Wafer Inspection Predictive Imaging Without Photoresist SEM Damage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wafer inspection methods using scanning electron microscopes (SEM) cause damage to the photoresist on the wafer, leading to altered key performance indicators and inaccurate predictive images of the etching process due to the limitations of physics-based models.
Innovation Solution
Employing a machine learning model trained with SEM images of the wafer before and after etching to generate predictive images without damaging the wafer, utilizing a convolutional neural network (CNN) to minimize SEM-induced damage and improve metrology accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a charged particle beam microscope (SEM) is used to image the wafer after photoresist development, then high resolution inspection is achieved, but the photoresist is damaged and key performance indicators are altered
Solution Approach 1:
The patent performs etching of the wafer before imaging, rather than imaging the developed wafer directly. By etching the wafer first to create a more robust structure, then imaging the etched wafer, the photoresist damage problem is avoided while still achieving high-resolution inspection of the critical features
Solution Approach 2:
The patent inverts the traditional inspection sequence by imaging the etched wafer instead of the developed wafer. This reversal allows the same high-resolution SEM imaging capability to be used while avoiding the harmful interaction between the electron beam and the sensitive developed photoresist
2Loss of information
If physics-based models are used to generate predictive images of the etching process, then process understanding is achieved, but accuracy is limited due to model simplifications
Solution Approach 1:
The patent uses machine learning models trained on actual SEM images of etched wafers to generate predictive images of what the developed wafer would look like. This creates a data-driven copy or representation of the etching process outcome, capturing real process variations and complexities that physics-based models cannot represent
Solution Approach 2:
The patent transitions from physics-based parameters to data-driven parameters by training machine learning models on actual imaging data. The model learns to map etched wafer images to predicted developed wafer images, capturing complex non-linear relationships that cannot be expressed through traditional physics model parameters
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 machine learning model allows for accurate, non-destructive predictive imaging of the wafer post-etching, enhancing inspection accuracy and throughput by avoiding SEM-induced alterations, thus improving process control and yield.
Implementation Method 1
train a machine learning model using the imaged portion of the developed wafer and the imaged segment of the etched wafer
Implementation Method 2
utilizing a convolutional neural network (CNN) to minimize SEM-induced damage and improve metrology accuracy
Implementation Method 3
electrons of a single primary electron beam, or electrons of a plurality of primary electron beams, can be focused at locations of interest of a wafer under inspection
Implementation Method 4
The primary electrons interact with the wafer and may be backscattered or may cause the wafer to emit secondary electrons
Implementation Method 5
The primary electrons interact with the wafer and may be backscattered or may cause the wafer to emit secondary electrons
Data Source
AI summary
A system and method for generating predictive images for wafer inspection using machine learning are provided. Some embodiments of the system and method include acquiring the wafer after a photoresist applied to the wafer has been developed; imaging a portion of a segment of the developed wafer; acquiring the wafer after the wafer has been etched; imaging the segment of the etched wafer; training a machine learning model using the imaged portion of the developed wafer and the imaged segment of the etched wafer; and applying the trained machine learning model using the imaged segment of the etched wafer to generate predictive images of a developed wafer. Some embodiments include imaging a segment of the developed wafer; imaging a portion of the segment of the etched wafer; training a machine learning model; and applying the trained machine learning model to generate predictive after-etch images of the developed wafer.