source
Figure 1~2
Figure 3A~4
Figure 5A~5B
AI summary
The effective spot size of a spectroscopic metrology device is reduced through deconvolution of a measurement spectra set acquired from a measurement target combined with a training spectra set obtained from a training target. The measurement spectra set may be obtained using sparse sampling of a grid scan of a measurement target. The training spectra set is obtained from a grid scan of a training target that is similar to the measurement target. The training spectra set and the measurement spectra set include spectra from different grid nodes. Deconvolution of the measurement spectra and the training spectra sets produces an estimated spectrum for the measurement target that is an estimate of a spectrum from the measurement target produced with incident light having an effective spot size that is smaller than the actual spot size. One or more characteristics of the measurement target may then be determined using the estimated spectrum.