Systems, methods, and non-transitory computer-readable media for inspecting holes formed in a semiconductor sample

Abstract

A system comprising a processor and memory circuit (PMC) is provided, the PMC being configured to: - obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one layer of a plurality of layers of the semiconductor sample, - segment the image into a plurality of regions, - generate at least one of: o data D pix_强度 providing information of one or more pixel intensities of one or more regions of the plurality of regions 几何形状 , o data D pix_强度 providing information of one or more geometric properties of one or more regions of the plurality of regions 几何形状 , and - feed at least one of D pix_强度 or D 几何形状 to a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates at a target layer of the plurality of layers.

Description

Technical Field

[0001] The currently published topics generally relate to the field of sample inspection, and more specifically to the inspection of automated samples. Background Technology

[0002] Current demands for high density and performance associated with the ultra-large-scale integration of manufactured devices require submicron features, increased transistor and circuit speeds, and improved reliability. These demands necessitate the formation of device features with high precision and uniformity, which in turn requires careful monitoring of the manufacturing process, including automated inspections of the devices while they are still in semiconductor wafer form.

[0003] Manufacturing processes may include forming holes using, for example, etching.

[0004] There is an increasing need to provide an accurate, effective, and robust way to evaluate holes formed by etching processes. Summary of the Invention

[0005] According to certain aspects of the currently disclosed subject matter, a semiconductor sample inspection system is provided, the system including a processor and memory circuitry (PMC), the PMC being configured to: acquire an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample; segment the image into a plurality of regions; and generate data D providing information on the pixel intensity of one or more regions among the plurality of regions. pix_强度 Data D, or data that provides information on one or more geometric properties of one or more of the plurality of regions. 几何形状 D pix_强度 Or D 几何形状 At least one of the layers is fed into a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

[0006] According to some implementations, the output of the trained classifier provides information to classify the hole into one of different categories representing different possible outcomes of the hole's manufacturing process.

[0007] According to some implementations, the different categories include at least one of the following: (i) properly etched holes; (ii) under-etched holes; (iii) first-order over-etched holes; and (iv) second-order over-etched holes, wherein the second-order over-etched holes are deeper than the first-order over-etched holes.

[0008] According to some embodiments, the hole is an inclined hole.

[0009] According to some implementations, the image is not radially symmetrical.

[0010] According to some embodiments, the system is configured to segment the image into multiple segments, on which most or all of the area of ​​the image on which the segmentation is performed corresponds to the image of the hole.

[0011] According to some implementation methods, data D pix_强度 Or data D 几何形状 At least one of them includes data that can be used to determine the depth of the hole.

[0012] According to some implementation methods, data D pix_强度 Or data D 几何形状 At least one of them includes data that provides information on one or more physical properties of the manufacturing process of the hole.

[0013] According to some implementation methods, data D pix_强度 Or data D 几何形状 At least one of them includes data that provides information on one or more physical properties of the manufacturing process of the hole in the NAND memory cell.

[0014] According to some implementations, the image is segmented into multiple N regions R1 to R2. N Among them, for the manufacturing process of the hole, a predefined region R i Provides information on over-etching, where i is between 1 and N, wherein the system is configured to determine the provision of the predefined region R. i Information data D 几何形状 Or data D pix_强度 At least one of them.

[0015] According to some implementations, the predefined region R that provides information on over-etching i This is based on the pixel intensity of the predefined region R relative to other regions. i The pixel intensity is used for selection.

[0016] According to some implementations, N is an integer equal to or greater than 2.

[0017] According to some implementation methods, D 几何形状 The data includes information providing a ratio between the following first data and second data, wherein the first data provides information on the average radius of pixels within a given region of the plurality of regions, and the second data provides information on the average radius of pixels within a second region of the plurality of regions, wherein the second region is different from the given region.

[0018] According to some implementation methods, data D pix_强度 This includes data that provides information on the average pixel intensity of a given region among the plurality of regions.

[0019] According to some implementation methods, D 几何形状 This includes data that provides information about the ratio between the area of ​​a given region and the boundary area of ​​one of the plurality of regions.

[0020] According to some implementation methods, D 几何形状 This includes data that provides information about the distance between the centroid of a given region and the pixel in the given region that is closest to the centroid.

[0021] According to some implementation methods, D pix_强度 This includes data that provides information on the contrast between a given region among the plurality of regions and one or more regions different from the given region.

[0022] According to some implementation methods, a given region among the plurality of regions includes the darkest region and the brightest region, wherein D pix_强度 This includes data that provides information about the contrast between the darkest and brightest areas.

[0023] According to some implementation methods, D pix_强度 This includes data that provides information on the contrast at the boundary between a given region in the plurality of regions and a second region in the plurality of regions that is different from the given region.

[0024] According to some implementation methods, D 几何形状 The data includes information providing ratios between: (i) the distance between the centroid of a given region and the pixel of the given region closest to the centroid, and (ii) the distance between the centroid of the given region and the pixel of the given region furthest from the centroid.

[0025] According to some implementation methods, D is determined for a given region. 几何形状 Provide information on whether the given region has at least a partial ring shape.

[0026] According to some implementation methods, D 几何形状 This includes data that provides information about the number of pixels in a predefined partition of one of the plurality of regions, wherein the region is associated with pixel intensity above a threshold.

[0027] According to some implementation methods, D 几何形状 This includes data that provides information on the side over-etching resulting from the etching process of the hole.

[0028] According to some embodiments, the aperture exhibits a high aspect ratio and has a nanoscale width, wherein the aperture exposes at least one of one or more sets of layers, wherein each set of layers includes layers with different electron yields, and wherein the sample is an intermediate product manufactured through one or more manufacturing stages of a three-dimensional NAND memory cell manufacturing process.

[0029] According to some embodiments, the depth of the pore is between 50 nm and 10,000 nm.

[0030] According to certain aspects of the subject matter of this disclosure, a method for examining a semiconductor sample is provided, the method comprising, via a processor and memory circuit (PMC),: acquiring an image of a hole formed in the semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample; segmenting the image into a plurality of regions; and generating data D providing information on the pixel intensity of one or more regions among the plurality of regions. pix_强度 Data D, or data that provides information on one or more geometric properties of one or more of the plurality of regions. 几何形状 D pix_强度 Or D 几何形状 At least one of the layers is fed into a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

[0031] According to some implementations, the method can achieve one or more of the features of the system described above.

[0032] According to other aspects of the currently disclosed subject matter, a non-transitory computer-readable medium is provided, the non-transitory computer-readable medium including instructions that, when executed by a computer, cause the computer to perform the methods described above.

[0033] According to some implementations, the proposed solution enables the evaluation of one or more characteristics of a hole formed in a semiconductor sample using an image of the hole, even if the image is not radially symmetrical.

[0034] According to some implementations, in the case of deep holes (e.g., holes formed through one hundred layers - this is not limiting), the proposed solution makes it possible to evaluate one or more characteristics of the holes using images of the holes formed in a semiconductor sample.

[0035] According to some implementations, the proposed solution enables the evaluation of one or more characteristics of a hole when the image of the hole has a low signal-to-noise ratio (SNR).

[0036] According to some implementations, when the hole is tilted, the proposed solution enables the evaluation of one or more characteristics of the hole using an image of the hole formed in a semiconductor sample.

[0037] According to some implementations, the proposed solution enables the evaluation of one or more characteristics of a hole formed in a semiconductor sample for one or more cases where prior art has been unsuccessful.

[0038] According to some implementations, the proposed solution detects one or more defects resulting from the manufacturing process of holes in semiconductor samples. Attached Figure Description

[0039] To understand this disclosure and how it can be implemented in practice, embodiments will now be described by way of non-limiting example only, with reference to the accompanying drawings, in which:

[0040] Figure 1 A general block diagram of an inspection system according to certain embodiments of the currently disclosed subject matter is shown.

[0041] Figure 2A A non-limiting example of pores formed in a layer of a sample is shown.

[0042] Figure 2B Another non-limiting example of a pore formed in a layer of a sample is shown, wherein the pore is angled.

[0043] Figure 2C A non-limiting example of a suitable hole for reaching the target layer is shown.

[0044] Figure 2D A non-limiting example of an underetched hole is shown.

[0045] Figure 2E A non-limiting example of a first-order over-etched hole is shown.

[0046] Figure 2F A non-limiting example of a double-over-etched hole is shown.

[0047] Figure 3 A general flowchart is shown for a method of using electronic images of pores to determine the type of pores formed in a sample.

[0048] Figure 4 It shows that it can be accessed through Figure 3 Non-limiting examples of segmented images produced by the method.

[0049] Figure 5A general flowchart is shown for a method of training a classifier using SEM images of holes formed in a sample to determine the category of the holes.

[0050] Figure 6A A general flowchart is shown for a method of determining data that provides information on one or more geometric properties of one or more regions among a plurality of regions present in a segmented image.

[0051] Figure 6B It shows Figure 6A Examples of methods.

[0052] Figure 7 An example of data is shown that provides information about one or more geometric properties of one or more regions among multiple regions present in a segmented image.

[0053] Figure 8 Another example of data is shown that provides information about one or more geometric properties of one or more regions among multiple regions present in a segmented image.

[0054] Figures 9 to 11 Various examples of data are shown that provide information about the intensity of one or more pixels in one or more regions of a segmented image.

[0055] Figure 12 Another example of data is shown that provides information about one or more geometric properties of one or more regions among multiple regions present in a segmented image.

[0056] Figure 13 An example of data used to identify defects in the manufacturing process is shown. Detailed Implementation

[0057] In the following detailed description, numerous specific details are set forth to provide a thorough understanding of this disclosure. However, those skilled in the art will understand that the subject matter of this disclosure can be practiced without these specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to obscure the subject matter of this disclosure.

[0058] Unless otherwise specifically stated, it should be understood, as will be apparent from the following discussion, that the use of terms such as “use,” “obtain,” “divide,” “determine,” “generate,” “output,” “feed,” etc., throughout the specification refers to the actions and / or processing of a computer that manipulates and / or converts data into other data, said data being represented as physical quantities (e.g., electronic quantities) and / or said data representing physical objects. The term “computer” should be broadly interpreted to encompass any kind of hardware-based electronic device with data processing capabilities, including, as a non-limiting example, system 103 and its corresponding portions disclosed in this application.

[0059] The term “sample” as used in this specification should be broadly interpreted to encompass any kind of wafers, masks, intermediate products and other structures, combinations and / or portions thereof, used in the manufacture of semiconductor integrated circuits, magnetic heads, flat panel displays, memories and other semiconductor articles.

[0060] The term "inspection" as used herein should be broadly interpreted to encompass any kind of metrology-related operation, as well as operations related to the detection and / or classification of defects in the sample during sample fabrication. As a non-limiting example, inspection processes may include runtime scanning (in a single or multiple scans), sampling, examination, measurement, classification, and / or other operations performed with respect to the sample or portions thereof using the same or different inspection tools. Similarly, inspection may be performed prior to the fabrication of the sample to be inspected and may include, for example, generating an inspection formula and / or other setup operations. It should be noted that, unless specifically stated otherwise, the term "inspection" or its derivatives as used herein are not limited in terms of the resolution or size of the inspection area. Various non-destructive inspection tools include, as non-limiting examples, scanning electron microscopes, atomic force microscopes, optical inspection tools, etc.

[0061] The term “defect” as used in this specification shall be interpreted broadly to encompass any kind of abnormal or undesirable feature formed on or within a sample.

[0062] It should be understood that, unless otherwise specifically stated, certain features of the currently disclosed subject matter described in the context of individual embodiments may also be provided in combination in individual embodiments. Conversely, various features of the currently disclosed subject matter described in the context of individual embodiments may also be provided individually or in any suitable sub-combination. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the methods and apparatus.

[0063] Keep this in mind, please pay attention. Figure 1 , Figure 1 A functional block diagram of an inspection system according to certain embodiments of the currently disclosed subject matter is shown. Figure 1The illustrated inspection system 100 can be used to inspect samples (e.g., samples of wafers and / or portions thereof) as part of a sample manufacturing process. The illustrated inspection system 100 includes a computer-based system 103 capable of automatically determining metrologically relevant and / or defect-related information using images acquired during sample manufacturing. System 103 can be operatively connected to one or more low-resolution inspection tools 101 and / or one or more high-resolution inspection tools 102 and / or other inspection tools. The inspection tools are configured to capture images and / or view the captured images and / or enable or provide measurements associated with the captured images.

[0064] System 103 includes a processor and memory circuitry (PMC) 104. PMC 104 is configured to provide all the processing required by operating system 103, as detailed below (see [link to documentation]). Figure 3 , Figure 5 The methods described herein (which may be executed at least partially by system 103) include a processor (not shown separately) and memory (not shown separately). The processor of PMC 104 may be configured to execute one or more functional modules according to computer-readable instructions implemented on a non-transitory computer-readable memory included in the PMC. Such functional modules are referred to below as being included in the PMC. The functional modules included in PMC 104 include classifier 112. Classifier 112 may include any adapted classifier, such as a neural network classifier (such as a deep neural network (DNN)), decision tree, or SVM (Support Vector Machine) algorithm. As explained below, classifier 112 is typically trained using supervised learning.

[0065] As a non-limiting example, the layers of classifier 112 can be organized according to a Convolutional Neural Network (CNN) architecture, a Recurrent Neural Network (RNN) architecture, a Generative Adversarial Network (GAN) architecture, or other methods. Optionally, at least some of these layers can be organized in multiple DNN subnetworks. Each layer of the DNN can include multiple basic computational elements (CEs), commonly referred to in the art as dimensions, neurons, or nodes.

[0066] Typically, computational elements of a given layer can be connected to CEs (Computational Elements) of the preceding and / or following layers. Each connection between a CE of a preceding layer and a CE of a following layer is associated with a weighted value. A given CE can receive input from a CE of a preceding layer through corresponding connections, each given connection being associated with a weighted value of the input that can be applied to the given connection. The weighted values ​​determine the relative strength of the connection and thus the relative impact of the corresponding input on the output of the given CE. A given CE can be configured to compute activation values ​​(e.g., a weighted sum of inputs) and further derive the output by applying an activation function to the computed activation values. The activation function can be, for example, an identity function, a deterministic function (e.g., a linear function, a sigmoid function, a threshold function, etc.), a random function, or other suitable function. The output from a given CE can be passed to CEs of subsequent layers through corresponding connections. Similarly, as described above, each connection at the output of a CE can be associated with a weighted value that can be applied to the output of the CE before being received as input to the CE of a subsequent layer. In addition to weighted values, there can also be thresholds (including constraint functions) associated with the connections and CEs.

[0067] The weights and / or thresholds of classifier 112 can be initially selected before training and can be further iteratively adjusted or modified during training to achieve an optimal set of weights and / or thresholds in the trained classifier 112. After each iteration, the difference (also known as the loss function) between the actual output produced by classifier 112 and the target output associated with the corresponding training dataset can be determined. This difference can be referred to as the error value. Training can be determined to be complete when the cost function or loss function indicating the error value is less than a predetermined value, or when a finite change in performance is achieved between iterations.

[0068] System 103 is configured to receive input data 121, 122. Input data 121, 122 may include data (and / or its derivatives and / or associated metadata) generated by inspection tools 101, 102. It should be noted that input data 121, 122 may include images (e.g., captured images, images derived from captured images, analog images, synthetic images, etc.) and associated digital data (e.g., metadata, handcrafted attributes, etc.). It should also be noted that image data may include data related to the layer of interest and / or one or more other layers of the sample.

[0069] System 103 is further configured to process at least a portion of the received input data 121, 122, and send the results (or a portion thereof) to storage system 107, to inspection tools, to a computer-based graphical user interface (GUI) 108 for presenting the results, and / or to an external system (e.g., FAB's Yield Management System (YMS)).

[0070] As a non-limiting example, the sample may be examined by one or more low-resolution inspection machines 101 (e.g., optical inspection systems, low-resolution SEMs, etc.). Result data 121, providing information about low-resolution images of the sample, may be transmitted directly to system 103 or via one or more intermediate systems. Alternatively or additionally, the sample may be examined by a high-resolution machine 102 (e.g., scanning electron microscope (SEM) or atomic force microscope (AFM)). Result data 122, providing information about high-resolution images of the sample, may be transmitted directly to system 103 or via one or more intermediate systems.

[0071] Those skilled in the art will readily understand that the teachings of the currently disclosed subject matter are not subject to Figure 1 The system constraints shown; equivalent and / or modified functionality may be combined or divided in another way and may be implemented in any suitable combination of software and firmware and / or hardware.

[0072] Without limiting the scope of this disclosure in any way, it should also be noted that the inspection tool can be implemented as various types of inspection machines, such as optical imaging machines, electron beam inspection machines, etc. In some cases, the same inspection tool can provide both low-resolution and high-resolution image data. In some cases, at least one inspection tool can have metrological capabilities.

[0073] It should be noted that, Figure 1 The inspection system shown can be implemented in a distributed computing environment. Figure 1 The functional modules described above can be distributed across several local and / or remote devices and can be connected via a communication network. It should also be noted that in other embodiments, at least some of the inspection tools 101 and / or 102, storage system 107 and / or GUI 108 can be external to the inspection system 100 and operate to communicate data with system 103. System 103 can be implemented as a stand-alone computer used in conjunction with the inspection tools. Alternatively, the corresponding functions of the system can be at least partially integrated with one or more inspection tools.

[0074] Now pay attention Figure 2A .

[0075] During the fabrication of a semiconductor sample, it may be necessary to form holes in one or more layers of the sample. These holes can be formed using an etching process. After the etching process, the holes can be filled with at least one conductive material. This enables the fabrication of structural elements, such as (but not limited to) conductors (e.g., contacts).

[0076] Non-limiting examples of semiconductor samples include three-dimensional NAND memory cells. It should be noted that the fabrication of semiconductor samples typically involves multiple fabrication stages. Forming vias through etching is one of these stages. The result of each fabrication stage is a semiconductor sample, often referred to as an "intermediate product." The output of the final fabrication stage provides the final semiconductor sample (e.g., a three-dimensional NAND memory cell).

[0077] Figure 2A A non-limiting example of a sample (e.g., an intermediate product) is shown, the sample comprising multiple layers. In some embodiments, the sample comprises pairs (see 200-208) of conductive layers (see 200(2) to 208(2)) and non-conductive layers (see 200(1) to 208(1)). Each pair of layers may include a non-conductive layer situated on top of a conductive layer. However, this is not limiting, and the sample may include layers not arranged according to the paired layers. Figures 2A to 2F The diagram depicts nine pairs of layers. This is not limiting, and different numbers of layers can be used. It should be noted that a set of layers may include more than two layers with different electron yields, and any set of layers may include multiple conductivity levels.

[0078] The nine layers include a first non-conductive layer 200(1), a first conductive layer 200(2), a second non-conductive layer 201(1), a second conductive layer 201(2), a third non-conductive layer 202(1), a third conductive layer 202(2), a fourth non-conductive layer 203(1), a fourth conductive layer 203(2), a fifth non-conductive layer 204(1), a fifth conductive layer 204(2), a sixth non-conductive layer 205(1), a sixth conductive layer 205(2), a seventh non-conductive layer 206(1), a seventh conductive layer 206(2), an eighth non-conductive layer 207(1), an eighth conductive layer 207(2), a ninth non-conductive layer 208(1), and a ninth conductive layer 208(2).

[0079] As described above, a non-limiting example of this sample is a three-dimensional NAND memory cell. Each pair can be considered as a set of layers with different electron yields. Therefore, more electrons are expected to be emitted from the conductive layers of a pair compared to electrons emitted from the non-conductive layers.

[0080] In some implementations, these layer pairs may be arranged in a stepped configuration. Each step comprises a single pair. The stepped configuration allows each pair to be connected to a conductor. In some implementations, these layers form a stepped structure in which each group of layers is wider than all groups of layers above it. This is not limiting.

[0081] Ideally, the formed conductor is vertical, but deviations in the manufacturing process can cause the hole to be non-vertical (tilted), which in turn causes the formed conductor to be non-vertical. Figure 2B A non-limiting example is shown. This type of deviation can be particularly prevalent when the hole is deep: because the hole must pass through numerous layers, the manufacturing process can cause the hole to tilt. Due to this tilt, the final image of the hole may be asymmetrical. In the non-limiting example, the tilt angle of the hole (relative to the vertical direction) is less than 1 degree.

[0082] In some implementations, the holes exhibit a high aspect ratio. The term "aspect ratio" refers to the ratio between the depth and width of an element. A high aspect ratio can be considered to be an aspect ratio exceeding, for example, 5:1.

[0083] In some implementations, the pores have a width on the nanoscale. The term "nanoscale" refers to a value ranging from 5 nm to 100 nm.

[0084] In some implementations, the height (also known as the depth) of the aperture is in the range of 50 nm to 10,000 nm (or any sub-interval within this range).

[0085] exist Figures 2A to 2E In this configuration, these layers are covered by a top region 220 (which may be, for example, an insulator or a top cover layer). These layers may be multiple regions of various shapes and / or sizes and / or covered by other layers.

[0086] Please note now. Figures 2C to 2F It shows an example of a first hole 230, a second hole 232, a third hole 234 and a fourth hole 236 formed in a portion of a sample (e.g., an intermediate product).

[0087] For simplicity of explanation, each diagram illustrates a single hole for a different target area. However, this is not limiting.

[0088] Figure 2C The first hole 230 is a suitable hole. In fact, it terminates at the target layer and does not expose layers located below the target layer. Figure 2CIn the example, the first hole 230 exposes the conductive second target layer 202(2) which belongs to the third group layer 202, and the second non-conductive layer 202(1) which is located above the second target layer 202(2) and also belongs to the third group layer 202.

[0089] Note Figure 2C The representation is illustrative: typically, the hole passes through multiple layers (e.g., a hundred - this is not limiting) until the hole reaches the target layer. Figure 2C Only the last two layers (the deepest layers) through which the hole passes are depicted.

[0090] supply Figure 2C The image 231 (electronic image) of the information of the first hole 230 includes a center 2311 representing the second target layer 202 (2) and a peripheral region 2312 representing the second non-conductive layer 202 (1).

[0091] In some implementations, the image of the hole is asymmetrical (non-radial symmetry). This is in Figure 2C As shown in the non-limiting example, image 231 in this figure is not radially symmetrical (due to the presence of feature 2322). In some embodiments, most of the image of the hole is asymmetrical.

[0092] Figure 2D The second hole 232 is an under-etched hole. In other words, the hole depth is insufficient, so the second hole 232 does not expose the target layer. The second hole 232 does not reach the fourth target layer 203(2), which is conductive and belongs to the fourth group of layers 203. The second hole 232 only exposes the fourth non-conductive layer, which is located above the fourth target layer 203(2) and also belongs to the fourth group of layers 203.

[0093] Note Figure 2D The representation is illustrative: typically, the hole passes through multiple layers (e.g., one hundred - this is not limiting) until it reaches layer 203 (1).

[0094] supply Figure 2D The image 233 (electronic image) of the information of the second hole 232 includes the center 2331 and the peripheral region 2312 representing the fourth non-conductive layer 203(1).

[0095] In some implementations, the image of the under-etched hole is asymmetrical.

[0096] Figure 2E The third hole, 234, is a first-order over-etched hole. This type of hole exposes the target layer, but also exposes a (single) additional layer located beneath the target layer. (As shown in...) Figure 2EAs can be seen, the third hole 234 exposes the fifth non-conductive layer 204(1) and the sixth non-conductive layer 205(1). The fifth non-conductive layer 204(1) is located above the fifth target layer 204(2) and belongs to the fifth group of layers 204. The fifth target layer 204(2) is conductive. The sixth non-conductive layer 205(1) is located below the fifth target layer 204(2) and belongs to the sixth group of layers 205. The third hole 234 penetrates the entire fifth target layer 204(2) but does not penetrate the entire sixth non-conductive layer 205(1).

[0097] Note Figure 2E The representation is illustrative: typically, the hole passes through multiple layers (e.g., a hundred - this is not limiting) until the hole reaches the target layer. Figure 2E Only the deepest layer through which the hole passes is depicted.

[0098] supply Figure 2E The image 235 (e.g., SEM image) of the third hole 234 includes a first (inner) portion 2351 representing an over-etched portion of the sixth non-conductive layer 205(1) exposed, a second portion 2352 representing the fifth target layer 204(2) exposed, and a peripheral region 2353 representing the fifth non-conductive layer 204(1).

[0099] In some implementations, the image of the first-stage over-etched hole is asymmetrical (lacking radial symmetry). This is in Figure 2E The non-restrictive examples are shown.

[0100] Figure 2F The fourth hole, 236, is a secondary over-etched hole. This hole exposes the target layer, but also two additional layers located beneath it. (As shown in...) Figure 2F As can be seen, the fourth hole 236 exposes the sixth non-conductive layer 205(1), the seventh non-conductive layer 206(1), and the seventh conductive layer 206(2). The sixth non-conductive layer 205(1) is located above the sixth target layer 205(2) and belongs to the sixth group of layers 205. The sixth target layer 205(2) is conductive. The seventh non-conductive layer 206(1) is located below the sixth target layer 205(2) and belongs to the seventh group of layers 206. The seventh conductive layer 206(2) is located below the seventh non-conductive layer and also belongs to the seventh group of layers 206.

[0101] Note Figure 2F The representation is illustrative: typically, the hole passes through multiple layers (e.g., a hundred - this is not limiting) until the hole reaches the target layer. Figure 2F Only the deepest layer through which the hole passes is depicted.

[0102] The fourth hole 236 penetrates the entire sixth target layer 205 (2) and also penetrates the entire seventh non-conductive layer 206 (1). If the fourth hole 236 is filled with conductive material, the sixth and seventh conductive layers will be short-circuited.

[0103] supply Figure 2F The image 237 (e.g., SEM image) of the fourth hole 236 includes a center 2371 (representing the seventh conductive layer 206(2)), a partially annular first intermediate region 2372 (representing the seventh non-conductive layer 206(1)), a second intermediate region 2373 (representing the sixth conductive layer 205(2)), and a peripheral region 2374 representing the sixth non-conductive layer 205(1).

[0104] In some implementations, the image of the secondary over-etched via is asymmetrical (lacking radial symmetry). This is in Figure 2F The non-limiting example shown is that, instead of obtaining the annular (loop) portions of the corresponding regions 2372 and 2374, only the portions of the loop are obtained.

[0105] Now pay attention Figure 3 .

[0106] Figure 3 The method includes obtaining (operation 300) an image (electronic image) of a hole formed in a semiconductor sample.

[0107] According to some embodiments, a large portion (or all) of the area of ​​the image obtained at operation 300 includes the image of the hole. According to some embodiments, an initial image of the sample may be obtained first, and then processed to extract an image from which only the hole (or a large portion of the hole) is included. In other words, the image of the hole is extracted from its background. Extracting the hole from the background in the initial image may rely on, for example, an initial segmentation process (different from the segmentation performed at operation 310 and described below).

[0108] As referenced above Figures 2A to 2E The sample comprises multiple layers. A hole exposes at least one layer from one or more sets of layers. The hole can expose one layer from a single set of layers, multiple layers from a single set of layers, or even layers from two or more sets of layers.

[0109] Image generation may include irradiating the hole with a beam of charged particles and detecting electrons emitted from the hole. The image may be provided, for example, by inspection tools 101 and / or 102.

[0110] Operation 300 may include acquiring images through or from an inspection tool (without performing electronic image acquisition processing).

[0111] Image acquisition and processing may include processing a detection signal representing the detection of a detected electron. Processing of the detection signal may include applying noise reduction, smoothing, and / or providing at least one of the following: an electron image in a certain format. The specific format may be a grayscale format, but other formats may also be provided. The detected electron may be a secondary electron, a backscattered electron, etc. Electron image acquisition and processing may be performed using a scanning electron microscope (SEM) (via a critical-size SEM or a defect inspection SEM), an electron beam imager, etc.

[0112] Figure 3 The method further includes (operation 310) segmenting the image into multiple regions. Image segmentation divides the image into groups of pixels based on pixel intensity distribution (e.g., gray-level intensity distribution) and spatial connectivity. Operation 310 may rely on algorithms such as edge detection, gray-level segmentation, clustering, K-means, etc. These examples are not limiting.

[0113] According to some implementations, most (or all) of the image to which segmentation (operation 310) is performed corresponds to the image of the hole. Therefore, the segmentation of operation 310 is performed primarily (or only) on the image of the hole.

[0114] In some implementations, the number of regions used to segment the image is stored as a predefined value. In some implementations, the number of regions is a parameter of the segmentation algorithm, which can be selected, for example, by the operator. In some implementations, the number of regions can vary depending on the type of sample, the type of hole, etc.

[0115] Figure 4 A non-limiting example of a segmented image 400 obtained as the output of operation 310 is shown.

[0116] exist Figure 4 In a non-limiting example, image 400 has been divided into five regions: 4001, 4002, 4003, 4004, and 4005. Region 4001 is the brightest region (with the lowest pixel intensity), and region 4005 is the darkest region (with the highest pixel intensity). Note that the number of regions can be greater than five. In some implementations, it can be chosen as an integer equal to or greater than 2. The value of N can be chosen to differentiate between regions.

[0117] Figure 3 The method may further include (operation 320) generating data D that provides information on the intensity (grayscale intensity) of one or more pixels in one or more regions of a plurality of regions. pix_强度 The segmented image generated at operation 310 can be used to obtain D. pix_强度 The following text provides D pix_强度 Various examples. In some implementations, Dpix_强度 Information on the pixel intensity of a given region among multiple regions can be provided. In some implementations, D pix_强度 Provides information on the contrast between a given area and other areas. These examples are not limiting.

[0118] Figure 3 The method may further include (operation 320) generating data D that provides information on one or more geometric properties of one or more regions among a plurality of regions. 几何形状 The segmented image generated at operation 310 can be used to obtain D. 几何形状 The following text provides D 几何形状 Various examples. In some implementations, D 几何形状 It can provide information about the shape of the region. For example, it can indicate whether the region tends to have a ring (circular) shape (or at least a partial ring). In other examples, D 几何形状 It can provide information about the integrity of the ring-shaped region: it can indicate whether the region is a complete ring or only a partial ring. In other examples, D 几何形状 Information about the size of the region can be provided (e.g., whether the region is a thin or thick annular region).

[0119] In some implementations, only D is calculated. pix_强度 In other implementations, only D is calculated. 几何形状 In other implementations, D is calculated. pix_强度 and D 几何形状 Both.

[0120] According to some implementation methods, data D pix_强度 Or data D 几何形状 At least one of these includes data that can be used to determine the depth of the hole. In fact, as explained below, data D... pix_强度 Or data D 几何形状 At least one of the data can be fed into a classifier, which outputs data providing information about the depth of the hole. Specifically, data D pix_强度 Or data D 几何形状 At least one of them can be used for indirect depth measurement of holes using classification processing (direct depth measurement of holes is not required).

[0121] According to some implementation methods, data D pix_强度 Or data D 几何形状 At least one of them includes data that provides information on one or more physical properties of the hole manufacturing process. Specifically, according to some embodiments, data D pix_强度 Or data D 几何形状At least one of them includes data that provides information about one or more physical properties of the manufacturing process of the holes in the NAND memory cell. Therefore, data D pix_强度 Or data D 几何形状 At least one of these may include data fitted to the manufacturing process used to manufacture NAND memory cells. Data D is given below. pix_强度 Or data D 几何形状 Different examples.

[0122] In other words, data D pix_强度 Or data D 几何形状 At least one of these includes data selected to have physical meaning. Such data may contain physical information related to the manufacturing process. Specifically, such data may provide information on the (physical) properties / nature of the manufacturing process (e.g., whether the manufacturing process enables the hole to reach the target layer, or alternatively, whether the manufacturing process produces over-etching or under-etching).

[0123] In some implementations, the fabrication process of the aperture (in the case of over-etching) and the analysis of the corresponding electronic images can reveal the process when the image is segmented into multiple N regions R1 to R2. N At that time, the predefined region R i (Where i is between 1 and N) is of particular significance for characterizing the pores. Specifically, this predefined region R... i Information on over-etching (e.g., primary over-etched holes or secondary over-etched holes) can be provided. Therefore, Figure 3 The method may include determining which predefined region R is provided. i Information data D 几何形状 and / or data D 灰度_等级 .

[0124] According to some implementations, N is an integer between 2 and 10.

[0125] In some implementations, a predefined region R provides information about over-etching. i It is selected based on the pixel intensity of that region relative to the pixel intensity of other regions. For example, suppose regions R1 to R... N It is sorted based on its brightness (R1 is the brightest region, R...). N If the region R is the darkest region, and N equals 5, then region R is the darkest region. i This can correspond to region R3 (the third dark region). Figure 4 In this context, region R3 corresponds to region 4003. This is not limiting and can vary depending on the manufacturing process. For different manufacturing processes, it is possible that region R provides information about over-etching. i It could be the brightest area R1, or the second darkest area R2, and so on. Area R iIt can also depend on the number N chosen for the segmentation. Relevant region R i This can be determined through analysis of the manufacturing process (in the case of over-etching) and the corresponding electronic images. This analysis can indicate which areas are most relevant to characterizing over-etching.

[0126] Figure 3 The method further includes (operation 330) D pix_强度 and / or D 几何形状 The data is fed into a trained classifier, such as trained classifier 112. In some implementations, trained classifier 112 has been trained to be based on D... pix_强度 and / or D 几何形状 Classify the holes into at least one of several categories.

[0127] In some implementations, for a given image of a hole, multiple different attributes D (of different types) pix_强度 and / or multiple different attributes D (of different types) 几何形状 It is fed into the trained classifier 112. D pix_强度 and / or D 几何形状 Various examples of different attributes are provided below. Therefore, the trained classifier 112 is able to determine the most suitable category of holes present in an image based on multiple different attributes.

[0128] In some implementations, the trained classifier 112 has been trained to perform information to classify holes into one of different categories representing different possible outcomes of the hole manufacturing process.

[0129] In some implementations, multiple categories may include appropriately etched holes (non-limiting examples of such holes are shown in...). Figure 2C (Depicted in the middle), under-etched holes (non-limiting examples of such holes are in) Figure 2D (Depicted in the middle), first-order over-etched holes (non-limiting examples of such holes are in...) Figure 2E (as depicted in the image), and secondary over-etched holes, wherein the secondary over-etched holes are deeper than the primary over-etched holes (non-limiting examples of such holes are shown in the image). Figure 2F (As depicted in the text). Note that different numbers / different types of categories can be used.

[0130] In some implementations, the multiple categories may further include categories indicating defects (also known as black hats, because they are characterized by clusters of dark pixels concentrated in a given area). This will be discussed below regarding... Figure 13 Further discussion.

[0131] The output of the trained classifier 112 can be used to determine whether a hole terminates at a target layer among multiple layers. In fact, if the trained classifier outputs a hole that belongs to the "properly etched hole" category, this can be used (e.g., to indicate to a user and / or a computerized system) that the hole terminates at a target layer among multiple layers.

[0132] If the trained classifier 112 outputs a hole that belongs to a category that is not "properly etched hole," this can be used (e.g., to indicate to a user and / or computerized system) that the hole does not terminate at a target layer in multiple layers (which corresponds to a defective hole). It should be noted that the output of the trained classifier 112 can provide details about the type of defective hole: the output of the trained classifier 112 can be used to indicate whether the defective hole is an under-etched hole, a first-order over-etched hole, or a second-order over-etched hole.

[0133] Figure 5 The method for training classifier 112 is shown.

[0134] Figure 5 The method includes obtaining (operation 500) training images (e.g., SEM images) of holes formed in a semiconductor sample. In some implementations, the training images may be obtained via an image acquisition device (see...). Figure 1 The images acquired (101 and 102 in the image) can be analog images.

[0135] Figure 5 The method further includes (operation 510) segmenting the image into multiple regions. Operation 510 is similar to operation 310 and will not be described again. In some embodiments, instead of acquiring and segmenting the image, Figure 5 The method involves obtaining an image of the already segmented holes.

[0136] In some implementations, the number of regions used to segment the image is predefined, or can be selected by the operator as a parameter of the segmentation algorithm.

[0137] Figure 5 The method may further include (operation 520) generating data D that provides information on the intensity of one or more pixels in one or more of a plurality of regions. pix_强度。 D can be obtained using the segmented image generated at operation 510. pix_强度 The following text provides D pix_强度 Various examples.

[0138] Figure 5 The method may further include (operation 520) providing data D that provides information on one or more geometric properties of one or more regions among a plurality of regions. 几何形状 The segmented image generated at operation 510 can be used to obtain D.几何形状 The following text provides D 几何形状 Various examples.

[0139] In some implementations, only D is calculated. pix_强度 In other implementations, only D is calculated. 几何形状 In other implementations, D is calculated. pix_强度 and D 几何形状 Both.

[0140] Figure 5 The method further includes (operation 525) obtaining labels that provide information about the type (category) of holes present in the image. The labels may be provided by an operator (supervised learning). In some implementations, the labels may indicate one of the following types (categories): properly etched holes, under-etched holes, first-order over-etched holes, and second-order over-etched holes, wherein second-order over-etched holes are deeper than first-order over-etched holes. In some implementations, the labels may indicate the presence of defects (present in dark areas of the segmented image) (such defects may be characterized by clusters of dark pixels concentrated at predefined locations).

[0141] Figure 5 The method further includes (operation 530) combining the label and D pix_强度 and / or D 几何形状 The data is fed into classifier 112 for its training. Thus, classifier 112 is trained to classify holes into one of different categories representing different possible outcomes of the hole's manufacturing process. Specifically, the category corresponds to the type of hole present in the label.

[0142] In some implementations, the multiple categories may include properly etched holes, under-etched holes, first-order over-etched holes, and second-order over-etched holes, wherein the second-order over-etched holes are deeper than the first-order over-etched holes. It should be noted that different numbers / types of categories may be used. In some implementations, the multiple categories may further include a category indicating defects (present in dark areas of the segmented image). Such defects may be characterized by clusters of dark pixels concentrated at predetermined locations.

[0143] In some implementations, a simplified list of categories may be used: a first category of indicator holes terminates at the target layer, and a second category of indicator holes does not terminate at the target layer.

[0144] It can be repeated for multiple training images Figure 5 The method. Note attribute D. pix_强度 and / or D 几何形状 The use of this allows classifiers to be trained without requiring excessively large training sets.112

[0145] As can be understood, this includes data D determined for each of the multiple training images.pix_强度 and / or D 几何形状 And a training set of labels (indicating the category of each hole in each of the multiple training images), enabling the training of a classifier to automatically generate a model (a computer-implemented model) that will assign attributes (D) pix_强度 and / or D 几何形状 (Link to the category of the hole)

[0146] Please note now. Figure 6A and Figure 6B .

[0147] According to some implementation methods, first data can be determined to provide information about the average radius 610 of pixels belonging to a given region among multiple regions. (Operation 605). This average radius can be calculated relative to the centroid of a given region. This given region can correspond to an interior region of the image. It should be noted that the value of radius 610 is smaller when the given region tends to be a continuous region, such as a disk (rather than a ring-shaped or partially ring-shaped region). In other words, this property can provide information about whether a given region tends to have a (at least partially) ring-shaped shape.

[0148] In some implementations, it is assumed that the image has been segmented into five regions 4001 to 4005, such as Figure 4 As shown. A given region can be selected as the third dark region 4003 (the third highest grayscale intensity) in this segmentation. In practice, in some embodiments, this given region 4003 can provide information about over-etching. It should be noted that this is not limiting, and in other embodiments, depending on the manufacturing process, another region may be considered.

[0149] In some implementations, the first data provides information about the radius 610. It can be scaled. Specifically, the method may include determining (operation 615) the second data. The second data provides information on the average radius 620 of pixels within a second region, which is different from the given region. In some implementations, the second region may be closer to the center of the image than the given region.

[0150] In some implementations, it is assumed that image 600 has been segmented into five regions 4001 to 4005. The second region can be selected as the brightest region 4001 (lowest gray level intensity) in the segmentation.

[0151] In some implementations, the following non-limiting formula can be used to estimate the average radius of the second region (region 4001). Second data information:

[0152]

[0153] In this formula, the area of ​​the second region 4001 is recorded as "the area of ​​4001".

[0154] supply and Information about the ratio between them can be determined and used as data D. 几何形状 .

[0155] For example, the ratio can be determined as follows:

[0156]

[0157] when When the value tends to be high (e.g., 100%), this tends to indicate that a given region (e.g., region 4003) has a ring-shaped shape, while when... When the value is small, this tends to indicate that a given area (e.g., area 4003) does not have a ring shape. As mentioned above, in some embodiments, the ring shape of a given area can be an indicator of the presence of over-etching. This property can be used by a trained classifier 112 to determine the type of hole.

[0158] According to some implementation methods, D pix_强度 This may include data that provides information on the average pixel intensity of a given region among the plurality of regions. In particular, in some embodiments, the data providing the average pixel intensity information is determined for regions that provide information on over-etching.

[0159] In some implementations, it is assumed that the image has been segmented into five regions 4001 to 4005, such as Figure 4 As shown, the average pixel intensity of the third dark region 4003 (the third highest grayscale intensity) in the segmentation can be determined. As mentioned above, in some embodiments, region 4003 provides information about possible over-etching. Therefore, the average pixel intensity of this region 4003 can be used by classifier 112 to determine the type of hole. It should be noted that this is not limiting, and in other embodiments, depending on the manufacturing process, another region may be considered.

[0160] If the average pixel intensity of region 4003 is high, this may indicate over-etching, and if the average pixel intensity of region 6003 is low, this may indicate that over-etching is not present. It should be noted that this is not restrictive, and the classification is performed by classifier 112.

[0161] Now pay attention Figure 7 .

[0162] According to some implementation methods, D 几何形状 This includes data that provides information about the ratio between the area of ​​a given region and the area of ​​its boundaries among multiple regions (of a segmented image). In some implementations, the given region may be a region that provides information about over-etching.

[0163] If this ratio is high, the likelihood of a given region having a ring or toroidal shape is low (in fact, this ratio indicates that the given region of interest fills most of the boundary area). This can indicate that the possibility of over-etching is low.

[0164] If the ratio is low, there is a high probability that a given region has a ring or annular shape (in effect, the ratio indicates that the given region of interest only fills a portion of the boundary area). This can indicate a high probability of over-etching (because such annular regions are often present when over-etching occurs). This is not limiting, and the decision regarding hole classification is performed by classifier 112.

[0165] exist Figure 7 In the example, the boundary area is a bounding box of 700 (e.g., a bounding rectangle or square). In this example, the given region is region 4003 (out of five regions) with the third highest pixel intensity. However, this is not limiting. Figure 7 In the example, the ratio is smaller, which indicates that region 4003 tends to have a shape that is at least partially annular.

[0166] Now pay attention Figure 8 .

[0167] According to some implementation methods, D 几何形状 Information is provided regarding the distance between the centroid of a given region and the pixel in the given region closest to the centroid. In other words, the minimum radius of the given region is determined. In some implementations, the given region may be a region that provides information about over-etching.

[0168] Figure 8 An example is provided. In this example, the given region is region 4003 (among the five regions output by segmentation) with the third highest pixel intensity. However, this is not limiting. The centroid of the given region 4003 is labeled 801. The distance between this center 801 and the nearest pixel of region 4003 is labeled 802. This distance can be calculated using the "distance transform" operator.

[0169] If this distance is large, there is a high probability that the given area has a ring or loop shape. This could indicate a high likelihood of over-etching.

[0170] If this distance is small, the probability that a given area has a ring or loop shape is low. This can indicate that the probability of over-etching is low. This is not limiting, and the final decision regarding hole classification is performed by classifier 112.

[0171] Now pay attention Figure 9 .

[0172] According to some implementation methods, D pix_强度 This includes data that provides information about pixel intensity contrast between a given region and one or more other regions different from the given region. In some implementations, the given region may be a region that provides information about over-etching.

[0173] Figure 9 An example is provided. Suppose a given region corresponds to region 4003 (among the five regions output by segmentation) with the third highest pixel intensity. The contrast between the given region 4003 and two other regions different from the given region (region 4001 with the lowest pixel intensity and region 4002 with the second highest pixel intensity) can be calculated. In other words, the contrast between the given region 4003 and its background (local background) is calculated. It should be noted that this is not limiting, and contrast can be calculated for different areas.

[0174] In some implementations, the contrast ratio can be calculated as follows:

[0175]

[0176] In this formula, The pixel intensity corresponding to the background (e.g., regions 4001 and 4002), and The pixel intensity corresponding to a given region (e.g., region 4003).

[0177] If the calculated contrast is high, this can indicate a high probability of over-etching. In fact, high contrast indicates that a given area 4003 is strongly different from its background, and therefore can indicate over-etching.

[0178] If the calculated contrast is low, this indicates a low probability of over-etching. This is not limiting, and the final decision regarding hole classification is performed by classifier 112.

[0179] Now pay attention Figure 10 .

[0180] Each region output by segmentation can include various levels of pixel intensity. Assume a given region among multiple regions includes the darkest region (the pixel cluster with the highest pixel intensity in this given region) and the brightest region (the pixel cluster with the lowest pixel intensity in this given region). According to some implementations, D... pix_强度 This may include data that provides information about the contrast between the darkest and brightest areas of a given region. In some implementations, the given region may be a region that provides information about over-etching.

[0181] Figure 10 Non-restrictive examples are provided.

[0182] Suppose a given region corresponds to region 4003 (among the five regions output by segmentation) which has the third highest pixel intensity. The contrast between the darkest region 1000 of given region 4003 and the brightest region 1010 of given region 4001 can be calculated. This is not limiting, and such calculation can be performed for different regions.

[0183] If the calculated contrast is high, the area is unlikely to have a ring shape, which in turn indicates a low probability of over-etching.

[0184] If the calculated contrast is low, then... area The likelihood of a ring shape is high, and consequently, this could indicate a high probability of over-etching. This is not limiting, and the final decision regarding hole classification is performed by classifier 112.

[0185] According to some implementation methods, D pix_强度 This includes data that provides information on the contrast at the boundary between a given region and a second region different from the given region. The contrast at the boundary can be calculated using the Kolmogorov-Smirnov method (e.g., available in the SciPy library).

[0186] In some implementations, the second region is located within a given region.

[0187] In some implementations, a given area may be an area that provides information about over-etching.

[0188] Figure 11 An example is provided. Assume a given region corresponds to region 4003 (out of five regions) with the third highest pixel intensity. However, this is not limiting. Assume the second region corresponds to region 4001 with the lowest pixel intensity. The boundary between region 4003 and region 4001 is labeled 1105. Calculate the pixel intensity contrast between region 10003 and region 10001 at boundary 1005.

[0189] If the calculated contrast is high, there is a high probability that the given region is a ring (i.e., annular). In effect, this indicates a strong difference between the inner second region and the outer given region. Therefore, there is a high probability of over-etching.

[0190] If the calculated contrast is small, the probability that a given region is a ring (i.e., annular) is low. In fact, this indicates that the inner second region and the outer given region are actually almost identical continuous regions. Therefore, the probability of over-etching is small. This is not limiting, and the final decision regarding hole classification is performed by classifier 112.

[0191] Now pay attention Figure 11 .

[0192] According to some implementation methods, D 几何形状 This includes data that provides information on the ratios between the following items:

[0193] - The distance between the centroid of a given region and the pixel in the given region closest to the centroid is 1200, and

[0194] - The distance between the centroid of a given region and the pixel furthest from the centroid of the given region is 1210.

[0195] This attribute indicates whether the shape of a given area tends to be disk-shaped or toroidal. This attribute can also provide an indication of the thickness of a toroidal shape (or a partially toroidal shape).

[0196] In some implementations, a given area may be an area that provides information about over-etching.

[0197] Assume a given region corresponds to region 4003 (among the five regions output by segmentation) with the third highest pixel intensity. It should be noted that this is not limiting, and another region may be selected in different implementations. This ratio may correspond to the ratio between distance 1200 and distance 1210, or vice versa.

[0198] If the calculated ratio is high, the likelihood of the area having a ring shape (or a partial ring shape) is low, which in turn indicates a low likelihood of over-etching.

[0199] If the calculated ratio is small, the region is highly likely to have a ring shape, which in turn may indicate a high probability of over-etching. This is not limiting, and the final decision regarding hole classification is performed by classifier 112.

[0200] Now pay attention Figure 13 .

[0201] In some implementations, data can be used to determine the number of pixels in a predefined partition (i.e., a partition with a predefined location) of a region among multiple regions, wherein the region is associated with a pixel intensity above a threshold.

[0202] In fact, for at least one manufacturing process, it has been found that pixel clusters indicate defects at predefined locations in the "dark" areas of the segmented image. Therefore, the aforementioned properties can be fed into a trained classifier 112, which can determine whether a defect actually exists. It should be noted that such defects may exist even without over-etching.

[0203] This defect may originate from the manufacturing process of the pores. In some cases, it may be due to a high concentration of oxygen in one or more layers. Operators of the manufacturing process wish to receive information about the presence of such defects in order to improve the manufacturing process.

[0204] In some implementations, the defect is lateral over-etching that occurs during the etching process. Lateral over-etching may occur in the channel walls during the etching process, for example, along a horizontal direction or along a direction orthogonal to the etching direction. Properties (such as those described above) that provide information about this defect can be generated so that classifier 112 can detect it.

[0205] exist Figure 13 The image shows a non-limiting example of this defect.

[0206] Assume region 4004 is selected as the region with pixel intensity above a threshold (in this example, region 4004 is the fourth dark region). It should be noted that this is not limiting, and in different implementations, another region can be selected. Both region 4004 and region 4005 can be selected.

[0207] Predefined partition 1200 corresponds to the partition of region 4004 located in the upper middle part of the image.

[0208] The number of pixels within a predefined partition 1200 is counted and stored as an attribute. This attribute can be fed into a trained classifier 112. If this number of pixels is large, it typically indicates a defect (also known as a "black hat") that can be detected by the trained classifier 112.

[0209] Despite Figures 6B to 13 In the image, each segmented region is clearly distinguishable from the others, but the actual image received by the system is usually associated with a low signal-to-noise ratio.

[0210] It should be understood that the application of the present invention is not limited to the details set forth in the description contained herein or shown in the accompanying drawings.

[0211] It should also be understood that the system according to the invention can be implemented, at least in part, on a suitably programmed computer. Similarly, the invention contemplates a computer program that can be read by a computer to perform the methods of the invention. The invention also envisions a non-transitory computer-readable storage medium that tangibly embodies a program of instructions executable by a computer to perform the methods of the invention.

[0212] This invention can be implemented in other ways and can be practiced and carried out in various manners. Therefore, it should be understood that the wording and terminology used herein are for descriptive purposes and should not be considered limiting. Consequently, those skilled in the art will understand that the concepts upon which this disclosure is based can readily be used as the basis for designing other structures, methods, and systems for performing several purposes of the subject matter of this disclosure.

[0213] Those skilled in the art will readily understand that various modifications and alterations can be applied to the embodiments of the invention as described above without departing from the scope defined by the appended claims.

Claims

1. A system comprising at least one processor and memory circuitry, the at least one processor and memory circuitry being configured to: Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. The image is divided into multiple N regions R1 to R2. N ,in, For the manufacturing process of the hole, a predefined region R is defined. i Provides information on over-etching, where i is between 1 and N. Generate at least one of the following: Provide the predefined region R i Data D containing information about the intensity of one or more pixels pix_强度 ,or Provide the predefined region R i Data D containing one or more geometric properties 几何形状 ,and D pix_强度 Or D 几何形状 At least one of the layers is fed into a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

2. The system of claim 1, wherein the output of the trained classifier provides information to classify the hole into one of different categories representing different possible outcomes of the manufacturing process of the hole.

3. The system of claim 2, wherein the different categories include at least one of the following: (i) Properly etched holes, (ii) Under-etched holes, (iii) First-order over-etched holes, or (iv) Secondary over-etched holes, wherein the secondary over-etched holes are deeper than the primary over-etched holes.

4. The system according to claim 1, wherein at least one of (i), (ii), or (iii) is satisfied: (i) The hole is an inclined hole. (ii) The image is not radially symmetric, or (iii) The system is configured to segment the image into multiple segments, wherein a majority or all of the area of ​​the image on which the segmentation is performed corresponds to the image of the hole.

5. The system of claim 1, wherein the predefined region R that provides information on over-etching i It is based on the predefined region R relative to the pixel intensity of other regions. i The pixel intensity is used for selection.

6. The system of claim 1, wherein at least one of (i), (ii), or (iii) is satisfied: (i) Data D pix_强度 Or data D 几何形状 At least one of them includes data that can be used to determine the depth of the hole; (ii) Data D pix_强度 Or data D 几何形状 At least one of them includes data that provides information on one or more physical properties of the manufacturing process of the hole; or (iii) Data D pix_强度 Or data D 几何形状 At least one of them includes data that provides information on one or more physical properties of the manufacturing process of the hole in the NAND memory cell.

7. The system according to claim 1, wherein data D pix_强度 This includes data that provides information on the average pixel intensity of a given region among the plurality of regions.

8. The system according to claim 1, wherein D 几何形状 This includes data that provides information about the ratio between the area of ​​a given region and the boundary area of ​​one of the plurality of regions.

9. The system according to claim 1, wherein D 几何形状 This includes data that provides information about the distance between the centroid of a given region and the pixel in the given region that is closest to the centroid.

10. The system according to claim 1, wherein D pix_强度 This includes data that provides information on the contrast between a given region among the plurality of regions and one or more regions different from the given region.

11. The system of claim 1, wherein a given region among the plurality of regions includes the darkest region and the brightest region, wherein D pix_强度 This includes data that provides information about the contrast between the darkest and brightest areas.

12. The system according to claim 1, wherein D pix_强度 This includes data that provides information on the contrast at the boundary between a given region in the plurality of regions and a second region in the plurality of regions that is different from the given region.

13. The system according to claim 1, wherein D 几何形状 This includes data that provides information on the ratios between the following items: The distance between the centroid of a given region and the pixel in the given region closest to the centroid, and The distance between the centroid of a given region and the pixel furthest from the centroid of the given region.

14. The system of claim 1, wherein the D determined for a given region 几何形状 Provide information on whether the given region has at least a partial ring shape.

15. The system according to claim 1, wherein at least one of the following (i) or (ii): (i) D 几何形状 This includes data providing information about the number of pixels in a predefined partition of one of the plurality of regions, wherein the region is associated with pixel intensities above a threshold, or (ii) D 几何形状 This includes data that provides information on the side over-etching resulting from the etching process of the hole.

16. The system of claim 1, wherein at least one of (i) or (ii) is satisfied: (i) The aperture exhibits a high aspect ratio and a nanoscale width, wherein the aperture exposes at least one layer from one or more sets of layers, wherein each set of layers comprises layers with different electron yields, wherein the sample is an intermediate product manufactured through one or more manufacturing stages of a three-dimensional NAND memory cell manufacturing process; or (ii) The depth of the pore is between 50 nm and 10,000 nm.

17. A method, the method comprising being performed via at least one processor and memory circuitry, the method comprising: Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. The image is divided into multiple N regions R1 to R2. N Among them, for the manufacturing process of the hole, a predefined region R i Provides information on over-etching, where i is between 1 and N. Generate at least one of the following: Provide the predefined region R i Data D containing information about the intensity of one or more pixels pix_强度 ,or Provide the predefined region R i Data D containing one or more geometric properties 几何形状 ,and D pix_强度 Or D 几何形状 At least one of the layers is fed into a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

18. A non-transitory computer-readable medium comprising instructions that, when executed by at least one processor and memory circuitry, cause the at least one processor and memory circuitry to perform an operation comprising: Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. The image is divided into multiple N regions R1 to R2. N ,in, For the manufacturing process of the hole, a predefined region R is defined. i Provides information on over-etching, where i is between 1 and N. Generate at least one of the following: Provide the predefined region R i Information data D of one or more pixel intensity pix_强度 ,or Provide the predefined region R i Data D containing one or more geometric properties 几何形状 ,and D pix_强度 Or D 几何形状 At least one of the layers is fed into a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

19. A system comprising at least one processor and memory circuitry, the at least one processor and memory circuitry being configured to: - Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. - Divide the image into multiple regions. -Generate data D 几何形状 The data D 几何形状 Provide the ratio between the following items: First data, which provides information on the average radius of pixels within a given region of the plurality of regions, and The second data provides information on the average radius of pixels within a second region located within the plurality of regions, wherein the second region differs from the given region. -The data D 几何形状 The feed is sent to a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

20. A method comprising: using at least one processor and memory circuitry. - Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. - Divide the image into multiple regions. -Generate data D 几何形状 The data D 几何形状 Provide the ratio between the following items: First data, which provides information on the average radius of pixels within a given region of the plurality of regions, and The second data provides information on the average radius of pixels within a second region located within the plurality of regions, wherein the second region differs from the given region. -The data D 几何形状 The feed is sent to a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

21. A non-transitory computer-readable medium comprising instructions that, when executed by at least one processor and memory circuitry, cause the at least one processor and memory circuitry to perform an operation comprising: Obtain an image of a hole formed in a semiconductor sample, wherein the hole exposes at least one of a plurality of layers of the semiconductor sample. The image is divided into multiple regions. Generate data D 几何形状 The data D 几何形状 Provide the ratio between the following items: First data, which provides information on the average radius of pixels within a given region of the plurality of regions, and The second data provides information on the average radius of pixels within a second region located within the plurality of regions, wherein the second region differs from the given region. The data D 几何形状 The feed is sent to a trained classifier to obtain an output, wherein the output of the trained classifier can be used to determine whether the hole terminates in a target layer among the plurality of layers.

Images