Neural-Network Eddy Current Monitoring for Thin-Layer CMP Edges
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In-situ monitoring systems, such as eddy current systems, face challenges in accurately measuring substrate thickness near the edge during chemical mechanical polishing due to signal distortions and the unreliability of neural networks trained on thick calibration substrates for thin conductive layers, leading to inaccuracies in endpoint detection and polishing parameter adjustments.
Innovation Solution
A method involving the use of a neural network trained with calibration substrates of lower conductivity materials to generate adjusted thickness values, compensating for signal distortions and improving accuracy in measuring thin conductive layers by scaling measurements based on conductivity ratios, and integrating this with real-time monitoring for endpoint detection and parameter adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If neural networks are trained on thick calibration substrates for in-situ monitoring, then the system can provide real-time thickness measurements during polishing, but the measurements become inaccurate for thin conductive layers near the substrate edge
Solution Approach 1:
The patent changes the material parameter (conductivity) of calibration substrates from high conductivity (copper) to low conductivity (titanium or aluminum). This parameter change allows the neural network to be trained on thick calibration substrates while accurately measuring thin conductive layers during polishing, resolving the contradiction between real-time monitoring capability and measurement accuracy for thin layers
Solution Approach 2:
The patent introduces low conductivity calibration substrates (titanium or aluminum) as an intermediary medium. These calibration substrates serve as a bridge between the thick calibration substrates needed for training and the thin conductive layers being measured, enabling the neural network to generalize accurately across different thicknesses and materials
2Reliability
If eddy current sensors are used for in-situ monitoring near substrate edges, then real-time endpoint detection is enabled, but signal distortions reduce measurement reliability
Solution Approach 1:
The patent changes the conductivity parameter of calibration substrates to match or be lower than the conductive layer being measured. This parameter matching reduces signal distortion effects at substrate edges, enabling reliable endpoint detection while maintaining signal accuracy for thin conductive layers
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
Enhances the accuracy of thickness measurements at the substrate edge, improves within-wafer and wafer-to-wafer non-uniformity, and enables more precise real-time profile control during polishing.
Implementation Method 1
an eddy current sensing system may be used to induce eddy currents in a conductive region on the substrate to determine parameters such as the local thickness of the conductive region
Implementation Method 2
A neutral network is trained to convert sensor measurements from the in-situ monitoring system to thickness measurements for a layer formed of the second material
Data Source
AI summary
A method of polishing a substrate includes polishing a conductive layer on the substrate at a polishing station, monitoring the layer with an in-situ eddy current monitoring system to generate a plurality of measured signals values for a plurality of different locations on the layer, generating thickness measurements the locations, and detecting a polishing endpoint or modifying a polishing parameter based on the thickness measurements. The conductive layer is formed of a first material having a first conductivity. Generating includes calculating initial thickness values based on the plurality of measured signals values and processing the initial thickness values through a neural network that was trained using training data acquired by measuring calibration substrates having a conductive layer formed of a second material having a second conductivity that is lower than the first conductivity to generated adjusted thickness values.