Wafer Metrology Sampling Plan for Faster Process Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face inefficiencies in wafer metrology sampling plans, which affect the effectiveness and efficiency of process control in semiconductor device production.
Innovation Solution
A method is introduced that defines a sampling plan with position information for a set of sample points on a process wafer, models measurement values using a wafer model, randomly selects a subset of these points, and iteratively refines the sampling plan by replacing sample points to generate a reduced set for further measurements, utilizing a phase_1 and phase_2 model to optimize the sampling process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of sample points are used in the sampling plan, then measurement precision and process control accuracy are improved, but measurement time and productivity are reduced
Solution Approach 1:
The patent extracts only the essential sample points needed for accurate process control by using iterative random sampling. The system identifies and retains a minimal subset of sample points that capture the critical process variations, removing redundant measurement points while maintaining control accuracy.
Solution Approach 2:
The patent applies partial action by using random sampling to select a sufficient but not exhaustive set of sample points. Instead of measuring all possible locations, the system uses probabilistic sampling to obtain adequate process control information with fewer measurements, balancing precision and efficiency.
2Reliability
If a large number of sample points are measured, then process control accuracy is improved, but the number of measurements and time consumption increase
Solution Approach 1:
The patent performs preliminary random sampling to identify which sample points provide the most valuable process control information. By pre-selecting an optimized subset of sample points before actual production measurement, the system prepares a efficient measurement plan that maintains reliability while reducing time consumption.
Solution Approach 2:
The system uses the measurement data itself to determine the optimal sampling plan. The iterative random sampling process allows the system to self-optimize which points to measure by analyzing the information content of measurements, eliminating the need for extensive pre-planning or external optimization.
3Loss of information
If all sample points in the sampling plan are measured, then complete process information is obtained, but measurement efficiency and resource utilization are reduced
Solution Approach 1:
The patent extracts the essential information content from a reduced set of randomly selected sample points. By using iterative sampling to identify the minimal subset of points that capture the critical process variations, the system obtains complete process information without measuring all original sample points.
Solution Approach 2:
The patent changes the sampling parameters by using random sampling with replacement and iterative refinement. This allows the system to find an optimized sampling configuration that maintains information completeness while reducing the total number of measurements required.
Data Source
AI summary
A method of manufacturing a semiconductor device includes defining a sampling plan in a process control system. Measurement values are obtained at the first number N of the sample points. The first number of measurement values are modelled using a wafer model to generate a first set of coefficients according to a reference model. A second number M of the first number N of sample points is randomly selected. The second number M of measurement values obtained at the second number M of sample points is modelled using the wafer model to generate a second set of coefficients according to a phase_1 model. One of the M sample points is randomly replaced by one of the N−M sample points to obtain a subsample. The measurement values of the subsample are modelled using the wafer model to generate a third set of coefficients according to a phase_2 model.