Pattern Matching for Semiconductor Memory Devices Using Spatial Frequency Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor memory device manufacturing, the high integration and miniaturization lead to challenges in precise pattern matching between GDS and CDSEM images, resulting in low matching rates and decreased reliability and productivity due to incomplete matching algorithms and increased time for re-measurement.
Innovation Solution
A pattern matching method that performs a two-dimensional Fourier transform on GDS images to analyze low spatial frequencies, limits the X/Y range for pattern matching based on whether the pattern is repeated or non-repeated, and adjusts the matching range to improve success rates, thereby enhancing the precision of CD measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If high integration and miniaturization are implemented to increase storage capacity and operational speed, then the integration level and performance of semiconductor memory devices are improved, but the process margin is reduced and alignment precision deteriorates
Solution Approach 1:
The patent applies preliminary action by performing pattern matching between GDS layout data and CDSEM measurement images before final manufacturing decisions are made. This allows potential alignment issues to be identified and corrected in advance, preventing defects in high-integration devices where process margins are minimal.
Solution Approach 2:
The patent implements feedback by using CDSEM to measure actual pattern dimensions and positions, then comparing these measurements with the original GDS design data. This closed-loop feedback enables real-time detection of alignment deviations and process variations, allowing corrective actions to maintain precision even as integration levels increase.
2Device complexity
If conventional pattern matching methods are used without spatial frequency analysis, then the matching process is simpler, but the matching rate decreases and re-measurement time increases
Solution Approach 1:
The patent transitions from conventional real-space pattern matching to frequency-space analysis by applying two-dimensional Fourier transforms. This dimensional transformation allows the matching algorithm to operate in the spatial frequency domain, where repetitive patterns can be efficiently identified and matched based on their frequency characteristics rather than detailed spatial configurations.
Solution Approach 2:
The patent changes the parameter space by analyzing low spatial frequency components of patterns rather than comparing raw pixel data. By transforming the matching criterion from direct image comparison to frequency domain correlation, the system achieves higher matching rates for repetitive patterns while reducing computational complexity for certain pattern types.
3Area of stationary object
If the X/Y range for pattern matching is not limited, then the matching covers more area, but the time required for matching increases and success rate decreases for repetitive patterns
Solution Approach 1:
The patent segments the pattern matching process by first analyzing the spatial frequency characteristics to identify repetitive structures, then dividing the matching area into smaller regions based on the repeating unit cell. This segmentation allows the system to perform matching on representative samples rather than exhaustively searching the entire large area, significantly reducing computation time while maintaining accuracy.
Solution Approach 2:
The patent applies partial action by limiting the X/Y search range to specific regions determined by low spatial frequency analysis. Instead of performing full-area matching, the system identifies critical regions where pattern deviations are most likely to occur and concentrates matching efforts there, achieving sufficient precision without the time cost of exhaustive full-area searching.
Data Source
AI summary
A pattern matching method for use in manufacturing a semiconductor memory device increases a pattern matching rate between a GDS image and an SEM image. The pattern matching method includes extracting a scanning electron microscope (SEM) image and a graphic data system (GDS) image to perform a pattern matching; performing a two-dimensional Fourier transform (FFT) for the extracted GDS image and analyzing a low spatial frequency; deciding whether or not a pattern is a repeated pattern or non-repeated pattern by using the analyzed low spatial frequency; and limiting an X/Y range for a pattern matching when the decision result is for the repeated pattern, and then performing the pattern matching between the SEM image and the GDS image.