Reverse Arrangement Test for Non-Monotonic SPC Trends
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional trend tests in statistical process control (SPC) for semiconductor manufacturing are limited in detecting non-monotonic trends, which can lead to undetected process variations and reduced device reliability and performance.
Innovation Solution
The implementation of a Reverse Arrangement Test (RAT) to identify both monotonic and non-monotonic increasing or decreasing trends, replacing or supplementing existing SPC trend tests, with corrected tables for accumulated probability and false alarm rate criteria to enhance sensitivity and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional trend tests are used in SPC, then the testing method is simple and easy to implement, but non-monotonic trends cannot be detected leading to reduced reliability
Solution Approach 1:
The patent applies the inversion principle by reversing the traditional approach of detecting trends. Instead of checking if points follow a monotonic pattern (increasing or decreasing), the Reverse Arrangement Test counts the number of times the pattern reverses direction. By inverting the detection logic from 'looking for consistent direction' to 'counting direction changes,' the method can identify both monotonic and non-monotonic trends effectively, thereby improving device reliability without excessive complexity.
2Measurement precision
If conventional monotonic trend tests are used, then false alarm rate is controlled, but sensitivity to detect actual process deviations is reduced
Solution Approach 1:
The patent changes the parameter being measured from 'number of monotonic points' to 'number of reverse arrangements.' By transforming the detection parameter from counting consistent directional changes to counting directional reversals, the test becomes sensitive to both monotonic and non-monotonic trends. This parameter transformation maintains controlled false alarm rates while significantly improving detection sensitivity for various trend patterns in semiconductor manufacturing processes.
Data Source
AI summary
A method for manufacturing semiconductor devices or other types of devices and/or entities. The method includes providing a process (e.g., etching, deposition, implantation) associated with a manufacture of a semiconductor device/ The method includes collecting a plurality information (e.g., data) having a non-monotonic trend of at least one parameter associated with the process over a determined period. The method includes processing the plurality of information having the non-monotonic trend. The method includes detecting an increasing or a decreasing trend from the processed plurality of information having the non-monotonic trend. The method includes performing an action based upon at least the detected increasing or decreasing trend.