Core set based mask inspection for semiconductor sample fabrication

Abstract

A system and method for a method of mask inspection are provided, comprising: obtaining a plurality of aerial images of a mask; generating a plurality of image core sets corresponding to the aerial images, the generating comprising, for each given aerial image: applying a printing threshold to the given aerial image to obtain a binary image representing printable features of the given aerial image; extracting a contour for each feature in a set of features of interest (FOI) from the printable features and generating descriptors characterizing the contours, thereby producing a set of contours associated with the respective descriptors; and creating an image core set of the group of contours based on the respective descriptors of the group of contours, the image core set comprising one or more families, each family comprising at least one representative contour representing one or more similar contours of a respective type from the group of contours. The plurality of image core sets may be merged to obtain a mask core set.

Description

Technical Field

[0001] The subject matter of the present disclosure generally relates to the field of mask inspection, and more particularly to defect detection regarding photomasks. Background Art

[0002] Current demands for high density and high performance associated with the very large scale integration of fabricated microelectronic devices require sub-micron features, increased transistor and circuit speeds, and improved reliability. As semiconductor processes advance, pattern dimensions (such as line widths) and other types of critical dimensions continue to shrink. Such demands require the formation of device features with high precision and high uniformity, which in turn requires careful monitoring of the manufacturing process, including automated inspection of the devices while they are still in the form of semiconductor wafers.

[0003] Photolithography masks (also known as photomasks, masks, or reticles) are typically used in the lithography process to fabricate semiconductor devices. The lithography process is one of the main processes for fabricating semiconductor devices and includes patterning the surface of a wafer according to the circuit design of the semiconductor device to be produced. Such a circuit design is first patterned on the mask. Thus, in order to obtain a functioning semiconductor device, the mask must be defect-free. Masks are fabricated by complex processes and may suffer from various defects and variations.

[0004] In addition, masks are typically used in a repetitive manner to create many die on one or more wafers. Thus, any defect on the mask will be repeated many times on the wafer and will result in multiple defective devices. Establishing a process worthy of production requires strict control of the entire lithography process. In the said process, critical dimension (CD) control is a decisive factor regarding device performance and yield.

[0005] Various mask inspection methods have been developed and are commercially available. According to certain conventional mask design and evaluation techniques, a mask is first created and used to expose a wafer, and then inspection is performed to determine whether the features / patterns of the mask have been transferred to the wafer according to the design. Any variation between the ultimately printed features and the intended design may require modifying the design, repairing the mask, creating a new mask, and / or exposing a new wafer.

[0006] In this regard, verifying the accuracy and quality of printed features allows for an indirect method of verifying a mask. However, since the final printed pattern on a wafer or die is formed after printing processes (such as resist development, substrate processing (such as material etching or deposition), etc.), it may be difficult to attribute, discern, or isolate errors in the final printed pattern as issues associated with the mask and / or resist deposition and / or development processes. Additionally, inspecting the final printed pattern on a wafer or die often provides a limited number of samples that can be used to detect, determine, and resolve any processing issues. The processes may also be laborious and require a significant amount of inspection and analysis time. Alternatively, various mask inspection tools can be used to directly inspect the mask. SUMMARY OF THE INVENTION

[0007] In certain aspects of the subject matter of the present disclosure, a computerized system for inspecting a mask that can be used to fabricate a semiconductor sample is provided. The system includes one or more processing circuitry configured to: obtain a plurality of aerial images, each aerial image capturing a respective portion of the mask; generate a plurality of image core sets corresponding to the plurality of aerial images. For each given aerial image, the generating includes: i) applying a printing threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) extracting the contour of each feature of interest (FOI) in a group of features of interest from the plurality of printable features and generating a descriptor characterizing the contour, thereby producing a group of contours corresponding to the group of FOIs and associated with the respective descriptors; and iii) creating an image core set of the group of contours based on the respective descriptors of the group of contours, the image core set including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the group of contours; and merging the plurality of image core sets to obtain a mask core set, wherein the mask core set includes one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the plurality of aerial images, and the at least one representative contour being indicated as normal or abnormal based on the number of one or more similar contours of the respective type from the plurality of aerial images.

[0008] In addition to the above features, the system according to the aspects of the subject matter of the present disclosure can include one or more of the following features (i) to (xiv) in any desired combination or arrangement that is technically possible:

[0009] (i). For a family in the mask core set that includes at least one representative contour indicated as an anomaly, one or more processing circuitry may be further configured to identify one or more similar contours of a corresponding type represented by the at least one representative contour, and report one or more FOIs corresponding to the one or more similar contours of the corresponding type as one or more defect candidates.

[0010] (ii). A defect candidate from one or more defect candidates may represent an edge displacement error, which indicates a relatively significant deviation of the contour of the FOI from the expected position of the FOI.

[0011] (iii). Each family in the image core set may be associated with an indication of being normal or abnormal at the image level for at least one representative contour of each family, the indication being obtained based on one or more similar contours of a corresponding type from the set of contours.

[0012] (iv). The image core set may be a subset of representative contours approximating the distribution of the set of contours.

[0013] (v). One or more processing circuitry may be configured to create the image core set by: initializing the image core set; for each given contour in the set of contours, sequentially: searching for one or more reference contours in the image core set based on a similarity measurement of descriptors applied to each given contour; in response to one or more reference contours being identified, measuring the deviation between the given contour and each of the one or more reference contours; and determining whether to add the given contour to a family in the core set based on the measured deviation.

[0014] (vi). In response to not finding one or more reference contours, one or more processing circuitry may be configured to add the given contour to a new family in the image core set.

[0015] (vii). One or more processing circuitry may be configured to measure the deviation between the given contour and each of the one or more reference contours by: registering the given contour with the one or more reference contours respectively, thereby generating one or more registered contour pairs; measuring the distance between corresponding points of each registered contour pair; and calculating the deviation based on the measured distance.

[0016] (viii). One or more processing circuitry may be configured to merge multiple image core sets by: measuring the deviation between representative contours from different image core sets among the multiple image core sets; and retaining at least one representative contour in the mask core set that represents one or more similar contours of the same type.

[0017] (ix). A plurality of image core sets may be generated in parallel by a plurality of processors at least in part and sent to a particular processor for merging.

[0018] (x). A descriptor of a contour may be based on one or more of the following: the region formed by the contour, the width of the region, the height of the region, the number of pixels along the contour, a chain code, a center of gravity, and polar coordinates of the contour.

[0019] (xi). A plurality of aerial images may be acquired sequentially by a photomask inspection tool configured to simulate the optical configuration of a lithography tool.

[0020] (xii). Merging of a plurality of image core sets may enable sharing of contour statistics between a plurality of aerial images across a mask, thereby allowing false alarm reduction and detection sensitivity improvement.

[0021] (xiii). A mask core set may be used as a mask core set model for subsequent inspection of one or more masks.

[0022] (xiv). A plurality of aerial images may be captured for a single-die mask or a multi-die mask.

[0023] In accordance with other aspects of the subject matter of the present disclosure, a method of inspecting a mask that may be used to fabricate a semiconductor sample is provided. The method includes: obtaining a plurality of aerial images, each aerial image capturing a corresponding portion of the mask; generating a plurality of image core sets corresponding to the plurality of aerial images, wherein for each given aerial image, the generating includes: i) applying a print threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) extracting a contour of each feature of interest (FOI) in a group of FOIs from the plurality of printable features and generating a descriptor characterizing the contour, thereby producing a group of contours corresponding to the group of FOIs and associated with the corresponding descriptors; and iii) creating an image core set of the group of contours based on the corresponding descriptors of the group of contours, the image core set including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the group of contours; and merging the plurality of image core sets to obtain a mask core set, wherein each family in the mask core set includes at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the plurality of aerial images, the at least one representative contour being indicated as normal or abnormal based on the number of one or more similar contours of a corresponding type from the plurality of aerial images.

[0024] With necessary modifications, aspects of the subject matter of the present disclosure can include any desired combination or arrangement technically possible of one or more of the features (i) to (xiv) listed above with respect to the system.

[0025] In accordance with other aspects of the subject matter of the present disclosure, there is provided a non-transitory computer-readable medium including instructions that, when executed by a computer, cause the computer to perform a method of inspecting a mask that can be used to fabricate a semiconductor sample. The method includes: obtaining a plurality of aerial images, each aerial image capturing a respective portion of the mask; generating a plurality of image core sets corresponding to the plurality of aerial images, and for each given aerial image, the generating includes: i) applying a printing threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) extracting the contour of each feature of interest (FOI) in a group of features of interest (FOI) from the plurality of printable features, and generating a descriptor characterizing the contour, thereby producing a group of contours corresponding to the group of FOIs and associated with respective descriptors; and iii) creating an image core set of the group of contours based on the respective descriptors of the group of contours, the image core set including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the group of contours; and merging the plurality of image core sets to obtain a mask core set, wherein each family in the mask core set includes at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the plurality of aerial images, and the at least one representative contour is indicated as normal or abnormal based on the number of one or more similar contours of a respective type from the plurality of aerial images.

[0026] With necessary modifications, aspects of the subject matter of the present disclosure can include any desired combination or arrangement technically possible of one or more of the features (i) to (xiv) listed above with respect to the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] To understand the present disclosure and to see how it may be carried out in practice, embodiments will now be described, by way of non-limiting example only, with reference to the accompanying drawings, in which:

[0028] Figure 1 A functional block diagram of a mask inspection system in accordance with certain embodiments of the subject matter of the present disclosure is illustrated.

[0029] Figure 2 A generalized flowchart of mask inspection in accordance with certain embodiments of the subject matter of the present disclosure is illustrated.

[0030] Figure 3Illustrated is a generalized flowchart of creating a core set of contour groups in accordance with certain embodiments of the subject matter of the present disclosure.

[0031] Figure 4 Illustrated is a generalized flowchart of measuring the deviation between a given contour and a reference contour of the given contour in accordance with certain embodiments of the subject matter of the present disclosure.

[0032] Figure 5 Illustrated is a schematic diagram of a photochemical inspection tool and a lithography tool in accordance with certain embodiments of the subject matter of the present disclosure.

[0033] Figure 6 Schematically illustrated are exemplary layouts of a single-die mask and a multi-die mask in accordance with certain embodiments of the subject matter of the present disclosure.

[0034] Figure 7 Illustrated is a schematic diagram showing a process of applying a printing threshold and examples of an aerial image and a corresponding binary image in accordance with certain embodiments of the subject matter of the present disclosure.

[0035] Figure 8A Illustrated are several examples of extracted contours of several structural features with different shapes in accordance with certain embodiments of the subject matter of the present disclosure.

[0036] Figure 8B Illustrated is an example of a reference contour of a given contour in accordance with certain embodiments of the subject matter of the present disclosure.

[0037] Figure 9 Illustrated are two examples of the measured deviation in accordance with certain embodiments of the subject matter of the present disclosure.

[0038] Figure 10 Illustrated is an example of a histogram representation of an image core set created for a given image in accordance with certain embodiments of the subject matter of the present disclosure.

[0039] Figure 11 Illustrated is a schematic diagram of mask core set generation based on an image core set in accordance with certain embodiments of the subject matter of the present disclosure.

[0040] Figure 12 Illustrated is a schematic diagram of false alarm reduction for mask inspection in accordance with certain embodiments of the subject matter of the present disclosure. Detailed Description

[0041] A photomask (also referred to as a mask or reticle) is used to fabricate semiconductors in a lithography process. The mask is fabricated in a complex process and may suffer from various defects and variations. One type of defect to be detected is related to the edge displacement of one or more structural features on the mask.

[0042] As used herein, the term "edge displacement" or "edge displacement error" refers to a relatively significant deviation of the edge / profile of a structural feature from its intended / desired position.

[0043] A structural feature (also referred to herein as a structural element) can refer to any primitive object on a mask having a contoured geometry / structure. In some cases, an object can be combined / superimposed with other object(s) to form a complex structural feature as a pattern. Examples of structural features can include features of a general shape (such as contacts, lines, etc.), and / or features having a complex structure / shape, and / or features combined with one or more other features. A structural feature can be a 2D or 3D element, and an image capturing the structural feature can reflect a 2D representation of the structural feature.

[0044] Defects in the edge displacement mentioned herein can be caused by various factors (such as (multiple) physical effects during the manufacturing process of the mask) and / or other factors (such as oxidation (which may occur gradually during the use of the mask), particles, scratches, crystal growth, electrostatic discharge (ESD), etc.). If such mask defects are not detected before mass-producing wafers, such mask defects will be repeated multiple times on the wafers and will result in multiple semiconductor devices being defective (e.g., affecting the functionality of the devices), thus significantly reducing the yield.

[0045] A mask includes a mask field that will be transformed onto a wafer. In some cases, the mask can contain a mask field that includes multiple die having the same design pattern (such a mask is referred to as a multi-die mask). In some other cases, the mask can contain a mask field that includes a single die (such a mask is referred to as a single-die mask). To detect whether there are defects associated with the structural features in a die, for the purpose of comparison in die-to-die inspection, a reference structural feature from another die is typically required. However, in the case of a single-die mask, there is no reference die available on the mask for comparison. Therefore, for the structural features in a single die (or the scribed area in a multi-die mask as Figure 6 illustrated), a reference for defect detection purposes needs to be obtained.

[0046] Now turning to Figure 6 , an exemplary layout of a single-die mask and an exemplary layout of a multi-die mask according to certain embodiments of the subject matter of the present disclosure are schematically illustrated.

[0047] As shown, the exemplary multi-die mask 604 includes a mask field with nine dies (one of the dies is labeled 611) having the same design pattern. For a structural feature in any die of the mask field, one or more reference structural features can always be found in one or more of the adjacent dies. However, for the single-die mask 602 with a mask field including a single die 606, there is no reference die on the mask that can be used for defect detection regarding the structural features in the single die.

[0048] In addition, the single-die mask 602 further includes a scribe region 608 between the die region 606 and the peripheral region 610 of the mask. The scribe region 608 contains auxiliary features such as alignment features, calibration features, etc. Such auxiliary features / structures can be printed on the wafer together with the patterns in the die region during the lithography process. Therefore, in addition to or instead of defect detection regarding the structural features in the die, it is also necessary to detect defects regarding these auxiliary features (if any). This also applies to the auxiliary features in the scribe region 612 of the multi-die mask 604, which is located between the die regions and the peripheral region of the nine dies 611 and between the dies. However, there is no reference feature for inspection regarding such auxiliary features on the mask.

[0049] Certain conventional techniques associated with photochemical inspection tools can acquire two images respectively from the transmission mode and the reflection mode of the tool and analyze the difference between the two images to estimate the presence of any defects. However, using the two imaging modes can be time-consuming both in the image acquisition process and the image processing process, thus affecting the inspection throughput (TpT).

[0050] Alternatively, some inspection tools can attempt to generate a simulated image based on the design data of the mask and use the simulated image as a reference image for defect detection on the mask image. However, this method requires obtaining the design data of the mask, which is not available in many cases. Additionally, the simulated image can be inaccurate due to the uncertainty of process variations during the mask manufacturing process, which may inevitably affect the accuracy of the inspection results.

[0051] In particular, due to the continuous development of advanced processes and complex features regarding photomasks that have increased the required sensitivity for defect detection on the mask, current mask inspection methods are insufficient to provide the desired process control of mask features. Therefore, an improved defect detection method that solves the above problems is needed to accurately detect defects regarding the structural features on the mask (e.g., the structural features in the die region of the single-die mask and the structural features in the scribe regions of both the single-die mask and the multi-die mask) with higher sensitivity and fewer false alarms without affecting the inspection throughput.

[0052] Accordingly, certain embodiments of the subject matter of the present disclosure present a mask inspection system and method for detecting defects related to edge displacement errors of structural features on a mask, the system and method being free of one or more of the above disadvantages.

[0053] The present disclosure proposes that for each aerial image, the contour of the printable features of the image is extracted and an image core set of the contour is created. The image core set is a smaller subset approximating the contour distribution and includes a family of representative contours of the corresponding type. In addition, the image core sets created for all aerial images are merged to obtain an overall mask core set so as to enable sharing of contour statistics across the entire mask. The mask core set includes families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of the corresponding type from all aerial images. The family and the at least one representative contour of the family are also associated with an indication of normal or abnormal. The proposed method has been shown to have a reduced false alarm rate and improved throughput for advanced process control of mask features, as will be described in detail below.

[0054] Based on this, please note Figure 1 , the figure illustrates a functional block diagram of a mask inspection system according to certain embodiments of the subject matter of the present disclosure.

[0055] Figure 1 The inspection system 100 illustrated in can be used to inspect a mask during or after a mask manufacturing process. As described above, the inspection mentioned herein can be interpreted to cover any type of operation regarding a mask or a portion of a mask, involving defect detection and / or various types of defect classification and / or metrology operations, such as, for example, critical dimension (CD) measurement. According to certain embodiments of the subject matter of the present disclosure, the inspection system 100 includes a computer-based system 101 capable of automatically detecting defects related to edge displacement of structural elements on a mask. The system 101 is thus also referred to as a mask defect detection system, and the mask defect detection system is a subsystem of the inspection system 100.

[0056] The system 100 includes a mask inspection tool 120, which is operably connected to the system 101 and is configured to scan the mask and capture one or more images of the mask for inspecting the mask. The term "mask inspection tool" used herein should be broadly interpreted to cover any type of inspection tool that can be used in mask inspection-related processes. As a non-limiting example, inspection-related processes include scanning (in single or multiple scans), imaging, sampling, detecting, measuring, classifying, and / or other processes performed on a mask or a portion of a mask.

[0057] Without in any way limiting the scope of the present disclosure, it should also be noted that the mask inspection tool 120 can be implemented as various types of inspection machines, such as optical inspection tools, electron beam tools, etc. In some cases, the mask inspection tool 120 can be a relatively low-resolution inspection tool (e.g., an optical inspection tool, a low-resolution scanning electron microscope (SEM), etc.). In some cases, the mask inspection tool 120 can be a relatively high-resolution inspection tool (e.g., a high-resolution SEM, an atomic force microscope (AFM), a transmission electron microscope (TEM), etc.). In some cases, the inspection tool can provide both low-resolution image data and high-resolution image data. In some embodiments, the mask inspection tool 120 has metrology capabilities and can be configured to perform metrology operations on the captured images. The resulting image data (low-resolution image data and / or high-resolution image data) can be transmitted to the system 101 directly or via one or more intermediate systems.

[0058] According to certain embodiments, the mask inspection tool can be implemented as a lithographic inspection tool configured to emulate / simulate the optical configuration of a lithographic tool (such as a scanner or a stepper) that can be used to fabricate semiconductor samples (e.g., by projecting the pattern formed in the mask onto a wafer).

[0059] Now turning to Figure 5 , a schematic diagram of a lithographic inspection tool and a lithographic tool according to certain embodiments of the subject matter of the present disclosure is shown.

[0060] Similar to the lithographic tool 520, the lithographic inspection tool 500 can include an illumination source 502 configured to generate light (e.g., a laser) at an exposure wavelength, illumination optics 504, a mask holder 506, and projection optics 508. The illumination optics 504 and the projection optics 508 can include one or more optical elements (such as, for example, lenses, apertures, spatial filters, etc.).

[0061] In the lithographic tool 520, the mask is positioned at the mask holder 506 and is optically aligned to project an image of the circuit pattern to be replicated onto a wafer placed on a wafer holder 512 (e.g., by using various stepping, scanning, and / or imaging techniques to generate or replicate the pattern on the wafer). Different from the lithographic tool 520, instead of placing a wafer holder 512, the lithographic inspection tool 500 places a detector 510 (such as, for example, a charge-coupled device (CCD)) at the position of the wafer holder. The detector 510 is configured to detect the light projected through the mask and to generate an image of the mask.

[0062] It can be seen that the actinic inspection tool 500 is configured to simulate the optical configuration of the lithography tool 520, including but not limited to, for example, illumination / exposure conditions such as wavelength, pupil shape, numerical aperture (NA), etc. Accordingly, it is expected that the mask image 514 acquired by the detector 510 is similar to the image 516 of the wafer fabricated using the mask via the lithography tool 520. The mask image acquired using such an actinic inspection tool is also referred to as an aerial image in the present disclosure. The aerial image is provided to the system 101 for further processing, as described below.

[0063] According to certain embodiments, in some cases, the mask inspection tool 120 can be implemented as a non-actinic inspection tool, such as, for example, a conventional optical inspection tool, an electron beam tool, etc. In such a case, the non-actinic inspection tool can be configured to acquire an image of the mask. A simulation can be performed on the acquired image to simulate the optical configuration of the lithography tool, thereby generating an aerial image. In some cases, the simulation can be performed by the system 101 (e.g., the functionality of the simulation can be integrated into the processing circuitry 102 of the system), while in some other cases, the simulation can be performed by a processing module of the mask inspection tool 120 or by a separate simulation unit operatively connected to the mask inspection tool 120 and the system 101.

[0064] The system 101 includes processing circuitry 102 and processing circuitry 103, which are operatively connected to the hardware-based I / O interface 126 and are configured to provide the processing required to operate the system, as described in further detail with reference to Figures 2 to 4 Any one of the processing circuitry 102 and processing circuitry 103 can include one or more processors (not shown separately) and one or more memories (not shown separately). One or more processors of the processing circuitry can be configured to execute several functional modules individually or in any suitable combination according to computer-readable instructions implemented on a non-transitory computer-readable memory included in the processing circuitry. Such functional modules are hereinafter referred to as being included in the processing circuitry.

[0065] One or more processors mentioned herein may represent one or more general-purpose processing devices, such as microprocessors, central processing units, etc. More specifically, a given processor may be one of the following: a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. One or more processors may also be one or more special-purpose processing devices, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a network processor, etc. One or more processors are configured to execute instructions for performing the operations and steps described herein.

[0066] The memory mentioned herein may include one or more of the following: internal memory (such as processor registers and caches, etc.), main memory (such as read only memory (ROM)), flash memory, dynamic random access memory (DRAM) (such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.).

[0067] According to certain embodiments, the functional modules included in the processing circuitry 102 of the system 101 may include an image processing module 104 and an image core set module 106 that are operably connected to each other. The processing circuitry 102 may be configured to obtain a plurality of aerial images via the I / O interface 126, each aerial image capturing a corresponding portion of the mask. As an example, the aerial images may be acquired by a mask inspection tool 120 (such as a lithography inspection tool).

[0068] The image core set module 106 may be configured to generate a plurality of core sets corresponding to the plurality of aerial images. Specifically, for each given aerial image, the image core set module 106 may be configured to apply a printing threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; extract the contour of each feature of interest (FOI) in the group of features of interest (FOI) from the plurality of printable features, and generate a descriptor characterizing the contour, thereby generating a group of contours corresponding to the group of FOIs and associated with the corresponding descriptors. The image core set module 106 may be further configured to create an image core set of the group of contours based on the corresponding descriptors of the group of contours. The image core set includes one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the group of contours.

[0069] The functional modules included in the processing circuitry 103 may include a mask core set module 108. The mask core set module 108 may be configured to combine a plurality of core sets into an overall mask core set. The mask core set includes one or more families. Each family in the mask core set includes at least one representative profile, and the at least one representative profile represents one or more similar profiles of a corresponding type from a plurality of aerial images. Based on the number of one or more similar profiles of a corresponding type from a plurality of aerial images, at least one representative profile of each family is indicated as normal or abnormal.

[0070] It should be noted that although some embodiments of the present disclosure relate to a processing circuitry 102 configured to perform the above operations, the functionality / operations of the foregoing functional modules may be performed in various ways by one or more processors in the processing circuitry 102. As an example, the operations of each functional module may be performed by a specific processor or a combination of processors. Thus, the operations of the various functional modules (such as various image processing operations and core set creation, etc.) may be performed by the corresponding processor (or combination of processors) in the processing circuitry 102, and optionally, these operations may be performed by the same processor. The present disclosure should not be construed as being limited to a single processor that always performs all operations. The above similarly applies to the processing circuitry 103.

[0071] It should also be noted that although the processing circuitry 102 and the processing circuitry 103 are Figure 1 illustrated in the figure as two separate processing circuitries operatively connected to each other, this is for illustrative purposes only and should not be regarded as limiting the present disclosure. In some cases, the functionality of the two processing circuitries may be (at least partially) combined and integrated into one processing circuitry or differently integrated / divided into multiple processing circuitries. In some cases, the functionality of the two processing circuitries may be integrated as part of an inspection tool 120, and in some other cases, at least some of the functionality (such as the functionality of the mask core set module 108) may be implemented in a separate device (such as a server operatively connected to the inspection tool (locally or remotely)).

[0072] According to some embodiments, system 100 may include a storage unit 122. The storage unit 122 may be configured to store any data required by operating systems 100 and 101, such as data related to the input and output of systems 100 and 101, and intermediate processing results generated by system 101. As an example, the storage unit 122 may be configured to store the (multiple) images and / or derivatives of the (multiple) images (e.g., images after preprocessing) generated by the mask inspection tool 120. Thus, the (multiple) images may be retrieved from the storage unit 122 and provided to the PMC 102 for further processing. The output of system 101 (such as, for example, multiple image core sets, mask core sets, etc.) may be sent to the storage unit 122 for storage.

[0073] In some embodiments, system 100 may optionally include a computer-based graphical user interface (GUI) 124 configured to enable user-specified input related to system 101. For example, a visual representation of the mask, including an image of the mask and / or an image representation of structural features, may be presented to the user (e.g., via a display forming part of the GUI 124). Options for defining certain operation parameters, such as, for example, print threshold, deviation threshold, detection threshold, etc., may be provided to the user via the GUI. In some cases, the user may also view operation results on the GUI, such as a family of core sets, measured deviations, detected defects, and / or further inspection results.

[0074] In some embodiments, in addition to system 101, the mask inspection system 100 may further include one or more inspection modules, such as additional (multiple) defect detection modules and / or an automatic defect review module (ADR) and / or an automatic defect classification module (ADC) and / or a metrology-related module and / or other inspection modules that may be used to perform additional inspections on the mask. The one or more inspection modules may be implemented as stand-alone computers, or the functionality (or at least some of the functionality) of the one or more inspection modules may be integrated with the mask inspection tool 120. In some embodiments, the output obtained from system 101 may be used by the mask inspection tool 120 and / or one or more inspection modules (or a part of the one or more inspection modules) to further inspect the mask.

[0075] Those skilled in the art will readily understand that the teachings of the subject matter of the present disclosure are not limited by Figure 1 the systems shown. Figure 1Each system component and module therein may be constituted by any combination of software, hardware, and / or firmware. Correspondingly, the software, hardware, and / or firmware are executed on one or more appropriate devices, and the one or more devices perform the functions defined and explained herein. The equivalent and / or modified functionality described for each system component and module may be combined or divided in another way. Thus, in some embodiments of the subject matter of the present disclosure, the system may include fewer, more, modified, and / or different components, modules, and functions than Figure 1 shown.

[0076] Figure 1 Each component therein may represent a plurality of specific components, and the plurality of specific components are adapted to operate independently and / or collaboratively to process various data and electrical inputs and to implement operations related to the computerized inspection system. In some cases, multiple instances of components may be utilized for reasons of performance, redundancy, and / or availability. Similarly, in some cases, multiple instances of components may be utilized for reasons of functionality or application. For example, different parts of a specific functionality may be placed in different instances of the component.

[0077] Note that the Figure 1 illustrated inspection system may be implemented in a distributed computing environment, where the foregoing functional modules included in the processing circuitry 102 and 103 may be distributed over several local and / or remote devices and may be connected via a communication network. As an example, the inspection tool 120 and the system 101 may be located at the same entity (hosted by the same device in some cases) or distributed over different entities. As another example, the processing circuitry 102 and 103 may be located at the same entity (hosted by the same device in some cases) or distributed over different entities. For example, in some cases, the processing circuitry 102 may be hosted by the inspection tool 120, while the processing circuitry 103 may be implemented at a separate server (local or remote) operably connected to the tool.

[0078] In some examples, certain components utilize cloud implementations, such as implemented in a private cloud or a public cloud. In cases where the various components of the inspection system are not entirely located at one location or one physical entity, the communication between the various components of the inspection system may be implemented via any signaling system or communication component, module, protocol, software language, and drive signal, and the communication may appropriately be wired and / or wireless.

[0079] It should also be noted that, in other embodiments, at least some of the inspection tool 120, the storage unit 122, and / or the GUI 124 may be external to the inspection system 100 and operate in data communication with the system 101 via the I / O interface 126. The system 101 may be implemented as one or more stand-along computers to be used in conjunction with the inspection tool. Alternatively, the corresponding functions of the system 101 may be at least partially integrated with the mask inspection tool 120, thereby enhancing and strengthening the functionality of the mask inspection tool 120 in inspection-related processes.

[0080] Although not necessarily so, the operation procedures of the systems 101 and 100 may correspond to some or all of the stages of the method described with respect to Figures 2 to 4 Similarly, the methods described with respect to Figures 2 to 4 and their possible implementations may be implemented by the systems 101 and 100. Therefore, note that, with necessary modifications, the embodiments discussed with respect to the methods described with respect to Figures 2 to 4 may also be implemented as various embodiments of the systems 101 and 100, and vice versa.

[0081] Now referring to Figure 2 , a generalized flowchart of mask inspection in accordance with certain embodiments of the subject matter of the present disclosure is illustrated.

[0082] A plurality of aerial images, each capturing a respective portion of the mask, may be obtained (202) (e.g., by the processing circuitry 102 via the I / O interface 126, from the mask inspection tool 120, or from the storage unit 122). As an example, the plurality of aerial images may be obtained by simulating the optical configuration of a lithography tool that may be used to fabricate a semiconductor sample.

[0083] In some embodiments, the mask to be inspected is a single-die mask. As illustrated in 602 of Figure 6 , the mask field of the single-die mask (including the die area of the single die 606 and the scribe region 608 (containing auxiliary features, such as alignment features, calibration features, etc.)) includes printable features / structures that will be transferred onto a wafer during a lithography process. Thus, the presently proposed inspection method is applicable to detecting defects with respect to any one of these regions / zones. As an example, the obtained aerial images may represent at least a portion of a single die area and / or at least a portion of the scribe region.

[0084] In some other embodiments, the mask to be inspected may be a multi-die mask, as illustrated in 604 of Figure 6 . In this case, the presently proposed inspection method is applicable to detecting defects with respect to at least a portion of the scribe region 612 (and, if needed, the die area 611) in the multi-die mask.

[0085] In some embodiments, the aerial image is acquired by a photochemical mask inspection tool, such as, for example, the Aera mask inspection tool of Applied Materials Inc. As described above with reference to Figure 5 The photochemical mask inspection tool is specifically configured to simulate the optical configuration of a lithography tool (e.g., a scanner or a stepper) used to fabricate a semiconductor wafer from a mask. The optical configuration to be simulated may include one or more of the following illumination / exposure conditions: such as, for example, wavelength, pupil shape, numerical aperture (NA), etc.

[0086] The aerial image expected to be acquired by such a photochemical inspection tool is similar to the image of a wafer fabricated using a mask via a lithography tool. In other words, the photochemical mask inspection tool is configured to capture a mask image that can simulate how the design pattern in the mask will actually appear in a physical wafer after the manufacturing process.

[0087] In some cases, a photochemical inspection tool may not be available for inspecting the mask. In such cases, a non - photochemical inspection tool (such as, for example, a conventional optical inspection tool, an electron beam tool, etc.) can be used to acquire an image (non - aerial image) of the mask. The acquired non - aerial image can be processed to simulate the optical configuration of a lithography tool, thereby generating an aerial image of the mask. Thus, in some embodiments, the mask inspection method as described with reference to Figure 2 may further include the following preliminary steps: obtaining an image acquired by a non - photochemical inspection tool, and (e.g., by an image processing module 104 or by a processing module of the mask inspection tool 120, etc.) processing the image to simulate the optical configuration of a lithography tool, thereby generating an aerial image.

[0088] In some embodiments, the acquired aerial image can be pre - processed before further processing, as described with reference to Figure 2 The pre - processing may include one or more of the following operations: interpolation (e.g., in the case where the first image has a relatively low resolution), noise filtering, focus correction, aberration compensation, image format conversion, etc.

[0089] It should be noted that the present disclosure is not limited to a specific modality of the mask inspection tool, the type of image acquired thereby, and / or the pre - processing operations required to process the image.

[0090] Multiple image coresets corresponding to multiple aerial images may be generated (204) (e.g., by the image coreset module 106). The image coresets generated for the aerial images are also referred to as image-level coresets (or simply coresets) as compared to the overall mask coreset (also referred to as the mask-level coreset) generated for the entire mask, as will be described in further detail below with reference to block 212. Specifically, for each given aerial image among the multiple aerial images, a corresponding image coreset may be created as follows. A print threshold may be applied (206) to the given aerial image to obtain a binary image representing multiple printable features of the given aerial image. Printable features refer to structural features on the mask that can be printed on a semiconductor sample (such as a wafer).

[0091] Now turning to Figure 7 , a schematic diagram showing a process of applying a print threshold and examples of an aerial image and a corresponding binary image according to certain embodiments of the subject matter of the present disclosure is shown.

[0092] As shown, illustration 700 shows a cross-sectional view of a portion of an exemplary mask that includes a transparent region 702 (e.g., made of quartz) and an opaque region 704 (e.g., made of chromium), where the transparent region transmits light upon illumination and the opaque region blocks light. The aerial image obtained as described above refers to an image captured by a detector that collects the transmitted light through the mask, as illustrated by image 710.

[0093] In fact, the actual wafer manufacturing process performed by a manufacturing tool (e.g., a scanner or a stepper) includes a resist process and an etching process after the lithography process. The wafer is coated with a photoresist as a photosensitive material. Exposure to light causes part of the resist to harden or soften, depending on the process. After exposure, the wafer is developed such that the photoresist dissolves in certain regions according to the amount of transmitted light (i.e., light intensity) received by the regions during exposure.

[0094] As an example, a waveform 705 representing the intensity of the transmitted light is illustrated. If the photoresist at a given region is exposed at an intensity below a specific intensity of the transmitted light, a pattern will be printed on the wafer. These regions with and without photoresist reproduce the design pattern on the mask. Thus, the specific intensity is referred to as the print threshold 706, as Figure 7 illustrated. The developed wafer is then exposed to a solvent that etches away the silicon in the portions of the wafer that are no longer protected by the photoresist coating, thereby producing a printed wafer 708 (for a given layer).

[0095] Thus, in a photochemical inspection tool that simulates the optical configuration of a wafer manufacturing tool, waveform 705 represents transmitted light that will be captured by a detector of the photochemical inspection tool to form an aerial image. Since, in the photochemical inspection tool, the detector replaces the wafer and there is no actual resist and etching process, in order to obtain an image similar to a printed wafer, a print threshold 706 needs to be applied to the aerial image to simulate the effects of the resist and etching processes, thereby generating a binary image that includes printable features on the wafer. Specifically, the binary image represents multiple structural features (i.e., printable features) of the mask that can be printed on the wafer.

[0096] Figure 7 An example of an aerial image 710 and a corresponding binary image 720 generated after applying the print threshold to a first image is also illustrated. As shown, binary image 720 is similar to the printed pattern on wafer 708. It should be noted that although in this example, patterns below the print threshold are illustrated as printable on the wafer (i.e., positive resist), this is not necessarily the case. In some other cases, it can be the opposite, i.e., patterns above the print threshold are printable on the wafer (i.e., negative resist). The present disclosure is not limited to a particular resist process for presenting printable features, nor to a particular application of the print threshold.

[0097] Continuing Figure 2 the description, for each given aerial image, the contour of each FOI in a group of features of interest (FOI) can be extracted (208) from a plurality of printable features (e.g., by the image processing module 104 of the processing circuitry 102).

[0098] In some embodiments, a group of features of interest (FOI) can be selected from a plurality of printable features, and contour extraction can be performed for each FOI in the group. As an example, the FOI group can be selected based on one or more of the following factors: the location of the printable features on the binary image, the type and / or shape of the printable features, printable features detected as defect candidates in a previous inspection, and customer input / feedback regarding the importance of certain printable features to be inspected, etc. In some cases, the selection can be skipped, and the FOI group can actually include the entire population of a plurality of printable features on the binary image.

[0099] The term "contour" may refer to the contour or boundary of a structural feature. In some embodiments of the present disclosure, the contour of a feature may be estimated by using a contour detection method. As an example, the contour detection method may be implemented using any type of contour detection / extraction algorithm, such as Canny, Sobel, or Moore neighborhood contour tracking algorithm, etc. Another example of an edge detection algorithm applicable to the present subject matter is described in U.S. Patent No. 9,165,376, titled "System, method and computer readable medium for detecting edges of a pattern", which is assigned to the assignee of the present patent application and incorporated herein by reference in its entirety. Figure 8A Illustrated are several examples of the extracted contours (marked by dashed lines) of several structural features having different shapes according to certain embodiments of the subject matter of the present disclosure.

[0100] Descriptors characterizing the extracted contours may be generated. The descriptors may be obtained / created in various ways. As an example, a contour descriptor may be generated based on one or more of the following properties of the contour: the area formed by the contour, the width of the area, the height of the area, the number of pixels along the contour (e.g., by counting the number of pixels belonging to the contour), chain code, centroid, and polar coordinates of the contour, etc. For example, a chain code may be used to represent a contour by a sequence of connected line segments of specified length and direction. Generally, the representation is based on 4 or 8 connectivity of segments. The direction of each segment is encoded by using a numbering scheme. The contour code formed as a sequence of such direction numbers is called a chain code, which indicates the shape of the contour. In another example, a contour may be converted into polar coordinates of the contour, and the polar coordinates may be represented as, for example, a θ radius histogram. In some cases, using polar coordinates (such as a histogram based on polar coordinates) as a descriptor (or part thereof) of a contour may enable improved comparison between contours and throughput (TpT) and efficiency of measurements.

[0101] In this way, a group of contours corresponding to the FOI group from multiple printable features may be obtained, and the group of contours is associated with the corresponding descriptors of the group of contours. The group of contours and the descriptors of the group of contours may be represented in different data formats, such as, for example, tables, vectors, lists, etc. As an example, a table representation may include N rows and one or more columns, each row representing a specific contour in the group of N contours, and one or more columns representing the descriptors associated with the contour.

[0102] An image core set of the contour group can be created (210) based on the corresponding descriptors of the contour group, e.g., by the image core set module 106 of the processing circuitry 102. The image core set includes one or more families, each family including at least one representative contour that represents one or more similar contours of a corresponding type from the contour group.

[0103] The core set of the original set generally refers to a smaller sample set that approximates the larger original set in terms of, e.g., the shape / distribution of the core set. This approximation ensures that solving a problem regarding the core set that is input as the core set can provably produce the same or similar results as solving the same problem regarding the original set. Since the core set selection takes into account the distribution of the entire population of the original set, the representative subset can preserve the sparsity of the original set. For example, the core set can well represent the dense and sparse regions of the distribution of the original set, with a much smaller number of selected samples compared to some other types of sampling, such as random sampling.

[0104] In some embodiments of the present disclosure, the core set of the contour group refers to a representative subset of contours that approximates the distribution of the contour group, such as, e.g., the distribution of the contours in the attribute space. Since the core set is created to represent the contour group extracted for a given aerial image (or a binary image of a given aerial image), such a core set is also referred to herein as an image-level core set or an image core set. The image core set includes one or more families, each family including at least one representative contour that represents one or more similar contours of a corresponding type from the contour group.

[0105] There are various ways to select / construct a core set for a contour group, such as, e.g., a greedy-based method or a statistics-based method, etc. Figure 3 FIG. illustrates a generalized flowchart of creating a core set of a contour group according to certain embodiments of the subject matter of the present disclosure.

[0106] The image core set can be initialized (302) first, e.g., by the image core set module 106 of the processing circuitry 102, which is a preliminary step in the construction process. The initial image core set can be constructed as an empty set before processing any contour in the group. For each given contour in the contour group, sequentially: a search for one or more reference contours in the image core set can be performed (304) based on a similarity measurement of the descriptors applied to each given contour (e.g., the descriptors of the given contour and any remaining contours in the core set, also referred to as candidate contours).

[0107] In some embodiments, one or more reference profiles can be identified by comparing the descriptor associated with a given profile with the corresponding descriptors associated with at least some of the remaining profiles in the core set. The comparison can be based on a similarity measure, and one or more profiles that satisfy the similarity criterion can be identified as the reference profiles of the given profile.

[0108] As an example, the similarity measure can be a distance-based metric such as, for example, Euclidean distance, Manhattan distance, cosine distance, Pearson correlation distance, Spearman correlation distance, etc. The similarity criterion can be a predetermined distance. In some cases, the number of reference profiles to be identified can also be predetermined.

[0109] In one example, the comparison can be performed, for example, by starting with the first candidate profile in the core set, calculating the distance between each pair of descriptors of the given profile and the candidate profiles from the remaining profiles in the core set, and when the number of similar profiles whose distance satisfies the similarity criterion is met, the search process ends and the identified similar profiles are provided.

[0110] Alternatively, in another example, the comparison can be performed by calculating the distance of each candidate profile in the remaining profiles relative to the given profile and selecting a predetermined number of reference profiles by sorting the calculated distances. In some cases, the number of reference profiles to be identified is not predetermined. All candidate profiles whose distance satisfies the similarity criterion can be identified as reference profiles.

[0111] Now refer to Figure 8B , which illustrates an example of the reference profile of a given profile according to certain embodiments of the subject matter of the present disclosure.

[0112] Assume that the number of reference profiles to be identified for each given profile is one. As an example, for profile 804, the candidate profile identified as satisfying the similarity criterion can be profile 802. For profile 806, the candidate profiles identified as satisfying the similarity criterion are profile 804 or 802. If the number of reference profiles is predetermined to be two, then both profiles 802 and 804 can be determined as the reference profiles of profile 806. For profile 810, the reference profile of the profile is profile 808.

[0113] A determination (306) can be made as to whether to identify reference profiles. In the case of identifying one or more reference profiles, the deviation between the given profile and each of the one or more reference profiles can be measured (308), and based on the measured deviation, it can be determined (310) whether to add the given profile to the family in the image core set. Otherwise, if no reference profiles are identified, the given profile can be directly added (312) to a new family in the image core set.

[0114] Figure 4Illustrated is a generalized flowchart for measuring the deviation between a given profile and a reference profile of the given profile, according to certain embodiments of the subject matter of the present disclosure.

[0115] As described above, the defect to be detected herein refers to edge displacement (or edge displacement error), and the edge displacement indicates a relatively significant deviation of a given profile from the expected position of the given profile (e.g., the position of the profile expected according to the original design data of the profile or according to the identified reference profile). It should be noted that the edge displacement is different from the edge roughness (which may be caused by different variations in the manufacturing process) at least in that: i) the edge displacement is local (present at a local position of the profile), while the edge roughness exists along the entire edge, and ii) the magnitude of the deviation of the edge displacement is relatively more significant (i.e., stronger / larger) compared to the magnitude of the fine roughness along the edge.

[0116] In some cases, such edge displacement may be caused by certain physical effects and / or other factors (such as, for example, oxidation, particles, scratches, crystal growth, electrostatic discharge (ESD), etc.) during the manufacturing process of the mask, and other factors may affect the electrical measurement of the manufactured device when printed on the wafer, thereby possibly causing a reduction in yield and degradation / failure of the device performance. Therefore, it is necessary to detect such displacement defects and measure the magnitude of the displacement defects.

[0117] Specifically, in order to measure the deviation indicating the magnitude of any possible displacement, the given profile may be registered (402) with one or more reference profiles, respectively, to generate one or more registered profile pairs. As an example, the registration may be performed by aligning the centroids and / or certain anchor points of the given profile and the reference profile. The distance between corresponding points of each registered profile pair may be measured (404). As an example, a distance metric (such as the Hausdorff distance metric) may be used to measure the distance. The deviation may be calculated (406) based on the measured distance in various ways. As an example, the maximum distance may be selected from the measured distances and used as the measured deviation between the given profile and the corresponding reference profile (of one or more similar profiles). As another example, the average distance of the measured distances may be derived and used as the measured deviation between the given profile and the corresponding reference profile.

[0118] The measured deviation between the given profile and each of one or more reference profiles may be used to determine whether to add the given profile to the core set, as described in reference box 310. In some embodiments, once one or more measured deviations between the given profile and one or more reference profiles are calculated, a combined deviation may be derived based on the one or more measured deviations, and a deviation threshold may be applied to the combined deviation.

[0119] In some cases, the deviation threshold can be predetermined based on, for example, a specific inspection application, the type of structural feature, the technology node used by the customer, and / or specifications, etc. For example, a combined deviation can be derived by averaging (or weighted averaging) one or more measurement deviations, such as the mean or median of the deviations or any other type of averaging calculation (with or without weights). As an example, in the case of identifying three reference profiles for a given profile, three deviations are measured respectively between the given profile and the three reference profiles. The three deviations can be averaged to generate a combined deviation that will be compared with the deviation threshold.

[0120] In the case where the combined deviation crosses the deviation threshold (e.g., the combined deviation exceeds the threshold), the given profile can be considered a different type of profile that is not representative in the core set, and thus will be added to the core set as a representative profile of a separate / new family. Otherwise, in the case where the combined deviation is below the deviation threshold, the given profile and the reference profile in the core set are considered similar profiles belonging to the same type of profile. Since the given profile already has a representative in the core set (i.e., the reference profile of the given profile), the given profile will not be added to the core set. In this case, for profiles of the same type, only one representative profile (or in some cases more than one, e.g., when the process variation between profiles of the same type is slightly prominent but still less than the deviation threshold) remains in the core set instead of many duplicate instances.

[0121] Now turning to Figure 9 , two examples of the measured deviations according to certain embodiments of the subject matter of the present disclosure are illustrated.

[0122] Graph 900 illustrates a profile 904 of a structural feature having an elliptical shape, and a reference profile 902 identified in the core set of profile 904 and representing the expected position of a profile of a defect-free structural feature. As shown, the two profiles are registered. In this example, the maximum distance 906 is measured as the deviation between the given profile 904 and the reference profile 902. In the case where the deviation is greater than a predetermined deviation threshold, profile 904 can be considered a different type of profile that is not representative in the core set, and thus will be added to the core set as a representative profile of a separate / new family. Graph 910 illustrates an example where the deviations at position 912 and position 914 are less than the predetermined deviation threshold, in which case the profiles are considered similar profiles to the reference profile and will not be added to the core set.

[0123] In some embodiments, when traversing profiles from a group of profiles extracted from a given image, as referenced Figure 3 and Figure 4As described in the core set construction process, each family in the image core set can be associated with a count of the number of similar contours of the corresponding contour type represented by at least one representative contour of the family. The count / number of such contours represented by each family can be continuously updated during the processing of a given contour. As an example, when processing a given contour in a contour group, if a reference contour for the given contour is found in a family from the core set and the measured deviation does not exceed the deviation threshold, then there is no need to add the given contour to the family because the given contour is already represented by the reference contour. The count associated with the family will be incremented by one, indicating an update to the number of similar contours represented by the family. In the case where no reference contour is found, or where a reference contour is found but the measured deviation exceeds the deviation threshold, the given contour will form a new family and the count associated with the new family will be initialized to one.

[0124] Figure 10 FIG. illustrates an example of a histogram representation of a contour group extracted from a given image according to certain embodiments of the subject matter of the present disclosure.

[0125] Assume that an aerial image representing a portion of a mask is obtained and the processing described with respect to blocks 206 and 208 is performed, thereby generating a contour group corresponding to the group of FOIs of printable features from the image. To create the image core set of the contour group, the process described with respect to Figure 3 can be performed. Specifically, an initial image core set can be first constructed, for example, as an empty subset. The first contour (such as contour 1002) in the contour group is to be processed. Since the current core set is empty, no reference contour for the first contour is identified when performing the search for the reference contour. Therefore, the first contour is directly added to the core set, forming family #1 that includes the representative contour 1002 of the first type of contour (e.g., a circular contour characterized by specific attributes).

[0126] Next, the second contour (such as contour 1004) in the group is to be processed. During the search for the reference contour, no reference contour is found according to the similarity measurement / similarity criterion because, for example, the distance between the first contour and the second contour does not satisfy the predefined similarity criterion. Therefore, the second contour is also directly added to the core set, forming family #2 that includes the representative contour 1004 of the second type of contour (e.g., a square contour characterized by specific contour attributes).

[0127] Assume that the third contour in the group is contour 1006. When searching for a reference contour, it is found that contour 1004 is similar to contour 1006. However, when measuring the deviation between the two contours, the measured deviation exceeds the deviation threshold. Therefore, contour 1006 is regarded as a contour of a different type and is assigned to the new family #3, and the new family includes the representative contour 1006 of the third type of contour (for example, a square contour with contour attributes different from the previous square contour). So far, the core set includes three families #1 to #3, and each family includes a representative contour.

[0128] Similarly, for any subsequent contour in the group, the above process will be repeated to determine whether any subsequent contour is added to the core set, or whether any subsequent contour is already represented by any family. For example, for any contour similar to contour 1002 in family #1, where the deviation between them is below the deviation threshold, such a contour will be regarded as a similar contour of the same type as contour 1002 and will not be added to the core set again. The count of the number of contours represented by family #1 will increase by one. Occasionally, in the case where a subsequent contour shows a significant change (still below the deviation threshold) from contour 1002, such a contour can be added to family #1 as an additional representative contour of the same type of contour.

[0129] As an example, for family #3, assume that the representative contour 1006 is located at the center of the distribution in the family. If the subsequent contour 1007 arrives, and the deviation of the subsequent contour from 1006 is slightly prominent (shown by the distance between 1006 and 1007 on the X-axis), although still below the deviation threshold, then contour 1007 can be added to family #3 as another representative of this type of contour. Similarly, for similar reasons, contour 1009 can be added to family #3 as an additional representative.

[0130] In some cases, once all the contours in the contour group have been processed, each family formed in the image core set can be associated with an indication of whether at least one representative contour of each family is normal or unique / abnormal at the image level. The indication can be derived based on the number of similar contours of the corresponding type represented by the family from the contour group.

[0131] In this example, one or more unique families can be identified, such as Family #4 and Family #5. These two families are considered unique because each of them includes a special type of contour, and the count of the number of contours represented by the family is very rare, indicating a sparse appearance of such features on a given part of the mask. As an example, when processing all the contours in a given image, the number of the special contour 1008 in Family #4 appears to be one, and the number of the special contour 1010 in Family #5 is also one. This is compared with the relatively large number of contours in Family #1, the number of contours in Family #1 totals several thousand, while the number of contours in Family #2 totals several hundred, and the number of contours in Family #3 totals several dozen. In this case, Family #1 to #3 or the representative contours therein are considered normal based on the frequent / abundant recurrence of the contour type in a given part of the mask, while Family #4 to #5 or the representative contours therein are considered unique / abnormal based on the rare / sparse occurrence of the contour type.

[0132] It should be noted that the created core set, as well as the normality and abnormality of the families and / or the contours of the families, are determined only within a given aerial image. Thus, as described above, such a core set is referred to as an image-level core set or an image core set. The families and / or the representative contours of the families are indicated as normal or abnormal at the image level.

[0133] It should also be noted that the Figure 3 exemplary method of creating a core set described with reference to is just one possible way of core set selection / construction. Other core set creation methods can be used instead of the above method, such as methods based on statistics and distribution, such as for example the K-means algorithm, the kernel density estimation (KDE) algorithm, etc.

[0134] Once the image core sets for each aerial image are created as described above, a plurality of image core sets corresponding to a plurality of aerial images are obtained. The plurality of image core sets can be merged (212) (e.g., by the mask core set module 1086 of the processing circuitry 103) to obtain a total core set of the mask. Each family in the mask core set includes at least one representative contour, and the at least one representative contour represents one or more similar contours of a corresponding type from a plurality of aerial images. Based on the number of similar contours of a corresponding type from a plurality of aerial images, at least one representative contour in a given family can be indicated as normal or abnormal. As an example, the mask core set can include at least one of the following: a family including at least one normal contour or a family including at least one abnormal contour.

[0135] In some embodiments, it can be related to the reference Figure 3 and Figure 4The multiple image core sets are combined into an overall mask core set in a manner similar to the described process. Generally, the deviation between representative contours from different image core sets among the multiple image core sets can be measured, and at least one representative contour can be retained in the mask core set, where the at least one representative contour represents one or more similar contours of the same type (e.g., when the measured deviation between the contours is below a predefined deviation threshold, the contours are considered similar).

[0136] In some embodiments, the multiple image core sets are generated in parallel by multiple processors (such as, for example, the multiple processors included in the processing circuitry 102). The multiple processors can be centralized or distributed. In some cases, the multiple processors can be hosted / integrated with the mask inspection tool. For example, the inspection tool can accommodate a queue of N processors, and whenever an aerial image is acquired, the aerial image will be sent to the first available processor in the queue for processing. Similarly, the next acquired image will be sent to the next idle processor in the queue. In this way, image acquisition and core set construction can be performed in parallel. The core set construction for different images can also be performed in parallel. Once the image core sets are created, the image core sets can be sent to a specific processor (such as, for example, the processor included in the processing circuitry 103), which is configured to merge / unify all the image core sets and generate a mask core set. The specific processor can be co-located with or separate from the multiple processors. As an example, the specific processor can be a separate processor located in the inspection tool or a processor located in a server separate from the tool.

[0137] Figure 11 A schematic diagram of mask core set generation based on image core sets according to certain embodiments of the subject matter of the present disclosure is shown.

[0138] Multiple aerial images (e.g., N individual images / frames 1102) are sequentially acquired by a mask inspection tool, and each aerial image captures a corresponding portion of the mask. For each individual image, when acquired by the tool, each individual image can be sent to a processor for image processing and core set construction, as referred to above Figures 2 to 4 described. In particular, a group of contours corresponding to the FOI group of multiple printable features from the individual image is extracted. The number of contours derived from an individual image can typically be on the order of thousands (i.e., O(1K)). When generating the image core set 1104, the magnitude of the representative contours in the core set can be sharply reduced to the order of dozens (i.e., O(10)). Since the search for reference contours and the deviation measurement for each given contour in the group are performed for the representative contours in the image core set rather than for all the remaining contours in the contour group, the construction and use of the image core set significantly improve the computational efficiency.

[0139] When generating N image core sets 1104 corresponding to N individual images 1102 respectively, they are sent for unification / merging into the mask core set. In some cases, the N image core sets can be sent together to a specific processor for core set unification 1106. In some other cases, especially in the case of parallel processing, each image core set can be sent individually when they are generated respectively. As an example, when receiving the first image core set sent by the first processor, the specific processor can store the first image core set. When receiving the second image core set sent by the second processor, the specific processor can start merging the two image core sets while waiting to receive other core sets. For example, assume the first image core set is similar to the core set example as Figure 10 shown, and the core set example includes five families #1 to #5. For each representative contour from the second image core set, the process described in boxes 304 to 312 of Figure 3 the reference can be performed with respect to the first image core set so as to integrate the two image core sets into an intermediate core set.

[0140] For example, starting from the first representative contour in the second image core set, a search for the reference contour can be performed in the first image core set, and if the reference contour is found, the deviation between the first representative contour and the reference contour can be measured to determine whether to add the first representative contour to the first image core set or whether there is already a representation of the first representative contour, similar to Figure 3 that described in boxes 304 to 312 of the reference.

[0141] Whenever a new image core set is received from the corresponding processor, the same process can be repeated, where the process in boxes 304 to 312 of Figure 3 the reference can be performed with respect to the intermediate core set. Finally, once all N image core sets have been received and processed, the overall mask core set can be generated. The mask core set can include one or more families, where each family includes at least one representative contour, and the at least one representative contour represents the corresponding type of similar contours from all N images.

[0142] Note that the families in the image core set and the families in the mask core set may not necessarily be the same in terms of the number of families, the number of representative contours in each family, and the number of contours / instances represented by each family. As an example, the first image core set may include families A, B, C, and D, where D may be temporarily indicated as abnormal / defective due to the single instance represented thereby, and the second image core set may include families A, D, E, and F. After merging two image core sets by contour comparison, the mask core set may include families A through F, where the count of instances represented by A is updated to the total number of instances from both image core sets, and the same is true for D. Additionally, D is updated to indicate a normal family since it is associated with a count of two instances rather than a single instance. It may also be the case that both image core sets include families A through D, and after merging, the mask core set also includes families A through D, where each family is associated with the total count of instances represented in the two image core sets. In some cases, for the purpose of differentiating the families in the image core set from the families in the mask core set after merging, the families in the image core set may also be referred to as image-level families, while the families in the mask core set may be referred to as mask-level families.

[0143] The magnitude of all representative contours from N image core sets can be on the order of millions or tens of millions (i.e., O(10M)). After merging and unifying, the overall mask core set may include on the order of thousands (i.e., O(1K)) of representative contours, where the unique / abnormal contours can be on the order of dozens (i.e., O(10)). The unique contours correspond to features that will be reported as defect candidates for edge displacement errors on the mask.

[0144] In addition to computational efficiency, the proposed core-set-based mask inspection can also effectively reduce false alarms detected across the entire mask. This is achieved through the merging and unifying of multiple image core sets, thus allowing for the sharing of contour statistics across the entire mask. Figure 12 A schematic illustration of false alarm reduction for mask inspection in accordance with certain embodiments of the subject matter of the present disclosure is shown.

[0145] During inspection, the mask holder and detector of the mask inspection tool may move in opposite directions during exposure (or one of them may move in steps relative to the other), and the mask may be scanned step by step by the inspection tool along the strips of the mask, where the inspection tool images a part / portion (within the strip) of the mask at a time. The strips are laid as parallel rows / columns adjacent to each other. In the present illustration 1200, the inspection tool may scan the mask from left to right along the strip, then switch to the next strip and scan the next strip in a similar manner (e.g., from right to left), and so on until the entire mask is scanned. Multiple aerial images of the mask may be obtained sequentially, each aerial image representing a corresponding part of the mask.

[0146] When each aerial image is inspected individually, in some cases, due to relatively low statistics and / or the presence of unique structural features in the aerial image, false alarms may exist among the detected defect candidates. As an example, there may be a structural feature in the aerial image having a unique contour shape that does not have a reference contour similar to the unique contour shape. In such a case, the inspection process may not find a reference for the structural feature and thus report the structural feature as a defect, or the inspection process may use a reference contour that is actually not similar to the feature for comparison and thus also report the structural feature as a defect. In such cases, these reported defects are false alarms because they actually represent the presence of a unique structural element rather than a real defect. In some cases, another instance of such a unique structural element may appear in some of the subsequent images. Figure 12 An example of such a situation is illustrated.

[0147] Assume that during inspection along the first strip, an aerial image 1202 capturing a portion of the capture mask is obtained. Through the image processing and core set construction process as described with respect to blocks 206, 208, and 210, an image core set of image 1202 is created, where one family is indicated as a unique family because the family includes a contour (illustrated in the figure in a triangular shape) that appears only once throughout the image. In the absence of the solution of the present disclosure, each image is processed independently, and the defect detection results of each image are reported regardless of the information of other images and their detection results. In this case, the contour of the triangle will be detected as a defect due to its uniqueness from other contours and its rare appearance.

[0148] However, as the inspection continues, during inspection along a subsequent strip, it is found that aerial image 1204 captures the same unique contour (i.e., the image core set of image 1204 also includes a unique family that includes a similar contour that appears only once in the image). Similarly, during inspection of the last strip, another image 1206 is obtained, and the other image 1206 also captures a similar contour.

[0149] When the image core sets of all aerial images on the mask are merged, all contour information and statistics at the image level can be centrally shared and compared. In this example, after merging, it can be found that the unique family includes a contour whose occurrence count throughout the mask is actually three, rather than one. Depending on the detection threshold (e.g., the number of occurrences of a feature on the mask, where the number of occurrences is represented by the number of contours of the feature represented by the unique family in the mask core set), such a contour and the corresponding feature can be understood as a false alarm indicating the presence of a unique structural feature rather than a defect. In this case, the indication of the family including the representative contour in the shape of a triangle will be updated from abnormal / unique to normal.

[0150] According to certain embodiments, for a family of at least one representative contour in a mask core set that is indicated as an anomaly, one or more similar contours of the corresponding type represented by the at least one representative contour can be identified, and one or more FOIs corresponding to the one or more similar contours of the corresponding type can be reported as one or more defect candidates. In some cases, for reporting purposes, a defect map of the mask can be provided, which indicates the presence and location of one or more defect candidates on the mask.

[0151] In some embodiments, optionally, when there are defect candidates and / or candidates for further review to determine if they are indeed real defects, it can be determined how to respond to any defects, e.g., by evaluating their printability, or evaluating whether these defects will affect the functionality of semiconductor samples fabricated using the mask when printed. As an example, the evaluation can include estimating the changes in printable structural elements / features associated with the defects when printed on the semiconductor sample. As an example, possible processing operations in response to the presence of defects can include: repairing the mask, defining the mask as a failed mask, defining the mask as functional, generating a repair indication for the mask, etc. For example, if these estimated changes are unacceptable, the mask can be sent to a mask shop for repair or rejection.

[0152] Additionally, in some embodiments, at least one or any combination of the following outputs / indications can be provided: (I) providing qualification criteria for the mask to be shipped out of the mask shop; (ii) providing input to the mask generation process; (iii) providing input to the semiconductor sample fabrication process; (iv) providing input to the simulation model used in the lithography process; (v) providing a correction map for the lithography tool; and (vi) identifying regions on the mask with characteristic parameter changes greater than expected.

[0153] In some embodiments, the mask core set generated for the entire mask can be used as a (214) mask core set model for subsequent inspection of one or more masks (e.g., the same mask or one or more new masks). As an example, the mask core set can be used as a core set model, and inspection methods such as die-to-history, die-to-model, etc. can be used to inspect one or more masks based on the mask core set. For example, for a new mask with a design similar to a previous mask, once the contour groups of each image of the new mask are extracted (as described in reference box 208), the effort of creating an image core set can be saved, and instead, each contour can be compared with the previously created core set model in a manner similar to that Figure 3 and Figure 4 described.

[0154] It should be noted that the mask applicable to the inspection method of the present disclosure can be any type of mask, and the mask may suffer from the types of edge displacement defects described herein, including but not limited to memory masks and / or logic masks, and / or Arf masks and / or EUV masks, etc. The present disclosure is not limited to a specific type or functionality of the mask to be inspected.

[0155] For exemplary and illustrative purposes, certain embodiments and / or examples of the subject matter of the present disclosure herein are described with respect to structural features having a specific type / shape and / or specific edge displacement. This is in no way intended to limit the present disclosure in any way. It should be understood that the proposed methods and systems can be applied to other types / shapes of structural elements having various edge displacements.

[0156] According to certain embodiments, the mask inspection process referred to above Figure 2 、 Figure 3 and Figure 4 described can be included as part of an inspection recipe, and the inspection recipe can be used by system 101 and / or inspection tool 120 for on-line inspection during runtime. Therefore, the subject matter of the present disclosure also includes systems and methods for generating an inspection recipe during the recipe setup phase, where the recipe includes steps as described with reference to Figure 2 、 Figure 3 and Figure 4 (and their various embodiments). It should be noted that the term "inspection recipe" should be broadly interpreted to cover any recipe that an inspection tool can use to perform operations related to any type of mask inspection, including the above embodiments.

[0157] It should be noted that the examples illustrated in the present disclosure (such as mask inspection tool architectures and configurations, mask layouts, exemplary structural features, specific ways of contour extraction, core set construction, and deviation measurement as described above, etc.) are shown for example purposes and should not be considered as limiting the present disclosure in any way. As an addition or alternative to the above, other suitable examples / implementations can be used.

[0158] One advantage of certain embodiments of the mask inspection process described herein is the ability to detect a specific type of defect (i.e., edge displacement) with respect to the structural features on the mask. The proposed process is specifically designed for single-die masks where there is no reference die available for die-to-die comparison. Additionally, the proposed process is also applicable to inspect non-die regions when needed, such as scribe regions and peripheral regions in both single-die masks and multi-die masks, as well as die regions in multi-die masks.

[0159] Among other advantages of certain embodiments of a mask inspection process as described herein, the proposed inspection process does not require acquiring images from different modalities for the purpose of providing a reference image, which can be time-consuming. Nor does it require acquiring design data of the mask (which may not be available in many cases) or simulations based on design data that tend to be inaccurate. The proposed process utilizes specific processing of aerial images to provide a reference within the image itself, which has proven to have improved accuracy and sensitivity in defect detection in advanced process control of mask features.

[0160] Another advantage of certain embodiments of the mask inspection process as described herein is that by creating an image-level image core set and, for each given contour from a group of contours extracted from an image, searching for a reference contour and deviation measurement in the image core set (where the magnitude of representative contours in the core set is on the order of tens (i.e., O(10))), rather than performing this operation with respect to all remaining contours in the group of contours (where the magnitude of contours from a single image is typically on the order of thousands (i.e., O(101K))), the computational efficiency of the inspection system can be significantly increased, thereby improving the system throughput (TpT).

[0161] Another advantage of certain embodiments of the mask inspection process as described herein is that the merging of multiple image core sets enables sharing of contour statistics among multiple aerial images across the mask, which, compared to processing each aerial image independently, allows for a reduction in false alarms in the detection results. This can further enable a reduction in the detection threshold and allow for a more sensitive scan that can reveal smaller defects and improve detection sensitivity.

[0162] It should be understood that the present disclosure is not limited in its application to the details set forth in the specification included herein or shown in the drawings.

[0163] In this detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, those skilled in the art will understand that the subject matter of the present disclosure may 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 the present disclosure.

[0164] Unless otherwise specifically stated, as will be apparent from this discussion, it should be understood that throughout the specification, discussions using terms such as "inspect", "obtain", "generate", "simulate", "apply", "estimate", "extract", "create", "merge", "identify", "report", "initialize", "search", "determine", "measure", "add", "acquire", "execute", "register", "calculate", "maintain", "share", "permit", etc. refer to the (multiple) actions and / or (multiple) processes of a computer that manipulates data and / or transforms data into other data, where the data is represented as physical quantities (such as electrical quantities), and / or the data represents physical objects. The term "computer" should be broadly interpreted to cover any type of hardware-based electronic device having data processing capabilities. By way of non-limiting example, computers include the mask inspection systems, mask defect detection systems, and their respective parts disclosed in this application.

[0165] The term "mask" as used in this specification is also referred to as "lithography mask" or "photomask" or "reticle". Such terms should be equivalently and expansively interpreted to cover a template that holds a circuit design (e.g., defining the layout of a particular layer of an integrated circuit) that will be patterned onto a semiconductor wafer in a lithography process. As an example, a mask can be implemented as a fused silica plate covered with a pattern of opaque, transparent, and phase-shifting regions that is projected onto the wafer in a lithography process. As an example, a mask can be an extreme ultraviolet (EUV) mask or an argon fluoride (ArF) mask. As another example, a mask can be a memory mask (usable for manufacturing memory devices) or a logic mask (usable for manufacturing logic devices).

[0166] As used in this specification, the term "inspection" or "mask inspection" should be broadly interpreted to cover any operation for evaluating the accuracy and integrity of a manufactured photomask relative to a circuit design and its ability to produce an accurate representation of the circuit design on a wafer. Inspection can include any type of operation related to defect detection, defect review, and / or various types of defect classification and / or metrology operations during and / or after the mask manufacturing process and / or during the use of the mask for semiconductor sample manufacturing. Inspection can be performed after mask manufacturing by using non-destructive inspection tools. As a non-limiting example, inspection processes can include one or more of the following operations: runtime scanning (single scan or multiple scans) of the mask or a portion of the mask using an inspection tool, imaging, sampling, detection, measurement, classification, and / or other operations. Similarly, mask inspection can also be interpreted to include, for example, generating (multiple) inspection recipes and / or other setup operations prior to the actual inspection of the mask. Note that unless otherwise specifically stated, the term "inspection" or derivatives of the term "inspection" used in this specification are not limited to the resolution or size of the inspection area. As a non-limiting example, various non-destructive inspection tools include optical inspection tools, scanning electron microscopes, atomic force microscopes, etc.

[0167] As used in this specification, the term "metrology operation" should be broadly interpreted to cover any metrology operation process for extracting metrology information related to one or more structural elements on a mask. In some embodiments, metrology operations can include measurement operations such as, for example, critical dimension (CD) measurements performed on certain structural elements on a sample, including but not limited to the following: dimensions (e.g., line width, line pitch, contact diameter, size of an element, edge roughness, gray level statistics, etc.); shape of an element; distance within or between elements; associated angles; overlap information associated with elements corresponding to different design levels, etc. For example, measurement results such as measured images are analyzed by employing image processing techniques. Note that unless otherwise specifically stated, the term "metrology" or derivatives of the term "metrology" used in this specification are not limited to measurement techniques, measurement resolution, or the size of the inspection area.

[0168] As used in this specification, the term "sample" should be broadly interpreted to cover any type of wafer, related structures, and their combinations and / or portions used for manufacturing semiconductor integrated circuits, magnetic heads, flat panel displays, and other semiconductor finished products.

[0169] The term "defect" as used in this specification should be broadly interpreted to cover any type of anomaly or undesired feature formed on a mask. In some cases, a defect can refer to a true defect or defect of interest (DOI) that, when printed on a wafer, can have an impact on the functionality of the manufactured device, and thus the customer is interested in detecting the defect. For example, any "killer" defect that may cause yield loss can be represented as a DOI. In some other cases, a defect can be a negligible impairment (also referred to as a "false alarm" defect) as the defect has no impact on the functionality of the completed device and does not affect the yield.

[0170] The term "defect candidate" as used in this specification should be broadly interpreted to cover a suspected defect location on a mask that is detected as having a high probability of becoming a defect of interest (DOI). Thus, when being reviewed / tested, a DOI candidate can actually be a DOI, or in some other cases, it can be an impairment, or it can be any noise caused by different variations (e.g., process variations, color variations, mechanical and electrical variations, etc.) during inspection.

[0171] The term "(a)n image" or "image data" as used in this specification should be broadly interpreted to cover any original image / frame of a mask captured by a mask inspection tool, derivatives of the captured image / frame obtained from various preprocessing stages, and / or computer-generated synthetic images. It should be noted that in some cases, in addition to the image (e.g., the captured image, the processed image, etc.), the image data referred to herein may also include numerical data associated with the image (e.g., metadata, manually crafted attributes, etc.). It should be further noted that the image data is related to the target layer of the semiconductor device to be printed on the wafer.

[0172] The terms "non-transitory memory" and "non-transitory storage medium" as used herein should be broadly interpreted to cover any volatile or non-volatile computer memory suitable for the subject matter of this disclosure. The terms should be considered to include a single medium or multiple media that store one or more instruction sets (e.g., a centralized or distributed database, and / or associated caches and servers). The terms should also be understood to include any medium that is capable of storing an instruction set executable by a computer or encoding the instruction set and causing the computer to execute any one or more of the methods of this disclosure. Thus, the terms should be considered to include, but not be limited to, read-only memory ("ROM"), random access memory ("RAM"), disk storage media, optical storage media, flash memory devices, etc.

[0173] It should be understood that, unless otherwise specifically stated, certain features of the subject matter of the present disclosure described in the context of separate embodiments may also be provided in combination in a single embodiment. Conversely, the various features of the subject matter of the present disclosure described in the context of a single embodiment may be provided separately or in any suitable sub-combination. In this detailed description, numerous specific details are set forth in order to provide a thorough understanding of the methods and apparatuses.

[0174] It should also be understood that the systems according to the present disclosure can be implemented at least in part on a suitably programmed computer. Similarly, the present disclosure contemplates a computer-readable computer program for performing the methods of the present disclosure. The present disclosure further contemplates a non-transitory computer-readable memory for performing the methods of the present disclosure, the non-transitory computer-readable memory tangibly embodying a program of instructions executable by a computer.

[0175] The present disclosure is capable of having other embodiments and of being practiced and carried out in various ways. Accordingly, it should be understood that the language and terminology used herein are for the purpose of description and should not be regarded as limiting. Thus, those skilled in the art will understand that the concepts on which the present disclosure is based can readily be used as a basis for designing other structures, methods, and systems for several purposes of implementing the subject matter of the present disclosure.

[0176] Those skilled in the art will readily understand that various modifications and changes can be made to the above-described embodiments of the present disclosure without departing from the scope of the present disclosure as defined by the appended claims.

Claims

1. A computerized system for inspecting a mask that can be used to fabricate a semiconductor sample, the system including one or more processing circuitry configured to: Obtain a plurality of aerial images, each aerial image capturing a respective portion of the mask; Generate a plurality of image core sets corresponding to the plurality of aerial images, and for each given aerial image, the generating includes: i) Apply a print threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) Extract the contour of each feature of interest (FOI) in the group of features of interest (FOI) from the plurality of printable features, and generate a descriptor characterizing the contour, thereby generating a group of contours corresponding to the group of FOIs and associated with the respective descriptors; and iii) Create an image core set of the group of contours based on the respective descriptors of the group of contours, the image core set including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the group of contours; and Merge the plurality of image core sets to obtain a mask core set, wherein each family in the mask core set includes at least one representative contour, the at least one representative contour representing one or more similar contours of a respective type from the plurality of aerial images, and the at least one representative contour is indicated as normal or abnormal based on the number of the one or more similar contours of the respective type from the plurality of aerial images.

2. The computerized system according to claim 1, wherein for a family in the mask core set that includes at least one representative contour indicated as abnormal, the one or more processing circuitry are further configured to: identify one or more similar contours of the respective type represented by the at least one representative contour, and report one or more FOIs corresponding to the one or more similar contours of the respective type as one or more defect candidates.

3. The computerized system according to claim 2, wherein a defect candidate from the one or more defect candidates represents an edge displacement error, the edge displacement error indicating a relatively significant deviation of the contour of the FOI from the expected position of the FOI.

4. The computerized system according to claim 1, wherein each family in the image core set is associated with an indication of being normal or abnormal at the image level for the at least one representative contour of each family, the indication being obtained based on the number of the one or more similar contours of the respective type from the group of contours.

5. The computerized system according to claim 1, wherein the image core set is a subset of representative contours approximating the distribution of the group of contours.

6. The computerized system according to claim 5, wherein the one or more processing circuitry are configured to create the image core set by: Initializing the image core set; For each given contour in the group of contours, sequentially: Searching for one or more reference contours in the image core set based on a similarity measure of the descriptors applied to each given contour; In response to the one or more reference contours being identified, measuring a deviation between the given contour and each of the one or more reference contours; And Based on the measured deviation, determining whether to add the given contour to a family in the image core set.

7. The computerized system according to claim 6, wherein in response to not finding the one or more reference contours, the one or more processing circuitry are configured to add the given contour to a new family in the image core set.

8. The computerized system according to claim 6, wherein the one or more processing circuitry are configured to measure the deviation between the given contour and each of the one or more reference contours by: Registering the given contour with each of the one or more reference contours, respectively, to produce one or more registered contour pairs; Measuring the distance between corresponding points of each registered contour pair; And Calculating the deviation based on the measured distance.

9. The computerized system according to claim 1, wherein the one or more processing circuitry are configured to merge the plurality of image core sets by: measuring a deviation between representative contours from different image core sets among the plurality of image core sets; and retaining at least one representative contour representing one or more similar contours of the same type in the mask core set.

10. The computerized system according to claim 1, wherein the plurality of image core sets are at least partially generated in parallel by a plurality of processors and are sent to a specific processor for merging.

11. The computerized system according to claim 1, wherein the descriptor of the contour is based on one or more of the following: the region formed by the contour, the width of the region, the height of the region, the number of pixels along the contour, a chain code, a center of gravity, and the polar coordinates of the contour.

12. The computerized system according to claim 1, wherein the plurality of aerial images are sequentially acquired by a photomask inspection tool configured to simulate the optical configuration of a lithography tool.

13. The computerized system according to claim 1, wherein the merging of the plurality of image core sets enables sharing of contour statistics between the plurality of aerial images across the mask, thereby allowing false alarms to be reduced and detection sensitivity to be increased.

14. The computerized system according to claim 1, wherein the mask core set can be used as a mask core set model for subsequent inspection of one or more masks.

15. The computerized system according to claim 1, wherein the plurality of aerial images are captured for a single-die mask or a multi-die mask.

16. A computerized method for inspecting a mask that can be used to fabricate a semiconductor sample, the method comprising: Obtaining a plurality of aerial images, each aerial image capturing a corresponding portion of the mask; Generate a plurality of image core sets corresponding to the plurality of aerial images. For each given aerial image, the generation includes: i) Apply a printing threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) Extract the contour of each FOI in the group of features of interest (FOI) from the plurality of printable features, and generate a descriptor characterizing the contour, thereby generating a group of contours corresponding to the group of FOIs and associated with the corresponding descriptors; and iii) Create an image core set based on the corresponding descriptors of the group of contours, the image core set including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the group of contours; and Merge the plurality of image core sets to obtain a mask core set, wherein each family in the mask core set includes at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the plurality of aerial images, and the at least one representative contour is indicated as normal or abnormal based on the number of the one or more similar contours of the corresponding type from the plurality of aerial images.

17. The computerized method according to claim 16, wherein the image core set is a subset of representative contours approximating the distribution of the group of contours.

18. The computerized method according to claim 17, wherein the image core set is created by: Initializing the image core set; For each given contour in the group of contours, sequentially: Search for one or more reference contours in the image core set based on a similarity measurement of the descriptors applied to each given contour; In response to the one or more reference contours being identified, measure the deviation between the given contour and each of the one or more reference contours; And Determine whether to add the given contour to the family in the core set based on the measured deviation.

19. The computerized method according to claim 16, wherein the merging of the plurality of image core sets comprises: Measure the deviation between representative contours from different image core sets among the plurality of image core sets; And retain at least one representative contour representing one or more similar contours of the same type in the mask core set.

20. A non-transitory computer-readable storage medium tangibly embodying a program of instructions that, when executed by a computer, cause the computer to perform a method for inspecting a mask capable of being used to fabricate a semiconductor sample, the method including: Obtain a plurality of aerial images, each aerial image capturing a corresponding portion of the mask; Generate a plurality of image core sets corresponding to the plurality of aerial images. For each given aerial image, the generation includes: i) Apply a printing threshold to the given aerial image to obtain a binary image representing a plurality of printable features of the given aerial image; ii) Extract the contour of each FOI in the group of features of interest (FOI) from the plurality of printable features, and generate a descriptor characterizing the contour, thereby generating a group of contours corresponding to the group of FOIs and associated with the corresponding descriptors; and iii) Create a core set of images of the group of contours based on the corresponding descriptors of the group of contours, the core set of images including one or more families, each family including at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the group of contours; and Merge the plurality of core sets of images to obtain a core set of masks, wherein each family in the core set of masks includes at least one representative contour, the at least one representative contour representing one or more similar contours of a corresponding type from the plurality of aerial images, the at least one representative contour being indicated as normal or abnormal based on the number of one or more similar contours of the corresponding type from the plurality of aerial images.

Images