Semiconductor Wafer Screening by Positional Variation Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor screening methods struggle to accurately detect defective products due to variations in physical characteristics caused by manufacturing conditions across a wafer, leading to false identification of non-defective products as defective.
Innovation Solution
A method that involves measuring electrical characteristic values of semiconductor devices, acquiring positional information, subtracting gently varying components related to the wafer surface, generating corrected distributions, and detecting outliers based on these post-correction values to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If physical quantities of semiconductor devices are measured and distributions are inspected to identify outliers, then defective products can be detected, but position-dependent variations in physical quantities cause false identification of non-defective products as defective
Solution Approach 1:
The patent segments the physical quantity variations into two distinct components: position-dependent variations (systematic gradients across the wafer) and position-independent variations (random defects). By separating these components through statistical analysis and spatial decomposition, the method can selectively eliminate position-dependent variations while preserving position-independent defect signals, thereby resolving the contradiction between detection reliability and measurement precision.
Solution Approach 2:
The patent transforms the physical quantity data by applying parameter changes - specifically, it converts raw physical quantity measurements into corrected values by subtracting position-dependent components. This parameter transformation allows the system to maintain sensitivity to actual defects while becoming insensitive to position-related variations, thus improving both detection accuracy and measurement precision simultaneously.
2Reliability
If variations in physical quantities are used to identify defective products, then defects can be detected, but position-dependent variations expand the distribution width making outlier detection difficult
Solution Approach 1:
The patent segments the total variation into position-dependent and position-independent components using statistical methods. By creating separate analysis streams for each component type, the system can analyze position-independent variations (defects) with reduced distribution width, while position-dependent variations are handled separately through spatial modeling, thus simplifying the overall detection complexity.
Solution Approach 2:
The patent performs preliminary action by pre-processing the physical quantity data to remove position-dependent variations before conducting outlier detection. This preliminary correction step narrows the distribution width of the corrected values, making subsequent outlier detection more effective and reducing the complexity of the main detection algorithm.
3Reliability
If all semiconductor devices on a wafer are screened using outlier detection, then comprehensive defect detection is achieved, but position-dependent variations cause non-defective devices to be misidentified
Solution Approach 1:
The patent applies preliminary action by pre-correcting all physical quantity measurements to remove position-dependent variations before performing outlier detection. This correction step is applied systematically to all devices on the wafer, ensuring that subsequent screening is based on corrected values that reflect true device characteristics rather than position-related artifacts, thereby maintaining high screening accuracy across all devices.
Solution Approach 2:
The patent transforms the screening process by changing the parameter basis from raw physical quantities to corrected physical quantities. This parameter change allows comprehensive screening of all devices while eliminating false positives caused by position-dependent variations, thus maintaining both high accuracy and efficient throughput across the entire wafer population.
Data Source
AI summary
This invention is to detect defective products of semiconductor devices with high accuracy even when the characteristics of the semiconductor devices vary according to their positions on each of wafers. A screening method includes the steps of measuring respective electrical characteristic values of a plurality of semiconductor devices included in a wafer; acquiring respective positional information of the semiconductor devices over the wafer; subtracting components relatively gently varying over the surface of the wafer, of variations in the electrical characteristic values of the semiconductor devices from the respective electrical characteristic values of the semiconductor devices to thereby correct the respective electrical characteristic values of the semiconductor devices; generating distributions of the post-correction electrical characteristic values with respect to the semiconductor devices; and detecting semiconductor devices in which the post-correction electrical characteristic values assume outliers, out of the semiconductor devices, based on the distributions.