Virtual Metrology Model Selection for Semiconductor Wafer Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As semiconductor feature sizes decrease, the increased need for monitoring and controlling in wafer fabrication leads to higher costs and manufacturing cycle delays, with existing virtual metrology techniques experiencing prediction errors that fail to meet tighter control specifications.
Innovation Solution
A method involving the collection of manufacturing data, optimization of functional transformations using algorithms like the Levenberg-Marquardt algorithm, and selection of the transformation with the least deviation to predict semiconductor process performance, allowing for controlled fabrication tool operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If more metrology tools and monitoring are used to maintain quality as feature sizes decrease, then manufacturing precision is improved, but device complexity and manufacturing cost increase
Solution Approach 1:
The patent creates a virtual copy of the semiconductor manufacturing process through computational modeling. Instead of using physical metrology tools to measure every wafer, the system creates a virtual model that replicates process behavior and predicts outcomes. This virtual copy allows quality control without the complexity of extensive physical measurement equipment.
Solution Approach 2:
The patent replaces physical metrology measurement systems with computational prediction systems. Instead of using mechanical/optical measurement tools to physically inspect wafers, the system uses mathematical models and algorithms to predict wafer outcomes based on process parameters, substituting mechanical measurement with computational analysis.
2Manufacturing precision
If more metrology tools and monitoring personnel are used to maintain quality as feature sizes decrease, then manufacturing precision is improved, but loss of time increases due to extended manufacturing cycle time
Solution Approach 1:
The patent performs virtual measurements and predictions before actual wafer processing is complete. By using computational models to predict outcomes in advance, the system eliminates the need for time-consuming post-processing measurements and inspections, allowing quality control decisions to be made earlier in the manufacturing cycle.
Solution Approach 2:
The virtual model allows the system to simulate and predict process outcomes without requiring physical measurement of every wafer. This copying approach eliminates time-consuming physical inspection steps while maintaining quality control, as the virtual model provides predictions instantaneously compared to physical measurement methods.
3Reliability
If existing virtual metrology modeling techniques are used, then some prediction capability is achieved, but measurement precision is insufficient to meet tighter control specifications at advanced technology nodes
Solution Approach 1:
The patent implements a dynamic model selection system that adapts to different process conditions. Instead of using a single static model, the system evaluates multiple functional transformations and selects the most appropriate model based on current process parameters and conditions. This dynamic adaptation allows the system to maintain high prediction accuracy across varying manufacturing conditions and advanced technology nodes.
Solution Approach 2:
The patent changes the parameters and structure of the modeling approach by using multiple functional transformations with different mathematical characteristics. The system varies model complexity, selection of independent variables, and functional forms to optimize prediction accuracy for specific process conditions, allowing adaptation to tighter control specifications at advanced technology nodes.
Data Source
AI summary
A method for use in semiconductor fabrication is provided that includes providing manufacturing data of a semiconductor process, providing a plurality of functional transformations, optimizing each of the functional transformations based on the manufacturing data, selecting one of the functional transformations that has a least deviation with respect to the manufacturing data, predicting performance of the semiconductor process using the selected transformation function, and controlling a fabrication tool based on the predicted performance.