Machine learning based image generation for model base alignments
A method for training a machine learning model to generate a predicted measured image, the method including obtaining (a) an input target image associated with a reference design pattern, and (b) a reference measured image associated with a specified design pattern printed on a substrate, wherein the input target image and the reference measured image are non-aligned images; and training, by a hardware computer system and using the input target image, the machine learning model to generate a predicted measured image.
1 . A method comprising:
obtaining a measured image associated with a design pattern on a substrate;
predicting, by a machine learning model, a predicted measured image from a target image associated with the design pattern;
aligning the measured image with the predicted measured image based an offset between the measured image and the predicted measured image; and
configuring, based on the alignment of the measured image with the predicted measured image, a computerized simulation model, separate from the machine learning model, that is configured to predict a characteristic associated with manufacture or measurement of a physical structure corresponding to the design pattern.
2 . The method of claim 1 , wherein the aligning the measured image includes determining the offset based on a cross correlation of the measured image and the predicted measured image.
3 . The method of claim 2 , wherein the determining the offset includes determining co-ordinates associated with the predicted measured image where a measure of the correlation exceeds a specified threshold.
4 . The method of claim 1 , wherein the offset is determined based on a normalized cross correlation of the measured image and the predicted measured image in a frequency domain.
5 . The method of claim 3 , wherein the determining the offset further includes determining the offset based on a normalized cross correlation of the measured image and the predicted measured image in a frequency domain.
6 . The method of claim 1 , wherein the obtaining the predicted measured image from the machine learning model includes:
obtaining a plurality of image pairs, wherein each image pair includes a) a training target image associated with the design pattern, and b) a training measured image that is obtained from an image capture device and aligned with the training target image, and
training the machine learning model by using the plurality of image pairs as training data.
7 . The method of claim 6 , wherein the training includes using the plurality of image pairs such that a cost function associated with the generation of the predicted measured image improves, the cost function determining a difference between the predicted measured image and the measured image.
8 . The method of claim 7 , wherein the training the machine learning model is an iterative process, wherein an iteration includes:
determining model parameters such that the cost function indicative of a difference between a training predicted measured image and the training measured image is minimized, wherein the training predicted measured image is predicted based on the training target image.
9 . The method of claim 7 , wherein the training the machine learning model includes:
determining the cost function, and
retraining the machine learning model with a second plurality of image pairs until the cost function is below a specified threshold.
10 . The method of claim 1 , wherein the machine learning model includes a convolutional neural network (CNN) model.
11 . The method of claim 1 , wherein the measured image is obtained via a metrology tool.
12 . The method of claim 11 , wherein the metrology tool is a scanning electron microscope (SEM) and the measured image is a SEM image, and wherein the predicted measured image is a predicted SEM image.
13 . The method of claim 1 , further comprising obtaining the target image from Graphic Database Systems (GDS) polygons associated with the design pattern.
14 . The method of claim 1 , further comprising performing a second alignment of the measured image with the predicted measured image by aligning a first contour associated with the measured image with a second contour associated with the predicted measured image.
15 . A non-transitory computer-readable medium having instructions therein, the instructions, when executed by a computer system, configured to cause the computer system to at least:
obtain a measured image associated with a design pattern on a substrate;
predict, by a machine learning model, a predicted measured image from a target image associated with the design pattern;
align the measured image with the predicted measured image based an offset between the measured image and the predicted measured image; and
configure, based on the alignment of the measured image with the predicted measured image, a computerized simulation model, separate from the machine learning model, that is configured to predict a characteristic associated with manufacture or measurement of a physical structure corresponding to the design pattern.
16 . The computer-readable medium of claim 15 , wherein the instructions configured to cause the computer system to align the measured image are further configured to cause the computer system to determine the offset based on a cross correlation of the measured image and the predicted measured image.
17 . The computer-readable medium of claim 15 , wherein the instructions configured to cause the computer system to align the measured image are further configured to cause the computer system to determine the offset based on a normalized cross correlation of the measured image and the predicted measured image in a frequency domain.
18 . The computer-readable medium of claim 15 , wherein the instructions configured to cause the computer system to obtain the predicted measured image from the machine learning model are further configured to cause the computer system to:
obtain a plurality of image pairs, wherein each image pair includes a) a training target image associated with the design pattern, and b) a training measured image that is obtained from an image capture device and aligned with the training target image, and
train the machine learning model by using the plurality of image pairs as training data.
19 . The computer-readable medium of claim 15 , wherein the instructions are further configured to cause the computer system to obtain the target image from Graphic Database Systems (GDS) polygons associated with the design pattern.
20 . The computer-readable medium of claim 15 , wherein the instructions are further configured to cause the computer system to perform a second alignment of the measured image with the predicted measured image by alignment of a first contour associated with the measured image with a second contour associated with the predicted measured image.