CMP In-Situ Defect Analysis for Real-Time Scratch Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chemical mechanical polishing (CMP) technologies struggle to efficiently detect and classify CMP-induced defects during polishing and cleaning process defects such as scratch, fall-on, residue, and other particles on semiconductor substrates, which are identified after the CMP process, with CMP-induced defects, which are not detected in real-time, causing additional defects and reducing product yield rate.
Innovation Solution
The use of an in-situ defect analyzer (DDA) with artificial intelligence (AI)-assisted defect database to identify and classify CMP-induced defects during polishing and cleaning processes, providing immediate defect warnings and improving defect detection efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional post-process defect identification is used, then defect classification can be performed with simple equipment, but defect detection time is delayed by more than 40 hours causing additional defects
Solution Approach 1:
The patent implements real-time defect detection during the CMP process itself, performing defect identification before the substrate leaves the polishing chamber. This preliminary action eliminates the 40+ hour delay by detecting defects at the moment of occurrence, allowing immediate process adjustment to prevent additional defects.
Solution Approach 2:
The patent replaces traditional mechanical/optical inspection systems with a combination of in-situ sensors, imaging systems, and AI-based analysis. This substitution enables real-time defect detection without requiring complex manual inspection equipment or post-process handling.
2Productivity
If real-time defect detection is implemented, then defect identification speed is improved, but system complexity and cost increase
Solution Approach 1:
The patent integrates defect detection functionality into the existing CMP system structure, allowing the same system to perform both polishing and real-time quality monitoring. The in-situ detection system uses the polishing process itself (slurry application, pad contact) to enable defect visualization without requiring separate dedicated inspection equipment.
Solution Approach 2:
The system enables the CMP process to self-monitor its own quality outcomes by incorporating sensors and imaging systems that detect defects during the polishing action. The AI-based analysis automatically classifies defects and provides feedback, allowing the system to self-adjust without external intervention.
3Reliability
If in-situ defect analysis is performed, then product yield rate is improved, but manufacturing process complexity increases
Solution Approach 1:
The patent implements a closed-loop feedback system where defects detected during CMP are immediately analyzed by AI algorithms, and the results are fed back to control the polishing process in real-time. This feedback mechanism allows immediate correction of process parameters to prevent defect propagation, directly improving yield rate.
Solution Approach 2:
By performing defect analysis during the polishing process rather than after completion, the system takes preliminary action to identify and address defects before they can lead to additional manufacturing issues. This prevents the compounding of defects that occurs during the 40+ hour post-process period.
Data Source
AI summary
Embodiments of the present disclosure relate to a CMP tool and methods for planarization a substrate. Particularly, embodiments of the present disclosure relate to an in-situ defect data analyzer to identify CMP induced defects during polishing processing and cleaning processing performed in the CMP tool. In some embodiments, the CMP tool includes an AI (artificial intelligence)-assisted defect database. The defect database may be used to identify and classify CMP related defects, such as scratch, fall-on slurry residuals, during polishing or cleaning process. As a result, defect warning cycle time for a CMP process is improved significantly.