Optical alignment correction using convolutional neural network evaluation of a beam image

Patent metadata

Field	Value
Publication number	US-10923318-B2
Application number	US-201816228201-A
Country	US
Kind code	B2
Filing date	Dec 20, 2018
Priority date	Dec 20, 2018
Publication date	Feb 16, 2021
Grant date	Feb 16, 2021

Abstract

A focused ion beam (FIB) is used to mill beam spots into a substrate at a variety of ion beam column settings to form a set of training images that are used to train a convolutional neural network. After the neural network is trained, an ion beam can be adjusted by obtaining spot image which is processed with the neural network. The neural network can provide a magnitude and direction of defocus, aperture position, lens adjustments, or other ion beam or ion beam column settings. In some cases, adjustments are not made by the neural network, but serve to indicate that the ion beam and associated ion column continue to operate stably, and additional adjustment is not required.

First claim

We claim:

1. A method, comprising: exposing a substrate to a charged particle beam (CPB) to obtain a plurality of training images, each of the training images associated with at least one CPB column characteristic; defining a neural network based on the training images so that the neural network is configured to indicate the at least one CPB column characteristic.

2. The method of claim 1 , wherein the training images are obtained by detecting CPB images with a scintillator and a photodetector array.

3. The method of claim 1 , wherein the CPB is an ion beam, the training images are training spot images produced by exposing the substrate to the ion beam, and further comprising:
segmenting the plurality of training spot images to form individual training spot images, wherein the neural network is based on the individual training spot images so that the neural network indicates at least one ion beam column characteristic.

4. The method of claim 3 , further comprising:
obtaining an exposure spot image associated with the least one characteristic of the ion beam; and
processing the exposure spot image with the defined neural network; and
adjusting the ion beam column based on the processing.

5. The method of claim 3 , wherein the neural network includes a convolution layer, and the training spot images are coupled to the convolution layer.

6. The method of claim 3 , wherein the at least one ion beam column characteristic is at least one of a lens focus, an ion beam current, a location of a beam defining aperture, or a stigmator.

7. The method of claim 3 , wherein the at least one ion beam column characteristic is associated with a setting of at least one beam shaping or beam deflection element.

8. The method of claim 7 , wherein the at least one ion beam column characteristic is an ion beam shape or an ion beam spot size.

9. The method of claim 3 , wherein the neural network includes an initial convolution layer, and one or more additional layers.

10. The method of claim 3 , wherein the training spots are arranged in an array.

11. The method of claim 10 , wherein the training spot images are H1 pixels wide by H2 pixels wide, and the convolution layer of the neural network includes N convolution kernels that map the training spot images to a J1 by J2 by N stack, wherein H1, H2, J1, J2, and N are positive integers, and J1 and J2 are less than H1 and H2, respectively.

12. The method of claim 11 , wherein J=J1=J2 and H=H1=H2, so that the convolution kernels map the J by J training spot images to a H by H by N stack.

13. The method of claim 3 , wherein the defining the neural network based on the individual training spot images comprises defining a plurality of convolution kernels.

14. The method of claim 3 , further comprising forming plurality of milled beam spots with the ion beam, wherein the training spot images are images of the milled beam spots.

15. The method of claim 3 , further obtaining the training spot images by imaging the focused ion beam.

16. A system, comprising:
a charged particle beam source;
a charged particle beam column; and
a processor coupled to the charged particle beam column, the processor coupled to process a spot image obtained with the charged particle beam source and the charged particle beam column with a neural network to determine at least one adjustment of the charged particle beam source and the charged particle beam column.

17. The system of claim 16 , further comprising a computer readable storage device having stored thereon computer-executable instructions defining the neural network.

18. The system of claim 17 , wherein the processor is further configured to adjust at least one of the charged particle beam source and the charged particle beam column based on the determined adjustment.

19. The system of claim 18 , wherein the processor is further configured to adjust the charged particle beam column based on the determined adjustment.

20. The system of claim 16 , wherein the processor is configured to receive a plurality of training spot images associated with the charged particle beam and the charged particle beam column, and define the neural network based on the training spot images.

21. The system of claim 20 , wherein the plurality of training spot images is based on exposure of a substrate to the charged particle beam.

22. The system of claim 16 , wherein the processor is coupled to adjust at least the charged particle beam column to produce the plurality of training spot images based on processing of milled beam spots on a substrate.

23. The system of claim 22 , wherein the processor is configured to produce the plurality of training spot images as an array, wherein the training spot images of the array are associated with corresponding charged particle beam column settings.

24. The system of claim 23 , wherein the processor is configured to segment the array of training spot images and provide the training spot images to the neural network as a stack of individual training spot images.

25. A method, comprising:
obtaining a plurality of images of a focused beam produced at a test substrate at corresponding focus settings;
training a convolutional neural network with the plurality of images; and
obtaining an operational image of the focused beam; and
processing the operational image of the focused beam to determine a focus adjustment.

26. The method of claim 25 , further comprising adjusting a beam focus based on the determined adjustment.

27. The method of claim 25 , wherein the beam is a charged particle beam.

28. The method of claim 25 , wherein the beam is an optical beam.

Assignees

• Fei Co

Inventors

• Gledhill Galen
• Maazouz Mostafa
• Mitchson Gavin

Classifications

• G06N3/048

  Activation functions · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title