Substrate Labeling via Decision-Tree Grouping for Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, existing methods struggle to effectively predict and control post-processing data based on pre-processing data due to the complexity of relating numerous pre-processing parameters to post-processing characteristics, especially in cases with many parameters and tools, leading to inefficient process control.
Innovation Solution
A method involving a decision tree algorithm to partition pre-processing data into subsets based on parameter values, creating representative object data fingerprints and analyzing them in a spectral domain to structure the context data, allowing for accurate grouping and labeling of substrates for improved process control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If sophisticated fine-tuning steps and tight control loops are applied to achieve low k1 lithography, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex pre-processing data into multiple groups based on similarity of post-processing characteristics. By dividing the data into manageable clusters, the system can apply targeted control parameters to each group rather than managing all data uniformly, thus reducing the effective complexity of the control system while maintaining high manufacturing precision through group-specific optimizations.
2Manufacturing precision
If large amounts of pre-processing data are used to predict post-processing data, then manufacturing precision may be improved, but the reliability of prediction models decreases due to parameter complexity
Solution Approach 1:
The patent extracts and utilizes only the most relevant features and parameters from the large pre-processing data set for each group. By selecting key parameters that have the strongest correlation with post-processing outcomes within each cluster, the system builds simpler and more reliable prediction models that avoid the pitfalls of overfitting and complexity associated with using all available parameters.
Solution Approach 2:
The patent applies different control parameters and prediction models to different groups of substrates based on their specific characteristics. Rather than using a single universal model, the system tailors the prediction approach to each group's local characteristics, improving both accuracy and reliability by matching the model complexity to the specific data characteristics of each group.
3Manufacturing precision
If context-based process control is implemented at wafer level, then manufacturing precision is improved, but loss of time increases due to data processing requirements
Solution Approach 1:
The patent performs clustering and grouping of substrates based on pre-processing data before the actual lithography process. By pre-organizing the data into groups with similar characteristics ahead of time, the system reduces the computational burden during real-time processing, thereby maintaining high manufacturing precision while minimizing time loss during execution.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A method of grouping data associated with substrates undergoing a process step of a manufacturing process is disclosed. The method comprises obtaining first data associated with substrates before being subject to the process step and obtaining a plurality of sets of second data associated with substrates after being subject to the process step, each set of second data being associated with a different value of a characteristic of the first data. A distance metric is determined which describes a measure of distance between the sets of second data; and the second data is grouped based on a property of the distance metric.