System and method for defect detection using conditionally shielded automatic encoders

By using a conditionally masked autoencoder (MAE) to segment and mask optical images, and combining them with design images to generate reference images, the problems of accuracy and computational cost in semiconductor device defect detection are solved, achieving efficient defect detection.

CN123125480APending Publication Date: 2026-11-17KLA CORP
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202580027355.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-12-23
Filing Date
2025-04-21
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to accurately predict the location, type, and structure of defects in semiconductor devices, and optical image modeling methods suffer from high computational costs and limitations in handling images of different modalities.

Method used

A conditionally masked automatic encoder (MAE) is used to selectively mask defect areas by segmenting optical images and combining them with design images to generate reference images for defect detection.

Benefits of technology

It improves the accuracy and efficiency of defect detection, can process images of different modalities, reduces computational costs, and improves downstream detection performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123125480A_ABST
    Figure CN123125480A_ABST
Patent Text Reader

Abstract

A method for defect detection includes receiving an optical image of a sample and segmenting the received optical image, comprising multiple image blocks. The method includes generating a masked optical image by selectively masking the image blocks based on characteristic data. The method includes providing the masked optical image to a first encoder. The method includes receiving a design image and providing it to a second encoder. The method includes generating a first set of feature vectors using the first encoder and generating a second set of feature vectors using the second encoder. The method includes concatenating the feature vectors from the first encoder and the feature vectors from the second encoder and generating a reference image of the sample using a decoder based on the concatenated feature vectors, wherein the generated reference image is a reconstructed image of the optical image and the design image.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Cross-reference of related applications

[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 648,172, filed May 16, 2024; and U.S. Provisional Application No. 63 / 708,759, filed October 18, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure generally relates to sample inspection, and more specifically, to systems and methods for detecting defects in samples using conditionally masked automatic encoders. Background Technology

[0004] The demand for electronic logic and memory devices with increasingly smaller footprints and features presents various manufacturing challenges beyond the expected scale of production. In the context of semiconductor manufacturing, predicting and removing defects from semiconductor devices is a crucial step in improving throughput and yield. One of the main challenges in defect identification is the inability to accurately predict the location, type, and structure of defects.

[0005] Optical image modeling is commonly used for defect identification. Existing methods for optical image modeling include physics-based methods, deep convolutional neural networks, or combinations thereof. These methods have limitations such as high computational cost and inability to handle images with different modalities (e.g., design images and inspection images).

[0006] Therefore, it is desirable to provide systems and methods for addressing one or more of the aforementioned defects. Summary of the Invention

[0007] According to one or more embodiments of this disclosure, a system for detecting defects on a sample is disclosed. In one embodiment, the system includes a controller comprising one or more processors configured to execute a set of program instructions stored in a memory, wherein a conditionally masked autoencoder is stored in the memory. In an embodiment, the set of program instructions is configured to cause the one or more processors to: receive an optical image; generate a plurality of image blocks, including the plurality of image blocks, by segmenting the received optical image; generate a masked optical image by selectively masking a first set of image blocks among the plurality of image blocks based on characteristic data, wherein the first set of image blocks corresponds to one or more defective regions, wherein the plurality of image blocks includes a second set of image blocks corresponding to one or more defect-free regions; provide the generated masked optical image to a first encoder of the conditional masking autoencoder; apply the first encoder to the second set of image blocks of the generated masked optical image to generate a first set of feature vectors; receive a design image; provide the design image to a second encoder of the conditional masking autoencoder; apply the second encoder to the design image to generate a second set of feature vectors; concatenate the first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder; and generate a reference image based on the concatenated feature vectors using a decoder of the conditional masking autoencoder, wherein the generated reference image is a reconstructed image of the optical image and the design image.

[0008] According to one or more embodiments of this disclosure, a system for detecting defects on a sample is disclosed. In one embodiment, the system includes an optical subsystem configured to acquire multiple optical images of the sample. In another embodiment, the system includes a controller communicatively coupled to the optical subsystem, wherein the controller includes one or more processors configured to execute a set of program instructions stored in a memory, and wherein a conditionally masked autoencoder is stored in the memory. In an embodiment, the set of program instructions is configured to cause the one or more processors to: receive an optical image from the plurality of optical images from the optical subsystem; generate a plurality of image blocks, including the plurality of image blocks, by segmenting the received optical images; generate a masked optical image by selectively masking a first set of image blocks among the plurality of image blocks based on characteristic data, wherein the first set of image blocks corresponds to one or more defective regions, wherein the plurality of image blocks includes a second set of image blocks corresponding to one or more defect-free regions; provide the generated masked optical image to a first encoder of the conditional masking autoencoder; apply the first encoder to the second set of image blocks of the generated masked optical image to generate a first set of feature vectors; receive a design image; provide the design image to a second encoder of the conditional masking autoencoder; apply the second encoder to the design image to generate a second set of feature vectors; concatenate the first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder; and generate a reference image based on the concatenated feature vectors using a decoder of the conditional masking autoencoder, wherein the generated reference image is a reconstructed image of the optical image and the design image.

[0009] According to one or more embodiments of this disclosure, a method for defect detection is disclosed. In one embodiment, the method includes receiving an optical image of a sample. In one embodiment, the method includes generating a plurality of image blocks by segmenting the received optical image, the plurality of image blocks comprising the plurality of image blocks. In one embodiment, the method includes generating a masked optical image by selectively masking a first set of image blocks among the plurality of image blocks based on characteristic data, wherein the first set of image blocks corresponds to one or more defect regions, and wherein the plurality of image blocks includes a second set of image blocks corresponding to one or more defect-free regions. In one embodiment, the method includes providing the generated masked optical image to a first encoder of a conditionally masked autoencoder. In one embodiment, the method includes applying the first encoder to the second set of image blocks of the generated masked optical image to generate a first set of feature vectors; receiving a design image. In one embodiment, the method includes providing the design image to a second encoder of the conditionally masked autoencoder; applying the second encoder to the design image to generate a second set of feature vectors. In one embodiment, the method includes concatenating the first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder. In one embodiment, the method includes using a decoder of the conditionally masked autoencoder to generate a reference image of the sample based on the concatenated feature vectors, wherein the generated reference image is a reconstructed image of the optical image and the design image.

[0010] It should be understood that the foregoing general description and the following detailed description are merely illustrative and explanatory and are not necessarily limiting of the invention as claimed. The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the invention and, together with the general description, serve to explain the principles of the invention. Attached Figure Description

[0011] Those skilled in the art will better understand the many advantages of this disclosure by referring to the accompanying drawings, among which:

[0012] Figure 1 A block diagram view illustrating a system for defect detection according to one or more embodiments of the present disclosure.

[0013] Figure 2 A simplified schematic view illustrating a SEM-based characterization subsystem of a system for defect detection according to one or more embodiments of the present disclosure.

[0014] Figure 3 The illustration shows a flowchart of a method for defect detection using conditional MAE according to one or more embodiments of the present disclosure.

[0015] Figure 4The illustration depicts a conceptual flowchart of a method for generating reference images using conditional MAE according to one or more embodiments of the present disclosure.

[0016] Figure 5 The illustration depicts a conceptual flowchart of a method for selective masking according to one or more embodiments of the present disclosure. Detailed Implementation

[0017] The disclosed subject matter will now be described in detail with reference to the accompanying drawings. This disclosure has been particularly shown and described with respect to certain embodiments and their specific features. The embodiments set forth herein should be considered illustrative rather than restrictive. It will be readily apparent to those skilled in the art that various changes and modifications in form and detail may be made without departing from the spirit and scope of this disclosure.

[0018] Embodiments of this disclosure relate to systems and methods for defect detection using a conditionally shielded automatic encoder (MAE). For example, the defect detection system and method may utilize a conditional MAE to generate reference images for defect detection (e.g., die-to-database (D2DB) defect detection), wherein the reference images are robust to process variations. The conditional MAE may include at least two encoders, wherein a first encoder is used for optical images (e.g., scanning electron microscope (SEM) images) and a second encoder is used for design images. The conditional MAE may utilize target shielding during the inference phase to specify which portions of the image should be generated. For example, target location information may be provided by an optical subsystem (e.g., a broadband plasma inspection subsystem) and used to specify which portions of the image should be generated and which should be considered defect-free references. In this respect, the conditional MAE may use defect-free regions to learn how to perform the design-to-SEM generation task and generate reference images corresponding to the inspection areas. The reconstructed SEM image may utilize optical images (or test images) of all unshielded blocks (e.g., defect-free regions), and shielded blocks (e.g., defective regions) may be replaced with the generated SEM data.

[0019] As considered herein, the systems and methods of this disclosure offer several advantages over prior art. For example, the systems and methods of this disclosure can take into account images with at least two modalities. By another example, the conditional MAE of this disclosure can be configured to learn to estimate color and shape variations from optical images to better perform the design-to-SEM mapping required for reference generation. By another example, the reference image generated using the systems and methods of this disclosure may differ only from the optical image at locations marked as potentially containing defects. Furthermore, the conditional MAE of this disclosure can natively utilize location filters available in the review subsystem to improve downstream defect detection performance.

[0020] For reference Figures 1 to 5The present disclosure describes in more detail a system and method for defect detection using a conditionally masked automatic encoder (MAE) according to one or more embodiments.

[0021] Figure 1 A block diagram view illustrating a system 100 for defect detection according to one or more embodiments of the present disclosure is provided. In one embodiment, system 100 includes an optical subsystem 102 and a controller 108 communicatively coupled to the optical subsystem 102.

[0022] Optical subsystem 102 can be configured to inspect and / or image a sample 104 fixed on a sample stage 106. Optical subsystem 102 can include any optical-based characterization system. For example, optical subsystem 102 can include a scanning electron microscopy (SEM) inspection system. For instance, the SEM inspection system can be configured for inspecting the optical wafers of sample 104 against a database, wherein the SEM inspection system can be configured to compare the optical image of the sample (or wafer) with a rendered image generated from the database to detect defects on the sample.

[0023] Controller 108 may include one or more processors 110 and memory 112. The one or more processors 110 may be configured to execute a set of program instructions maintained in memory 112. The one or more processors 110 of controller 108 may be configured to apply conditional MAE 113 to generate a reference image for die-to-database inspection to detect defects on sample 104. For example, conditional MAE 113 may be stored in memory 112.

[0024] As considered in this paper, Conditional MAE 113 can be any type of deep learning model. For example, Conditional MAE 113 can be a self-supervised deep learning model that includes at least two Vision Transformer (ViT) encoders and decoders. For example, at least two ViT encoders can consider images of different modalities (e.g., design images and optical (or SEM) images).

[0025] In embodiments, one or more processors 110 may be configured to receive one or more optical images 115 of sample 104. For example, one or more processors 110 may be configured to receive one or more optical images 115 from optical subsystem 102. For example, one or more processors 110 may be configured to receive one or more SEM images 115 from SEM inspection subsystem 200. For the purposes of this disclosure, "optical image," "test image," "SEM image," or variations thereof may refer to an optical image of a physical sample (e.g., a wafer) acquired using an optical subsystem (e.g., an SEM inspection system).

[0026] In an embodiment, one or more processors 110 may be configured to receive one or more design images 117 of the sample 104. The one or more design images 117 may contain design data, which typically refers to the physical design (or layout) of the sample and / or data derived from the physical design through simulation or modeling. For the purposes of this disclosure, "design image" may refer to a design image corresponding to the physical design (or layout) of a semiconductor device.

[0027] In embodiments, one or more processors 110 may be configured to selectively mask one or more optical images 115 of sample 104. For example, one or more processors 110 may be configured to decompose one or more optical images 115 into multiple blocks and selectively mask corresponding blocks that correspond to one or more defect regions. For example, one or more processors 110 may be configured to receive inspection data from an inspection subsystem, wherein the inspection data indicates the location of one or more defects. In this regard, blocks having pixels marked as defects by the inspection subsystem may correspond to defect regions, such that corresponding blocks can be selectively masked.

[0028] In embodiments, one or more processors 110 may be configured to generate a reference image using a conditional MAE. For example, one or more processors 110 may be configured to apply a first encoder of the conditional MAE to a shielding optical image and a second encoder of the conditional MAE to a design image. By another example, one or more processors 110 may be configured to map the design and optical images based on the outputs from the first and second encoders and then use a decoder to generate a reference image corresponding to the inspection area.

[0029] Sample 104 may include any sample known in the art, including but not limited to wafers, photomasks, photomasks, flat panel displays, etc. In an embodiment, sample 104 is positioned on stage assembly 106 to facilitate movement of sample 104. For example, stage assembly 106 may include an actuable stage. For example, stage assembly 106 may include, but is not limited to, one or more translation stages adapted to selectively translate sample 104 along one or more linear directions (e.g., the x-direction, y-direction, and / or z-direction). By another example, stage assembly 106 may include, but is not limited to, one or more rotation stages adapted to selectively rotate sample 104 along a rotational direction. By another example, stage assembly 106 may include, but is not limited to, rotation stages and translation stages adapted to selectively translate sample 104 along linear directions and / or rotate sample 104 along a rotational direction. It should be noted herein that system 100 may operate in any scanning mode known in the art.

[0030] Figure 2 A simplified schematic view illustrating an optical subsystem 102 according to one or more embodiments of the present disclosure.

[0031] In an embodiment, the optical subsystem 102 includes an SEM inspection subsystem 200 configured to perform inspection of the sample 104. In this regard, the SEM inspection subsystem 200 may be configured to acquire one or more optical images of the sample 104. The SEM inspection subsystem 200 may include, but is not limited to, an electron beam source 202, one or more electron optical elements 204, one or more collecting electron optical elements 206, and an electron detector assembly 208 including one or more electron sensors 210. In an embodiment, the electron beam source 202 is configured to guide one or more electron beams 201 onto the sample 104. The electron optical elements 204 may form electron optical columns. In an embodiment, the system 100 may include one or more additional and / or alternative electron optical elements 204 configured to focus and / or guide one or more electron beams 201 onto the surface of the sample 104. One or more electron optical elements 206 may be configured to collect secondary and / or backscattered electrons 203 emitted from the surface of sample 104 in response to one or more electron beams 201. It should be noted herein that one or more electron optical elements 204 and one or more electron optical elements 206 may comprise any electron optical element configured to guide, focus and / or collect electrons, including but not limited to one or more deflectors, one or more electron optical lenses, one or more condenser lenses (e.g., magnetic condenser lenses), one or more objectives (e.g., magnetic condenser lenses), etc.

[0032] It should be noted that the electro-optical assembly of the SEM inspection subsystem 200 is not limited to the components provided for illustrative purposes only. Figure 2 The electro-optical elements depicted herein. It should be further noted that system 100 may include any number and type of electro-optical elements required to guide / focus one or more electron beams 201 onto sample 104 and, in response, collect and image emitted secondary and / or backscattered electrons 203 onto electron detector assembly 208.

[0033] The SEM inspection subsystem is generally discussed in U.S. Patent No. 11,239,048, published February 1, 2022; U.S. Patent No. 11,410,830, published August 9, 2022; U.S. Patent Publication No. 2024 / 0194440, published June 13, 2024; U.S. Patent Publication No. 2022 / 0108862, published April 7, 2022; and U.S. Patent No. 11,880,193, published January 23, 2024. The entire contents of all the patents and publications mentioned are incorporated herein by reference.

[0034] Figure 3The illustration depicts a flowchart of a method 300 for defect detection according to one or more embodiments of the present disclosure. It should be noted herein that the embodiments and enabling techniques previously described in the context of system 100 should be interpreted as extensions of method 300. However, it should be further noted that method 300 is not limited to the architecture of system 100.

[0035] In an embodiment, method 300 includes step 302 of receiving one or more optical images of a sample. For example, one or more processors 110 may be configured to receive one or more optical images from optical subsystem 102 (or other systems). For example, one or more processors 110 may be configured to receive one or more SEM images from SEM inspection subsystem 200.

[0036] In an embodiment, method 300 includes a step 304 of generating a reference image of the sample using a conditional MAE, as will be discussed in more detail below. For example, one or more processors 110 may be configured to generate a reference image of the sample using a conditional MAE stored in memory 112. The conditional MAE may include at least two vision transformer encoders and decoders. For example, the at least two encoders may include at least a first encoder utilizing an optical (or SEM) image and a second encoder utilizing a design image, such that the at least two encoders take into account images of different modalities. In this respect, the resulting generated reference image (e.g., an SEM image) may be robust to process variations and generated based on a test region.

[0037] In an embodiment, method 300 includes a step 306 of performing defect detection on a sample. For example, one or more processors 110 may be configured to compare a generated reference image with an optical image to identify one or more defects on sample 104.

[0038] Figure 4 This describes a conceptual flowchart of a method 304 for generating reference images under MAE conditions according to one or more embodiments of the present disclosure.

[0039] In an embodiment, method 304 includes a step 402 of receiving one or more optical images 115. For example, one or more processors 110 may be configured to receive one or more optical images 115 of sample 104. For example, one or more processors 110 may be configured to receive one or more SEM images 115 from SEM inspection system 200. As considered herein, the optical images received in step 302 may be the same as or different from the optical images 115 received in step 402.

[0040] In an embodiment, method 304 includes a step 404 of selectively masking one or more received optical images 115. For example, one or more processors 110 may be configured to divide each optical image 115 into blocks (e.g., regular, non-overlapping blocks). Each optical image 115 may have a block size between approximately 16×16 and 32×32 pixels. For example, in a non-limiting example, each optical image 115 may have a block size of 16×16. By another example, in a non-limiting example, each optical image 115 may have a block size of 14×14. By another example, in a non-limiting example, each optical image 115 may have a block size of 28×28. By another example, in a non-limiting example, each optical image 115 may have a block size of 32×32.

[0041] In one embodiment, one or more processors 110 may be configured to selectively mask multiple blocks based on characteristic data. Figure 5 The illustration depicts a conceptual flowchart of a method for selective masking according to one or more embodiments of the present disclosure. In an embodiment, a first set of image blocks corresponding to one or more defect regions may be selectively masked. For example, one or more processors 110 may be configured to receive one or more defect markers 502 from an inspection system (e.g., a broadband plasma (BBP) inspection system), wherein the respective defect marker 502 corresponds to the location (e.g., defect region) of a defect 504 on a sample 104. The one or more processors 110 then selectively mask the corresponding blocks 506 associated with the defect markers 502. In this respect, blocks 506 having pixels marked as defect 504 may be selectively masked.

[0042] In one embodiment, method 304 includes step 406 of providing a shielded optical image to a first encoder and applying the first encoder to the shielded optical image to generate a plurality of feature vectors. For example, one or more processors 110 may be configured to apply the first encoder to a second set of image blocks corresponding to one or more defect-free regions (e.g., unshielded portions of an image). In this regard, the resulting generated reference image (in step 416) utilizes optical images (or SEM images) of all unshielded blocks, and the shielded blocks are replaced with generated SEM data.

[0043] In an embodiment, method 304 includes the step 408 of receiving one or more design images. For example, one or more processors 110 may be configured to receive one or more design images 117 of sample 104. The one or more design images 117 may contain design data, which typically refers to the physical design (or layout) of the sample and / or data derived from the physical design through simulation or modeling.

[0044] In one embodiment, method 304 includes step 410 of providing a design image to a second encoder and applying the second encoder to the design image. For example, one or more processors 110 may be configured to apply the second encoder (different from the first encoder) to the design image to generate multiple feature vectors.

[0045] In one embodiment, method 304 includes step 412 of concatenating feature vectors from the first encoder with feature vectors from the second encoder. For example, one or more processors 110 may be configured to concatenate feature vectors generated by the two encoders. For example, one or more processors 110 may be configured to generate a latent representation of the feature vectors, wherein the latent representation comprises a single matrix that combines features from the SEM and the designed feature vectors.

[0046] In one embodiment, method 304 includes step 414 of applying the decoder to the concatenated feature vector. For example, one or more processors 110 may be configured to receive the concatenated feature vector generated by two encoders and provide it as input to the decoder.

[0047] In an embodiment, method 304 includes a step 416 of generating a reference image (or a reconstructed image). For example, one or more processors 110 may be configured to generate a reference image by using optical images of all unshielded blocks and replacing the shielded blocks with SEM data generated by a decoder. For example, one or more processors 110 may use a decoder to reconstruct an SEM image by utilizing feature vectors derived from both the design and the shielded SEM. In this respect, the decoder is capable of performing accurate image reconstruction.

[0048] Refer again Figure 1 Additional components of system 100 are described in more detail according to one or more embodiments of this disclosure.

[0049] One or more processors 110 of controller 108 may generally comprise any processor or processing element known in the art. For the purposes of this disclosure, the terms “processor” or “processing element” may be broadly defined to encompass any device having one or more processing or logic elements (e.g., one or more microprocessor devices, one or more application-specific integrated circuit (ASIC) devices, one or more field-programmable gate arrays (FPGAs), or one or more digital signal processors (DSPs)). In this sense, one or more processors 110 may comprise any device configured to execute algorithms and / or instructions (e.g., program instructions stored in memory). In one embodiment, one or more processors 110 may be embodied as a desktop computer, mainframe computer system, workstation, graphics computer, parallel processor, networked computer, or any other computer system configured to execute a program configured to operate or in conjunction with system 100, as described throughout this disclosure. Furthermore, different subsystems of system 100 may include processors or logic elements suitable for performing at least a portion of the steps described in this disclosure. Therefore, the above description should not be construed as limiting the embodiments of this disclosure but is merely illustrative. Furthermore, the steps described throughout this disclosure can be performed by a single controller or alternatively by multiple controllers. Additionally, controller 108 may comprise one or more controllers housed within a common housing or multiple housings. In this way, any controller or combination of controllers can be separately packaged as modules suitable for integration into the metering system 100. Furthermore, controller 108 can analyze or otherwise process process data received from the optical subsystem 102 and feed the data to additional components within or outside the system 100.

[0050] Furthermore, memory device 112 may comprise any storage medium known in the art suitable for storing program instructions executable by one or more associated processors 110. For example, memory device 112 may comprise a non-transitory memory medium. As additional examples, memory device 112 may comprise, but is not limited to, read-only memory, random access memory, magnetic or optical storage devices (e.g., magnetic disks), magnetic tape, solid-state drives, etc. It should be further noted that memory device 112 may be housed together with one or more processors 110 in a common controller housing.

[0051] In this respect, controller 108 can perform any of the various processing steps associated with characterization. For example, controller 108 can be configured to generate control signals to guide or otherwise control the characterization subsystem 102 or any of its components. For example, controller 108 can be configured to guide stage 106 to translate sample 104 along one or more measurement paths or scan bands. By another example, controller 108 can be configured to receive images from optical subsystem 102. By another example, controller 108 can generate correction terms for one or more additional manufacturing subsystems based on measurements from optical subsystem 102 as feedback and / or feedforward control for one or more additional manufacturing subsystems.

[0052] Those skilled in the art will recognize that the components (e.g., operations), devices, objects, and accompanying discussions described herein are used as examples for clarity of concept and to account for various configuration modifications. Therefore, as used herein, the specific examples and accompanying discussions are intended to represent their more general categories. In general, the use of any specific example is intended to represent its category, and the omission of specific components (e.g., operations), devices, and objects should not be considered limiting.

[0053] Those skilled in the art will understand that various tools (e.g., hardware, software, and / or firmware) exist to implement the processes and / or systems and / or other technologies described herein, and the preferred tools will vary depending on the context in which the processes and / or systems and / or other technologies are deployed. For example, if the implementer determines that speed and accuracy are paramount, then the implementer may choose a primary hardware and / or firmware tool; alternatively, if flexibility is paramount, then the implementer may choose a primary software implementation; or again, alternatively, the implementer may choose some combination of hardware, software, and / or firmware. Thus, several possible tools exist to implement the processes and / or apparatus and / or other technologies described herein, none of which is inherently superior to another, because any tool to be used depends on the context in which the tool will be deployed and the implementer's specific considerations (e.g., speed, flexibility, or predictability), any of which can vary.

[0054] The foregoing description is presented to enable those skilled in the art to make and use the invention as provided in the context of a particular application and its requirements. As used herein, directional terms such as “top,” “bottom,” “above,” “below,” “up,” “down,” “below,” and “down” are intended to provide relative positions for descriptive purposes and are not intended to specify an absolute frame of reference. Those skilled in the art will understand that various modifications to the described embodiments are possible, and that the general principles defined herein can be applied to other embodiments. Therefore, the invention is not intended to be limited to the specific embodiments shown and described, but is to be accorded the broadest scope consistent with the principles and novel features disclosed herein.

[0055] Regarding the use of virtually any plural and / or singular terms in this document, those skilled in the art may convert plural to singular and / or singular to plural as needed by the context and / or application. For clarity, various singular / plural arrangements are not explicitly described herein.

[0056] All methods described herein may include storing the results of one or more steps of the method embodiments in memory. The results may include any of the results described herein and may be stored in any manner known in the art. The memory may include any memory described herein or any other suitable storage medium known in the art. After the results have been stored, they may be accessed in memory and used by any of the method or system embodiments described herein, formatted for display to a user, used by another software module, method, or system, etc. Furthermore, the results may be stored "permanently," "semi-permanently," "temporarily," or for a period of time. For example, the memory may be random access memory (RAM), and the results may not necessarily be stored in memory indefinitely.

[0057] Upon further consideration, each of the embodiments of the methods described above may include any other step of any other method described herein. Furthermore, each of the method embodiments described above may be performed by any of the systems described herein.

[0058] The topics described herein sometimes refer to different components contained within or connected to other components. It should be understood that such depicted architectures are merely exemplary, and many other architectures can in fact be implemented to achieve the same functionality. Conceptually, any arrangement of components achieving the same functionality is effectively “associated” to achieve the desired functionality. Therefore, any two components in this document combined to achieve a particular functionality can be considered “associated” with each other to achieve the desired functionality, regardless of the architecture or intermediate components. Similarly, any two such associated components can also be considered “connected” or “coupled” to each other to achieve the desired functionality, and any two components that can be suchly associated can also be considered “coupleable” to each other to achieve the desired functionality. Specific examples of coupleability include, but are not limited to, components that can be physically mated and / or physically interact, and / or components that can wirelessly interact and / or wirelessly interact, and / or components that can logically interact and / or logically interact.

[0059] Furthermore, it should be understood that the invention is defined by the appended claims. Those skilled in the art will understand that, generally, the terms used herein, and especially in the appended claims (e.g., the body of the appended claims), are generally intended to be “open-ended” terms (e.g., the term “comprising” should be interpreted as “comprising but not limited to,” the term “having” should be interpreted as “at least having,” the term “comprising” should be interpreted as “comprising but not limited to,” etc.). Those skilled in the art will further understand that if it is intended to introduce a particular number of claim statements, then this intention will be explicitly stated in the claims, and the absence of such a statement will not indicate the presence of such an intention. For example, to aid understanding, the appended claims may contain the introductory phrases “at least one” and “one or more” to introduce claim statements. However, the use of such phrases should not be construed as implying that the introduction of a claim statement by the indefinite article “a” or “an” limits any particular claim containing such an introduction to an invention containing only one such statement, even if the same claim contains the introductory phrases “one or more” or “at least one” and indefinite articles such as “a” or “an” (e.g., “a” and / or “an” should generally be interpreted as meaning “at least one” or “one or more”); the same applies to the use of definite articles used to introduce a claim statement. Furthermore, even if a specific number of claim statements is explicitly stated to be introduced, a person skilled in the art will recognize that such a statement should generally be interpreted as meaning at least the number stated (e.g., a simple statement of “two statements” without other modifiers generally means at least two statements or two or more statements). Furthermore, in examples where conventions such as "at least one of A, B, and C" are used, generally, those skilled in the art should understand that this construction is intended in the sense of the convention (e.g., "a system having at least one of A, B, and C" includes, but is not limited to, systems having only A, only B, only C, both A and B, both A and C, both B and C, and / or both A, B, and C). In examples where conventions such as "at least one of A, B, or C" are used, generally, those skilled in the art should understand that this construction is intended in the sense of the convention (e.g., "a system having at least one of A, B, or C" includes, but is not limited to, systems having only A, only B, only C, both A and B, both A and C, both B and C, and / or both A, B, and C). Those skilled in the art will further understand that any transitional words and / or phrases that present two or more alternative terms (whether in this description, claims, or drawings) should be understood to include the possibility of including one, any, or both of the terms. For example, the phrase "A or B" should be understood to include the possibility of including "A" or "B" or "A and B".

[0060] It is believed that this disclosure and its many accompanying advantages will be understood from the foregoing description, and it should be understood that various changes can be made to the form, construction, and arrangement of the components without departing from the disclosed subject matter or without sacrificing all its significant advantages. The described forms are merely illustrative, and the following claims are intended to cover and encompass such changes. Furthermore, it should be understood that the invention is defined by the appended claims.

Claims

1. A system for detecting defects on a sample, the system comprising: A controller comprising one or more processors configured to execute a set of program instructions stored in memory, wherein a condition-masked autoencoder is stored in the memory, and the set of program instructions is configured to cause the one or more processors to: Receive optical images; Multiple image blocks are generated by segmenting the received optical image, including the multiple image blocks; A shielded optical image is generated by selectively shielding a first group of image blocks among the plurality of image blocks based on characteristic data, wherein the first group of image blocks corresponds to one or more defective regions, and wherein the plurality of image blocks includes a second group of image blocks corresponding to one or more defect-free regions. The generated shielded optical image is provided to the first encoder of the conditionally shielded autoencoder; The first encoder is applied to the second set of image blocks that have generated shielded optical images to generate a first set of feature vectors; Receive design images; The design image is provided to the second encoder of the conditionally masked autoencoder; The second encoder is applied to the design image to generate a second set of feature vectors; The first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder are concatenated. and The decoder using the conditionally masked autoencoder generates a reference image based on the concatenated feature vectors, wherein the generated reference image is a reconstructed image of the optical image and the design image.

2. The system of claim 1, wherein the set of program instructions is further configured to cause the one or more processors: One or more defects on the sample are identified by comparing the generated reference image with the sample image.

3. The system according to claim 1, wherein each of the plurality of image blocks has a block size of 16×16 pixels.

4. The system according to claim 1, wherein each of the plurality of image blocks has a block size of 14×14 pixels.

5. The system according to claim 1, wherein each of the plurality of image blocks has a block size of 28×28 pixels.

6. The system of claim 1, wherein each of the plurality of image blocks has a block size of 32×32 pixels.

7. The system of claim 1, wherein the characterization data includes defect inspection data from a broadband inspection system.

8. The system of claim 1, wherein the conditionally masked autoencoder comprises a self-supervised deep learning model.

9. The system of claim 8, wherein the first encoder and the second encoder comprise a vision transformer encoder, and wherein the decoder comprises a vision transformer decoder.

10. The system of claim 1, further comprising: An optical subsystem configured to acquire the optical image of the sample.

11. The system of claim 10, wherein the optical subsystem comprises a scanning electron microscopy subsystem.

12. The system of claim 11, wherein the optical image comprises a scanning electron microscope (SEM) image.

13. A system for detecting defects on a sample, the system comprising: An optical subsystem configured to acquire multiple optical images of the sample; and A controller communicatively coupled to the optical subsystem includes one or more processors configured to execute a set of program instructions stored in memory, wherein the conditionally masked autoencoder is stored in the memory, and the set of program instructions is configured to cause the one or more processors to: Receive an optical image from the plurality of optical images from the optical subsystem; Multiple image blocks are generated by segmenting the received optical image, including the multiple image blocks; A shielded optical image is generated by selectively shielding a first group of image blocks among the plurality of image blocks based on characteristic data, wherein the first group of image blocks corresponds to one or more defective regions, and wherein the plurality of image blocks includes a second group of image blocks corresponding to one or more defect-free regions. The generated shielded optical image is provided to the first encoder of the conditionally shielded autoencoder; The first encoder is applied to the second set of image blocks that have generated shielded optical images to generate a first set of feature vectors; Receive design images; The design image is provided to the second encoder of the conditionally masked autoencoder; The second encoder is applied to the design image to generate a second set of feature vectors; The first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder are concatenated. and The decoder using the conditionally masked autoencoder generates a reference image based on the concatenated feature vectors, wherein the generated reference image is a reconstructed image of the optical image and the design image.

14. The system of claim 13, wherein the set of program instructions is further configured to cause the one or more processors: One or more defects on the sample are identified by comparing the generated reference image with the optical images in the plurality of optical images.

15. The system of claim 13, wherein each of the plurality of image blocks has a block size of 16×16 pixels.

16. The system of claim 13, wherein each of the plurality of image blocks has a block size of 14 × 14 pixels.

17. The system of claim 13, wherein each of the plurality of image blocks has a block size of 28 × 28 pixels.

18. The system of claim 13, wherein each of the plurality of image blocks has a block size of 32×32 pixels.

19. The system of claim 13, wherein the characterization data comprises defect inspection data from a broadband inspection system.

20. The system of claim 13, wherein the conditionally masked autoencoder comprises a self-supervised deep learning model.

21. The system of claim 20, wherein the first encoder and the second encoder comprise a vision transformer encoder, and wherein the decoder comprises a vision transformer decoder.

22. The system of claim 13, wherein the optical subsystem comprises a scanning electron microscopy subsystem.

23. The system of claim 22, wherein the optical image comprises a scanning electron microscope (SEM) image.

24. A method comprising: Optical images of the received sample; Multiple image blocks are generated by segmenting the received optical image, including the multiple image blocks; A shielded optical image is generated by selectively shielding a first group of image blocks among the plurality of image blocks based on characteristic data, wherein the first group of image blocks corresponds to one or more defective regions, and wherein the plurality of image blocks includes a second group of image blocks corresponding to one or more defect-free regions. The generated shielded optical image is provided to the first encoder of the conditionally shielded automatic encoder; The first encoder is applied to the second set of image blocks that have generated shielded optical images to generate a first set of feature vectors; Receive design images; The design image is provided to the second encoder of the conditionally masked autoencoder; The second encoder is applied to the design image to generate a second set of feature vectors; The first set of feature vectors from the first encoder and the second set of feature vectors from the second encoder are concatenated. and The decoder using the conditionally masked autoencoder generates a reference image of the sample based on the concatenated feature vectors, wherein the generated reference image is a reconstructed image of the optical image and the design image.

25. The method of claim 24, further comprising: One or more defects on the sample are identified by comparing the generated reference image with the optical images in the plurality of optical images.

Citation Information

Patent Citations

  • Arrayed column detector

    US11239048B2

  • Defect inspection and review using transmissive current image of charged particle beam system

    US11410830B1

  • System and method for rendering SEM images and predicting defect imaging conditions of substrates using 3D design

    US11880193B2

  • Electron source with magnetic suppressor electrode

    US20220108862A1

  • System and method for multi-beam electron microscopy using a detector array

    US20240194440A1