Retaining Ring Classification for Uniform CMP Polishing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing chemical mechanical polishing (CMP) processes face challenges in achieving uniform polishing rates across substrates due to variations in the bottom surface profiles of retaining rings, leading to non-uniformities and increased production costs and downtime.
Innovation Solution
High-resolution measurements of retaining ring bottom surfaces using a coordinate measurement machine (CMM) are combined with unsupervised machine learning algorithms to classify retaining rings into categories based on their profiles, and a neural network model is trained to predict polishing profiles, enabling efficient detection and adjustment of polishing processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional CMP processes are used without classification, then the polishing process is simple and fast, but polishing uniformity deteriorates due to variations in retaining ring bottom surface profiles
Solution Approach 1:
The patent applies preliminary action by measuring and classifying retaining rings before they are used in polishing. The bottom surface profiles of retaining rings are measured using a coordinate measurement machine, and unsupervised machine learning algorithms classify the rings into categories based on their profiles. This preliminary classification allows for targeted polishing process adjustments to achieve uniform polishing rates across different substrate regions.
Solution Approach 2:
The patent implements feedback by using machine learning models that are trained on historical data linking retaining ring classifications to polishing outcomes. The system continuously learns from polishing results and updates its classification models, creating a closed-loop system that improves polishing uniformity over time while adapting to variations in retaining ring profiles.
2Manufacturing precision
If retaining rings with varying bottom surface profiles are used, then production throughput is maintained, but polishing quality deteriorates due to non-uniform polishing rates
Solution Approach 1:
The patent applies local quality by tailoring polishing parameters to specific regions and retaining ring types. Based on the classification of retaining rings into categories (e.g., Category 1, Category 2, Category 3) based on their bottom surface profiles, the system adjusts polishing pressure, speed, and other parameters locally for different substrate regions and ring types, ensuring optimal polishing quality for each combination while maintaining overall production throughput.
3Manufacturing precision
If machine learning classification is implemented, then polishing uniformity improves, but measurement and processing time increases
Solution Approach 1:
The patent uses copying by creating a digital representation (model) of the retaining ring's bottom surface profile through coordinate measurement. Instead of physically adjusting each ring, the system creates a digital copy of the ring's geometry, classifies it using machine learning, and uses this virtual model to determine polishing parameters, thereby reducing physical trial-and-error time while improving uniformity.
4Measurement precision
If unsupervised learning algorithms are used to classify retaining rings, then classification accuracy improves without prior labels, but computational complexity increases
Solution Approach 1:
The patent applies self-service by using unsupervised learning algorithms that automatically classify retaining rings based on their measured bottom surface profiles without requiring pre-defined categories or labeled training data. The algorithms (such as k-means clustering or hierarchical clustering) autonomously identify patterns and group rings into categories based on their geometric characteristics, enabling the system to adapt to new ring designs without manual reconfiguration while maintaining high classification accuracy.
Data Source
AI summary
A method for optimizing polishing includes, for each respective retaining ring of a plurality of retaining rings mounted on a particular carrier head, performing measurements for a bottom surface of the respective retaining ring mounted on the particular carrier head using a coordinate measurement machine and collecting a respective removal profile of a substrate polished using the respective retaining ring. A machine learning model is trained based on the measurements of the bottom surface of the retaining ring and the respective removal profiles.