SEM Roughness Metrology for Low-SNR Microfabrication Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The reduction in resist film thickness in high-NA EUVL leads to reduced signal-to-noise ratio (SNR) in e-beam based metrology, hindering accurate estimation of pattern roughness parameters such as line-edge roughness (LER) and linewidth roughness (LWR), which are crucial for resist pattern performance.
Innovation Solution
A method involving generating multiple sets of frames with varying numbers of frames, estimating feature data, fitting a model equation to preliminary roughness parameters, and computing a final estimate as an asymptotic value to improve accuracy in low-SNR regimes, using techniques like spatial frequency density representation and composite image generation to reduce noise bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the resist film thickness is reduced to enable smaller feature sizes in high-NA EUVL, then the manufacturing precision is improved, but the signal-to-noise ratio in e-beam metrology deteriorates
Solution Approach 1:
The method performs preliminary metrology measurements on the pattern at multiple different focus positions before final analysis. By collecting data across a focus range and then analyzing the focus-dependent behavior, the method can extract accurate roughness parameters even when individual measurements are noisy. This preliminary multi-focus data collection enables subsequent computational separation of focus effects from true roughness signals.
Solution Approach 2:
The method changes the focus position parameter during measurement, acquiring images at multiple focus offsets. By systematically varying this parameter and analyzing how roughness measurements change with focus, the method can distinguish between apparent roughness caused by defocus and actual pattern roughness. This parameter variation approach transforms a single noisy measurement into a multi-dimensional dataset from which accurate roughness can be extracted.
2Stability of the object's composition
If the resist film thickness is reduced to avoid large aspect ratios, then the pattern stability is improved, but the metrology accuracy deteriorates due to low SNR
Solution Approach 1:
The method creates multiple copies of the pattern image at different focus positions instead of relying on a single measurement. By acquiring multiple focus copies and analyzing their collective behavior, the method can statistically separate signal from noise. The computational analysis of these multiple copies enables accurate roughness extraction even when each individual copy is noisy, effectively using redundancy to overcome the low SNR problem.
Solution Approach 2:
The method performs preliminary data collection across multiple focus positions before final roughness calculation. This preliminary multi-focus measurement campaign builds a dataset that captures the focus-dependent characteristics of the pattern. Subsequent computational analysis of this pre-collected data enables accurate roughness determination without requiring high SNR in any single measurement.
3Measurement precision
If multiple frames are averaged to improve SNR, then the measurement precision is improved, but the device complexity increases
Solution Approach 1:
The method segments the measurement process into distinct focus positions rather than averaging multiple frames at a single focus. By dividing the measurement space into multiple focus slices and analyzing each separately, the method avoids the need for complex frame-by-frame averaging while still achieving SNR improvement through the statistical power of multiple measurements. Each focus position provides independent information that contributes to the final roughness estimate.
Solution Approach 2:
The method adds the focus position dimension to the measurement space, transforming a two-dimensional image analysis problem into a three-dimensional analysis involving focus offset. By moving into this additional dimensional space and analyzing roughness as a function of focus, the method extracts more information from the same set of measurements without requiring complex processing of multiple frames at each position. The focus dimension provides a natural way to separate signal from noise.
Data Source
AI summary
A method includes generating, by a SEM, sets of frames corresponding to regions of a microfabrication pattern, for each set of frames, estimating feature data representing edge positions, linewidths, or centerline positions of one or more features of each region of the pattern, and computing a preliminary estimate of a roughness parameter from the feature data. The roughness parameter is indicative of a line edge roughness, a linewidth roughness, or a pattern placement roughness of the one or more features. The method further includes fitting a model equation to the preliminary estimates of the roughness parameter using a model parameter dependent on the number of frames of each set of frames, the model equation relating the model parameter to the roughness parameter; and computing a final estimate of the roughness parameter as an asymptotic value of the fitted model equation.