Lithography Placement Monitoring Using Yield-Based EPE Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring edge placement error (EPE) in lithographic processes are inefficient, often requiring destructive SEM inspections and fail to accurately account for local stochastic variations, such as local overlay and CD uniformity, leading to potential device failures.
Innovation Solution
A computer program is used to reconstruct EPE by defining a statistical model based on placement error contributor parameters and yield data, using a generalized linear model to predict yield metrics, allowing for flexible inclusion of multiple contributors and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If SEM inspection is used to measure EPE, then measurement accuracy is improved, but the process becomes destructive and time-consuming
Solution Approach 1:
The patent performs EPE measurements on metrology targets before the actual lithographic patterning process. By measuring placement errors in advance using the same lithographic apparatus, the system can predict and correct EPE issues before they affect production wafers, eliminating the need for time-consuming post-processing SEM inspections
Solution Approach 2:
The patent uses metrology targets that are copies or representations of the actual device features. These targets are patterned alongside the real features and serve as proxies for EPE measurement. By measuring the metrology targets instead of the actual device features, the system obtains EPE data without requiring destructive SEM inspection of the finished device
2Device complexity
If traditional overlay and CD monitoring are used separately, then process control is simplified, but EPE prediction accuracy deteriorates
Solution Approach 1:
The patent combines multiple EPE contributor measurements (overlay, CD, LWR, LER) into a unified statistical model that predicts overall EPE. Instead of controlling each parameter separately, the system measures all contributors and uses a combined model (such as root-sum-square or machine learning approaches) to predict the net EPE effect, providing both simplified process control and accurate prediction
3Ease of operation
If global overlay and CD measurements are used, then measurement simplicity is improved, but local stochastic variations are not captured
Solution Approach 1:
The patent divides the wafer into multiple local regions and performs separate EPE contributor measurements in each region. Instead of a single global measurement, the system collects overlay, CD, LWR, and LER data from multiple locations and uses statistical models to predict local EPE for each region, capturing spatial variations that global measurements would miss
Solution Approach 2:
The patent applies different measurement and analysis approaches to different local regions based on their specific characteristics. Each region's EPE is predicted using local measurements of overlay, CD, LWR, and LER, allowing the system to account for local stochastic variations and spatial non-uniformities in the lithographic process
Data Source
Figure 1~2
Figure 3(a)~4
Figure 5(a)~5(b)
AI summary
Disclosed is a computer implemented method of determining a placement metric relating to placement of one or more features on a substrate in a lithographic process. The method comprises obtaining setup data comprising placement error contributor data relating to a plurality of placement error contributor parameters and yield data representative of yield and defining a statistical model for predicting a yield metric, the statistical model being based on a placement metric, the placement metric being a function of said placement error contributor parameters, and associated model coefficients. The model coefficients are fitted based on said setup data; and the placement metric determined from said fitted model coefficients.