KR-102965367-B1 - Artificial Intelligence-Enabled Process End-Pointing

Abstract

A method and system for implementing an artificial intelligence-based preparation termination instruction are disclosed. An exemplary method comprises the steps of acquiring an image of one surface of a sample comprising a plurality of features, analyzing the image to determine whether a termination point has been reached based on a feature of interest among the plurality of features observable in the image, and removing a layer of material from the surface of the sample based on whether a termination point has not been reached.

Inventors

• 밀러 토마스 개리
• 플라나간 4세 존 에프.
• 루스 주니어 브라이언
• 영 리차드
• 라슨 브래드
• 쉬로트르 아디테

Assignees

• 에프이아이 컴파니

Dates

Publication Date

20260513

Application Date

20231123

Priority Date

20190510

Claims (20)

1. A step of acquiring an image of the surface of a sample including a plurality of feature parts by means of a charged particle beam; A step of determining the relative position of the milled surface of the sample to a feature of interest among a plurality of feature parts observable in the image by analyzing the image using a machine learning system; A method comprising the step of controlling the milling of the surface of the sample based on the relative position above.
2. In claim 1, the machine learning system is an artificial neural network, method.
3. A method according to paragraph 2, wherein the artificial neural network is a convolutional neural network, a multi-layer perceptron, or a recurrent neural network.
4. In claim 1, the step of determining the relative position by analyzing the image by a machine learning system is: A step of determining the relative rotation of the milled surface with respect to the feature of interest; and A method comprising the step of tracking the amount of relative rotation across a plurality of acquired images of the surface of the sample.
5. In claim 4, the step of controlling the milling of the surface of the sample comprises the step of controlling the milling to minimize or eliminate the amount of relative rotation.
6. In claim 1, the step of determining the relative position by analyzing the image by the machine learning system is: A step of determining a first class probability and a second class probability for at least one of the plurality of features by the machine learning system; and A method comprising the step of determining the rotation of the sample from the imaged surface based on the first class probability and the second class probability.
7. A method according to claim 6, wherein the step of determining the rotation of the sample from the imaged surface based on the first class probability and the second class probability comprises the step of determining the rotation based on the oscillation of the first class probability and the second class probability across a plurality of images of the surface of the sample.
8. In claim 6, the step of controlling the milling of the surface of the sample comprises the step of limiting the surface area of the milled sample.
9. A method according to claim 1, wherein the plurality of feature parts are circuit structures.
10. A method according to claim 1, wherein the sample is a lamella.
11. As a charged particle microscope system for processing samples, An ion column that provides a focused ion beam; An electron column providing an electron beam; and Includes a controller containing code, When the above code is executed, the above charged particle microscope: By the above electron beam, an image of the surface of a sample including a plurality of feature parts is obtained, and The machine learning system analyzes the image to determine the relative position of the milled surface of the sample with respect to a feature of interest among a plurality of feature parts observable in the image, and Based on the above relative position, to control the milling of the surface of the sample by the focused ion beam, The system that causes it.
12. In Clause 11, the machine learning system is a system that is an artificial neural network.
13. In paragraph 12, the artificial neural network is a system that is a convolutional neural network, a regional convolutional neural network, a full convolutional neural network, a multilayer perceptron, or a recurrent neural network.
14. In paragraph 11, the code that causes the charged particle microscope to determine the relative position by analyzing the image by the machine learning system is such that, when executed, the charged particle microscope: Determining the relative rotation of the milled surface with respect to the above feature of interest; To track the amount of relative rotation across a plurality of acquired images of the surface of the sample, A system containing additional code that causes.
15. In claim 14, the code causing the charged particle microscope to control the milling further comprises code causing the charged particle microscope to control the milling to minimize or eliminate the relative amount of rotation when executed, a system.
16. In paragraph 11, the code that causes the charged particle microscope to analyze the image and determine the relative position by the machine learning system, when executed, causes the charged particle microscope: The machine learning system above determines a first class probability and a second class probability for at least one of the plurality of features; To determine the rotation of the sample from the imaged surface based on the above first class probability and second class probability, A system containing additional code that causes.
17. A system according to claim 16, wherein the code causing the charged particle microscope to determine the rotation further comprises code causing the charged particle microscope to determine the rotation based on the oscillation of the first class probability and the second class probability for a plurality of images of the surface of the sample during execution.
18. A system according to claim 16, wherein the code causing the charged particle microscope to control the milling further includes code causing the charged particle microscope to limit the surface area of the milled sample during execution.
19. In Clause 11, the above plurality of feature parts are circuit structures, a system.
20. In paragraph 11, the above sample is a system that is a lamella.

Description

Artificial Intelligence-Enabled Process End-Pointing The present invention generally relates to artificial intelligence (AI)-based process control, and specifically to AI-based preparation end-pointing for use in sample preparation in a charged particle microscope. In many fields of industry and research, the analysis and measurement of small structures are performed for product/process development, quality control, medical evaluation, and the like. Such analysis and measurement can be carried out using various types of inspection tools that may involve forming images of one or more structures of interest. For example, in the semiconductor industry, charged particle microscopes are used to image circuit structures at the nanometer scale, which typically forms the basis of analysis and measurement operations. In these examples, before measurements are performed, images of the circuit structures in question must be acquired from portions of the wafer or chip where the circuit components are located and removed from them. However, such removal usually requires a highly skilled operator to determine the location to remove the portion containing the features of interest. After the removal of the portion, it can usually be subjected to additional processing, such as thinning, to ensure that the desired circuit structures are visible for subsequent imaging, for example, with a transmission electron microscope. While parts of this processing can be automated, additional processing is difficult to automate due to changes in circuit structure geometry and layout that render conventional pattern recognition unreliable when the technology is not suitable. Over the years, there have been many attempts to standardize and/or automate these processes, but these attempts have generally failed to provide the desired results and still require highly skilled technicians. As such, more robust automated control of at least the additional processing required to form these imaging structures, e.g., lamellae, is required. A method and system for implementing an artificial intelligence-based preparation termination instruction are disclosed. An exemplary method comprises the steps of: acquiring an image of one surface of a sample comprising a plurality of features; analyzing the image to determine whether a termination point has been reached based on a feature of interest among the plurality of features observable in the image; and removing one layer of material from the surface of the sample based on the fact that a termination point has not been reached. An exemplary system may be a charged particle microscope comprising an ion column for providing a focused ion beam, an electron column for providing an electron beam, and a controller. The controller may include code that, when executed by the controller, causes the charged particle microscope to acquire an image of one surface of a sample comprising a plurality of features, analyzes the image to determine whether a termination point has been reached based on a feature of interest among a plurality of features observable within the image, mills the surface of the sample with a focused ion beam to remove a layer of material based on the fact that a termination point has not been reached, and stops the removal of the material based on the fact that a termination point has been reached, or may be connected to a memory containing said code. FIG. 1 is an example of a charged particle microscope system according to one embodiment of the present disclosure. FIG. 2 is an exemplary method for determining a process end point using artificial intelligence according to one embodiment of the present disclosure. FIG. 3 is an exemplary method for training an artificial neural network used for termination point detection according to one embodiment of the present disclosure. FIG. 4 is an exemplary image sequence including associated ANN analysis of images according to one embodiment disclosed herein. FIG. 5 is an exemplary functional block diagram of a computing system in which an embodiment of the present invention can be implemented. The same reference number refers to corresponding parts across multiple points in the drawing. Embodiments of the present invention relate to AI-enhanced termination point detection. In some embodiments, the AI aspect helps determine when a desired processing termination point has been reached, which may be based on structures visible in an image. For example, after a layer of material is removed by milling one surface of a sample, the surface is imaged, and the image is analyzed by a neural network to determine whether the structure indicates a termination point for the milling process. If the neural network determines that the structure indicates a termination point, the process is terminated; otherwise, the milling process may be repeated. However, it should be understood that the method described herein is generally applicable to a wide range of different AI-en