Semiconductor Substrate Sampling Scheme for Overlay Fingerprint Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Higher order correction models for overlay error in lithographic processes require more measurements and computing power, affecting throughput and are limited by spatial resolution, while existing methods like WO2016/146217 need dense sampling across sub-fields, which is inefficient.
Innovation Solution
A method for determining a sampling scheme that combines global and local fingerprint models to reduce the number of measurements by identifying optimal sampling points, allowing for efficient control of lithographic processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If higher order correction models are used to account for non-linear distortions, then manufacturing precision is improved, but productivity deteriorates due to increased measurements and computing requirements
Solution Approach 1:
The patent segments the substrate surface into multiple regions with different fingerprint models (first fingerprint model for first region, second fingerprint model for second region). This segmentation allows each region to be modeled independently with appropriate complexity, reducing the overall computational burden while maintaining correction accuracy across the entire substrate.
Solution Approach 2:
The patent applies local quality by using different fingerprint models for different regions of the substrate. Each region receives a customized modeling approach based on its specific characteristics, allowing for accurate local correction without requiring a uniformly complex model across the entire substrate, thus reducing overall computational requirements.
2Manufacturing precision
If higher order correction models are used, then manufacturing precision is improved, but device complexity increases due to more measurements and computing power requirements
Solution Approach 1:
The patent divides the substrate into multiple regions, each with its own fingerprint model. This segmentation reduces the complexity of the overall system by breaking down the large-scale complex modeling problem into several smaller, more manageable regional modeling problems.
Solution Approach 2:
The patent applies partial action by using different orders of fingerprint models for different regions based on the specific requirements of each region. Not all regions require the same level of modeling complexity, allowing the system to apply appropriate computational effort selectively rather than uniformly across the entire substrate.
3Measurement precision
If dense sampling across sub-fields is used as in WO2016/146217, then measurement precision is improved, but productivity deteriorates due to the large number of measurements required
Solution Approach 1:
The patent segments the substrate into multiple regions and applies different fingerprint models to each segment. This approach maintains measurement precision within each region while reducing the total number of measurements needed compared to uniform dense sampling across the entire substrate.
Solution Approach 2:
The patent changes the parameters of the fingerprint models (different orders and types of models) to match the specific characteristics of different substrate regions. This parameter adaptation allows for accurate local modeling with fewer measurements by using the appropriate level of model complexity for each region's specific needs.
Data Source
AI summary
A method for determining a sampling scheme, the method including: obtaining a first fingerprint model relating to a first spatial distribution of a performance parameter over a first portion of a semiconductor substrate and a second fingerprint model relating to a second spatial distribution of the performance parameter over a second portion of the semiconductor substrate; and determining a sampling point corresponding to a measuring location on the semiconductor substrate for generating measurement data based on an expected reduction of a first uncertainty metric associated with evaluation of the first fingerprint model over the first portion and an expected reduction of a second uncertainty metric associated with evaluation of the second fingerprint model over the second portion.