Substrate Alignment Correction Using Multi-Model Sampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing substrate treatment systems face challenges in achieving precise alignment of substrates, particularly in miniaturized semiconductor devices, due to high equipment replacement costs and the need for accurate alignment using existing equipment.
Innovation Solution
A substrate treatment method and system that utilizes a bonding apparatus to align substrates by determining multiple models based on alignment error data, including full-sampling and subsampling methods, to minimize alignment errors through a series of correction models, enabling accurate alignment even with limited measurement data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple models are determined using full-sampling alignment error data to improve alignment accuracy, then alignment precision is improved, but device complexity and measurement time increase
Solution Approach 1:
The patent segments the alignment error correction process into multiple distinct models: a first model based on full-sampling alignment error data, a second model based on subsampling data, and a third model that combines both. This segmentation allows each model to handle specific aspects of alignment correction, improving overall accuracy without requiring the entire system to be overly complex. The segmentation of data processing into different sampling strategies (full-sampling vs. subsampling) enables targeted correction approaches.
Solution Approach 2:
The patent performs preliminary alignment error measurement and model determination before the actual substrate alignment process. By pre-determining the first model using full-sampling data and the second model using subsampling data, the system prepares correction parameters in advance. This preliminary action allows the bonding apparatus to apply pre-calculated correction values during the bonding process, improving alignment accuracy without adding complexity to the real-time bonding operation.
2Measurement precision
If full-sampling alignment error data is collected from all alignment marks to improve measurement precision, then alignment accuracy is improved, but measurement time and processing complexity increase
Solution Approach 1:
The patent applies partial action by using subsampling of alignment error data in the second model. Instead of requiring measurements from all alignment marks, the system selectively samples a subset of alignment error data to determine the second model. This partial sampling approach provides sufficient measurement precision for correction purposes while significantly reducing the time and resources required compared to full-sampling of all alignment marks across both substrates.
Solution Approach 2:
The patent merges full-sampling data (first model) and subsampling data (second model) to create a comprehensive third model for alignment correction. By combining the results from both sampling strategies, the system achieves high measurement precision that leverages the thoroughness of full-sampling while incorporating the efficiency of subsampling. This merging allows the system to achieve accurate alignment error characterization without the complete time cost of full-sampling alone.
3Ease of manufacture
If existing equipment is used for substrate alignment instead of replacing with new equipment, then cost is reduced, but alignment precision deteriorates
Solution Approach 1:
The patent changes the parameters of the existing bonding apparatus by determining and applying multiple correction models (first model, second model, and third model) that compensate for the limitations of the existing equipment. By modifying the alignment process parameters through these mathematical models rather than changing the physical equipment, the system achieves improved alignment accuracy while continuing to use existing cost-effective equipment. The parameter changes in the correction models effectively enhance the capability of the existing apparatus.
Solution Approach 2:
The patent creates virtual models (first model, second model, third model) that represent and compensate for the alignment characteristics of the existing equipment. These software-based correction models act as virtual copies or representations of the ideal alignment behavior, allowing the existing hardware to achieve precision comparable to or better than what might be expected from new equipment. The correction models effectively copy the desired alignment performance and apply it through computational correction.
Data Source
AI summary
A substrate treatment method includes determining a first model of an upper substrate and a lower substrate based on alignment error data taken by measuring positions of a plurality of alignment marks of each of the upper substrate and the lower substrate, determining a second model of the upper substrate and the lower substrate based on first sampling alignment error data regarding at least one alignment mark from the alignment error data, determining a third model of the upper substrate and the lower substrate based on second sampling alignment error data taken by measuring positions of the at least one alignment mark, determining a fourth model by correcting the third model based on a difference between the second model and the first model, and aligning a position of a substrate selected between the upper substrate and the lower substrate according to the fourth model.