Semiconductor Process Monitoring for Real-Time Defect Index Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor manufacturing processes, particularly those involving high-aspect ratio contacts (HARC), face challenges in detecting process defects non-destructively and in real-time, leading to inefficiencies in product development and facility ramp-up due to the inability to accurately assess hole profile deformations until after the entire process is completed.
Innovation Solution
An electronic device equipped with a communication circuit, processor, and memory that processes real-time facility data to generate factors quantifying the process state, allowing for the prediction of defect indices during the target process without destroying the product, thereby enabling immediate defect diagnosis and process improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If destructive testing or electrical property evaluation is performed on fab-out wafers to determine hole profile defects, then defect detection reliability is improved, but product loss and time delay increase
Solution Approach 1:
The system performs real-time monitoring and prediction of hole profile defects during the HARC process itself, rather than waiting until fab-out. By analyzing process data as it occurs and predicting final hole profiles using machine learning models, the system identifies defects before the wafer leaves the fabrication line, eliminating the time delay associated with post-process evaluation while maintaining detection reliability
Solution Approach 2:
The patent replaces destructive physical testing and electrical property evaluation with a non-destructive computational approach. Machine learning models predict hole profile defects by analyzing process data, substituting the need for physical destruction or electrical testing of the finished product, thereby preventing product loss while maintaining defect detection capability
2Reliability
If real-time monitoring is implemented during the target process, then defect prediction capability is improved, but system complexity increases
Solution Approach 1:
The system uses a unified machine learning framework that handles multiple defect types and process variations through a single predictive model architecture. This multi-functional approach consolidates what would otherwise require multiple separate monitoring and analysis systems, improving defect prediction capability while controlling overall system complexity through consolidation
Solution Approach 2:
The patent introduces an intermediary computational layer that processes raw process data and translates it into predictive defect information. This intermediary machine learning system acts as a mediator between the complex manufacturing process and the decision-making system, simplifying the interface and data flow while enhancing prediction capability
3Measurement precision
If the entire process is completed before defect evaluation, then comprehensive assessment is improved, but productivity decreases
Solution Approach 1:
The system performs comprehensive hole profile assessment during the HARC process itself using real-time data analysis and machine learning prediction, rather than waiting until the entire fabrication process is complete. This preliminary assessment provides comprehensive defect evaluation while enabling immediate identification of problematic wafers, thereby improving productivity by preventing downstream processing of defective items
Data Source
AI summary
An electronic device and a method of operating the same are provided. The electronic device includes a communication circuit, at least one processor, and a memory, wherein the memory stores instructions that, when executed by the at least one processor, cause the electronic device to obtain real-time facility data while a target process for a semiconductor wafer is in progress, post-process the real-time facility data, generate at least one factor that quantifies a process state for each time section of the target process based on the processed real-time facility data, and predict a defect index of the target process based on the at least one factor.