MetroLER is our new software product that provides the most accurate pattern roughness analysis available.
We employ unique physics-based models and algorithms (computational metrology) to calculate Line-Edge Roughness and Linewidth Roughness from SEM images using a proprietary methodology. Tests on images from both extreme ultraviolet (EUV) and 193-nm immersion processes have demonstrated significantly better accuracy than typical metrics.
MetroLER has been developed for:
- Semiconductor manufacturing process monitoring and control
- 193 process development, improvement, and control
- EUV process development
- Photoresist development, selection, and optimization
- Prediction of the device impact of roughness
- Metrology tool matching
- Lithography research
MetroLER will be released in the summer of 2017. Please contact us if you are interested in becoming an early-access partner.
Robust edge detection without filtering or smoothing
It is difficult to accurately and robustly detect edges in a noisy SEM image. That is why most SEM edge detection algorithms perform some level of smoothing or filtering (such as a Gaussian filter) to reduce the noise and make edge detection more reliable. But filtering smoothes out more than just the noise – it smoothes the rough edges of the feature as well. Thus, one is faced with an unpleasant trade-off: robust edge detection produces biased answers.
MetroLER uses a different solution: the Analytical Linescan Model (ALM). The ALM is a physics-based algorithm for edge detection that rejects noise without resorting to statistical filtering. The result is robust edge detection without modifying the true roughness characteristics of the feature edge you are detecting.
Removing SEM errors to produce unbiased roughness measurements
The noise in an SEM image adds to the roughness of the feature to produce an inflated, biased roughness measurement. MetroLER fixes this problem by measuring the noise in an SEM image and then statistically subtracting it out to produce a far more accurate and unbiased measurement. Among other benefits, this allows measurement tool matching when different tools have different levels of noise.
Another SEM error that affects roughness measurement is SEM image field distortion. Even sub-nanometer levels of field distortion over a micron field of view can cause significant increases in the measured low-frequency roughness. MetroLER provides features to help characterize and eventually subtract out systematic field distortion.
Fully characterizing roughness across the frequency spectrum allows prediction of the roughness impact on device features
Roughness measurement and characterization is based on the measurement of very long (typically about 2 μm) lines. But real device features are generally much shorter. These shorter device features suffer from two different aspects of stochastic-induced roughness: within-feature roughness and feature-to-feature roughness (called local CD uniformity, or LCDU). The Conservation of Roughness principle says the 3-sigma roughness of the very long line will be divided up between within-feature roughness and feature-to-feature roughness depending on a second important parameter of roughness: the correlation length. MetroLER uses frequency analysis to provide not just an unbiased 3-sigma roughness measurement, but also an additional measurement of the correlation length. With this, we provide predictions of the stochastic behavior of device features as well.
Software designed with semiconductor manufacturing in mind
Measurements in a semiconductor environment are used for understanding the behavior of a tool or process, optimizing that behavior, monitoring it, and eventually controlling it. MetroLER has been designed from the ground up to satisfy these multiple needs. It can serve as an engineering tool in the early days of discovery and process development, and as a manufacturing tool for process monitoring and control.
Understanding stochastic-induced roughness
Feature sizes continue to shrink faster than the roughness of their edges, so that the problem of roughness is growing significantly. Today’s best performance is not good enough for the next process technology node. Reduction of the negative consequences of stochastic-induced roughness requires a better understanding of the fundamental principles of stochastic-induced variation in lithography and etch. And better understanding begins with better metrology. Only by seeing more clearly what is happening in our patterning processes can we hope to understand what can, and can’t, be done to make things better.
Predicting the impact of stochastic-induced roughness on device features
What level of LWR is tolerable? How should stochastic effects be included in my overlay budget? How will stochastic-induced LCDU affect CD control in the upcoming process node? MetroLER, through its unbiased measurement of the full range of roughness metrics through frequency, can help to answer all of these questions, and more.
Process monitoring and control in manufacturing
As roughness effects grow in proportion to feature size, so does the need to monitor and control roughness in manufacturing. Today’s biased and limited roughness metrics provide too little information to properly monitor and control roughness at the 10-nm node and below. MetroLER subtracts out metrology tool effects and provides the needed accuracy and stability of the metrics so that production monitoring has the sensitivity needed.
Metrology tool optimization and tool matching
Metrology tool recipe optimization and tool matching are common tasks of the CD-SEM metrologist. But roughness measurements add additional challenges since SEM errors that have little or no impact on the measurement of mean CD can have a major impact on roughness measurement. The variation of SEM noise and SEM image field distortion over time and from tool to tool could have a significant impact on process monitoring of roughness metrics. MetroLER has the tools needed to measure, monitor, and subtract out these SEM measurement artifacts.
Resist characterization and optimization
Resist manufacturers must optimize their formulations with many goals in mind: resolution, process window, dose sensitivity, and low roughness. MetroLER gives resist manufacturers the information they need to perform focus-exposure matrix analysis or formulation design of experiments with improved accuracy and trust.
Process and Tool Characterization
How does roughness change from before to after etch? In a complex SADP process, what steps need more care to keep roughness low? Are various post-lithography smoothing techniques effective? How does a smoothing process step affect in-device features rather than just long line/space features? Use MetroLER to help answer these questions, and more.