Neural Network Image Conversion for Semiconductor Wafer Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing miniaturization of semiconductor devices requires high-resolution imaging for accurate monitoring, which increases processing time and costs due to the need for more detailed electron beam scanning and higher resolution images, often necessitating multiple electron microscopes and longer fabrication times.
Innovation Solution
A non-transitory computer-readable medium with an image conversion model using an artificial neural network that receives low-resolution images of semiconductor wafers and converts them into high-resolution images, reducing the need for high-resolution image generation by electron microscopes, thereby decreasing processing time and costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution imaging is used to monitor shrinking semiconductor devices, then measurement precision is improved, but loss of time increases due to more extensive electron beam scanning and processing
Solution Approach 1:
The patent creates a virtual copy of the high-resolution image through super-resolution reconstruction algorithms that synthesize detailed images from multiple low-resolution measurements, avoiding the need for direct high-resolution electron beam scanning and thereby reducing processing time while maintaining measurement precision
Solution Approach 2:
The patent performs preliminary low-resolution scanning to identify regions of interest before conducting targeted high-resolution scanning only in those specific areas, reducing the overall scanning time and processing requirements while maintaining adequate measurement precision for the critical dimensions
2Measurement precision
If more frames and processing iterations are used to improve signal to noise ratio, then measurement precision is improved, but loss of time increases due to additional scanning and processing
Solution Approach 1:
The patent uses computational image processing to create enhanced signal-to-noise ratio images through frame averaging and noise filtering algorithms, achieving the desired measurement precision without requiring multiple physical scanning passes, thereby reducing total scanning time
Solution Approach 2:
The patent applies noise reduction and signal enhancement processing to only the critical regions and essential image components rather than processing the entire image at full resolution, achieving adequate signal-to-noise ratio for measurement purposes while minimizing processing time
3Measurement precision
If the size of electron beam is reduced to monitor shrinking materials, then measurement precision is improved, but productivity decreases due to increased scanning time and processing requirements
Solution Approach 1:
The patent divides the wafer scanning into multiple passes, with a first pass using larger electron beam for rapid overview scanning and a second pass using smaller electron beam for detailed critical dimension measurement only in identified regions of interest, thereby maintaining measurement precision while improving overall productivity
Solution Approach 2:
The patent performs preliminary low-magnification scanning to locate and identify critical features and defects before conducting high-magnification scanning with reduced electron beam size, ensuring that detailed measurement is performed only where necessary, thus maintaining productivity while achieving required measurement precision
Data Source
AI summary
A non-transitory computer-readable medium for monitoring a semiconductor fabrication process includes an image conversion model having an artificial neural network. The image conversion model, when executed, causes the processor to receive a first image and a second image of a semiconductor wafer. The artificial neural network is trained by inputting a dataset representing the first image and the second image, generating a conversion image of the semiconductor wafer and calibrating weights and biases of the artificial neural network to match the conversion image to the second image. A third image of the semiconductor wafer is generated based on the calibrated weights and biases of the artificial neural network. The image conversion model with the trained artificial neural network may be transmitted to another device for image conversion of low resolution images.