SiC Substrate Scratch Detection via Contrast Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for manufacturing silicon carbide (SiC) substrates fail to effectively distinguish between harmless and harmful scratches, leading to reduced yield and potential device failures due to incorrect classification of substrates as defective or non-defective.
Innovation Solution
A method involving chemical mechanical polishing (CMP) treatment followed by image capture and analysis to detect scratches with high contrast values, determining substrates as good only if scratch lengths meet specific criteria relative to substrate diameter and device chip size, thereby identifying and preventing epitaxial defects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If detection sensitivity of scratches is increased to identify all potential defects, then more harmful scratches are detected, but SiC substrates with harmless scratches are misclassified as defective, reducing yield
Solution Approach 1:
The invention applies different evaluation criteria to different scratches based on their local characteristics. Specifically, scratches are classified by their contrast values in captured images, with high contrast value scratches (≥threshold) considered harmful and low contrast value scratches considered harmless. This local differentiation allows precise identification of truly harmful defects while preserving harmless substrates, resolving the contradiction between detection sensitivity and yield.
2Productivity
If detection sensitivity is lowered to increase yield, then more substrates are classified as good, but substrates with harmful scratches are misclassified as good, causing device failure
Solution Approach 1:
The invention changes the detection parameter from simple scratch presence to scratch contrast value. By capturing images and calculating contrast values between scratch regions and surrounding areas, the system can distinguish harmful scratches (high contrast) from harmless ones (low contrast). This parameter transformation enables reliable identification of harmful defects while maintaining high yield through the formula-based acceptance criterion.
3Reliability
If all scratches are treated as defective regardless of their harmfulness, then device failure is prevented, but yield is significantly reduced due to misclassification of harmless scratches
Solution Approach 1:
The invention segments the scratch population into harmful and harmless categories based on contrast value thresholds. By dividing scratches into these distinct groups and applying different handling rules (rejection for high contrast, acceptance for low contrast), the system prevents device failure from harmful scratches while maintaining high yield by not rejecting harmless ones. This segmentation resolves the contradiction between reliability and productivity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the accurate identification of harmful scratches, reducing the misclassification of substrates and preventing yield reduction and device failures by establishing a standard for acceptable scratch lengths, ensuring only problematic substrates are flagged as defective.
Implementation Method 1
performing a CMP treatment on an SiC substrate
Implementation Method 2
after the CMP treatment, capturing an image of a surface of the SiC substrate to detect a scratch
Data Source
AI summary
A method for manufacturing an SiC substrate includes: performing a CMP treatment on an SiC substrate; after the CMP treatment, capturing an image of a surface of the SiC substrate to detect a scratch; determining the SiC substrate as a good article when a length L of the scratch having a contrast value equal to or larger than a threshold value is not more than π(D/2)2/A×F/100, wherein the scratch having the contrast value equal to or larger than the threshold value in the image serves as a starting point of an epitaxial defect, a diameter of the SiC substrate is represented by D, a length of a long side of a device chip to be formed on the SiC substrate is represented by A, and an allowable defective rate caused by scratches is represented by F.