A method, device and storage medium for reducing custom defect location information in a wafer

Abstract

The present application relates to a kind of methods for restoring custom defect position information in wafer, comprising the steps of: obtaining at least one wafer to be measured at least one set of first defect image;At least one set of first defect image is input to at least one corresponding custom defect extraction model, custom defect extraction model is based on pixel point information to identify the custom defect pattern in first defect image, and output the first position information of custom defect pattern and the second defect image including corresponding custom defect pattern;Second position information of custom defect pattern is determined based on the size information of first position information and wafer to be measured.The present application also provides a device for restoring custom defect position information in wafer and computer readable storage medium.The method, device and computer readable storage medium of the present application can accurately position custom defect image in wafer.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing, and in particular to a method, apparatus and storage medium for restoring custom defect location information in a wafer. Background Technology

[0002] With the development of semiconductor device technology, there are more and more processes used to manufacture semiconductor devices, and each process has a certain degree of complexity. Each process may produce some unexpected structures in the processing of wafers. Among them, those that cause the circuits on the chip to malfunction are called wafer defects.

[0003] In chip manufacturing, wafer defects are diverse, especially those with complex shapes and unique distributions. Existing AI classification models often struggle to accurately identify and classify these defects, frequently necessitating manual visual inspection. For instance, after critical process nodes, manual inspection is required. Workers, based on experience, identify defect patterns with specific distributions on the wafer (i.e., special defect patterns), manually mark them, and extract the defect data. However, this manual marking method has relatively large errors, low efficiency, and is highly dependent on worker experience. When there are many wafers requiring visual inspection or a large number of defects, manual inspection struggles to guarantee timeliness. Furthermore, worker fatigue can impair the accuracy and reliability of the inspection over extended periods. This is particularly true for complex special defect patterns, such as linear, arc-shaped, ring-shaped, composite, wavy, and dispersed defects. Even experienced workers struggle to accurately mark and extract these defects and correlate them with specific defect data. Especially when there are many defects on the wafer or many overlapping areas of defects, it is more difficult to mark the pattern of a single defect. Summary of the Invention:

[0004] To address the aforementioned technical problems, the present invention aims to provide a method for restoring custom defect location information in a wafer, comprising the following steps:

[0005] Acquire at least one set of first defect images of at least one wafer under test;

[0006] At least one set of the first defect images is input into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic.

[0007] Based on the first location information and the size information of the wafer under test, the second location information of the custom defect pattern is determined, wherein the second location information includes: the actual coordinate information of the custom defect pattern in the wafer under test;

[0008] The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics.

[0009] In some embodiments, the first defect image includes at least one type of custom defect graphic.

[0010] In some embodiments, when the first defect image includes multiple custom defect graphics, after the custom defect extraction model identifies the custom defect graphics in the first defect image based on pixel information, the method further includes the step of: the custom defect extraction model splits the multiple custom defect graphics based on pixel information, and outputs multiple second defect images based on the splitting results.

[0011] In some embodiments, custom defects include: linear defects, and / or arcuate defects, and / or annular defects, and / or dispersed defects, and / or composite defects, and / or wavy defects, and / or aggregated defects.

[0012] In some embodiments, the steps further include:

[0013] Collect the second defect image output by the custom defect extraction model and the corresponding first location information;

[0014] The custom defect extraction model is periodically updated based on the second defect image and the corresponding first location information.

[0015] In some embodiments, the steps further include:

[0016] Based on the second location information, the corresponding engineering data of the custom defect graphic is obtained.

[0017] In some embodiments, the custom defect extraction model extracts only one type of custom defect graphic; correspondingly, the method further includes the step of:

[0018] The second defect image is subjected to secondary detection, and when a custom defect graphic on the second defect image is detected that does not belong to the extraction category of the corresponding custom defect extraction model, the second defect image is manually labeled.

[0019] In some embodiments, the steps further include:

[0020] The custom defect extraction model is updated based on the manually labeled second defect image and the corresponding first location information.

[0021] In some embodiments, the pixel coordinates are (X1, Y1), and correspondingly, determining the second position information of the custom defect pattern based on the first position information and the size information of the wafer to be tested includes the following steps:

[0022] Based on at least one pixel coordinate information, determine the relative coordinates (X3, Y3) of at least one pixel, where X3 is the X percentage coordinate of the pixel and Y3 is the Y percentage coordinate of the pixel.

[0023] The actual coordinate information (X2, Y2) of the custom defect graphic is determined based on the relative coordinates and the size information;

[0024] Where X3 = X2 / D, Y3 = Y2 / D, and D is the diameter of the wafer to be tested.

[0025] In some embodiments, the step of acquiring at least one set of first defect images of at least one wafer under test includes:

[0026] Acquire multiple images of at least one of the wafers to be tested for classification;

[0027] The multiple images to be classified are classified to obtain at least one set of first defect images, and each of the first defect images in the set of first defect images includes at least one custom defect graphic of the same category.

[0028] A second aspect of the present invention also provides an apparatus for restoring custom defect location information in a wafer, comprising:

[0029] The image acquisition module is used to acquire at least one type of first defect image of at least one wafer under test;

[0030] The defect identification module is used to input at least one type of the first defect image into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic.

[0031] The defect location module determines the second location information of the custom defect pattern based on the first location information and the size information of the wafer under test. The second location information includes the actual coordinate information of the custom defect pattern in the wafer under test.

[0032] The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics.

[0033] In some embodiments, the first defect image includes at least one type of custom defect graphic.

[0034] In some embodiments, the device further includes:

[0035] The data storage module is used to store the first defect image collected and input into the custom defect extraction model, as well as the corresponding first location information;

[0036] The model update module periodically updates the custom defect extraction model based on the first defect image and the corresponding first location information.

[0037] A third aspect of the present invention is that a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the methods described in the above embodiments.

[0038] Beneficial technical effects

[0039] This invention relates to a method for restoring the location information of a custom defect in a wafer, which performs feature recognition on the corresponding custom defect pattern in a first defect image based on pixel information.

[0040] Existing technologies typically involve manual identification and extraction of macroscopic, custom-defined defect patterns. This approach is inefficient and prevents the direct use of the acquired graphic data (e.g., for correlating with engineering data or performing defect analysis). Furthermore, the extraction accuracy is poor; even if defect location information is obtained from the extracted defect image (which is prone to distortion during extraction), the accuracy of the defect location information cannot be guaranteed. Unlike existing technologies, this application does not identify or extract defects from the macroscopic graphic structure. Instead, it identifies the features of individual pixels (microscopic information in the first defect image) in the first defect image and identifies the set of pixels that combine to form a custom-defined defect pattern. During this process, information about the set of pixels identified as defects (such as pixel coordinates) is recorded, and the actual position of the defect pattern on the wafer under test is then located based on the pixel coordinates of the defect pattern. The pixel identification method used in this method can more directly and accurately identify defects, thereby truly and accurately reconstructing the defect location information.

[0041] Another aspect of this invention is its excellent versatility, particularly for a type of wafer defect with industrial mechanisms and high reproducibility. Specifically, for this type of custom defect, information (such as graphic information) corresponding to one or more custom defects can be obtained by collecting historical experience data from previous work. Based on this information, a corresponding custom defect extraction model can be constructed. Furthermore, when these defects are rediscovered during wafer manufacturing, the pre-constructed custom defect extraction model can quickly reconstruct the defect location (such as actual coordinate information), facilitating further identification and judgment of the defect location by work personnel, and analyzing the cause of the custom defect. For example, it can determine whether the defect is caused by equipment failure, random error, or debris (such as dust), enabling rapid resolution and troubleshooting.

[0042] Furthermore, by extracting the set of pixels marked as defects, a second defect image including the corresponding custom defect graphic is obtained. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale. Obviously, the drawings described below are some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of a method flow of an exemplary embodiment of the present invention;

[0045] Figure 2a This is a structural schematic diagram of some custom defect graphics in a specific embodiment of the present invention;

[0046] Figure 2b This is a schematic diagram of a dispersed defect structure.

[0047] Figure 3a This is a schematic diagram of the extraction result of a custom defect graphic in a specific embodiment of the present invention;

[0048] Figure 3b This is a schematic diagram of the extraction result of the arc-shaped defect pattern in another specific embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of the classification result of an image to be classified in an exemplary embodiment of the present invention;

[0050] Figure 5This is a schematic diagram showing the disassembly result of an arc-shaped defect pattern in an exemplary embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of the module structure of the device in an exemplary embodiment of the present invention;

[0052] Figure 7 This is a block diagram of an electronic device in an exemplary embodiment of the present invention.

[0053] Among them, 1 is an arc-shaped defect, 2 is a linear defect, 3 is a composite defect, 4 is an aggregated defect, 5 is a ring-shaped defect, 6 is a wavy defect, and 7 is a dispersed defect. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] In this document, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" may be used interchangeably.

[0056] Definition of noun:

[0057] In this article, "custom defect graphics" refers to defect graphics with special shapes and structures (such as irregularities) or complex structural distributions, such as... Figure 2a , Figure 2b The defects shown include arc-shaped defects (1), linear defects (2), compound defects (3), clustered defects (4), ring-shaped defects (5), wavy defects (6), and dispersed defects (7), etc. Due to the unique shape and structure of these defects, they often can only be identified visually by experienced workers. Furthermore, there are numerous such defects, and changes in processing technology or conditions may generate several new, customized defect patterns. In such cases, workers often need to combine long-term historical data and experience to identify and analyze them.

[0058] The terms "same" or "similar" in this article do not mean that the difference between the two or more specified items is zero, but rather that the difference between the specified items is very small and can be ignored in practical engineering applications.

[0059] Example 1

[0060] The first aspect of this invention is that it provides a method for restoring custom defect location information in a wafer, such as... Figure 1 As shown, the method includes the following steps:

[0061] S02: Acquire at least one set of first defect images of at least one wafer under test;

[0062] S04: Input at least one set of first defect images into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic.

[0063] The first location information includes: the pixel coordinates of the custom defect graphic;

[0064] S06: Determine the second position information of the custom defect pattern based on the first position information and the size information of the wafer under test. The second position information includes: the actual coordinate information of the custom defect pattern in the wafer under test.

[0065] The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics.

[0066] In some embodiments, the pixel coordinate data used for model training can be provided by an image, i.e., model training is performed based on an image labeled with custom defects (such as a first defect image). Alternatively, the pixel coordinate data used for model training can be obtained by training on a set of pixel coordinate data of custom defects, for example, by training on a table of pixel coordinate data of custom defects.

[0067] Furthermore, in some embodiments, the custom defect extraction model targets a type of wafer defect with industrial mechanisms and high repeatability. For example, when a certain type of custom defect (such as an arc-shaped defect) occurs in the current process step, this type of custom defect may reappear on the wafer when the same or similar process steps are used next time.

[0068] These custom defects are often related to the design of machine components or their operation. For example, ring-shaped defects may be related to rotating equipment parts; for instance, a fixed hard object above a rotating wafer can cause ring-shaped scratches. Linear defects, on the other hand, are related to linearly moving parts, such as scratches caused by a robotic arm transporting the wafer within the equipment. The causes of these custom defect patterns are varied (e.g., machine malfunctions requiring quick machine shutdown and troubleshooting, or even production line shutdown; or debris like dust on the machine requiring quick removal). Analysis must be combined with other data. To shorten defect analysis time, staff regularly collect discovered abnormal patterns (i.e., custom defect patterns) and their causes for identification and matching analysis.

[0069] For example, in some embodiments, certain custom defects are often related to the aging of equipment components. This aging-related failure can cause significant yield losses, requiring rapid detection and handling by staff. If component aging failures are not addressed promptly, these custom defect patterns will appear on subsequent batches of wafers, adversely affecting wafer production. However, the images obtained from the production line often cannot directly determine the cause of the defect. It could be due to component aging or accidental factors (such as dust). The former requires immediate shutdown of the equipment or even the production line, while the latter can be handled quickly without stopping production. In wafer production, stopping equipment, especially the production line, is extremely costly; therefore, timely defect analysis is crucial in this situation. Based on the second location information provided in this embodiment, staff can quickly locate the actual location of the custom defect, identify and analyze the specific defect area, determine the cause of the custom defect, and facilitate rapid troubleshooting.

[0070] Preferably, in this embodiment, when workers discover one or more custom defects, they periodically collect a large amount of information about one or more custom defect images (such as multiple test images containing the custom defect image) based on historical experience data, thereby performing targeted modeling for various custom defects. Therefore, in subsequent wafer manufacturing processes, for various known custom defect images, the second location information of the custom defect can be specifically restored using a pre-built custom defect extraction model. That is to say, the method for restoring location information provided in this embodiment has good versatility (especially for a type of wafer defect with industrial mechanisms and recurring occurrences).

[0071] Preferably, in some embodiments, the pixel information includes: the signal intensity of the pixel (such as grayscale value or brightness value), and the correlation distribution characteristics of pixels with the same or similar signal intensity. For example, they may be continuously distributed as linearly; or they may be distributed in a dispersed manner; or they may be distributed in a regularly dispersed manner (such as...). Figure 2a Composite defect 3 in Figure (b).

[0072] The custom defect location information restoration method proposed in this embodiment accurately identifies various defects to be tested based on the morphological distribution characteristics (specifically pixel features) of custom defect graphics (or special defect graphics), and simultaneously records the pixel information identified as defects.

[0073] Preferably, in this embodiment, a second defect image containing only the corresponding custom defect is obtained based on the pixel information identified as a defect. For example, a first defect image including a ring defect is input into a ring defect image extraction model, and the ring defect extraction model outputs a second defect image, which contains only the ring defect image.

[0074] It is understood that the second defect image output in this embodiment is mainly intended to facilitate subsequent data verification and analysis by staff. The second defect image is the result data after the defect information has been identified and extracted.

[0075] There are many types of custom defect graphics; for example, in some embodiments, such as... Figure 2a , Figure 2b As shown, custom defects include one or more of the following types of defect graphics: linear defect 2, arc defect 1, ring defect 5, dispersed defect 7, composite defect 3, and wavy defect 6.

[0076] For example, in some embodiments, the composite defect 3 is a composite defect composed of two or more custom defects, or in other words, the composite defect 3 includes the shape distribution characteristics of two or more custom defects. For example, Figure 2a Figures (a) and (e) show that the composite defect 3 is a composite structure of a corrugated defect and a ring defect. For example, the composite defect 3 shown in Figure (b) is a composite structure of a dispersed defect (i.e., a defect that is distributed in a point-like pattern) and a ring defect.

[0077] To further improve the accuracy and efficiency of custom defect identification, in some embodiments, the first defect image includes at least one first defect image. Specifically, at least one set of acquired first defect images all include the same type of custom defect, such as an arc-shaped defect. Further, the first defect image is input into a corresponding custom defect extraction model; in some embodiments, such as... Figure 4 As shown, at least one set of first defect images including arc-shaped defects are input into an arc-shaped defect extraction model, and the extraction model outputs at least one set of second defect images, wherein the second defect images only include arc-shaped defect graphics.

[0078] In this embodiment, the step of obtaining at least one set of first defect images in step S02 is as follows:

[0079] Acquire multiple images (or images to be classified) of at least one wafer to be tested;

[0080] Multiple images to be classified are classified to obtain at least one set of first defect images, and each of the first defect images in the set includes at least one identical custom defect graphic. In this embodiment, the images to be tested are first classified for defects, and then the first defect images obtained after classification are input into the custom defect extraction models of the corresponding categories.

[0081] For example, the images to be tested are input into at least one custom defect classification model (or custom defect graphic classification model), and the custom defect classification model outputs at least one set of first defect images, wherein each set of first defect images includes the same type of custom defect images. Of course, in some other embodiments, the first defect images can be directly selected from the images to be tested.

[0082] For example, in some embodiments, the custom defect classification model is an arc-shaped defect classification model. When multiple images to be tested are input into the arc-shaped defect classification model, the arc-shaped defect classification model outputs a set of multiple first defect images including arc-shaped defects.

[0083] For example, in some embodiments, the steps for constructing a custom defect classification model include:

[0084] Acquire multiple target defect images, and each of the multiple target defect images includes at least one type of the same custom defect graphic;

[0085] The deep convolutional neural network (CNN) method is used to extract features and model multiple target defect images, and a corresponding custom defect image classification model is obtained.

[0086] For example, in some embodiments, the custom defect graphic classification model can classify input test images that include multiple types of custom defect graphics. For instance, multiple input test images can be divided into multiple groups of first defect images. Specifically, based on the included custom defect categories, they are divided into a first group of first defect images (including a first category of custom defects), a second group of first defect images (including a second category of custom defects), ..., an Nth group of first defect images (including an Nth category of custom defects). It is understood that some test images may be classified repeatedly due to the presence of two or more types of custom defect graphics.

[0087] For example, in other embodiments, the custom defect image classification model can specifically classify a certain type of custom defect. Specifically, the custom defect image classification model can be a clustered defect image classification model. After inputting multiple test images into this clustered defect image classification model, the model can output a set of first defect images, and each of the first defect images in this set includes at least one clustered defect image. In other words, in some embodiments, the custom defect image classification model can be a set of one or more specific custom defect images, thereby enabling the classification of one or more custom defect images. Figure 4 As shown, after inputting multiple images to be tested into a custom defect image classification model, multiple sets of first defect images are output, such as the first set of images in the first set, the second set, the third set, the fourth set, and the fifth set.

[0088] Specifically, in some embodiments, the corresponding image to be tested is acquired using an image acquisition device such as a scanning electron microscope (SEM), an optical microscope, or an electron beam microscope.

[0089] For example, in one specific embodiment, such as Figure 3a As shown, Figure 3a Figure (a) shows a test image (or defect image) obtained for a wafer under test, which includes at least four types of custom defect patterns (i.e., clustered defects, linear defects, arc defects, and ring defects). The test image contains multiple custom defects that intersect or overlap, such as areas where clustered defects, arc defects, and linear defects intersect, or where some linear defects partially overlap with arc defects. In this case, manually identifying and extracting defects from the test image (e.g., edge extraction) is very difficult, especially for areas where two or more defect patterns overlap, making accurate segmentation and extraction of each defect pattern even more challenging.

[0090] In this embodiment, before accurately identifying and extracting defect patterns in the image under test, the image under test is first classified into defect images. At this time, the image under test shown in this embodiment will be simultaneously classified into a first group of first defect images (including clustered defect patterns), a second group of first defect images (including linear defect patterns), a third group of first defect images (including arc-shaped defect patterns), and a fourth group of first defect images (including ring-shaped defect patterns). Subsequently, the image under test is input into the clustered defect extraction model, the linear defect extraction model, the arc-shaped defect extraction model, and the ring-shaped defect extraction model, respectively. The extraction models correspondingly output the clustered defect image, the linear defect image, the arc-shaped defect image, and the ring-shaped defect image (i.e., the second defect image) on the image under test, respectively. Figure 3a (See Figures (b), (c), (d), and (e) in the original text). In this embodiment, custom defect graphics with complex shapes and structures can be accurately and specifically identified and extracted based on multiple custom defect extraction models. Unlike existing identification and extraction methods, this method does not directly observe the graphic structure to identify and extract defect graphics (e.g., visually identify and manually extract macroscopic defect graphics by workers). Instead, it uses a custom defect extraction model to perform feature recognition on the most original pixel information in the first defect image and marks the pixel information with defect features as the corresponding defect graphics. This embodiment uses pixel-based information for defect identification and extraction, which greatly improves the accuracy of defect identification.

[0091] Furthermore, based on the pixel information marked as a custom defect, the first position information of the corresponding custom defect graphic can be directly obtained, avoiding the potential accuracy loss during information conversion. This first position information includes the pixel coordinates of the custom defect graphic. Subsequently, based on the first position information and the corresponding size information of the wafer under test, the actual position information (or actual coordinate information) of the defect graphic on the wafer under test can be located. The actual coordinate information can be the absolute position coordinates of the defect graphic on the wafer, or the relative position coordinates (such as the x and y percentage coordinates of the defect graphic on the wafer). In this embodiment, by performing feature recognition on the pixel information of the defect, various custom defect graphics can be accurately located.

[0092] Furthermore, to facilitate the identification, classification, and subsequent data analysis of the generated defects by staff, in some embodiments, when the first defect image includes multiple custom defect graphics, after the custom defect extraction model identifies the custom defect graphics in the first defect image based on pixel information, the method further includes the step of: the custom defect extraction model splitting the multiple custom defect graphics based on pixel information, and outputting multiple second defect images based on the splitting results. Figure 5 As shown, Figure 5 The first defect image, which contains multiple arc-shaped defect patterns, is divided into several (e.g., 4) second defect images, each of which contains only one arc-shaped defect pattern. This facilitates subsequent verification and analysis of the second defect images by the staff.

[0093] Specifically, in some embodiments, the information (or set of pixels) of each pixel identified (or determined) as a custom defect is recorded separately, and the corresponding second defect image is output based on the information of each pixel.

[0094] Preferably, in some embodiments, when there are two or more custom defect images in the output second defect image, the two or more custom defect images do not intersect or overlap.

[0095] Preferably, in some embodiments, the output second defect image includes a custom defect graphic.

[0096] Furthermore, to improve the accuracy of identifying and classifying custom defects, some embodiments also include the following steps:

[0097] Verify the second defect image that is labeled with a defect category (such as being labeled as an arc defect, ring defect, etc.).

[0098] For example, in some embodiments, the staff verifies the acquired multiple second defect images. When the staff detects a second defect image with an incorrect defect category label (e.g., when a linear defect graphic is output in an arc-shaped defect extraction model), the staff relabels the acquired second defect image.

[0099] To improve the applicability and accuracy of the custom defect extraction model, some embodiments further include the following steps:

[0100] The custom defect extraction model is updated (or optimized) based on the relabeled (or secondary-labeled) second defect image.

[0101] For example, in some embodiments, when the staff finds that a second defect image including a linear defect graphic is output in the arc-shaped defect extraction model, the second defect image is re-labeled and collected for use in updating the custom defect extraction model.

[0102] Of course, in other embodiments, to improve the applicability and accuracy of the custom defect extraction model, the method further includes the following steps:

[0103] Collect the second defect images output by the custom defect extraction model;

[0104] The custom defect extraction model is updated based on the collected second defect image. In this embodiment, the second defect image used to update the model can be part or all of the results output by the custom defect extraction model.

[0105] Specifically, in some embodiments, the method further includes the step of:

[0106] Collect the second defect image output by the custom defect extraction model and the corresponding first location information;

[0107] The custom defect extraction model is periodically updated based on the second defect image and the corresponding first location information.

[0108] In this embodiment, the custom defect extraction model is preferably updated and optimized using pixel information obtained during the defect extraction process.

[0109] The custom defect extraction model selected in this embodiment can actively learn and optimize based on the output recognition results (such as the second defect image and the corresponding pixel information).

[0110] In some embodiments, the custom defect extraction model extracts only one type of custom defect graphic; correspondingly, the method further includes the step of:

[0111] The second defect image is subjected to secondary detection, and when a custom defect graphic on the second defect image is detected that does not belong to the extraction category of the corresponding custom defect extraction model, the second defect image is manually labeled.

[0112] In some embodiments, the secondary inspection is performed by manual visual inspection, or the second defect image is input into the corresponding custom defect extraction model and secondary inspection is performed by the extraction model. If an unrecognizable defect graphic (or second defect image) is found in the secondary inspection, it is manually marked.

[0113] To facilitate more refined analysis of wafer defect information by staff, some embodiments specifically include the following steps:

[0114] The custom defect extraction model is updated based on the manually labeled second defect image and the corresponding first location information.

[0115] In some embodiments, the pixel coordinates are (X1, Y1), and correspondingly, determining the second position information of the custom defect pattern based on the first position information and the size information of the wafer to be tested includes the following steps:

[0116] Based on at least one pixel coordinate information, determine the relative coordinates (X3, Y3) of at least one pixel, where X3 is the X percentage coordinate of the pixel and Y3 is the Y percentage coordinate of the pixel.

[0117] The actual coordinate information (X2, Y2) of the custom defect graphic is determined based on the relative coordinates and the size information;

[0118] Where X3 = X2 / D, Y3 = Y2 / D, and D is the diameter of the wafer to be tested.

[0119] In this embodiment, in order to ensure the matching accuracy between the actual coordinate information (X2, Y2) and the pixel coordinate information (X3, Y3) during the matching process, the actual coordinate information is restored by X3 = X2 / D and Y3 = Y2 / D.

[0120] It is understood that the coordinate information in this embodiment is described using a Cartesian coordinate system. In other embodiments, the coordinate information may also be described using a polar coordinate system.

[0121] In this embodiment, because the location of the defect is accurately located, it is convenient for staff to obtain engineering information (or engineering data) related to the defect based on the defect location information.

[0122] For example, in some embodiments, the method includes the steps of:

[0123] Based on the second location information, the corresponding engineering data of the custom defect graphic is obtained.

[0124] In some embodiments, engineering data includes: circuit diagrams, structures, and related measurement data (such as dimensions, thickness, resistance values, etc.) corresponding to the locations of custom defects.

[0125] In some embodiments, the engineering data also includes the final yield test results of the chip to which the corresponding custom defect image belongs (i.e., the wafer under test).

[0126] In some embodiments, the engineering data may also include other device parameters such as the location of the defect or the chip to which it belongs, such as the pressure, voltage, gas flow rate, etc., applied during the processing at the corresponding location of the defect.

[0127] In this embodiment, by accurately locating the pixel information (such as pixel coordinates) of the defect pattern, and based on the pixel information, accurately locating the actual coordinates of the defect pattern in the wafer under test, it is possible to associate the defect pattern with specific engineering data, enabling staff to monitor and optimize the wafer manufacturing process based on the correlation between the defect pattern and the engineering data.

[0128] In other words, this embodiment associates unstructured data (defect images) with structured data (such as engineering data) by recognizing the features of pixels. For example, in some implementations, some engineering data is shown in Table 1 below. Here, DEFE CITD is the defect number, XREL is the actual X-axis coordinate of the defect in the chip (grid) coordinate system, YREL is the actual Y-axis coordinate of the defect in the chip (grid) coordinate system, XINDEX is the column number of the chip (grid) containing the defect on the wafer, YINDEX is the row number of the chip (grid) containing the defect on the wafer, XSIZE is the X-axis dimension of the wafer, YSIZE is the Y-axis dimension of the wafer, DEFECT AREA is the defect area, DSIZE is the equivalent size of the defect, CLASS NUMB ER is the defect category number, TEST is the test used to discover this defect, IMAGE COUNT is the number of photos associated with this defect, and IMAGE LIST is the number of photos associated with this defect. By obtaining the actual coordinate information of the custom defect image, the relevant parameters corresponding to the defect can be quickly retrieved.

[0129] Table 1

[0130]

[0131] Example 2

[0132] A second aspect of the present invention is that, based on the method in Embodiment 1 above, a corresponding apparatus for restoring custom defect location information in a wafer is also provided, such as... Figure 6 As shown, it includes:

[0133] Image acquisition module 02 is used to acquire at least one type of first defect image of at least one wafer under test;

[0134] The defect identification module 04 is used to input at least one type of the first defect image into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic.

[0135] Defect location module 06 determines the second location information of the custom defect pattern based on the first location information and the size information of the wafer under test. The second location information includes the actual coordinate information of the custom defect pattern in the wafer under test.

[0136] The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics.

[0137] In some embodiments, the first defect image includes at least one type of custom defect graphic.

[0138] Since custom defects can arise from a variety of causes, rapid troubleshooting is essential to minimize wafer yield losses. This method can accurately and quickly locate custom defects, providing workers with easily identifiable defect location information, such as actual coordinates. For example, based on the acquired coordinates, workers can pinpoint specific areas on the wafer to further determine the cause of the custom defect and thus identify a suitable solution.

[0139] In some embodiments, the custom defect extraction model includes one or more of the following defect extraction models: a first custom defect extraction model for extracting linear defects, a second custom defect extraction model for extracting arc-shaped defects, a third custom defect extraction model for extracting annular defects, a fourth custom defect extraction model for extracting dispersed defects, a fifth custom defect extraction model for extracting clustered defects, a sixth custom defect extraction model for extracting wavy defects, and a seventh custom defect extraction model for extracting composite defects.

[0140] In some embodiments, it also includes:

[0141] The data storage module is used to store the first defect image collected and input into the custom defect extraction model, as well as the corresponding first location information;

[0142] The model update module periodically updates the custom defect extraction model based on the first defect image and the corresponding first location information.

[0143] In some embodiments, the defect identification module includes:

[0144] The defect extraction unit is used to input at least one set of the first defect images into at least one corresponding custom defect extraction model, wherein the custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic.

[0145] The defect splitting unit is used to split the multiple custom defect graphics based on pixel information when the first defect image includes multiple custom defect graphics, and output multiple second defect images based on the splitting results.

[0146] In some embodiments, the device further includes:

[0147] The secondary detection module is used to perform secondary detection on the second defect image, and when it is detected that the custom defect graphic on the second defect image does not belong to the extraction category of the corresponding custom defect extraction model, the second defect image is manually marked.

[0148] Furthermore, in some embodiments, the model update module is also configured to update the corresponding custom defect extraction model based on the manually labeled second defect image and the corresponding first location information.

[0149] Furthermore, in some embodiments, the defect location module 06 includes:

[0150] The first positioning unit is used to determine the relative coordinates (X3, Y3) of at least one pixel point based on the coordinate information of at least one pixel point, wherein X3 is the X percentage coordinate of the pixel point and Y3 is the Y percentage coordinate of the pixel point;

[0151] The second positioning unit is used to determine the actual coordinate information (X2, Y2) of the custom defect graphic based on the relative coordinates and the size information;

[0152] Where X3 = X2 / D, Y3 = Y2 / D, and D is the diameter of the wafer to be tested.

[0153] Furthermore, in some embodiments, it also includes: a defect information association module, used to obtain engineering data of the corresponding custom defect graphic based on the second location information.

[0154] Example 3

[0155] A third aspect of the present invention is that, based on the method in Embodiment 1 above, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the methods described in Embodiment 1.

[0156] For example, in some embodiments, when the computer program is executed by a processor, it performs the following steps: acquiring at least one set of first defect images of at least one wafer under test, the first defect images including at least one type of custom defect pattern;

[0157] At least one set of the first defect images is input into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic.

[0158] Based on the first location information and the size information of the wafer under test, the second location information of the custom defect pattern is determined, wherein the second location information includes: the actual coordinate information of the custom defect pattern in the wafer under test;

[0159] The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics.

[0160] Based on the methods in the above embodiments, the present invention also provides an electronic device, including a memory 502, a processor 501, and a computer program stored in the memory 502 and executable on the processor 501. When the processor 501 executes the program, it implements the steps of the methods described above. For ease of explanation, only the parts related to the embodiments of this specification are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of this specification. This electronic device can be any electronic device, including PCs, cloud servers, and even mobile phones, tablets, PDAs (Personal Digital Assistants), POS (Point of Sales), in-vehicle computers, desktop computers, etc.

[0161] Specifically, Figure 7This is a block diagram illustrating the structural composition of an electronic device according to an exemplary embodiment of the present invention. Bus 500 may include any number of interconnected buses 500 and bridges, linking various circuits including one or more processors represented by processor 501 and memory represented by memory. Bus 500 may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Communication interface 503 provides an interface between the bus and receiver and / or transmitter 504, which may be separate, independent receivers or transmitters 504 or a single element such as a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 501 is responsible for managing bus 500 and general processing, while memory 502 may be used to store data used by processor 501 during operation.

[0162] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, or network device, etc.) to execute the methods described above according to the embodiments of this disclosure.

[0163] The computer-readable storage medium may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0164] Program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0165] The aforementioned computer-readable medium carries one or more programs, which, when executed by a device, cause the computer-readable medium to perform the following functions: acquiring at least one set of first defect images of at least one wafer under test, the first defect images including at least one type of custom defect pattern; inputting the at least one set of first defect images into at least one corresponding custom defect extraction model, the custom defect extraction model identifying the custom defect pattern in the first defect image based on pixel information, and outputting first position information of the custom defect pattern and a second defect image including the corresponding custom defect pattern, wherein the first position information includes: pixel coordinates of the custom defect pattern; determining second position information of the custom defect pattern based on the first position information and the size information of the wafer under test, the second position information including: actual coordinate information of the custom defect pattern in the wafer under test; wherein the custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect patterns.

[0166] Those skilled in the art will understand that the above modules can be distributed in the device as described in the embodiments, or they can be modified accordingly and placed in one or more devices that are unique to this embodiment. The modules in the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0167] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a computer terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0168] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0169] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for restoring the location information of a custom defect in a wafer, characterized in that, Including the following steps: Acquire at least one set of first defect images of at least one wafer under test; At least one set of the first defect images is input into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic. Based on the first location information and the size information of the wafer under test, a second location information of the custom defect pattern is determined; the second location information includes: the actual coordinate information of the custom defect pattern in the wafer under test; the pixel coordinates are (X1, Y1). Correspondingly, determining the second location information of the custom defect pattern based on the first location information and the size information of the wafer under test includes the steps of: determining the relative coordinates (X3, Y3) of at least one pixel based on at least one of the pixel coordinate information, where X3 is the X percentage coordinate of the pixel and Y3 is the Y percentage coordinate of the pixel; determining the actual coordinate information (X2, Y2) of the custom defect pattern based on the relative coordinates and the size information; where X3 = X2 / D, Y3 = Y2 / D, where D is the diameter of the wafer under test. The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics. The method further includes: collecting the second defect image and the corresponding first location information; The custom defect extraction model is periodically updated based on the second defect image and the corresponding first location information.

2. The method according to claim 1, characterized in that, When the first defect image includes multiple custom defect graphics, after the custom defect extraction model identifies the custom defect graphics in the first defect image based on pixel information, the method further includes the step of: the custom defect extraction model splits the multiple custom defect graphics based on pixel information, and outputs multiple second defect images based on the splitting results. And / or, custom defects include: linear defects, and / or arc-shaped defects, and / or ring-shaped defects, and / or dispersed defects, and / or composite defects, and / or wavy defects, and / or clustered defects.

3. The method according to claim 1, characterized in that, It also includes the following steps: Based on the second location information, the corresponding engineering data of the custom defect graphic is obtained.

4. The method according to claim 1, characterized in that, The custom defect extraction model extracts only one type of custom defect image. Accordingly, the method further includes the following steps: The second defect image is subjected to secondary detection, and when a custom defect graphic on the second defect image is detected that does not belong to the extraction category of the corresponding custom defect extraction model, the second defect image is manually labeled.

5. The method according to claim 4, characterized in that, It also includes the following steps: The custom defect extraction model is updated based on the manually labeled second defect image and the corresponding first location information.

6. The method according to claim 1, characterized in that, The steps of acquiring at least one set of first defect images of at least one wafer under test include: Acquire multiple images of at least one of the wafers to be tested for classification; The multiple images to be classified are classified to obtain at least one set of first defect images, and each of the first defect images in the set of first defect images includes at least one custom defect graphic of the same category.

7. An apparatus for restoring the location information of a custom defect in a wafer, characterized in that, include: The image acquisition module is used to acquire at least one type of first defect image of at least one wafer under test; The defect identification module is used to input at least one type of the first defect image into at least one corresponding custom defect extraction model. The custom defect extraction model identifies the custom defect graphic in the first defect image based on pixel information and outputs the first position information of the custom defect graphic and a second defect image including the corresponding custom defect graphic. The first position information includes the pixel coordinates of the custom defect graphic. The defect location module determines the second location information of the custom defect pattern based on the first location information and the size information of the wafer under test; the second location information includes: the actual coordinate information of the custom defect pattern in the wafer under test; The pixel coordinates are (X1, Y1). Correspondingly, determining the second position information of the custom defect pattern based on the first position information and the size information of the wafer to be tested includes: determining the relative coordinates (X3, Y3) of at least one pixel based on at least one of the pixel coordinate information, where X3 is the X percentage coordinate of the pixel and Y3 is the Y percentage coordinate of the pixel; determining the actual coordinate information (X2, Y2) of the custom defect pattern based on the relative coordinates and the size information; where X3 = X2 / D, Y3 = Y2 / D, and D is the diameter of the wafer to be tested. The custom defect extraction model is trained based on the pixel coordinate data of several corresponding custom defect graphics. The device is further configured to: collect the second defect image and the corresponding first location information; The custom defect extraction model is periodically updated based on the second defect image and the corresponding first location information.

8. The apparatus according to claim 7, characterized in that, Also includes: The data storage module is used to store the first defect image collected and input into the custom defect extraction model, as well as the corresponding first location information; The model update module periodically updates the custom defect extraction model based on the first defect image and the corresponding first location information.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images