Two-Step Quality Distribution Prediction for Semiconductor Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, it is challenging to predict the quality distribution of products with reduced delay due to the complexity of the manufacturing process flow and the variability of process parameters, leading to difficulties in responding to customer demands and detecting disturbances in the manufacturing environment.
Innovation Solution
A high-dimensional regression technique is used to model the final electrical characteristics of semiconductor devices based on measurement data from various manufacturing stages, enabling the prediction of quality distribution and yield, which can be applied at any intermediate stage without additional final electrical characteristic measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional quality measurement methods are used at intermediate manufacturing stages, then measurement precision is improved, but loss of time increases due to the need to wait until final electrical characteristic measurements
Solution Approach 1:
The patent creates a virtual copy of the final quality distribution by using a prediction model that replicates the relationship between intermediate measurement data and final electrical characteristics. This virtual copy allows quality assessment at intermediate stages without waiting for actual final measurements, thereby reducing time loss while maintaining measurement precision through the predictive relationship established by the model.
Solution Approach 2:
The patent performs preliminary quality prediction at intermediate manufacturing stages by establishing a prediction model beforehand that can estimate final quality distribution from intermediate measurement data. This preliminary action enables early detection of quality issues and timely responses to customer demands without having to wait until the final measurement stage.
2Manufacturing precision
If comprehensive final electrical characteristic measurements are performed, then manufacturing precision is improved, but productivity decreases due to the inability to provide timely quality information
Solution Approach 1:
The patent creates a virtual representation of final quality distribution through prediction modeling, allowing manufacturers to access quality information at intermediate stages without performing time-consuming final electrical characteristic measurements on all devices. This virtual copy enables timely quality assessment and customer demand response while maintaining the accuracy benefits of comprehensive final measurements when actually performed.
Solution Approach 2:
The patent performs preliminary quality assessment at intermediate manufacturing stages using the prediction model, enabling early identification of quality trends and potential issues. This preliminary action allows manufacturers to respond to customer demands much earlier than waiting for final measurements, thereby improving productivity while the actual final measurements continue to provide the gold standard for manufacturing precision.
3Adaptability or versatility
If frequent process recipe changes are implemented to meet customer demands, then adaptability is improved, but reliability decreases due to increased variability in manufacturing processes
Solution Approach 1:
The patent implements a feedback mechanism where the prediction model continuously monitors the relationship between process parameters and quality outcomes. When process recipes are changed to meet customer demands, the model quickly predicts the impact on quality distribution, allowing manufacturers to assess the reliability impact of these changes before implementation and make adjustments to maintain process stability while achieving the desired adaptability.
Solution Approach 2:
The patent performs preliminary prediction of quality distribution before implementing process recipe changes. This allows manufacturers to anticipate the impact of frequent recipe changes on reliability and make proactive adjustments to maintain process stability, thereby enabling adaptability to customer demands while minimizing the negative impact on reliability.
Data Source
AI summary
By performing a two-step approach for predicting a quality distribution during the fabrication of semiconductor devices, enhanced flexibility and efficiency may be accomplished. The two-step approach first models electrical characteristics on the basis of measurement data, such as inline measurement data, and, in a second step, an appropriate distribution for the electrical characteristics may be established, thereby obtaining modeled wafer sort data which may then be used for predicting a quality distribution of the semiconductor devices under consideration.