Wafer Yield Prediction Using Defect Density Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The inability to accurately predict the yield rate of new products in semiconductor integrated circuit wafer manufacturing leads to prolonged decision-making times and increased costs, due to uncertainty and subjective empirical factors in the manufacturing process.
Innovation Solution
A yield rate prediction method based on the functional relationship between random defect density and yield rate, using regression analysis and a data model established from mature product yield rate data, involving steps to select candidate reference products, obtain influence factors, and create a new product prediction model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If subjective empirical factors are used in yield rate prediction, then the prediction process is simple, but the prediction accuracy is low
Solution Approach 1:
The patent replaces subjective empirical judgment with an objective data model based on the Poisson distribution formula. The yield rate is calculated using the formula Y = e^(-λ), where λ represents defect density. This substitution of mechanical/mathematical calculation for subjective estimation directly improves prediction accuracy while maintaining computational simplicity.
Solution Approach 2:
The patent transforms the prediction approach by changing from qualitative empirical parameters to quantitative defect density parameters. By measuring and calculating actual defect density values from manufacturing data and substituting them into the Poisson formula, the system achieves accurate yield rate predictions without complex subjective judgment processes.
2Loss of time
If no yield rate prediction is performed, then decision-making is fast, but production costs increase due to uncertainty
Solution Approach 1:
The patent performs yield rate prediction in advance before production decisions are made. By calculating expected yield rates using historical defect density data and the Poisson formula, the system provides preliminary information that enables faster and more confident decision-making, reducing both time loss and costly uncertainties in production planning.
3Measurement precision
If complex prediction models are used, then prediction accuracy improves, but the model complexity increases
Solution Approach 1:
The patent extracts and isolates the key factor affecting yield rate - defect density (λ) - from complex manufacturing variables. By focusing solely on measuring and calculating defect density and substituting it into the Poisson formula Y = e^(-λ), the system achieves accurate predictions without requiring complex multi-variable models, thus maintaining simplicity while improving accuracy.
Data Source
AI summary
The invention provides a yield rate prediction method for manufacture of integrated circuit wafers, which includes the following steps: obtaining candidate reference product models; obtaining parameters of the candidate reference products; obtaining functions of candidate reference products; predicting candidate reference products; selecting a final reference product; obtaining a new product prediction model; and predicting yield rate of new product. The yield rate prediction method for manufacture of integrated circuit wafer provided by the disclosure is performed based on the functional relationship between random defect density and yield rate in wafer manufacturing. By referring to the yield rate data of mature products on the production line, establishing data model and performing regression analysis, a more accurate yield rate prediction value of a new product can be obtained, thus providing a new product yield rate prediction method for manufacture of semiconductor integrated circuit wafers.