KR20230075369A - Mask inspection for semiconductor specimen fabrication - Google Patents

Abstract

A system and method for mask inspection is provided, comprising: obtaining, for an inspection area of a mask, a plurality of inspection images and a set of reference images, one or more defect candidates located in the inspection area of each inspection image. Generating a plurality of defect maps each including, arranging one or more defect candidates in a list of defect candidates of interest (DCI), for at least one given DCI in the list, a plurality of corresponding to a plurality of inspection images. generating difference patches of , calculating a class based on the plurality of difference patches, and applying a detection threshold to the class to determine whether a given DCI is a defect of interest (DOI).

Description

Mask inspection for semiconductor specimen manufacturing {MASK INSPECTION FOR SEMICONDUCTOR SPECIMEN FABRICATION}

The presently disclosed subject matter relates generally to the field of mask inspection, and more specifically to defect detection for photomasks.

Current demands for high density and performance associated with ultra-large scale integrated circuits of fabricated microelectronic devices require submicron features, increased transistor and circuit speeds, and improved reliability. As semiconductor processes advance, pattern dimensions, eg, line width, and other types of critical dimensions continue to shrink. Such demands require forming device features with high precision and uniformity, which in turn requires careful monitoring of the manufacturing process, including automated inspection of devices while they are still in the form of semiconductor wafers. .

Semiconductor devices are often fabricated using photolithography masks (also referred to as photomasks or masks or reticles) in a photolithography process. The photolithography process is one of the main processes in the manufacture of semiconductor devices, and includes patterning the surface of a wafer according to the circuit design of semiconductor devices to be manufactured. Such a circuit design is first patterned on a mask. Therefore, in order to obtain working semiconductor devices, the mask must be free of defects. Masks are manufactured by a complex process and can suffer from various defects and variations.

Additionally, the mask is often used in an iterative manner to create many dies on a wafer. Thus, any defect on the mask will be repeated multiple times on the wafer, leaving multiple devices defective. Establishing a productive process requires tight control of the overall lithography process, especially considering the large-scale circuit integration and decreasing size of semiconductor devices.

Various mask inspection methods have been developed and utilized. According to certain conventional techniques for designing and evaluating masks, a mask is created and used to expose a wafer therethrough, and then inspection is performed to determine whether the features/patterns of the mask have been transferred to the wafer according to the design. is carried out Any variations in the final printed features from the intended design may require modifying the design, repairing the mask, creating a new mask, and/or exposing a new wafer.

Alternatively, the mask can be directly inspected using various mask inspection tools. An inspection process may include a plurality of inspection steps. During the fabrication process of the mask, the inspection steps may be performed multiple times, eg after fabrication or processing of certain layers, etc. Additionally or alternatively, each inspection step may be repeated multiple times, for example for different mask positions or for the same mask positions with different inspection settings.

Mask inspection generally involves generating specific inspection output (eg, images, signals, etc.) for a mask by directing light or electrons towards the mask and detecting the light or electrons from the mask. Once the output is generated, defect detection is typically performed by applying a defect detection method and/or algorithm to the output. Very often, the goal of inspection is to increase the effectiveness of suppressing the detection of false alarms/disturbings and noises, while reducing the defects of interest (which, if not corrected, cause the end device to fail to meet the desired performance or malfunction). and, therefore, to provide high sensitivity for detection of yield).

According to certain aspects of the presently disclosed subject matter, a computerized system for inspecting a mask usable for fabricating a semiconductor specimen is provided, the system including a processing and memory circuit (PMC), the PMC comprising: an inspection area of the mask. , acquire a plurality of inspection images having a plurality of fields of view (FOVs) overlapped at least by the inspection area, and for each inspection image, a plurality of references of each reference area among one or more corresponding reference areas. obtain a set of reference images comprising images; generating a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more defect candidates located in an inspection area of each inspection image; Sort and create a list of defect candidates of interest (DCI); For at least one given DCI in the list, generate a plurality of difference patches - the PMC: extracts an image patch surrounding the location of the given DCI from the set of test image and reference images, respectively, the test patch and the reference patch generating a set of s; For each reference patch, compute a filter optimized to minimize the difference between the corrected reference patch obtained using the filter and the inspection patch, and a set of corrected reference patches and a set of filters corresponding to the set of reference patches. to generate; and a difference corresponding to each inspection image of the plurality of inspection images by combining the set of calibrated reference patches to obtain a composite reference patch and comparing the inspection patch with the composite reference patch to obtain a difference patch. configured to create patches -; and calculate a grade based on the plurality of difference patches and apply a detection threshold to the grade to determine whether a given DCI is a defect of interest (DOI).

In addition to the above features, a system according to this aspect of the disclosed subject matter may include one or more of the features (i) to (xii) listed below, in any desired combination or permutation technically possible.

(i). The mask is a multi-die mask. The inspection area is located on the inspection die on the mask. Each of the one or more reference areas is from one or more reference dies of the inspection dies on the mask.

(ii). The mask is a single die mask. The inspection area and one or more reference areas are from a single die on the mask and share the same design pattern.

(iii). The PMC is configured to respectively register the test patch with each reference patch to correct for each offset between the test patch and each reference patch from the set prior to computing the filter.

(iv). The filter is computed to correct at least one of the following noises of the reference patch: registration residual error, intensity gain and offset, defocus, or field of view (FOV) distortion.

(v). The filter includes a set of filter elements for correcting the respective noises of the reference patch.

(vi). The ranking is calculated by calculating a score for each of the plurality of difference patches based on the highest pixel value in the difference patch, generating a plurality of scores corresponding to the plurality of difference patches, and calculating the plurality of scores to obtain the ranking. It is calculated by averaging.

(vii). The PMC performs generating a plurality of difference patches, calculating a rank and applying a detection threshold to each DCI in the list of DCIs to determine whether the DCI is a DOI, and by determining a detected and to provide an updated defect map corresponding to the inspection area.

(viii). PMC acquires multiple inspection images, creates multiple defect maps, aligns one or more defect candidates, creates multiple difference patches, calculates a grade, and performs one or more additional inspections on a mask. and further configured to repeat applying the detection threshold to the region.

(ix). A plurality of inspection images are acquired sequentially by the actinic inspection tool having a predefined step size. The actinic inspection tool is configured to emulate the optical configuration of a lithography tool usable for fabrication of semiconductor specimens.

(x). The system further includes an actinic inspection tool.

(xi). The plurality of inspection images are: sequentially acquiring a plurality of images using a non-actinic inspection tool with a predefined step size, and a plurality of images to simulate the optical configuration of a lithography tool usable for fabrication of a semiconductor specimen. It is obtained by performing simulation on the images of and generating a plurality of inspection images.

(xii). The list of defect candidates of interest (DCI) includes one or more defect candidates common to at least a majority of the plurality of inspection images.

According to other aspects of the presently disclosed subject matter, a method of inspecting a mask usable for fabricating a semiconductor specimen is provided, the method performed by a processing and memory circuit (PMC), the method comprising: for an inspection area of the mask. , Acquiring a plurality of inspection images having a plurality of fields of view (FOVs) overlapped by at least the inspection area, and for each inspection image, a plurality of reference images of each reference area among one or more corresponding reference areas. obtaining a set of reference images comprising; generating a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more defect candidates located in an inspection area of each inspection image; sorting and generating a list of defect candidates of interest (DCI); For at least one given DCI in the list, generating a plurality of difference patches, the generating step: extracting, respectively, an image patch surrounding the location of the given DCI from the test image and the set of reference images, the test patch and creating a set of reference patches; For each reference patch, compute a filter optimized to minimize the difference between the corrected reference patch obtained using the filter and the inspection patch, and a set of corrected reference patches and a set of filters corresponding to the set of reference patches. generating; and combining the set of calibrated reference patches to obtain a composite reference patch, and comparing the test patch with the composite reference patch to obtain a difference patch, thereby corresponding to each inspection image of the plurality of inspection images. including generating a difference patch; and calculating a rank based on the plurality of difference patches and applying a detection threshold to the rank to determine whether a given DCI is a defect of interest (DOI).

This aspect of the disclosed subject matter may include one or more of the features (i) through (xii) listed above in relation to a system, mutatis mutandis, in any desired combination or permutation technically possible.

According to other aspects of the disclosed subject matter, a non-transitory computer-readable medium containing instructions is provided, which instructions, when executed by a computer, cause the computer to inspect a method for inspecting a mask usable for fabricating a semiconductor specimen. and the method: acquires, for an inspection area of the mask, a plurality of inspection images having at least a plurality of fields of view (FOVs) overlapped by the inspection area, and for each inspection image, one or more corresponding obtaining a set of reference images including a plurality of reference images of each reference area of the reference area; generating a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more defect candidates located in an inspection area of each inspection image; sorting and generating a list of defect candidates of interest (DCI); For at least one given DCI in the list, generating a plurality of difference patches, the generating step: extracting, respectively, an image patch surrounding the location of the given DCI from the test image and the set of reference images, the test patch and creating a set of reference patches; For each reference patch, compute a filter optimized to minimize the difference between the corrected reference patch obtained using the filter and the inspection patch, and a set of corrected reference patches and a set of filters corresponding to the set of reference patches. generating; and combining the set of calibrated reference patches to obtain a composite reference patch, and comparing the test patch with the composite reference patch to obtain a difference patch, thereby corresponding to each inspection image of the plurality of inspection images. including generating a difference patch; and calculating a rank based on the plurality of difference patches and applying a detection threshold to the rank to determine whether a given DCI is a defect of interest (DOI).

This aspect of the disclosed subject matter may include one or more of the features (i) through (xii) listed above in relation to a system, mutatis mutandis, in any desired combination or permutation technically possible.

In order to understand the present disclosure and see how it can be practiced, embodiments will now be described, by way of non-limiting example only, with reference to the accompanying drawings, in which:
1 illustrates a functional block diagram of a mask inspection system in accordance with certain embodiments of the presently disclosed subject matter.
2 illustrates a generalized flow diagram of mask inspection for a mask usable for fabricating a semiconductor specimen, in accordance with certain embodiments of the presently disclosed subject matter.
3 illustrates a generalized flow diagram for computing and applying a filter in accordance with certain embodiments of the presently disclosed subject matter.
4 illustrates a generalized flow diagram for calculating a rating in accordance with certain embodiments of the presently disclosed subject matter.
5 illustrates a schematic diagram of an actinic inspection tool and lithography tool, in accordance with certain embodiments of the presently disclosed subject matter.
6 schematically illustrates an example of a multi-die mask as well as an inspection region and a plurality of inspection images and reference images thereof, in accordance with certain embodiments of the presently disclosed subject matter.
7 schematically illustrates an example of a single die mask as well as an inspection area and a plurality of inspection images and reference images thereof, in accordance with certain embodiments of the presently disclosed subject matter.
8 illustrates an example of a plurality of difference patches in accordance with certain embodiments of the presently-disclosed subject matter.

In the detailed description that follows, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, one of ordinary skill in the relevant art will understand that the disclosed subject matter may be practiced without these specific details. In other instances, well known methods, procedures, components and circuits have not been described in detail in order not to obscure the subject matter disclosed.

As will be apparent from the discussions below, the terms “examine”, “acquire”, “generate”, “sort”, “extract”, “compute”, “combine”, “compute”, “combine”, Discussions utilizing terms such as "compare", "acquisition", "compute", "apply", "match", "calibrate", "average", "perform", "provide", "repeat", "acquire", etc. , the operation(s) and/or process(s) of a computer that manipulates and/or transforms data, wherein the data is physically represented by, eg, electronic, quantities, and/or the data represents physical objects, into other data. ) is understood to refer to The term "computer" is broadly used to encompass any kind of hardware-based electronic device having data processing capabilities, including, by way of non-limiting example, the mask inspection system, defect detection system, and respective portions thereof disclosed herein. should be interpreted

The term "mask" as used herein is also referred to as "photolithography mask" or "photomask" or "reticle". Such terms should be interpreted equivalently and broadly to encompass a template bearing circuit design (eg, defining the layout of a particular layer of an integrated circuit) to be patterned on a semiconductor wafer in a photolithography process. As an example, the mask can be implemented as a fused silica plate covered with a pattern of opaque, transparent and phase shifting regions that are projected onto wafers in a lithography process. As an example, the mask may be an extreme ultraviolet (EUV) mask or an argon fluoride (ArF) mask. As another example, the mask can be a memory mask (usable to fabricate a memory device) or a logic mask (usable to fabricate a logic device).

As used herein, the term "inspection" or "mask inspection" refers to any operation for evaluating the accuracy and integrity of a fabricated photomask with respect to a circuit design, and its production of an accurate representation of the circuit design on a wafer. It should be interpreted broadly to encompass competence. Inspection is any kind of operation associated with various types of defect detection, defect review and/or defect classification, and/or metrology operations during and/or after the mask fabrication process and/or during the use of the mask for semiconductor specimen fabrication. can include Inspection may be provided using non-destructive inspection tools after fabrication of the mask. By way of non-limiting example, the inspection process may include the following operations: scanning (single or multiple scans), imaging, sampling, detection, measurement, classification, and provisioning of a mask or portions thereof, using an inspection tool. /or one or more of the other operations. Likewise, mask inspection may also be interpreted to include, for example, generating inspection recipe(s) and/or other setup operations prior to actual inspection of the mask. Note that unless specifically stated otherwise, the term "inspection" or derivatives thereof used herein are not limited with respect to the size or resolution of the inspection area. Various non-destructive inspection tools include, by way of non-limiting example, optical inspection tools, scanning electron microscopes, atomic force microscopes, and the like.

The term "metrology operation" as used herein should be interpreted broadly to encompass any metrology operation procedure used to extract metrology information about one or more structural elements, such as a mask, on a semiconductor specimen. . In some embodiments, metrology operations may include, for example: dimensions (eg, line widths, line spacing, contact diameters, size of an element, edge roughness, gray level statistics, etc.), shapes of elements , distances within or between elements, angles involved, overlay information associated with elements corresponding to different design levels, etc. (CD) measurements. Measurement results, eg measured images, are analyzed eg by employing image processing techniques. Note that unless specifically stated otherwise, the term "metrology" or derivatives thereof used herein is not limited with respect to measurement technique, measurement resolution, or size of an inspection area.

As used herein, the term "specimen" refers to any type of wafers, related structures, and combinations thereof used to fabricate semiconductor integrated circuits, magnetic heads, flat panel displays, and other semiconductor manufactured articles. and/or parts should be construed broadly.

The term "defect" as used herein should be interpreted broadly to encompass any kind of abnormality or undesirable feature/functionality formed on a mask. In some cases, a defect is a real defect that, when printed on a wafer, has certain effects on the functionality of a manufactured device, and thus, what is detected may be a defect of interest (DOI) of interest to a customer. For example, any “killer” defects that can cause yield loss may be marked with a DOI. In some other cases, the fault may be a negligible disturbance (also referred to as a “false alarm” fault) because it has no impact on the functionality of the finished device.

The term "defect candidate" as used herein should be interpreted broadly to encompass a suspected defect location on a mask that is detected as having a relatively high probability of being a defect of interest (DOI). Therefore, a defect candidate, when reviewed, may actually be a DOI, or in some other cases may be caused by different variations during inspection (eg, process variations, color variations, mechanical and electrical variations, etc.) It can be a nuisance or random noise.

The terms "non-transitory memory" and "non-transitory storage medium" as used herein should be interpreted broadly to encompass any volatile or non-volatile computer memory suitable for the presently disclosed subject matter. The terms should be taken to include a single medium or multiple mediums (eg, a centralized or distributed database, and/or associated caches and servers) that store one or more sets of instructions. The terms should also be construed to include any medium capable of storing or encoding a set of instructions for execution by a computer and causing the computer to perform any one or more of the methodologies of this disclosure. Accordingly, the terms should be construed as including, but not limited to read only memory (“ROM”), random access memory (“RAM”), magnetic disk storage media, optical storage media, flash memory devices, and the like.

Unless specifically stated otherwise, it is understood that certain features of the disclosed subject matter that are described in the context of separate embodiments can also be provided in combination in a single embodiment. Conversely, various features of the disclosed subject matter that are described in the context of a single embodiment can also be provided individually or in any suitable subcombination. In the detailed description that follows, numerous specific details are set forth in order to provide a thorough understanding of the methods and apparatus.

With this in mind, attention is drawn to FIG. 1 , which illustrates a functional block diagram of a mask inspection system in accordance with certain embodiments of the presently disclosed subject matter.

The inspection system 100 illustrated in FIG. 1 may be used for inspection of a mask using a mask, during or after a mask manufacturing process, and/or during a semiconductor specimen manufacturing process. As described above, inspection referred to herein relates to various types of defect inspection/detection, defect classification, and/or metrology operations, e.g., critical dimension (CD) measurements, on a mask or parts thereof. It can be interpreted to encompass any kind of operation. According to certain embodiments of the presently disclosed subject matter, the illustrated inspection system 100 includes a computer-based system 101 capable of automatically inspecting and detecting defects on a mask. As explained above, a defect to be detected herein may refer to any kind of anomaly or undesirable feature/functionality formed on a mask. As an example, in some cases, the defects to be detected may be related to edge positioning displacement (EPD), which refers to the difference between the actual and intended positions of the edges of printed features on the mask. In some other cases, defects to be detected relate to CD measurements and/or CD uniformity (ie, dispersions of a CD measurement across a mask or portion of a mask), or any other types of defects formed on a mask. It can be. System 101 is also referred to as a mask defect detection system, which is a subsystem of inspection system 100 .

System 101 may be operatively connected to a mask inspection tool 120 configured to scan a mask and capture one or more images of a mask for inspection of the mask. As used herein, the term "mask inspection tool" refers to, by way of non-limiting example, scanning, imaging, sampling, detection, measurement, classification (of a single or multiple scans) provided in connection with a mask or portions thereof. and/or other processes, including any type of inspection tools that may be used in mask inspection related processes.

It should also be noted that the mask inspection tool 120 may be implemented as various types of inspection machines, such as optical inspection tools, electron beam tools, etc., without limiting the scope of the present disclosure in any way. In some cases, mask inspection tool 120 may be a relatively low resolution inspection tool (eg, optical inspection tool, low resolution scanning electron microscope (SEM), etc.). In some cases, mask inspection tool 120 may be a relatively high-resolution inspection tool (eg, high-resolution SEM, atomic force microscope (AFM), transmission electron microscope (TEM), etc.). In some cases, an inspection tool may provide both low resolution image data and high resolution image data. In some embodiments, mask inspection tool 120 may have metrology capabilities and be configured to perform metrology operations on captured images. The resulting image data (low-resolution image data and/or high-resolution image data) may be transmitted to system 101 - either directly or via one or more intermediate systems. This disclosure is not limited to any particular type of mask inspection tools and/or resolution of image data resulting from inspection tools.

According to certain embodiments, a mask inspection tool is a lithography tool usable for fabrication of a semiconductor specimen, for example by projecting a pattern formed in a mask onto a wafer, as described in more detail below with respect to FIG. 5 . (eg, a scanner or stepper, for example) configured to emulate/emulate optical configurations.

Referring now to FIG. 5 , a schematic diagram of an actinic inspection tool and lithography tool is shown, in accordance with certain embodiments of the presently disclosed subject matter.

Similar to the lithography tool 520, the actinic inspection tool 500 includes an illumination source 502, illumination optics 504, and a mask holder 506 configured to generate light (eg, a laser) at an exposure wavelength. , and projection optics 508. Illumination optics 504 and projection optics 508 may include one or more optical elements (eg, lenses, apertures, spatial filters, etc.).

In a lithography tool 520, a mask is placed in a mask holder 506, and an image of a circuit pattern to be replicated is placed on a wafer disposed on a wafer holder 512 (e.g., to create or replicate a pattern on a wafer). optically aligned to project (by employing various stepping, scanning and/or imaging techniques to do so). Unlike the lithography tool 520, instead of placing the wafer holder 512, the actinic inspection tool 500 places a detector 510 (eg, a charge coupled device (CCD)) at the location of the wafer holder; Here detector 510 is configured to detect light that is projected through the mask and creates an image of the mask.

As can be seen, the actinic inspection tool 500 is used in an actual lithography process to expose photoresist during semiconductor device fabrication, eg, the illumination/exposure conditions, eg, wavelength, portion of the exposure light It is configured to emulate the optical configurations of the lithography tool 520, including but not limited to coherence, pupil shape, illumination aperture, numerical aperture (NA), and the like. Therefore, mask image 514 acquired by detector 510 is expected to be similar to image 516 of a wafer to be fabricated using a mask via a lithography tool. Mask images acquired using such actinic inspection tools are also referred to as aerial images. The aerial image is provided to system 101 for further processing, as described below.

According to certain embodiments, in some cases mask inspection tool 120 is implemented as a non-actinic inspection tool, such as, for example, a regular optical inspection tool, an electron beam tool (eg, SEM), or the like. It can be. In such cases, the inspection tool's detector may interface with the specific type of microscope used and digitize image information from the microscope, thereby obtaining an image of the mask.

To simulate the optical configuration of the lithography tool, simulation may be performed on the acquired image, thereby creating an aerial image. In some cases, image simulation may be performed by system 101 (eg, functionality of the simulation may be incorporated into PMC 102 by including an image simulation model into PMC 102 of the system). On the other hand, in some other cases, image simulation may be performed by a processing module of mask inspection tool 120, or by a separate simulation engine/unit operatively connected to mask inspection tool 120 and system 101. can

For illustrative purposes only, certain embodiments of the following description are provided for images acquired by an actinic mask inspection tool. Those skilled in the relevant art will understand that the teachings of the presently disclosed subject matter are likewise applicable to images acquired by any other suitable technique and inspection tool, and further converted to aerial images using an appropriate simulation model. you will easily understand The term “aerial image” should be interpreted broadly to encompass aerial images simulated from images acquired by actinic mask inspection tools and images captured by non-actinic inspection tool(s).

System 101 includes a processor and memory circuit (PMC) 102 operably coupled to a hardware-based I/O interface 126 . PMC 102 is configured to provide processing necessary to operate the system as described in more detail with reference to FIGS. 2, 3 and 4, and includes a processor (not individually shown) and memory (not individually shown). includes A processor of PMC 102 may be configured to execute several functional modules according to computer readable instructions embodied on a non-transitory computer readable memory included in the PMC. Such functional modules are hereinafter referred to as being included in the PMC.

A processor as referred to herein may refer to one or more general purpose processing devices, such as microprocessors, central processing units, and the like. More specifically, a processor implements a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction set (VLIW) microprocessor, a processor that implements other instruction sets, or a combination of instruction sets. It may be processors that do. A processor may also be one or more special purpose processing devices, such as application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), digital signal processors (DSPs), network processors, and the like. A processor is configured to execute instructions for performing the actions and steps discussed herein.

Memory as referred to herein includes main memory (e.g. read only memory (ROM), flash memory, dynamic random access memory (DRAM) such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.), and static memory. (eg, flash memory, static random access memory (SRAM), etc.).

As mentioned above, in some embodiments, system 101 may be configured to detect defects on a mask. Mask defects, if not detected prior to mass production of wafers, will be repeated many times on production wafers, thus leaving many semiconductor devices defective (e.g., affecting functionality of the devices and not meeting desired performance and ), adversely affecting the yield.

Because of the massive circuit integration in advanced processes of photomasks and the decreasing size of semiconductor devices, mask inspection is becoming increasingly sensitive to different types of fluctuations and noises. To detect smaller defects, it is desirable to inspect the mask with sensitive scans (i.e., scans with relatively high sensitivity), where most of the suspected defects reflected in the defect map are more likely to be false alarms or noises. . In such cases, DOIs can be buried in false alarms and noise, thus affecting detection sensitivity and causing an increase in the false alarm rate (FAR).

In mask inspection, die-to-die (D2D) inspections are often used for defect detection using an inspection image from one die and one or more reference images from one or more reference dies. The inspection image and reference images are usually compared to create a difference image representing potential defects on the mask. Since conventional D2D inspection utilizes one or more adjacent dies as references, the number of reference images is significantly limited. Additionally, as described above, inspection images and reference images are acquired using sensitive scans, and therefore suffer from the aforementioned increase in false alarms and noise and the difficulty of separating DOIs from false alarms and noise.

According to certain embodiments of the presently-disclosed subject matter, obtaining sufficient reference images to be further optimized and used to create a combined optimal reference image, thereby obtaining a more reliable difference with a higher signal-to-noise ratio (SNR). An improved mask inspection system and method configured to detect DOIs by rendering an image and thus improving detection sensitivity while reducing FAR is proposed.

To generate sufficient fiducials, multiple fiducial locations (e.g., from multiple fiducial dies, or from a single die) will be used, and overlapping scans will be used to capture multiple images for each fiducial location. can Overlapping images of multiple fiducials are used together to reduce noise and increase detection reliability. Multiple reference images are each further calibrated using image filtering techniques to create an optimal reference image, thereby enabling better separation between DOIs and false alarms or random noises. The proposed process has been demonstrated to have improved sensitivity for defect detection of advanced process control of mask features without affecting inspection throughput.

According to certain embodiments, functional modules included in PMC 102 of system 101 may include image processing module 104 and defect detection module 106 . The PMC 102 acquires (through the I/O interface 126) a plurality of inspection images (via the I/O interface 126) for an inspection area of the mask, each having a plurality of fields of view (FOVs) overlapping at least by the inspection area, and each For the inspection image of , obtain a set of reference images including a plurality of reference images of each corresponding reference area of the one or more corresponding reference areas. As an example, inspection images and reference images may be acquired by mask inspection tool 120, such as, for example, an actinic inspection tool.

The defect detection module 106 may be configured to generate a plurality of defect maps corresponding to a plurality of inspection images. Each defect map includes one or more defect candidates located in an inspection area of each inspection image. One or more defect candidates of respective inspection images may be sorted, creating a list of defect candidates of interest (DCI).

For each given DCI in the list, image processing module 104 may be configured to generate a plurality of difference patches. Specifically, the image processing module 104 performs at least the following steps: extracting an image patch surrounding a location of a given DCI from a set of test images and reference images, respectively, and generating a set of test patches and reference patches. ; For each reference patch, compute a filter optimized to minimize the difference between the corrected reference patch obtained using the filter and the inspection patch, and a set of corrected reference patches and a set of filters corresponding to the set of reference patches. generating; and combining the set of calibrated reference patches to obtain a composite reference patch, and comparing the test patch with the composite reference patch to obtain a difference patch, corresponding to each inspection image of the plurality of inspection images. It can be configured to create a difference patch that

The defect detection module 106 may be further configured to calculate a rating based on the plurality of difference patches and apply a detection threshold to the rating to determine whether a given DCI is a defect of interest (DOI).

Operations of systems 100, 101, PMC 102 and functional modules thereof will be described in more detail with reference to FIGS. 2, 3 and 4.

According to certain embodiments, system 100 may include storage unit 122 . Storage unit 122 includes any data required to operate systems 100 and 101, such as data related to inputs and outputs of systems 100 and 101, as well as data generated by system 101. may be configured to store intermediate processing results. As an example, storage unit 122 may be configured to store inspection images and reference images and/or derivatives thereof (eg, images after preprocessing) generated by mask inspection tool 120 . Accordingly, images may be retrieved from storage unit 122 and provided to PMC 102 for further processing.

In some embodiments, system 100 may optionally include a computer-based graphical user interface (GUI) 124 configured to enable user specific inputs related to system 101 . For example, a user may be presented with (eg, by a display forming part of GUI 124 ) a visual representation of the mask, including images of the mask or parts thereof. Via the GUI, the user may be provided with options to define certain operating parameters, such as, for example, sensitive scan parameters, number of reference regions/locations, list of defect candidates (DCI) of interest, detection threshold, etc. there is. In some cases, the user may also view operational results on the GUI, eg, composite reference patch, difference patch(s), detected DOI(s), defect map(s), and/or additional inspection results. there is.

As described above, system 101 is configured to receive a plurality of images of a mask (eg, inspection images and/or reference images) via I/O interface 126 . The images may include image data generated by mask inspection tool 120 (and/or derivatives thereof) and/or image data stored in storage unit 122 or one or more data archives. In some cases, image data may refer to images captured by a mask inspection tool, and/or preprocessed images derived from captured images, as obtained by various preprocessing stages, etc. . Note that in some cases, images may include associated numerical data (eg, metadata, manual attributes, etc.). It is further noted that in some embodiments the image data relates to a target layer of a semiconductor device to be printed on a wafer.

The system 101 processes the received images and, via an I/O interface 126, transfers inspection results (eg, detected DOIs, defect maps, complex reference patches, etc.) to a storage unit 122, and/or GUI 124 for rendering, and/or mask inspection tool 120.

In some embodiments, in addition to system 101, mask inspection system 100 includes one or more inspection modules, such as, for example, additional defect detection module(s) usable to perform additional inspection of a mask and / or an automatic defect review module (ADR) and / or an automatic defect classification module (ADC) and / or metrology related modules and / or other inspection modules may be further included. One or more inspection modules may be implemented as standalone computers, or their functionalities (or at least a portion thereof) may be integrated with mask inspection tool 120 . In some embodiments, output as obtained from system 101 may be used by mask inspection tool 120 and/or one or more inspection modules (or portions thereof) for further inspection of a mask.

Those skilled in the art will appreciate that the teachings of the presently disclosed subject matter are not limited by the system illustrated in FIG. 1; It will be readily appreciated that equivalent and/or modified functionality may be integrated or divided in different ways and may be implemented in any suitable combination of hardware and/or firmware and software.

The mask inspection system illustrated in FIG. 1 may be implemented in a distributed computing environment in which the aforementioned functional modules, such as included in the PMC 102, may be distributed to several local and/or remote devices and linked through a communication network. Note that you can In other embodiments, one or more of mask inspection tool 120, storage unit 122, and/or GUI 124 may be external to system 100 and may be external to system 101 via I/O interface 126. ) and can operate in data communication. System 101 may be implemented as standalone computer(s) to be used in conjunction with a mask inspection tool. Alternatively, functions of each of system 101 may be integrated, at least in part, with mask inspection tool 120, thereby facilitating and enhancing the functionalities of mask inspection tool 120 in inspection-related processes. let it

Although not necessarily, the process of operation of systems 101 and 100 may correspond to some or all of the stages of the methods described with respect to FIGS. 2-4. Likewise, the methods described with respect to FIGS. 2-4 and possible implementations thereof may be implemented by systems 101 and 100 . Therefore, it should be noted that the embodiments discussed in connection with the methods described in connection with FIGS. 2-4 may also be implemented as mutatis mutandis various embodiments of systems 101 and 100, and vice versa. pay attention to

Referring now to FIG. 2 , a generalized flow diagram of mask inspection for a mask usable for fabricating a semiconductor specimen is illustrated, in accordance with certain embodiments of the presently disclosed subject matter.

For an inspection area of the mask, a plurality of inspection images may be acquired (e.g., by PMC 102 via I/O interface 126, from mask inspection tool 120, or from storage unit 122). can (202). A plurality of inspection images have a plurality of fields of view (FOVs) overlapped by at least an inspection area. For each inspection image, a set of reference images including a plurality of reference images of each corresponding reference region of the one or more corresponding reference regions may be obtained. The inspection images and reference images are aerial images as described above.

In some embodiments, the plurality of inspection images (and/or reference images) are combined with an actinic inspection tool, such as, for example, Aira of Applied Materials Inc. (Aera) can be acquired sequentially by the mask inspection tool. The actinic inspection tool is configured to emulate the optical configuration of a lithography tool (eg, a scanner or stepper) usable in the manufacture of semiconductor wafers according to a mask, as described above with reference to FIG. 5 . Images may be acquired with a predefined step size such that the FOVs of the images overlap at least by the inspection area.

Images acquired by such an actinic inspection tool (ie, aerial images) are expected to be similar to images of a wafer fabricated using a mask through a lithography tool. In other words, the actinic mask inspection tool is configured to capture mask images that can mimic how design patterns in the mask will actually appear on a physical wafer after the fabrication process.

In some cases, an actinic inspection tool may not be available to inspect the mask. In such cases, a non-actinic inspection tool, such as, for example, a regular optical inspection tool, an electron beam tool, or the like, may be used to acquire non-aerial images of the mask. To simulate the optical configurations of the lithography tool, simulation may be performed on the acquired non-aerial images, thereby creating aerial images of the mask. Accordingly, in some embodiments, a mask inspection method as described with reference to FIG. 2 includes acquiring a plurality of images acquired by a non-actinic inspection tool, to simulate an optical configuration of a lithography tool. , simulate the images (e.g., by the image processing module 104 of the PMC 102, or by the processing module of the mask inspection tool 120, etc.), and generate a plurality of inspection images ( That is, preliminary steps of generating aerial images) may be further included.

In some embodiments, during inspection, the mask may move a step size relative to the detector of the mask inspection tool during exposure (or the mask and tool may move in opposite directions), and the mask may be moved at a time ( It can be scanned step by step along the swaths of the mask by a mask inspection tool that only images a part/part (also referred to as the tool or field of view (FOV) of the image) within the swath. For example, at each step, light may be detected from a rectangular portion of the mask, and such detected light is converted to multiple intensity values at multiple points of the portion, thereby making the portion/portion of the mask form an image corresponding to The FOV or the size and dimensions of the image corresponding to the FOV may vary depending on certain factors such as different tool configurations. In one example, each image corresponding to the rectangular FOV of the mask may be about 1000 pixels long and 1000 pixels wide. In another example, an image corresponding to a rectangular FOV may be approximately 800 pixels by 1600 pixels in size.

Therefore, multiple images of the mask may be acquired sequentially during sequential scanning along the swaths of the mask, each representing a respective part/portion of the mask. For example, a first swath of the mask can be scanned from left to right, a second swath from right to left, and so on, until the entire mask (or region of interest in the mask) has been scanned.

In some cases, multiple images may be acquired by a mask inspection tool with a predefined step size such that the FOVs of the multiple images may partially overlap according to the step size. As an example, in cases where the step size is predefined as 1/3 of the length of the FOV, three sequentially captured images overlap by 1/3 of the FOV. In other words, a given inspection area of overlapped area can be captured three times in three sequential images. Therefore, using overlapping imaging acquisition, each inspection area on the mask can be captured multiple times in multiple overlapping images.

A set of reference images including a plurality of reference images of each corresponding reference region of the one or more corresponding reference regions may be obtained for each inspection image. For each inspection area, one or more reference areas may be identified and used as criteria for comparison. As an example, in cases where the mask to be inspected is a multi-die mask (its mask field contains multiple dies with the same/similar design patterns) and the inspection area is located at an inspection die on the mask, one of the inspection dies on the mask One or more reference areas (corresponding to the location of the inspection area) from one or more reference dies (eg, neighboring dies of the inspection die) may be used as references (eg, in D2D inspection).

As another example, in cases where the mask is a single die mask (its mask field contains only one die), the inspection region and one or more reference regions are located on the same die of the mask, and the one or more reference regions are located on the same die as the inspection region. They share the same/similar design patterns. For example, one or more reference regions may be identified based on the design data of the mask, using any suitable algorithms available for identifying similar patterns, as described in more detail below.

In some embodiments, the plurality of inspection images (and/or reference images) as acquired may be pre-processed prior to further processing, as will be described with reference to FIG. 2 . Preprocessing may include one or more of the following operations: interpolation (eg, when images have relatively low resolution), noise filtering, focus corrections, aberration compensation, image format conversion, and the like.

It should be noted that the present disclosure is not limited to the particular aspect of the mask inspection tool, and/or the type of images acquired thereby, and/or the pre-processing operations required to process the images.

A plurality of defect maps corresponding to the plurality of inspection images may be generated (eg, by defect detection module 106 of PMC 102) (204). Each defect map may be created using at least one reference image (eg, one reference image from a set of reference images) and may represent a distribution of defect candidates over each inspection image. For example, at least one difference image may be generated based on differences between pixel values of the inspection image and pixel values of the at least one reference image. A defect map may be created by determining locations of suspected defects (ie, defect candidates) based on at least one difference image using a detection threshold. In some embodiments, the defect map may further indicate one or more defect characteristics of the defect candidates, such as, for example, locations, intensity and size of the defect candidates, and the like. Defect candidates, as revealed by the defect map, can be located in the corresponding inspection image based on their locations.

As described above, in some embodiments, an inspection image and one or more reference images may be acquired using sensitive scans. As an example, sensitive scans may be made possible by configuring the inspection tool to specific parameters with higher sensitivity. The configuration parameters may include one or more of the following: lighting conditions, polarization, and noise level per area, and the like.

In some embodiments, in addition to or instead of sensitive scans, a detection threshold may be configured according to sensitive detection requirements. For example, a relatively low threshold may be used to reveal more suspected defects in the defect map, thus resulting in defect detection with higher sensitivity. Defect candidates in such a defect map resulting from sensitive scan and/or sensitive detection are most likely noises and/or false alarms (because DOIs are rare). Such a defect map is therefore also referred to as a noise map.

In particular, each defect map may include one or more defect candidates located in an inspection area of each inspection image. One or more defect candidates of respective inspection images may be sorted, creating a list of defect candidates of interest (DCI). As an example, alignment between the defect candidates of each inspection image may be performed based on its defect characteristics (eg, defect locations). For example, a defect map can be converted to mask coordinates, whereby each defect candidate can be described by a list of coordinates in a mask coordinate system. Multiple defect maps of overlapping inspection images may report identical defects located in overlapping inspection areas between FOVs of the inspection images. Once the defect candidates are reported using the mask coordinates, integration between the defect candidates of the inspection images can be done. Integration may be performed by matching the defect candidate coordinates in the mask coordinate system with a matching criterion (eg, by applying a dilation to the location of the defect candidates). In some embodiments, in addition to matching coordinates, integration may further consider the number of occurrences of a defect candidate in the defect maps. For example, as illustrated further below, a configuration parameter may be defined for further filtering of matched defect candidates, in a certain number of overlapping defect maps (eg, in all of the plurality of defect maps). Request detection of defect candidates. In some embodiments, optionally, once defect candidates are integrated, additional filtering may be performed based on the strength of the candidates being ranked within each defect map, eg, integration with a relatively higher rating. selected candidates will be selected as a list of DCIs.

Referring now to FIG. 6 , an example of a multi-die mask as well as an inspection area, a plurality of inspection images and reference images thereof is schematically illustrated, in accordance with certain embodiments of the presently disclosed subject matter.

As shown, the multi-die mask 600 has a mask field containing nine dies that share the same design pattern. For a given inspection area 602 of the inspection die of the mask 600, three overlapping inspection images 603, 604 and 605 are acquired sequentially. The three images are captured with a particular step size such that their FOVs overlap by, for example, 1/3 of the image's FOV. Thus, the inspection area 602 is acquired three times in three images (e.g., the inspection area 602 is located on the right side of the first image, in the middle of the second image, and on the left side of the third image). do).

For an inspection area 602 of an inspection die of mask 600, two reference areas 606 and 608 from two reference dies (e.g., two neighboring dies) of an inspection die on the mask (corresponding to the position of the inspection area 602) can be used as criteria for comparison. For any one reference area, three reference images can similarly be acquired, and the FOVs of those images overlap by 1/3 of the image's FOV. Therefore, a set of six reference images is obtained for each of the inspection images 603, 604 and 605.

During defect detection, a defect map may be created for each inspection image using at least one of the six reference images. Accordingly, three defect maps corresponding to the three inspection images 603, 604 and 605 are generated. Each defect map includes one or more defect candidates as revealed within the inspection area of the respective inspection image. The defect candidates of the respective inspection images may be sorted/matched, thereby creating a list of defect candidates of interest (DCI).

As an example, assuming that the same number of defect candidates (e.g., 3 defect candidates) are revealed at corresponding locations of the inspection area of 3 inspection images and they are properly aligned between the images, the list of DCIs is 3 It may contain defect candidates. As another example, in some cases, different numbers of defect candidates may appear in different inspection images due to various reasons, such as noises and/or variations, and the like. For example, say that three defect candidates are revealed in the inspection area of the first inspection image, three defect candidates are revealed in the inspection area of the second inspection image, and only two defect candidates are revealed in the inspection area of the third inspection image. Assume. Per alignment, two defect candidates are merged as being common defect candidates among the three inspection images, while the remaining defect candidates appear in the two images but are missing anyway from the third inspection image. In some cases, a configuration parameter as noted above may be defined as requiring generation of a defect candidate in a majority (eg, 2 out of 3) of the inspection images. In such cases, the list of DCIs may include three defect candidates, such that the image information from the corresponding location of the missing defect candidate in the third image also determines whether this defect candidate is a DOI or a false alarm. It can be obtained and processed in further processing to assist in determining. In some other cases, alternatively, a configuration parameter as mentioned above may be defined as requiring generation of a defect candidate in all inspection images. In such cases, the list of DCIs may contain only two common defect candidates appearing in all images. In still other cases, the list of DCIs may be determined based on additional or alternative factors such as, for example, a predetermined number of DCIs to be selected, ratings of defect candidates for particular defect characteristics, and the like.

Continuing the description of FIG. 2 , for at least one given DCI in the list of DCIs as determined above, a plurality of difference patches may be generated (e.g., by image processing module 104 of PMC 102). can (206). Specifically, a difference patch corresponding to each of the plurality of inspection images may be created in a process as described below with reference to blocks 208-214.

Specifically, for each inspection image, an image patch surrounding the location of a given DCI may be extracted from the inspection image and the set of reference images, respectively, creating a test patch and a set of reference patches (208). In the example of FIG. 6 , a DCI that is common in the three inspection images is illustrated (eg, M selected DCIs, only one of which is illustrated in FIG. 6 (as a black dot), but assumed to exist). For example, for the DCI of the inspection image 603, a peripheral image patch having a square shape surrounding the DCI may be extracted from the inspection image 603 and each of the six reference images, respectively, and the inspection patch ( For example, a left test patch among test patches 610) and a set of reference patches 612 (six reference patches in this example) are created. Thus, for each of the DCIs in the list of DCIs, a corresponding test patch and set of reference patches may be created.

Optionally, in some embodiments, each test patch may be matched 210 with a respective reference patch from the set of reference patches. The two image patches (inspection patch and respective reference patch) are registered to perform an accurate comparison. Registration may include measuring an offset between the two image patches, and shifting one image patch relative to the other image patch to correct for the offset. The offset results from various factors such as, for example, tool drifts (eg, scanner and/or stage drift), since the two patches are extracted from different images acquired for different dies. may be caused by navigation errors, etc.

Matching may be implemented according to any suitable matching algorithms known in the art. As an example, matching may be performed using one or more of the following algorithms: area-based algorithm, feature-based matching, or phase correlation matching. An example of an area-based method is registration using optical flow, such as the Lucas Kanade algorithm (LK). Feature-based methods are based on finding discrete points of information ("features") in two images and calculating the necessary transformation between each pair based on the correspondences of the features. This allows elastic mating (ie, non-rigid mating) where the different regions are moved individually. Phase correlation matching is done using frequency domain analysis (a phase difference in the Fourier domain is converted to a match in the image domain).

Alternatively, in some embodiments, matching may be omitted. For example, in cases where it can be assumed that there may be no substantial offset between the test patch and the respective reference patches, the registration can be omitted.

For each reference patch, a filter may be computed 212 for the purpose of being applied to the reference patch to produce a calibrated reference patch with better quality (eg, less noise with higher SNR). In some embodiments, a filter is a difference between an inspection patch and a corresponding corrected reference patch (obtained by applying the filter to the corresponding reference patch) to improve the SNR of the defect signal in the resulting difference patch. It can be obtained using an optimization method to minimize . The filter thus obtained is also referred to herein as the optimal filter for each reference patch.

In some embodiments, the optimal filter thus created can correct for certain variations and transformations between the test patch and the reference patch, which can be attributed to various noise types/sources, such as, for example: matching Residual error (e.g., rigid-registered residuals, or elastic-registered residuals, etc.), intensity gain and offset, defocus, and field of view (FOV) distortion (also referred to as image sensor uniformity or CCD uniformity noise), etc. can be caused by one or more of

와 같은 함수는 검사 패치 및 기준 패치의 픽셀들(x, y)의 그레이 레벨들 간의 강도 이득 및 오프셋 변동을 표현할 수 있고, 여기서 계수(a^k)는 이득 인자를 표현하고, b^k는 오프셋을 표현한다. 일부 다른 경우들에서, 잡음들 중 일부는 비선형 모델들, 예컨대, 예를 들어, FOV 왜곡, 탄성 정합 잔여들 등으로 표현될 수 있다.Some of such noise can be represented by a linear model, such as, for example, rigid-body matched residuals, intensity gain and offset, defocus, and the like. As an example, for example

A function such as can represent the intensity gain and offset variation between the gray levels of pixels (x, y) of the test patch and reference patch, where the coefficient a ^k represents the gain factor and b ^k represents the offset express In some other cases, some of the noise may be represented by non-linear models such as, for example, FOV distortion, elastic match residuals, and the like.

In particular, FOV distortion, as mentioned above, refers to image intensity fluctuations and non-uniformity at different locations within the FOV of an image. This is due to certain optical system aberrations, including but not limited to, for example, astigmatism, uneven illumination in the field, distortions due to lens shape, specks, etc., as described in more detail below. can be caused

As an example, FOV distortion can result from uneven lighting (eg, lighting non-uniformity), which becomes particularly significant in low light conditions and can increase noises in certain areas of the image. Astigmatism and field curvature are known aberrations of the optical system, which can cause a cylindrical effect, where the contrast and focus in the x and y directions of images are not uniform and vary along the FOV. These aberrations can dramatically affect patch similarity and cause differences in pattern appearance at different locations of the FOV.

Additionally, distortion due to lens shape can lead to different magnification along the FOV (which can lead to variations in pixel sizes). Speckles refer to high-frequency noise that varies from frame to frame (eg, noise may not appear in the same location in different images).

As explained, while the aforementioned aberrations can vary between different FOV locations, in some cases they can also be pattern dependent. It should be noted that although FOV distortion is typically represented by a non-linear filter, when processing it on relatively small image patches, its behavior can be assumed to be linear, and thus can be corrected using a linear filter in some cases. do.

To deal with various noises as described above, different filtering methods may be used (alone or in any suitable combination), such as, for example, linear regression, nonlinear regression, matched filters, and the like. In particular, in some embodiments, a specific calibration is needed for every individual reference patch, since different patches may experience different noises. For example, different FOV distortions may occur when using reference patches from different locations of the FOV. Therefore, a unique filter is required to individually calibrate each reference patch to address one or more of the noises and aberrations mentioned above.

와 같이 공식화될 수 있고 여기서, h는 필터를 나타내고, ref는 기준 패치의 픽셀 값들을 나타내고, ins는 검사 패치의 픽셀 값들을 나타낸다. 예를 들어, 필터는 고정된 크기(예를 들어, 5*5, 또는 7*7)를 가질 수 있고, 이는, 기준 패치에 적용될 때, 기준 패치의 대응하는 부분과 순차적으로 콘볼빙할 수 있고, 이에 의해, 보정된 기준 패치를 도출한다. 그러한 컴퓨팅된 필터는 다음: 정합 잔여 오류, 강도 이득 및 오프셋, 디포커스, 및 FOV 왜곡 등 중 하나 이상에 의해 야기되는 위에서 설명된 변동들을 보정할 수 있다.In some embodiments, a filter may be obtained using an optimization method to minimize the difference between a test patch and a corresponding corrected reference patch (when applying the filter to the corresponding reference patch). As an example, the objective loss function of the optimization process is:

where h denotes the filter, ref denotes the pixel values of the reference patch, and ins denotes the pixel values of the test patch. For example, a filter can have a fixed size (e.g., 5*5, or 7*7), which, when applied to a reference patch, can sequentially convolve with corresponding portions of the reference patch; , whereby a corrected reference patch is derived. Such a computed filter can correct for the variations described above caused by one or more of the following: match residual error, intensity gain and offset, defocus, FOV distortion, and the like.

In some embodiments, the filter may be computed using least squares (LS) optimization, in cases where the necessary corrections can be considered a linear transformation. An exemplary implementation of such optimization is described below.

이도록 검색된다. f는 기준 패치와 이미지 패치 사이의 차이를 최소화하는 필터이다. 최적화 알고리즘은 I 및 R * f(f에 의해 필터링된 R)가 LS 측면에서 가장 유사하도록 필터(f)를 찾는다. 이 식은 LS 최적화가 필터 변수들에 대해 계산되도록, 선형 연립방정식들으로서 조직될 수 있다. 알고리즘의 출력은 추정된 최적 필터(f) 및 필터링된 이미지(R_보정(corrected))를 포함한다. 이미지(R)에 (예를 들어, 3x3 크기의) 알려지지 않은 필터(f)를 적용하고 이를 이미지(I)와 비교함으로써, 근사화된 선형 식들의 세트가 생성될 수 있다. 이 선형 식들의 세트는 행렬 표기법(x = Ab)으로 변환되고, 한편:For a given image (I) and reference image (R), the filter (f) is

is searched for f is a filter that minimizes the difference between the reference patch and the image patch. The optimization algorithm finds filter f such that I and R * f(R filtered by f) are most similar in terms of LS. This equation can be organized as a system of linear equations such that the LS optimization is computed over the filter variables. The output of the algorithm includes the estimated optimal filter (f) and the filtered image ( R_corrected ). By applying an unknown filter f (e.g. of size 3x3) to image R and comparing it to image I, a set of approximated linear equations can be generated. This set of linear equations is converted to matrix notation ( x = A b ), while:

이다.

am.

로 변환된다. 그리고, LS 해는:

에 의해 주어진다(여기서, b는 알려진 관측들의 벡터이고, A는 알려진 행렬이고, x는 알려지지 않은 파라미터들의 벡터이다).Discrete 2D convolution can be expressed as matrix multiplication, and the problem is the LS squared problem of:

is converted to And, the LS solution is:

(where b is a vector of known observations, A is a known matrix, and x is a vector of unknown parameters).

In some other embodiments, in cases where the necessary corrections can be considered a non-linear transformation, the filter can be computed using, for example, an iterative optimization method. As an example, the optimization method may be implemented as a first-order (gradient-based) line search optimization method. A first coarse fit using the initial values of the filter can be computed and used as an initial starting point for optimization. At each iteration, the gradient at the starting point can be computed, a descending direction (eg, an anti-gradient direction) is identified, and along this descending direction the loss function will be sufficiently reduced. A step size is computed that determines how far to move along that direction. Upon moving along the downward direction with the determined step size, a new starting point is derived and used to start the next iteration. Iterations may be repeated until convergence is reached (i.e., until the error of the loss function reaches its minimum value). The filter with optimized values that allows minimization of the loss function is the optimal filter to be used for correcting the reference patch.

Optionally, in some cases, a filter may additionally include one or more image masks. As an example, a mask may be used to avoid invalid image information, eg, blemish pixels (eg, burned pixels) of image patches. As another example, a weight mask can be used for the purpose of applying different weights to different regions of an image patch. For example, a weight mask can be configured such that weights for important regions of an image patch can be increased while weights for less important regions can be decreased. Examples of areas that may require special specific weights are: suspected defective pixels, the area around the suspected defect (not including the defect itself), surrounding pixels, edges in the image (typically high slopes) have), including (but not limited to). Such masks may be used individually or in combination.

Optionally, in some cases, the size of the filter may be adjusted according to noise characteristics (eg, noise levels) of the reference images.

According to certain embodiments, as illustrated in FIG. 3 , for each reference patch, one or more noise to be corrected (e.g., noises of different types/causes), e.g., registration residual error (including rigid registration residuals and/or elastic registration residuals, etc.), intensity gain and offset, defocus, and/or field of view (FOV) distortion (302). For each reference patch, a filter comprising a set of filter components can be specifically computed (304) to correct the respective noises (using any suitable optimization method). Each filter element may be linear or non-linear as described above, and may optionally process one or more of the noise types described above, along with one or more image masks. The computed filter may be applied to the reference patch to obtain a calibrated reference patch (306).

Once each filter has been computed for each reference patch, a set of filters corresponding to the set of reference patches is obtained (eg, by applying the set of filters to the corresponding reference patches) and a corrected reference patch is obtained. A set of patches may be created. The set of calibrated reference patches may be combined to obtain a composite reference patch (214). A test patch can be compared to a composite reference patch to obtain a difference patch.

As an example, a difference patch may be created based on the difference between the pixel values of the test patch and the pixel values of the composite reference patch. In some cases, a difference patch may be further normalized (for at least some of the pixels of the patch) using one or more difference normalization factors. By way of example, difference normalization factors may be determined based on the behavior of a normal population of pixel values of the difference patch and/or the composite reference patch. For example, to reduce the shot noise effect, the difference patch can be normalized according to the gray levels of the corresponding pixel values of the composite reference patch (e.g., pixels with higher gray level values are typically are noisier than pixels with lower gray level values, and thus different normalization factors may be assigned to different pixels). In some cases, pixel values of a difference patch may be normalized as a ratio between a corresponding original pixel value of the difference patch and a corresponding difference normalization factor.

In the example of Figure 6, each filter is computed and applied to each of the six reference patches 612, generating six corresponding corrected reference patches. The six calibrated reference patches are combined to create a composite reference patch. By way of example, a combination can be used to calculate corresponding pixel values of corrected reference patches (or at least a portion thereof) using any of the following calculations or combinations thereof: average, weighted average, minimum, maximum, and arithmetic average, and the like. It can be done by combining/aggregating.

Once the above process of blocks 208-214 is performed for each inspection image and a difference patch corresponding to the inspection patch is obtained, a plurality of difference patches (for a given DCI) corresponding to the plurality of inspection images are obtained. is obtained In the example of FIG. 6 , for DCI as illustrated, the above process of blocks 208-214 is repeated for each of three inspection images 603, 604 and 605, and three inspection patches 610 ) are obtained.

A grade may be computed based on a plurality of difference patches, and a detection threshold is added to the grade (e.g., defect detection module 106 of PMC 102) to determine whether a given DCI is a defect of interest (DOI). by) can be applied (216).

As illustrated in FIG. 4 , in some embodiments, the ranking is calculated by calculating a score for each of the plurality of difference patches based on the highest pixel value in the difference patch, and by calculating a plurality of differences corresponding to the plurality of difference patches. It can be calculated (402) by generating scores and averaging (404) a plurality of scores to obtain a rating. As an example, the score for each differential patch may be the patch's highest pixel value, indicating the strength of the suspected defect signal. As another example, the score may be derived as an average value (eg, average, weighted average, or arithmetic average) of the pixel values of the patch.

8 illustrates an example of a plurality of difference patches in accordance with certain embodiments of the presently-disclosed subject matter. As shown, the three difference patches 802, 804 and 806 corresponding to the three test patches 610 of FIG. 6, for example, each test patch (based on the six reference patches) generated by comparison with each composite reference patch). The three difference patches are further normalized as described above. As illustrated, a score is computed for each difference patch based on the patch's highest pixel value, generating three scores. A rating can be derived by averaging the three scores.

In some embodiments, generation 206 of a plurality of difference patches, calculation of a class and application of a detection threshold 216 may be performed for each DCI in the list of DCIs, thereby determining whether the DCI is a DOI. decide An updated defect map corresponding to the inspection area, including one or more DOIs as detected by the decision, may be provided. In the example of FIG. 6, assuming that there are M DCIs in the set of DCIs (although only one of them is illustrated in FIG. 6) as selected in the process described with reference to block 204, the decision as described above This may be done for each of the M DCIs, and it may be determined that N of the M DCIs are DOIs. An updated defect map including N DOIs corresponding to the inspection area may be created.

The process as described with reference to FIG. 2 may be repeated for one or more additional inspection areas on the mask. In some embodiments, a region of interest (ROI) to be inspected on the mask may be predefined, and one or more inspection regions within the ROI may be inspected using the process described above. In some cases, an ROI may be defined as an entire mask, while in some other cases, an ROI may be defined as part of a mask.

Although the mask inspection process as described with reference to FIG. 2 is illustrated using the example of a multi-die mask as illustrated in FIG. 6 , it should be noted that this is in no way intended to limit the present disclosure in any way. It should be noted. It is understood that the proposed methods and systems can be similarly applied to a single die mask. 7 illustrates an example of a single die mask as well as an inspection area, a plurality of inspection images and reference images thereof, in accordance with certain embodiments of the presently disclosed subject matter.

As shown, single die mask 700 has a mask field composed of a single die. For a given inspection area 702 of a single die, three inspection images 703, 704 and 705 are acquired sequentially. Similarly as described above with reference to FIG. 6 , the three images are captured with a particular step size such that their FOVs overlap by, for example, 1/3 of the image's FOV. Thus, the inspection area 702 is acquired three times in three images (e.g., the inspection area 702 is located on the right side of the first image, in the middle of the second image, and on the left side of the third image). do).

For inspection area 702, three reference areas 706, 708 and 709 from the same die sharing the same design pattern as the inspection area can be used as criteria for comparison. For each reference region, three reference images can similarly be acquired, and the FOVs of those images overlap by 1/3 of the image's FOV. Therefore, a set of nine reference images is obtained for each of the inspection images 703, 704 and 705.

The reference regions of a single die mask can be identified in a variety of ways. The design data of a die (or portion(s) of a die) may include various design patterns with specific geometries and arrangements. A design pattern may be defined as being composed of one or more structural elements each having a contoured geometric shape (eg, one or more polygons).

In some embodiments, design data of a single die mask may be received, and a plurality of design groups, each corresponding to one or more die regions having the same design pattern, may be retrieved. Therefore, regions of the die that correspond to the same design pattern can be identified. It should be noted that design patterns can be considered "identical" when they are the same, or when they are highly correlated, or when they are similar to each other. Various similarity criteria and algorithms may be applied to match and cluster similar design patterns, and this disclosure should not be construed as limited by any particular criteria used to derive design families. Clustering of design groups (ie, division of CAD data into multiple design groups) may be performed by PMC 102 beforehand or as a preliminary step in the present inspection process.

Once reference areas are identified and reference images are acquired in a single die scenario as illustrated in FIG. 7, the generation of defect maps, difference patches, as well as the classification calculation and determination of DOIs are described above with reference to FIGS. 2 and 6. Similar to multi-die scenarios such as

According to certain embodiments, the output as resulting from the inspection process described with respect to FIG. 2 is: the DOIs as determined, an updated defect map for the inspection area, and/or a composite for each inspection patch. may include one or more of a reference patch, and the like. The output may include mask inspection tool 120 and/or further inspection of the mask, such as, for example, additional defect detection, defect review, defect classification, metrology related operations (eg, CD measurement), and/or any It may be used by one or more inspection modules as included in mask inspection system 100 for other inspection operations. As an example, a composite reference patch may be used for defect detection related to EPD.

It should be noted that the mask applicable to the presently disclosed inspection process may be any kind of mask, including but not limited to memory masks and/or logic masks, and/or Arf masks and/or EUV masks, etc. It should be noted. The present disclosure is not limited to a specific type or functionality of masks to be inspected.

According to certain embodiments, the mask inspection process as described above with reference to FIGS. 2, 3 and 4 is an inspection recipe usable by system 101 and/or inspection tool 120 for runtime online mask inspection. may be included as part of Therefore, the presently disclosed subject matter also includes a system and method for generating an inspection recipe during a recipe setup step, wherein the recipe steps as described with reference to FIGS. 2, 3 and 4 (and various embodiments thereof). include them Note that the term "inspection recipe" should be interpreted broadly to encompass any recipe that may be used by an inspection tool to perform operations related to mask inspection of any kind, including embodiments as described above. should pay attention to

For example, examples illustrated in this disclosure, such as, for example, mask inspection tool architectures and configurations, mask types and/or layouts, illustrated image patches, specific noise types/sources, and It should be noted that filters, as well as optimization methods and the like as described above, are illustrated for illustrative purposes and should not be considered limiting the disclosure in any way. Other suitable examples/implementations may be used in addition to or in place of the above.

Among the advantages of certain embodiments of the mask inspection process as described herein is the ability to detect defects of interest (DOIs) on the mask prior to mass production of wafers at the FAB, with improved detection sensitivity, thereby: Achieve higher DOI capture rates while suppressing FAR.

This, at least, increases the number of references by acquiring overlapped images, effectively removes various noises from each reference patch by computing specific filters customized for each reference patch, and generates multiple corrected reference patches. .

Additionally, the above process is repeated for a number of superimposed inspection images capturing the same inspection area, and multiple results (i.e., multiple difference patches) are all taken into account when making a decision on the existence of DOIs ( For example, a rank is generated based on the scores of a plurality of difference patches), which can further eliminate false alarms and effectively improve detection sensitivity without the need to adjust the detection threshold. Additional advantages exist in that working on small image patches rather than full inspection and reference images allows more accurate calibration of the references and thus higher quality reference creation.

Among the advantages of certain embodiments of the mask inspection process as described herein are application of a multi-criteria inspection process as described above to a single die mask, thereby enabling detection of DOIs on a single die mask with increased detection sensitivity. There is something to do.

It is to be understood that the present disclosure is not limited in this application to the details set forth in the description or illustrated in the drawings contained herein.

It will also be appreciated that a system according to the present disclosure may be implemented, at least in part, on a suitably programmed computer. Likewise, the present disclosure contemplates a computer program readable by a computer for carrying out the methods of the present disclosure. The present disclosure further contemplates a non-transitory computer readable memory tangibly embodying a program of instructions executable by a computer to carry out the methods of the present disclosure.

The disclosure is capable of other embodiments and of being practiced or of being carried out in various ways. Therefore, it should be understood that the phraseology and terminology employed herein is for the purpose of description and should not be regarded as limiting. As such, those skilled in the art can readily utilize the concept on which this disclosure is based as a basis for designing other structures, methods, and systems for carrying out some of the purposes of the disclosed subject matter. you will understand that

Those skilled in the relevant art will understand that various modifications and changes may be applied to the embodiments of the present disclosure as described above and as defined in and by the appended claims without departing from the scope of the disclosure. You will easily understand that you can.

Claims (15)

A computerized system for inspecting a mask usable for fabricating semiconductor specimens, the system comprising:
Processing and Memory Circuit (PMC)
Including, the PMC is:
Acquiring, for an inspection area of the mask, a plurality of inspection images having at least a plurality of fields of view (FOVs) overlapped by the inspection area, and for each inspection image, each of one or more corresponding reference areas obtain a set of reference images including a plurality of reference images of a corresponding reference area;
generate a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more defect candidates located in the inspection area of each inspection image; sort the defect candidates and create a list of defect candidates of interest (DCI);
For at least one given DCI in the list, generate a plurality of difference patches - the PMC:
extracting an image patch surrounding the location of the given DCI from the test image and the set of reference images, respectively, and creating a test patch and a set of reference patches;
For each reference patch, compute a filter optimized to minimize the difference between the test patch and the corrected reference patch obtained using the filter, and compute a set of filters and a corrected set corresponding to the set of reference patches. creating a set of reference patches; and
combining the set of calibrated reference patches to obtain a composite reference patch and comparing the test patch to the composite reference patch to obtain the difference patch.
configured to generate a difference patch corresponding to each inspection image among the plurality of inspection images by;
calculate a rank based on the plurality of difference patches, and apply a detection threshold to the rank to determine whether the given DCI is a defect of interest (DOI);
A structured, computerized system. According to claim 1,
wherein the mask is a multi-die mask, the inspection areas are located at inspection dies on the mask, and the one or more reference areas are each from one or more reference dies of the inspection dies on the mask. According to claim 1,
wherein the mask is a single die mask, and wherein the inspection area and the one or more reference areas are from the single die on the mask and share the same design pattern. According to claim 1,
wherein the PMC is configured to respectively register the test patch with each reference patch to correct for each offset between the test patch and each reference patch from the set prior to computing the filter. According to any one of claims 1 to 4,
wherein the filter comprises a set of filter components for correcting each of the following noises: registration residual error, intensity gain and offset, defocus, or field of view (FOV) distortion. , a computerized system. According to any one of claims 1 to 4,
Wherein the filter is computed using least squares optimization. According to any one of claims 1 to 4,
The rank is configured to calculate a score for each of the plurality of difference patches based on a highest pixel value in the difference patch, generate a plurality of scores corresponding to the plurality of difference patches, and obtain the rank. calculated by averaging the plurality of scores. According to any one of claims 1 to 4,
the PMC performs generating the plurality of difference patches, calculating the class and applying a detection threshold to each DCI in the list of DCIs to determine whether the DCI is a DOI; The computerized system is further configured to provide an updated defect map comprising one or more DOIs as detected by the determination and corresponding to the inspection area. According to any one of claims 1 to 4,
The plurality of inspection images are sequentially acquired by an actinic inspection tool having a predefined step size, and the actinic inspection tool emulates an optical configuration of a lithography tool usable for manufacturing the semiconductor specimen. A computerized system configured to: A computerized method for inspecting a mask usable for fabricating a semiconductor specimen, comprising:
The method is performed by a processing and memory circuit (PMC), the method comprising:
Acquiring, for an inspection area of the mask, a plurality of inspection images having at least a plurality of fields of view (FOVs) overlapped by the inspection area, and for each inspection image, each of one or more corresponding reference areas obtaining a set of reference images including a plurality of reference images of a corresponding reference area;
generate a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more defect candidates located in the inspection area of each inspection image; sorting the defect candidates and generating a list of defect candidates of interest (DCI);
For at least one given DCI in the list, generating a plurality of difference patches comprising:
extracting an image patch surrounding the location of the given DCI from the test image and the set of reference images, respectively, and generating a test patch and a set of reference patches;
For each reference patch, compute a filter optimized to minimize the difference between the test patch and the corrected reference patch obtained using the filter, and compute a set of filters and a corrected set corresponding to the set of reference patches. creating a set of reference patches; and
combining the set of calibrated reference patches to obtain a composite reference patch and comparing the test patch to the composite reference patch to obtain the difference patch.
generating a difference patch corresponding to each inspection image among the plurality of inspection images by;
calculating a rank based on the plurality of difference patches and applying a detection threshold to the rank to determine whether the given DCI is a defect of interest (DOI);
Including, computerized method. According to claim 10,
wherein the mask is a multi-die mask, the inspection areas are located at inspection dies on the mask, and the one or more reference areas are each from one or more reference dies of the inspection dies on the mask. According to claim 10 or 11,
wherein the filter comprises a set of filter elements for correcting noises of each of the following noises: match residual error, intensity gain and offset, defocus, or field of view (FOV) distortion. . According to claim 10 or 11,
generating the plurality of difference patches, calculating a rank and applying a detection threshold to each DCI in the list of DCIs to determine whether the DCI is a DOI; and the determining and providing an updated defect map corresponding to the inspection area and comprising one or more DOIs as detected by the computerized method. According to claim 10 or 11,
The plurality of inspection images are sequentially acquired by an actinic inspection tool having a predefined step size, the actinic inspection tool being configured to emulate an optical configuration of a lithography tool usable for manufacturing the semiconductor specimen. computerized method. A non-transitory computer-readable storage medium tangibly embodying a program of instructions,
The instructions, when executed by a computer, cause the computer to perform the method of claim 10 or 11.

Images