Liquid Curtain Profile Detection for Semiconductor Processing Errors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor fabrication processes face challenges in accurately detecting potential processing errors, particularly related to the uniformity of the liquid curtain profile during the dispensing of processing liquids, which can lead to uneven treatment of wafers and increased risk of defects or damage.
Innovation Solution
A system utilizing an image sensor to capture images of the liquid distribution process and a trained machine learning model to classify curtain profiles, grouping images based on these classifications, and determining potential processing errors without manual inspection, thereby reducing false positives and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual inspection is used to detect processing errors, then false positives increase and accuracy decreases, but automation reduces false positives and improves accuracy
Solution Approach 1:
The patent replaces manual inspection (mechanical/human system) with an automated image-based detection system using machine learning models. The system captures images of the liquid distribution component and uses trained models to automatically classify curtain profile integrity, eliminating human subjectivity and improving detection accuracy while reducing false positives.
Solution Approach 2:
The patent introduces an intermediary computational system that processes images between the physical liquid distribution process and the final error detection decision. The machine learning model acts as an intermediary that translates visual data into reliable error classifications, bridging the gap between physical processes and quality control decisions.
2Reliability
If automated image-based detection is implemented, then detection accuracy improves, but system complexity increases
Solution Approach 1:
The patent segments the error detection task into distinct stages: image capture, image preprocessing, feature extraction, machine learning classification, and error determination. Each stage is handled by specialized components, which improves reliability through focused optimization while managing overall system complexity through modular design.
Solution Approach 2:
The patent performs preliminary actions by training machine learning models in advance with labeled data before deployment. The models are pre-trained to recognize patterns of curtain profile integrity and damage, so that during actual operation, the system can reliably classify errors without requiring complex real-time analysis, thereby improving reliability while keeping operational complexity manageable.
3Loss of time
If continuous monitoring of liquid curtain profile is performed, then processing errors are detected earlier, but measurement precision requirements increase
Solution Approach 1:
The patent applies partial monitoring by focusing image capture and analysis only on critical regions of the liquid distribution component where curtain profile integrity is most important. Rather than measuring every aspect of the liquid flow with high precision, the system targets specific areas that most affect wafer processing, reducing measurement precision requirements while maintaining early error detection capability.
Data Source
AI summary
An exemplary method includes, capturing, during a semiconductor fabrication process performed using a semiconductor processing device including a liquid distribution component configured to dispense a liquid flowing with an intact curtain profile, first images of a view of a chamber of the semiconductor processing device. The method includes determining curtain profile classifications of the first images. A curtain profile classification of the curtain profile classifications indicates a first value indicating that an image exhibits the liquid flowing with the intact curtain profile, or a second value indicating that the image does not exhibit the liquid flowing with the intact curtain profile. The method includes determining a plurality of groups of images based upon an order of the first images and the curtain profile classifications of the first images. The method includes determining, based upon the plurality of groups, whether the semiconductor fabrication process is associated with a potential processing error.