Semiconductor Substrate Categorization Using Physics-Driven Decision Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing processes face challenges in efficiently categorizing substrates for rework or further processing, leading to significant throughput penalties and yield loss due to the reliance on statistical control techniques that are blind to the physics of the process and require sensitive data sharing between manufacturers and users.
Innovation Solution
A physics-driven fault detection and classification method that uses scanner data to derive categorical indicators based on functional indicators, allowing for decision-making without sharing sensitive information, and employs machine learning techniques to train models that adapt to the specific process conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical control techniques are used for substrate categorization, then decision-making can be performed, but false positives and negatives increase and prediction accuracy decreases
Solution Approach 1:
The patent replaces statistical control techniques with a physics-driven model that incorporates domain knowledge of semiconductor manufacturing processes. This substitution enables more accurate predictions by using physically meaningful parameters and relationships rather than purely statistical correlations, thereby reducing false positives and negatives while improving prediction accuracy.
Solution Approach 2:
The patent transforms the categorization approach by changing from statistical parameters to physics-based parameters. The decision model uses functional indicators derived from process physics and equipment measurements, fundamentally altering the parameter space to achieve better prediction reliability and reduce information loss.
2Manufacturing precision
If tight control loops based on optical measurements are implemented, then pattern reproduction quality improves, but process complexity increases
Solution Approach 1:
The patent extracts and utilizes measurements and data already generated during normal semiconductor manufacturing operations. By leveraging existing process data and measurements rather than adding new complex measurement systems, the method maintains high manufacturing precision while avoiding increased process complexity.
Solution Approach 2:
The decision model uses data and measurements that are already being collected by the manufacturing system for other purposes. The system essentially serves itself by repurposing existing data streams for quality assessment and substrate categorization, eliminating the need for additional complex measurement infrastructure.
3Reliability
If rework decisions are made to avoid yield loss, then substrate quality improves, but throughput decreases due to significant throughput penalty
Solution Approach 1:
The patent performs substrate categorization and quality assessment before final processing decisions are made. By using physics-driven predictions to identify substrates likely to fail early in the process, the system enables targeted rework only for problematic substrates rather than blanket rework, thereby maintaining substrate quality while preserving throughput.
Solution Approach 2:
The decision model incorporates feedback from process measurements and functional indicators to dynamically adjust substrate categorization and rework decisions. This feedback mechanism enables real-time quality assessment and targeted intervention, improving substrate quality outcomes while minimizing unnecessary rework that would reduce throughput.
Data Source
AI summary
A method for categorizing a substrate subject to a semiconductor manufacturing process including multiple operations, the method including: obtaining values of functional indicators derived from data generated during one or more of the multiple operations on the substrate, the functional indicators characterizing at least one operation; applying a decision model including one or more threshold values to the values of the functional indicators to obtain one or more categorical indicators; and assigning a category to the substrate based on the one or more categorical indicators.