Neural-Network Spectral Monitoring for CMP Endpoint Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Chemical mechanical polishing (CMP) processes face challenges in achieving consistent material removal rates due to variations in substrate thickness, slurry distribution, polishing pad conditions, and load, making it difficult to determine the polishing endpoint and achieve desired profiles.
Innovation Solution
An in-situ optical monitoring system that uses an artificial neural network to process measured spectra of light reflected from the substrate, reducing dimensionality and generating characterizing values to adjust processing parameters and determine endpoint detection, thereby improving endpoint detection reliability and reducing within-wafer and wafer-to-wafer non-uniformity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If in-situ optical monitoring is used to measure substrate thickness during CMP, then endpoint detection capability is improved, but system complexity and data processing requirements increase
Solution Approach 1:
The patent extracts only the essential spectral information needed for thickness measurement by reducing the full spectrum dimensionality. Principal component analysis identifies and extracts the most significant spectral features while discarding redundant information, simplifying the data processing burden while maintaining measurement accuracy.
Solution Approach 2:
The patent introduces an artificial neural network as an intermediary between the optical monitoring system and endpoint detection. The neural network acts as a intelligent mediator that automatically learns the complex relationship between spectral data and substrate thickness, handling the complexity of data processing while providing accurate thickness measurements.
2Productivity
If dimensional reduction of spectral data is performed using principal component analysis, then data processing speed is improved, but information loss may occur
Solution Approach 1:
The patent employs dynamic adaptation in the neural network that can adjust to variations in polishing conditions. The neural network learns from training data to dynamically identify the most relevant spectral features for thickness measurement under different polishing scenarios, optimizing the balance between data reduction and information retention.
Solution Approach 2:
The patent changes the parameter representation of spectral data from the original high-dimensional wavelength space to a lower-dimensional principal component space. This parameter transformation maintains the essential information needed for thickness measurement while reducing computational complexity, with the neural network compensating for any information loss through learned relationships.
3Reliability
If artificial neural network is used to generate characterizing values from spectral data, then endpoint detection reliability is improved, but computational requirements and processing time increase
Solution Approach 1:
The patent performs preliminary training of the neural network offline using training spectra obtained during polishing. This preliminary action allows the neural network to learn the complex mapping between spectral features and substrate thickness beforehand, so that during actual polishing operations, the network can quickly process spectral data in real-time without extensive computational delays.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution enables more accurate and rapid thickness measurement of substrate layers, reduces non-uniformity, and enhances the reliability of endpoint detection in CMP processes.
Implementation Method 1
measuring during the processing with an in-situ optical monitoring system a measured spectrum of light reflected from the substrate undergoing processing
Data Source
AI summary
A method of processing a substrate includes subjecting a substrate to processing that modifies a thickness of an outer layer of the substrate, measuring a spectrum of light reflected from the substrate during processing, reducing the dimensionality of the measured spectrum to generate a plurality of component values, generating a characterizing value using an artificial neural network, and determining at least one of whether to halt processing of the substrate or an adjustment for a processing parameter based on the characterizing value. The artificial neural network has a plurality of input nodes to receive the plurality of component values, an output node to output the characterizing value, and a plurality of hidden nodes connecting the input nodes to the output node.