Mask inspection for semiconductor sample manufacturing

Abstract

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

Description

Technical Field

[0001] The presently disclosed subject matter generally relates to the field of mask inspection and, more particularly, to defect detection with respect to photomasks. Background Art

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

[0003] Semiconductor devices are typically fabricated using a photolithography mask (also referred to as a photomask or mask or reticle) in a photolithography process. The photolithography process is one of the main processes for fabricating semiconductor devices and includes patterning the surface of a wafer according to the circuit design of the semiconductor device to be produced. This circuit design is first patterned on the mask. Thus, in order to obtain an operational semiconductor device, the mask must be defect-free. Masks are fabricated through complex processes and may have various defects and variations.

[0004] In addition, masks are typically used in a repetitive manner to fabricate many die on a wafer. Thus, any defect on the mask will be repeated many times on the wafer and can cause defects in multiple devices. Establishing a process with production value requires strict control of the entire photolithography process, especially considering large-scale circuit integration and the reduced size of semiconductor devices.

[0005] A variety of mask inspection methods have been developed and put into use. According to certain conventional techniques for designing and evaluating masks, masks are fabricated and used to expose wafers through the masks, and then inspections are performed to determine whether the features / patterns of the masks have been transferred to the wafers according to the design. Any differences between the ultimately printed features and the intended design may require modifying the design, repairing the mask, fabricating a new mask, and / or exposing a new wafer.

[0006] Alternatively, various mask inspection tools can be used to directly inspect the masks. The inspection process can include multiple inspection steps. During the fabrication process of the masks, the inspection steps can be performed multiple times, such as after the fabrication or processing of certain layers, etc. Additionally or alternatively, each inspection step can be repeated multiple times, such as for different mask positions or for the same mask position with different inspection settings.

[0007] Mask inspection generally involves generating certain inspection outputs (e.g., images, signals, etc.) for a mask by directing light or electrons onto the mask and detecting the light or electrons from the mask. Once the output has been generated, defect detection is typically performed by applying defect detection methods and / or algorithms to the output. The goal of the inspection is often to provide high-sensitivity detection of defects of interest (which, if uncorrected, may cause the final device to fail to meet the expected performance or cause malfunctions, thus having an adverse effect on the yield), while improving the efficiency of suppressing the detection of false alarms / nuisances and noise. SUMMARY OF THE INVENTION

[0008] According to certain aspects of the presently disclosed subject matter, there is provided a computerized system for inspecting a mask that can be used to fabricate a semiconductor sample, the system including a processing and memory circuitry (PMC) configured to: obtain, for an inspection region of the mask, a plurality of inspection images having a plurality of fields of view (FOVs) that at least overlap the inspection region, and for each inspection image, obtain a set of reference images, the set of reference images including a plurality of reference images for each corresponding reference region in one or more corresponding reference regions; generate a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more candidate defects located in the inspection region of the corresponding inspection image, and align the one or more candidate defects of the corresponding inspection image, thereby generating a list of defects of interest (DCIs); for at least one given DCI in the list, generate a plurality of differential patches, wherein the PMC is configured to generate a differential patch corresponding to each of the plurality of inspection images by: extracting an image patch around the location of the given DCI from the inspection image and from a set of reference images, respectively, thereby generating an inspection patch and a set of reference patches; for each reference patch, calculate a filter that is optimized to minimize the difference between the inspection patch and a corrected reference patch obtained using the filter, thereby generating a set of filters and a set of corrected reference patches corresponding to the set of reference patches; and combine the set of corrected reference patches to obtain a composite reference patch, and compare the inspection patch with the composite reference patch to obtain a differential patch; calculate a grade based on the plurality of differential patches, and apply a detection threshold to the grade to determine whether the given DCI is a defect of interest (DOI).

[0009] In addition to the above features, the system according to this aspect of the presently disclosed subject matter may include one or more of the following features (i) to (xii) in any desired combination or arrangement that is technically possible:

[0010] (i). The mask is a multi-die mask. The inspection area is located in an inspection die on the mask. One or more reference areas are respectively from one or more reference dies of the inspection die on the mask.

[0011] (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.

[0012] (iii). The PMC is configured to register an inspection patch with each reference patch from the group to correct a corresponding offset between the inspection patch and each reference patch before calculating the filter.

[0013] (iv). The filter is calculated to correct at least one of the following noises in the reference patch: registration residuals, intensity gain and offset, defocus, or field of view (FOV) distortion.

[0014] (v). The filter includes a set of filter components for correcting the corresponding noises of the reference patch.

[0015] (vi). The grade is calculated by: calculating a score for each of the plurality of differential patches based on the highest pixel value in the differential patch, thereby generating a plurality of scores corresponding to the plurality of differential patches, and averaging the plurality of scores to obtain the grade.

[0016] (vii). The PMC is further configured to perform generating a plurality of differential patches, calculating the grade, and applying a detection threshold for each DCI in the DCI list to determine whether the DCI is a DOI, and providing an updated defect map corresponding to the inspection area and including one or more DOIs detected by the determination.

[0017] (viii). The PMC is further configured to repeat the following operations: obtaining a plurality of inspection images, generating a plurality of defect maps, aligning one or more candidate defects, generating a plurality of differential patches, calculating the grade, and applying a detection threshold for one or more additional inspection areas on the mask.

[0018] (ix). The plurality of inspection images are sequentially acquired by a photochemical inspection tool with a predefined step size. The photochemical inspection tool is configured to simulate the optical configuration of a lithography tool that can be used to manufacture a semiconductor sample.

[0019] (x). The system further includes a photochemical inspection tool.

[0020] (xi). The plurality of inspection images are obtained by: sequentially acquiring a plurality of images using a non-photochemical inspection tool with a predetermined step size, and performing a simulation on the plurality of images to simulate the optical configuration of a lithography tool that can be used to manufacture a semiconductor sample, thereby generating a plurality of inspection images.

[0021] (xii). The list of candidate defects of interest (DCI) includes one or more candidate defects common to at least a majority of the plurality of inspection images.

[0022] According to other aspects of the presently disclosed subject matter, there is provided a method for inspecting a mask that can be used to fabricate a semiconductor sample, the method being executed by a processing and memory circuitry (PMC) and the method comprising: for an inspection region of the mask, obtaining a plurality of inspection images having a plurality of fields of view (FOV) that at least overlap the inspection region, and for each inspection image, obtaining a set of reference images, the set of reference images including a plurality of reference images for each corresponding reference region in one or more corresponding reference regions; generating a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more candidate defects located in the inspection region of the corresponding inspection image, and aligning the one or more candidate defects of the corresponding inspection image, thereby generating a list of candidate defects of interest (DCI); for at least one given DCI in the list, generating a plurality of differential patches, including generating differential patches corresponding to each of the plurality of inspection images by: extracting image patches surrounding the location of the given DCI from the inspection image and the set of reference images, respectively, thereby generating an inspection patch and a set of reference patches; for each reference patch, calculating a filter that is optimized to minimize the difference between the inspection patch and a corrected reference patch obtained using the filter, thereby generating a set of filters and a set of corrected reference patches corresponding to the set of reference patches; and combining the set of corrected reference patches to obtain a composite reference patch, and comparing the inspection patch with the composite reference patch to obtain a differential patch; and calculating a score based on the plurality of differential patches, and applying a detection threshold to the score to determine whether the given DCI is a defect of interest (DOI).

[0023] With necessary modifications, this aspect of the presently disclosed subject matter may include one or more of the above-listed features (i) to (xii) in any desired combination or arrangement that is technically possible with respect to the system.

[0024] In other aspects of the presently disclosed subject matter, there is provided a non - transitory computer - readable medium including instructions that, when executed by a computer, cause the computer to perform a method for inspecting a mask that can be used to fabricate a semiconductor sample. The method includes: for an inspection region of the mask, obtaining a plurality of inspection images having a plurality of fields of view (FOVs) that at least overlap the inspection region, and for each inspection image, obtaining a set of reference images, the set of reference images including a plurality of reference images for each corresponding reference region in one or more corresponding reference regions; generating a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more candidate defects located in the inspection region of the corresponding inspection image, and aligning the one or more candidate defects of the corresponding inspection image to produce a list of defects of interest (DCI); for at least one given DCI in the list, generating a plurality of differential patches, including generating differential patches corresponding to each of the plurality of inspection images by: respectively extracting image patches around the location of the given DCI from the inspection image and a set of reference images, thereby producing an inspection patch and a set of reference patches; for each reference patch, calculating a filter that is optimized to minimize the difference between the inspection patch and a corrected reference patch obtained using the filter, thereby producing a set of filters and a set of corrected reference patches corresponding to the set of reference patches; and combining the set of corrected reference patches to obtain a composite reference patch, and comparing the inspection patch with the composite reference patch to obtain a differential patch; and calculating a score based on the plurality of differential patches, and applying a detection threshold to the score to determine whether the given DCI is a defect of interest (DOI).

[0025] With necessary modifications, this aspect of the presently disclosed subject matter may include one or more of the above - listed features (i) to (xii) in any desired combination or arrangement that is technically possible with respect to the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To understand the present disclosure and to see how it may be implemented in practice, embodiments will now be described by way of non - limiting examples only with reference to the accompanying drawings, in which:

[0027] Figure 1 A functional block diagram of a mask inspection system in accordance with certain embodiments of the presently disclosed subject matter is shown.

[0028] Figure 2 A general flowchart of mask inspection of a mask that can be used to fabricate a semiconductor sample in accordance with certain embodiments of the presently disclosed subject matter is shown.

[0029] Figure 3 A general flowchart of calculating and applying a filter in accordance with certain embodiments of the presently disclosed subject matter is shown.

[0030] Figure 4 Shows a general flowchart of a computing hierarchy according to certain embodiments of the presently disclosed subject matter.

[0031] Figure 5 Shows a schematic diagram of a photochemical inspection tool and a lithography tool according to certain embodiments of the presently disclosed subject matter.

[0032] Figure 6 Schematically shows an example of a multi-die mask and inspection regions, multiple inspection images, and a reference image of the multi-die mask according to certain embodiments of the presently disclosed subject matter.

[0033] Figure 7 Schematically shows an example of a single-die mask and inspection regions, multiple inspection images, and a reference image of the single-die mask according to certain embodiments of the presently disclosed subject matter.

[0034] Figure 8 Shows an example of multiple differential patches according to certain embodiments of the presently disclosed subject matter. Detailed Description

[0035] In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the present disclosure. However, those skilled in the art will understand that the presently 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 to avoid obscuring the presently disclosed subject matter.

[0036] Unless otherwise specifically stated, it will be apparent from the following discussion that, throughout the specification discussion, terms such as "inspect," "obtain," "generate," "align," "extract," "calculate," "combine," "compare," "acquire," "compute," "apply," "register," "correct," "average," "perform," "provide," "repeat," "acquire," etc. are to be understood to refer to actions and / or processes of a computer that manipulate and / or transform data into other data, where the data is represented as physical (such as electronic) quantities and / or the data represents physical objects. The term "computer" should be broadly interpreted to cover any type of hardware-based electronic device having data processing capabilities, by way of non-limiting example, including the mask inspection systems, defect detection systems, and their corresponding parts disclosed in the present application.

[0037] As used in this specification, the term "mask" is also referred to as "photolithography mask" or "photomask" or "reticle". These terms should be interpreted equivalently and broadly to cover the template that holds the circuit design to be patterned on a semiconductor wafer in a photolithography process (e.g., defining the layout of a particular layer of an integrated circuit). For example, a mask can be implemented as a fused silica plate that is covered with a pattern of opaque, transparent, and phase-shifting regions that are projected onto the wafer during a photolithography process. For example, a mask can be an extreme ultraviolet (EUV) mask or an argon fluoride (ArF) mask. As another example, a mask can be a memory mask (usable for manufacturing memory devices) or a logic mask (usable for manufacturing logic devices).

[0038] As used in this specification, the term "inspection" or "mask inspection" should be interpreted broadly to cover any operation for evaluating the accuracy and integrity of a manufactured photomask in terms of circuit design and its ability to produce an exact representation of the circuit design onto a wafer. Inspection can include any type of operation related to various types of defect detection, defect review, and / or defect classification, and / or metrology operations during and / or after the mask manufacturing process and / or during the use of the mask for semiconductor sample manufacturing. Inspection can be performed using non-destructive inspection tools after the mask is manufactured. As a non-limiting example, the inspection process can include one or more of the following operations: scanning (in a single or multiple scans), imaging, sampling, detecting, measuring, classifying, and / or providing other operations on the mask or a portion thereof using an inspection tool. Similarly, mask inspection can also be interpreted to include, for example, generating (multiple) inspection recipes and / or other setup operations before actually inspecting the mask. It should be noted that unless otherwise specifically stated, the term "inspection" or its derivatives as used in this specification is not limited in terms of the resolution or size of the inspection area. Various non-destructive inspection tools include optical inspection tools, scanning electron microscopes, atomic force microscopes, etc.

[0039] As used in this specification, the term "metrology operation" should be interpreted broadly to cover any metrology operation process for extracting metrology information related to one or more structural elements on a semiconductor sample such as a mask. In some embodiments, a metrology operation can include measurement operations such as, for example, critical dimension (CD) measurements of certain structural elements on the sample, including but not limited to the following: dimensions (e.g., line width, line pitch, contact diameter, element size, edge roughness, gray-scale statistics, etc.), element shape, distances within an element or between elements, relevant angles, overlap information associated with elements corresponding to different design levels, etc. For example, the measurement results such as measurement images are analyzed by using image processing techniques. It should be noted that unless otherwise specifically stated, the term "metrology" or its derivatives as used in this specification is not limited to measurement techniques, measurement resolution, or inspection area size.

[0040] As used herein, the term "sample" should be broadly construed to cover any type of wafer, related structures, combinations thereof, and / or components used in the manufacture of semiconductor integrated circuits, magnetic heads, flat panel displays, and other semiconductor devices.

[0041] As used herein, the term "defect" should be broadly construed to cover any type of abnormality or undesired feature / function formed on a mask. In some cases, a defect may be a defect of interest (DOI), which is a real defect that, when printed on a wafer, has a certain impact on the function of the manufactured device, and thus detecting a DOI is in the customer's interest. For example, any "fatal" defect that may cause a yield loss can be represented as a DOI. In some other cases, a defect may refer to a nuisance (also known as a "false alarm" defect), which can be ignored because it has no impact on the function of the completed device.

[0042] As used herein, the term "candidate defect" should be broadly construed to cover a suspected defect location on a mask that has a relatively high probability of being detected as a defect of interest (DOI). Thus, after reviewing a candidate defect, the candidate defect may actually be a DOI, or in some other cases, the candidate defect may be a nuisance or random noise caused by various variations during inspection (e.g., process variations, color variations, mechanical and electrical variations, etc.).

[0043] As used herein, the terms "non-transitory memory" and "non-transitory storage medium" should be broadly construed to cover any volatile or non-volatile computer memory suitable for the presently disclosed subject matter. These terms should be considered to include a single medium or multiple media storing one or more instruction sets (e.g., a centralized or distributed database, and / or associated caches and servers). These terms should also be considered to include any medium capable of storing or encoding an instruction set for execution by a computer and causing the computer to perform any one or more of the methods of the present disclosure. Thus, these terms should include, but are not limited to, read-only memory ("ROM"), random access memory ("RAM"), magnetic disk storage media, optical storage media, flash devices, etc.

[0044] It should be understood that certain features of the presently disclosed subject matter described in the context of separate embodiments may also be provided in combination in a single embodiment. Conversely, various features of the presently disclosed subject matter described in the context of a single embodiment may also be provided separately or in any suitable sub-combination. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the methods and apparatuses.

[0045] With this in mind, note thatFigure 1 , Figure 1 shows a functional block diagram of a mask inspection system in accordance with certain embodiments of the presently disclosed subject matter.

[0046] Figure 1 The inspection system 100 shown in can be used to inspect a mask during or after a mask manufacturing process and / or during use of the mask for semiconductor sample manufacturing. As described above, the inspection referred to herein can be interpreted to cover any type of operation related to various types of defect inspection / detection, defect classification, and / or metrology operations, such as, for example, critical dimension (CD) measurement of a mask or a portion thereof. According to certain embodiments of the presently disclosed subject matter, the inspection system 100 shown includes a computer-based system 101 that is capable of automatically inspecting and detecting defects on a mask. As described above, the defects to be detected herein can refer to any type of abnormality or undesired feature / function formed on the mask. For example, in some cases, the defect to be detected may be related to an edge placement displacement (EPD), which indicates a difference between an expected position on the mask and an actual position of a printed feature edge. In some other cases, the defect to be detected may be related to CD measurement and / or CD uniformity (i.e., variation in CD measurement across the mask or a portion thereof) or any other type of defect formed on the mask. The system 101, also referred to as a mask defect detection system, is a subsystem of the inspection system 100.

[0047] The system 101 is operatively connected to a mask inspection tool 120 that is configured to scan a mask and capture one or more images of the mask to inspect the mask. The term "mask inspection tool" as used herein should be broadly interpreted to cover any type of inspection tool that can be used in a process related to mask inspection, as non-limiting examples, including scanning (in a single or multiple scans), imaging, sampling, detecting, measuring, classifying, and / or other processes of a mask or a portion thereof.

[0048] Without limiting the scope of the present disclosure in any way, it should also be noted that the mask inspection tool 120 can be implemented as various types of inspection machines, such as optical inspection tools, electron beam tools, etc. In some cases, the mask inspection tool 120 can be a relatively low-resolution inspection tool (e.g., optical inspection tool, low-resolution scanning electron microscope (SEM), etc.). In some cases, the mask inspection tool 120 can be a relatively high-resolution inspection tool (e.g., high-resolution SEM, atomic force microscope (AFM), transmission electron microscope (TEM), etc.). In some cases, the inspection tool can provide both low-resolution image data and high-resolution image data. In some embodiments, the mask inspection tool 120 has metrology capabilities and can be configured to perform metrology operations on the captured images. The resulting image data (low-resolution image data and / or high-resolution image data) can be transmitted directly or via one or more intermediate systems to the system 101. The present disclosure is not limited to any particular type of mask inspection tool and / or the resolution of the image data generated by the inspection tool.

[0049] According to certain embodiments, the mask inspection tool can be implemented as a photochemical inspection tool configured to simulate / imitate the optical configuration of a lithography tool (e.g., a scanner or stepper) that can be used to fabricate semiconductor samples, e.g., by projecting the pattern formed in the mask onto a wafer, as further described in detail below with respect to Figure 5 the following.

[0050] Turning now to Figure 5 , Figure 5 FIG. shows a schematic diagram of a photochemical inspection tool and a lithography tool according to certain embodiments of the presently disclosed subject matter.

[0051] Similar to the lithography tool 520, the photochemical inspection tool 500 can include an illumination source 502 configured to generate light (e.g., a laser) at an exposure wavelength, illumination optics 504, a mask holder 506, and projection optics 508. The illumination optics 504 and the projection optics 508 can include one or more optical elements (such as, for example, lenses, apertures, spatial filters, etc.).

[0052] In the lithography tool 520, the mask is positioned at the mask holder 506 and optically aligned to project an image of the circuit pattern to be replicated onto a wafer placed on a wafer holder 512 (e.g., by using various stepping, scanning, and / or imaging techniques to generate or replicate a pattern on the wafer). Different from the lithography tool 520, instead of placing the wafer holder 512, the photochemical inspection tool 500 places a detector 510 (such as, for example, a charge-coupled device (CCD)) at the position of the wafer holder, where the detector 510 is configured to detect the light projected through the mask and generate an image of the mask.

[0053] It can be seen that the actinic inspection tool 500 is configured to simulate the optical configuration of the lithography tool 520, including but not limited to, for example, illumination / exposure conditions such as wavelength, partial coherence of the exposure light, pupil shape, illumination aperture, numerical aperture (NA), etc., and the optical configuration is used to expose photoresist in an actual lithography process during semiconductor device manufacturing. Therefore, the mask image 514 acquired by the detector 510 is expected to be similar to the image 516 of the wafer manufactured using the mask via the lithography tool. The mask image acquired using such an actinic inspection tool is also referred to as an aerial image. As described below, the aerial image is provided to the system 101 for further processing.

[0054] According to certain embodiments, in some cases, the mask inspection tool 120 can be implemented as a non-actinic inspection tool, such as, for example, a conventional optical inspection tool, an electron beam tool (e.g., SEM), etc. In such a case, the detector of the inspection tool is capable of being connected to the specific type of microscope used and digitizing the image information from the microscope, thereby acquiring an image of the mask.

[0055] Simulations can be performed on the acquired image to simulate the optical configuration of the lithography tool, thereby generating an aerial image. In some cases, the image simulation can be performed by the system 101 (e.g., by incorporating an image simulation model in the PMC 102, the simulation function can be integrated into the PMC 102), while in some other cases, the image simulation can be performed by a processing module of the mask inspection tool 120 or by a separate simulation engine / unit operatively connected to the mask inspection tool 120 and the system 101.

[0056] For illustrative purposes only, certain embodiments described below are provided for images acquired by an actinic mask inspection tool. Those skilled in the art will readily understand that the teachings of the presently disclosed subject are equally applicable to images acquired by any other suitable techniques and inspection tools, and further converting the images into aerial images using suitable simulation models. The term "aerial image" should be broadly interpreted to cover images acquired by actinic mask inspection tools and aerial images simulated from images captured by (multiple) non-actinic inspection tools.

[0057] The system 101 includes a processor and a memory circuit system (PMC) 102 operatively connected to a hardware-based I / O interface 126. The PMC 102 is configured to provide the processing required to operate the system, as referred to in Figure 2 , Figure 3 and Figure 4Further detailed description, and PMC 102 includes a processor (not shown separately) and a memory (not shown separately). The processor of PMC 102 can be configured to execute several functional modules according to computer-readable instructions implemented on a non-transitory computer-readable memory included in the PMC. Such functional modules are hereinafter referred to as being included in the PMC.

[0058] The processor referred to herein can represent one or more general-purpose processing devices, such as a microprocessor, a central processing unit, etc. More specifically, the processor can be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor implementing other instruction sets, or a processor implementing a combination of instruction sets. The processor can also be one or more dedicated processing devices, such as an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), a network processor, etc. The processor is configured to execute the instructions for performing the operations and steps discussed herein.

[0059] The memory referred to herein can include a 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 (e.g., flash memory, static random access memory (SRAM), etc.).

[0060] As previously mentioned, in some embodiments, system 101 can be configured to detect defects on a mask. If mask defects are not detected before wafer mass production, the mask defects will be repeated many times on the produced wafers, resulting in defects in multiple semiconductor devices (e.g., affecting the function of the device and failing to achieve the expected performance), and having an adverse impact on the yield.

[0061] Given the large-scale circuit integration and the reduced size of semiconductor devices in the advanced process of photomasks, mask inspection has become increasingly sensitive to different types of variations and noises. To detect smaller defects, it is expected to inspect the mask by sensitive scanning (i.e., scanning 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 this case, the DOI may be hidden in the false alarms and noises, thus affecting the detection sensitivity and resulting in an increase in the false alarm rate (FAR).

[0062] In mask inspection, die-to-die (D2D) inspection is typically 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 is typically compared with the reference images to create a differential image indicative of potential defects on the mask. Conventional D2D inspection uses one or more adjacent dies as references, so the number of reference images is very limited. In addition, as described above, the inspection image and the reference images are acquired using sensitive scanning, so they are affected by the above-mentioned false positives and increased noise, and it is difficult to separate the DOI from the false positives and the noise.

[0063] According to certain embodiments of the presently disclosed subject matter, improved mask inspection systems and methods are provided that are configured to detect DOI by acquiring a sufficient number of reference images that are further optimized and used to create a combined best reference image, thereby presenting a more reliable differential image with a higher signal-to-noise ratio (SNR), thereby improving detection sensitivity while reducing the FAR.

[0064] To create a sufficient number of references, multiple reference positions (e.g., from multiple reference dies, or from a single die) are used, and overlapping scans can be used to capture multiple images for each reference position. The overlapping images of the multiple references are used together to reduce noise and increase detection confidence. Image filtering techniques are used to further correct the multiple reference images separately in order to create the best reference image, thereby enabling better discrimination between the DOI and false positives or random noise. The proposed process has been shown to improve the sensitivity of defect detection for advanced process control of mask features without affecting the inspection throughput.

[0065] According to certain embodiments, the functional modules included in the PMC 102 of the system 101 may include an image processing module 104 and a defect detection module 106. The PMC 102 may be configured to obtain, for an inspection region of a mask (via the I / O interface 126), multiple inspection images having a plurality of fields of view (FOVs) that at least overlap the inspection region, and for each inspection image, obtain a set of reference images, the set of reference images including multiple reference images for each corresponding reference region in one or more corresponding reference regions. For example, the inspection images and the reference images may be acquired by a mask inspection tool 120 such as, for example, a lithography inspection tool.

[0066] The defect detection module 106 may be configured to generate multiple defect maps corresponding to the multiple inspection images. Each defect map includes one or more candidate defects located in the inspection region of the corresponding inspection image. The one or more candidate defects of the corresponding inspection image may be aligned to produce a list of defects of interest (DCI).

[0067] For each given DCI in the list, the image processing module 104 can be configured to generate a plurality of differential fill blocks. Specifically, the image processing module 104 can be configured to generate differential fill blocks corresponding to each of the plurality of inspection images by at least the following steps: extracting image fill blocks around the position of the given DCI from the inspection image and a set of reference images respectively, thereby generating an inspection fill block and a set of reference fill blocks; for each reference fill block, calculating a filter that is optimized to minimize the difference between the inspection fill block and the corrected reference fill block obtained using the filter, thereby generating a set of filters corresponding to the set of reference fill blocks and a set of corrected reference fill blocks; and combining the set of corrected reference fill blocks to obtain a composite reference fill block, and comparing the inspection fill block with the composite reference fill block to obtain a differential fill block.

[0068] The defect detection module 106 can be further configured to calculate a rank based on the plurality of differential fill blocks and apply a detection threshold to the rank to determine whether the given DCI is a defect of interest (DOI).

[0069] Reference will be made to Figure 2 、 Figure 3 and Figure 4 for a further detailed description of the operation of the systems 100, 101, PMC 102 and the functional modules therein.

[0070] According to certain embodiments, the system 100 can include a storage unit 122. The storage unit 122 can be configured to store any data required by the operating systems 100 and 101, such as data related to the input and output of the systems 100 and 101, and intermediate processing results generated by the system 101. For example, the storage unit 122 can be configured to store inspection images and reference images generated by the mask inspection tool 120 and / or their derivatives (e.g., pre-processed images). Thus, images can be retrieved from the storage unit 122 and provided to the PMC 102 for further processing.

[0071] In some embodiments, the system 100 can optionally include a computer-based graphical user interface (GUI) 124 configured to enable user-specified input related to the system 101. For example, a visual representation of the mask can be presented to the user (e.g., via a display forming part of the GUI 124), the visual representation including an image of the mask or a portion thereof. Options for defining certain operating parameters (such as, for example, sensitive scan parameters, the number of reference regions / positions, a list of candidate defects of interest (DCI), detection thresholds, etc.) can be provided to the user via the GUI. In some cases, the user can also view operation results on the GUI, such as composite reference fill blocks, (multiple) differential fill blocks, (multiple) detected DOIs, (multiple) defect maps, and / or further inspection results.

[0072] As described above, system 101 is configured to receive multiple images of a mask (e.g., inspection images and / or reference images) via I / O interface 126. The images may include image data (and / or derivatives thereof) generated by mask inspection tool 120 and / or image data stored in storage unit 122 or one or more data repositories. In some cases, the image data may refer to images captured by the mask inspection tool and / or pre-processed images derived from the captured images obtained through various pre-processing stages, etc. It should be noted that in some cases, the images may include associated digital data (e.g., metadata, manually crafted attributes, etc.). It should also be noted that in some embodiments, the image data is related to a target layer of a semiconductor device to be printed on a wafer.

[0073] System 101 is further configured to send inspection results (e.g., detected DOIs, defect maps, composite reference patches, etc.) to storage unit 122, and / or GUI 124 for rendering, and / or mask inspection tool 120 by processing the received images and via I / O interface 126.

[0074] In some embodiments, in addition to system 101, mask inspection system 100 may further include one or more inspection modules, such as, for example, (multiple) additional defect detection modules and / or automatic defect review modules (ADR) and / or automatic defect classification modules (ADC) and / or metrology-related modules and / or other inspection modules available for performing additional inspections on the mask. One or more inspection modules may be implemented as stand-alone computers, or the functions of one or more inspection modules (or at least some of them) may be integrated with mask inspection tool 120. In some embodiments, the output obtained from system 101 may be used by mask inspection tool 120 and / or one or more inspection modules (or parts thereof) for further inspecting the mask.

[0075] Those skilled in the art will readily understand that the teachings of the presently disclosed subject matter are not limited by Figure 1 the system shown; equivalent and / or modified functions may be combined or divided in another manner and may be implemented in any suitable combination of software with firmware and / or hardware.

[0076] It should be noted that Figure 1The mask inspection system shown can be implemented in a distributed computing environment, where the above-described functional modules included in the PMC 102 can be distributed across multiple local and / or remote devices and can be linked via a communication network. It should also be noted that in other embodiments, one or more of the mask inspection tool 120, the storage unit 122, and / or the GUI 124 can be external to the system 100 and communicate data with the system 101 via the I / O interface 126. The system 101 can be implemented as a (multiple) stand-alone computer for use with the mask inspection tool. Alternatively, the various functions of the system 101 can be at least partially integrated with the mask inspection tool 120, thereby facilitating and enhancing the functions of the mask inspection tool 120 in the inspection process.

[0077] Although not necessarily so, the operating procedures of the system 101 and the system 100 can correspond to some or all of the stages of the method described with respect to Figures 2 to 4 Similarly, the method described with respect to Figures 2 to 4 and its possible implementation manners can be implemented by the system 101 and the system 100. Therefore, it should be noted that, with necessary modifications, the various embodiments of the system 101 and the system 100 can implement the discussed embodiments related to the method described with respect to Figures 2 to 4 and vice versa.

[0078] Now referring to Figure 2 , Figure 2 shows a general flowchart of the mask inspection of a mask that can be used to fabricate a semiconductor sample according to certain embodiments of the presently disclosed subject matter.

[0079] For the inspection area of the mask, a plurality of inspection images can be obtained (202) (e.g., by the PMC 102 via the I / O interface 126, from the mask inspection tool 120, or from the storage unit 122). The plurality of inspection images have a plurality of fields of view (FOV) that at least overlap with the inspection area. For each inspection image, a set of reference images can be obtained, and the set of reference images includes a plurality of reference images for each corresponding reference area in one or more corresponding reference areas. The inspection images and the reference images are aerial images as described above.

[0080] In some embodiments, the plurality of inspection images (and / or reference images) can be sequentially acquired by a photochemical inspection tool (such as, for example, the Aera mask inspection tool of Applied Materials Inc.). The photochemical inspection tool is configured to simulate the optical configuration of a lithography tool (e.g., a scanner or a stepper) that can be used to fabricate a semiconductor wafer according to a mask, as described above with reference to Figure 5 . The images can be acquired at a predefined step size such that the FOV of the images at least overlaps with the inspection area.

[0081] Images obtained by such a photochemical inspection tool (i.e., aerial images) are expected to be similar to images of wafers fabricated using a mask via a lithography tool. In other words, the photochemical mask inspection tool is configured to capture a mask image that can simulate how the design pattern in the mask will actually appear in a physical wafer after the manufacturing process.

[0082] In some cases, a photochemical inspection tool may not be available for inspecting the mask. In such cases, non-photochemical inspection tools such as, for example, a conventional optical inspection tool, an electron beam tool, etc. can be used to obtain a non-aerial image of the mask. The obtained non-aerial image can be simulated to mimic the optical configuration of the lithography tool, thereby generating an aerial image of the mask. Thus, in some embodiments, the mask inspection method described Figure 2 may further include a preparatory step of obtaining a plurality of images obtained by a non-photochemical inspection tool and performing a simulation on 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.) to mimic the optical configuration of the lithography tool, thereby generating a plurality of inspection images (i.e., aerial images).

[0083] In some embodiments, during inspection, the mask may be moved in steps relative to the detector of the mask inspection tool during exposure (or the mask and the tool may be moved in opposite directions with respect to each other), and the mask may be scanned step by step along a strip of the mask by the mask inspection tool, which images only a component / portion (also referred to as the field of view (FOV) of the tool or the image) of the mask within the strip at a time. For example, in each step, light may be detected from a rectangular portion of the mask, and the detected light may be converted into a plurality of intensity values at multiple points within the portion, thereby forming an image corresponding to the component / portion of the mask. The size and dimensions of the FOV or the image corresponding to the FOV may vary depending on certain factors such as different tool configurations. In one example, each image corresponding to a rectangular FOV of the mask may be approximately 1000 pixels in length and approximately 1000 pixels in width. In another example, the image corresponding to the rectangular FOV may be of a size of approximately 800 pixels × 1600 pixels.

[0084] Thus, a plurality of images of the mask can be sequentially obtained during sequential scanning along a strip of the mask, each image representing a corresponding component / portion of the mask. For example, the first strip of the mask may be scanned from left to right, then the second strip may be scanned from right to left, and so on until the entire mask (or the region of interest of the mask) is scanned.

[0085] In some cases, multiple images can be obtained by a mask inspection tool in steps, where the step size is predefined such that the FOVs of the multiple images can partially overlap according to the step size. For example, in the case where the step size is predefined as 1 / 3 of the FOV length, three sequentially captured images overlap by 1 / 3 of the FOV. In other words, a given inspection area of the overlapping region can be captured three times in three consecutive images. Thus, using overlapping imaging acquisition, each inspection area on the mask can be captured multiple times in multiple overlapping images.

[0086] A set of reference images can be obtained for each inspection image, where the set of reference images includes multiple reference images of each corresponding reference area in one or more corresponding reference areas. For each inspection area, one or more reference areas can be identified and used as a reference for comparison. For example, in the case where the mask to be inspected is a multi-die mask (where the mask field includes multiple dies with the same / similar design patterns) and the inspection area is located in the inspected die on the mask, one or more reference areas (corresponding to the position of the inspection area) from one or more reference dies (e.g., adjacent dies to the inspected die) on the mask can be used as a reference (such as, for example, in D2D inspection).

[0087] As another example, in the case where the mask is a single-die mask (where the mask field includes only one die), the inspection area and one or more reference areas are located in the same die on the mask, where the one or more reference areas share the same / similar design patterns with the inspection area. As will be described in more detail below, for example, based on the design data of the mask, any suitable algorithm available for identifying similar patterns can be used to identify the one or more reference areas.

[0088] As will be referred to Figure 2 As described, in some embodiments, the multiple inspection images (and / or reference images) obtained can be preprocessed before further processing. The preprocessing can include one or more of the following operations: interpolation (e.g., in the case where the image has a relatively low resolution), noise filtering, focus correction, aberration compensation, and image format conversion, etc.

[0089] It should be noted that the present disclosure is not limited to a specific modality of the mask inspection tool, and / or the type of images obtained thereby, and / or the preprocessing operations required for processing the images.

[0090] Multiple defect maps corresponding to a plurality of inspection images can be generated (204) (e.g., by the defect detection module 106 in the PMC 102). At least one reference image (e.g., one reference image from a set of reference images) can be used to generate each defect map, and each defect map can indicate the distribution of candidate defects on the corresponding inspection image. For example, at least one difference image can be generated based on the difference between the pixel values of the inspection image and the pixel values of at least one reference image. The defect map can be generated by determining the locations of suspicious defects (i.e., candidate defects) based on the at least one difference image using a detection threshold. In some embodiments, the defect map can further indicate one or more defect characteristics of the candidate defects, such as, for example, the location, intensity, and size of the candidate defects, etc. The candidate defects shown in the defect map can be located in the corresponding inspection image based on the location of the candidate defects.

[0091] As described above, in some embodiments, sensitive scanning can be used to acquire the inspection image and one or more reference images. For example, sensitive scanning can be enabled by using an inspection tool with a specific parameter configuration having higher sensitivity. The configuration parameters can include one or more of the following: lighting conditions, polarization, and noise level of each region, etc.

[0092] In some embodiments, in addition to or instead of sensitive scanning, the detection threshold can be configured according to sensitive detection requirements. For example, a relatively low threshold can be used to display more suspicious defects in the defect map, resulting in defect detection with higher sensitivity. The candidate defects in the defect map generated by such sensitive scanning and / or sensitive detection are most likely to be noise and / or false alarms (since DOIs are rare). Such a defect map is thus also referred to as a noise map.

[0093] Specifically, each defect map may include one or more candidate defects located in the inspection region of the corresponding inspection image. One or more candidate defects of the corresponding inspection image may be aligned to generate a list of Defects of Interest (DCI). For example, the candidate defects of the corresponding inspection image may be aligned based on defect features of the candidate defects (e.g., defect location). For example, the defect map may be converted to mask coordinates such that each candidate defect is described by a list of coordinates in the mask coordinate system. Multiple defect maps of the overlapping inspection images may report the same defect located in the overlapping inspection region between the FOVs of the inspection images. Once the candidate defects are reported using the mask coordinates, they may be unified among the candidate defects of the inspection image. The unification may be performed by matching the candidate defect coordinates in the mask coordinate system with a matching criterion (e.g., by applying dilation to the location of the candidate defect). In some embodiments, in addition to matching the coordinates, the unification may also take into account the number of occurrences of the candidate defects in the defect map. For example, configuration parameters may be defined for additional filtering of the matched candidate defects, requiring the detection of the candidate defects in a certain number of overlapping defect maps (e.g., in all of the multiple defect maps), as further illustrated below. In some embodiments, optionally, once the candidate defects are unified, additional filtering may be performed based on the intensity of the candidates ranked in each defect map. For example, the unified candidates with relatively high rankings will be selected for the DCI list.

[0094] Reference is now made to Figure 6 , Figure 6 which schematically shows an example of a multi-die mask and inspection regions, multiple inspection images, and a reference image of the multi-die mask according to certain embodiments of the presently disclosed subject matter.

[0095] As shown, the multi-die mask 600 has a mask field including nine dies sharing the same design pattern. For a given inspection region 602 in the inspected die of the mask 600, three overlapping inspection images 603, 604, and 605 are sequentially acquired. These three images are captured at a specific step size such that their FOVs overlap, for example, by one third of the FOV of the image. Thus, the inspection region 602 is acquired three times in the three images (e.g., the inspection region 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).

[0096] For the inspection region 602 in the inspection die for the mask 600, two reference regions 606 and 608 (corresponding to the position of the inspection region 602) from two reference dies (e.g., two adjacent dies) on the mask of the inspection die can be used as a reference for comparison. For any reference region, three reference images can be similarly obtained, where the FOV overlaps by one-third of the FOV of the image. Thus, a set of six reference images is obtained for each of the inspection images 603, 604, and 605.

[0097] During defect detection, at least one of the six reference images can be used to generate a defect map for each inspection image. Thus, three defect maps corresponding to the three inspection images 603, 604, and 605 are generated. Each defect map includes one or more candidate defects shown within the inspection region of the corresponding inspection image. The candidate defects of the corresponding inspection images can be aligned / registered, thereby generating a list of candidate defects of interest (DCI).

[0098] For example, assume that the same number of candidate defects (e.g., three candidate defects) are shown at corresponding positions in the inspection regions of the three inspection images and the candidate defects are correctly aligned between the images. Then the DCI list can include three candidate defects. Another example, in some cases, due to various reasons such as noise and / or variations, different numbers of candidate defects may be shown in different inspection images. For example, assume that three candidate defects are shown in the inspection region of the first inspection image, three candidate defects are shown in the inspection region of the second inspection image, and only two candidate defects are shown in the inspection region of the third inspection image. Each time alignment is performed, two candidate defects are unified as common candidate defects in the three inspection images, while the remaining one candidate defect appears in two images but somehow is missing from the third inspection image. In some cases, the configuration parameters as described above can be defined to require that the candidate defects appear in most of the inspection images (e.g., two-thirds). In this case, the DCI list can include three candidate defects, and thus the image information of the position of the missing candidate defect in the third image can also be obtained and processed in further processing to assist in determining whether this candidate defect is a DOI or a false alarm. In some other cases, alternatively, the above configuration parameters can be defined to require that the candidate defects appear in all inspection images. In this case, the DCI list can only include two common candidate defects that appear in all images. In a further case, the DCI list can be determined based on additional or alternative factors (such as, for example, a predetermined number of DCIs to be selected, the ranking of candidate defects relative to certain defect characteristics, etc.).

[0099] Continue Figure 2For the description of, for at least one given DCI in the DCI list determined as above, multiple differential complementary blocks can be generated (206) (e.g., by the image processing module 104 in the PMC 102). Specifically, the differential complementary blocks corresponding to each inspection image among the multiple inspection images can be generated in the process described with reference to frames 208 to 214 below.

[0100] Specifically, for each inspection image, image patches surrounding the position of the given DCI can be extracted (208) from the inspection image and a set of reference images respectively, thereby generating an inspection patch and a set of reference patches. In Figure 6 the example of, a DCI common to three inspection images is shown (e.g., assuming there are M selected DCIs, although only one of them (as a black dot) is shown in Figure 6 ). For example, for the DCI in inspection image 603, square-shaped surrounding image patches surrounding the DCI can be extracted from inspection image 603 and each of the six reference images respectively, thereby generating an inspection patch (e.g., the left inspection patch in inspection patch 610) and a set of reference patches 612 (six reference patches in this example). Therefore, for each DCI in the DCI list, a corresponding inspection patch and a set of reference patches can be generated.

[0101] Optionally, in some embodiments, the inspection patch is respectively registered (210) with each reference patch from a set of reference patches. Two image patches (the inspection patch and the corresponding reference patch) are registered to perform an accurate comparison. Registration can include measuring the offset between the two image patches and moving one image patch relative to the other image patch to correct the offset. The offset may be caused by various factors (such as, for example, navigation errors caused by tool drift (e.g., scanner and / or platform drift)), because the two patches are extracted from different images acquired for different dies.

[0102] Registration can be implemented according to any suitable registration algorithm known in the art. For example, one or more of the following algorithms can be used to perform registration: region-based algorithms, feature-based registration, or phase correlation registration. An example of a region-based method is to use optical flow such as the Lucas Kanade algorithm (LK) for registration. Feature-based methods are based on finding different information points ("features") in two images and calculating the required transformation between each pair of images according to the correspondence of the features. This allows elastic registration (i.e., non-rigid registration), where different regions move separately. Phase correlation registration is completed using frequency domain analysis (where the phase difference in the Fourier domain is converted into registration in the image domain).

[0103] Alternatively, in some embodiments, registration may be skipped. For example, registration may be omitted in cases where it can be estimated that there may be no substantial offset between the inspection patch and the corresponding reference patch.

[0104] For each reference patch, a filter (212) may be calculated to be applied to the reference patch to generate a corrected reference patch with better quality (e.g., less noise, higher SNR). In some embodiments, an optimization method may be used to obtain the filter to minimize the difference between the inspection patch and the corresponding corrected reference patch (the corrected reference patch is obtained by applying the filter to the corresponding reference patch), thereby increasing the SNR of the defect signal in the resulting differential patch. The filter thus obtained is also referred to herein as the optimal filter for the corresponding reference patch.

[0105] In some embodiments, the optimal filter thus generated is capable of correcting certain variations and transformations between the inspection patch and the reference patch, which may be caused by various noise types / sources, such as one or more of the following: registration residuals (e.g., rigid registration residuals or elastic registration 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.

[0106] Some of such noise can be represented by a linear model, such as rigid registration residuals, intensity gain and offset, defocus, etc. For example, a function such as can represent the variation of intensity gain and offset between the gray levels of pixels (x, y) in the inspection patch and the reference patch, where the coefficient a k represents the gain factor and b k represents the offset. In some other cases, some of the noise can be represented by a non - linear model, such as FOV distortion, elastic registration residuals, etc.

[0107] Specifically, the FOV distortion as described above refers to the image intensity variations and non - uniformities at different positions within the FOV of the image. This may be caused by certain optical system aberrations, including but not limited to, for example, astigmatism, non - uniform illumination in the field, distortion due to lens shape and speckle, etc., as will be described in further detail below.

[0108] For example, FOV distortion may be caused by non-uniform illumination (e.g., illumination non-uniformity), which becomes particularly critical under low illumination conditions and may increase the noise in certain regions of the image. Astigmatism and field curvature are known aberrations of the optical system, which may result in cylindrical effects where the contrast and focus are non-uniform in the x and y directions of the image and vary along the FOV. These aberrations can significantly affect the similarity of the patches and cause differences in the appearance of the pattern at different positions in the FOV.

[0109] In addition, distortion caused by the lens shape results in different magnifications along the FOV (which may cause changes in pixel size). Speckles refer to high-frequency noise that varies between frames (e.g., the noise may not appear at the same position in different images).

[0110] As described, the aforementioned aberrations can vary between different FOV positions, and in some cases, the aberrations also depend on the pattern. It should be noted that although FOV distortion is typically represented by a non-linear filter, when solving FOV distortion on relatively small image patches, the behavior of FOV distortion can be estimated as linear, and thus a linear filter can be used to correct FOV distortion in some cases.

[0111] To address the various noises described above, different filtering methods can be used (individually or in any suitable combination), such as, for example, linear regression, non-linear regression, matched filters, etc. In particular, in some embodiments, each individual reference patch requires specific correction because different patches may be subject to different noises. For example, different FOV distortions may occur when using reference patches from different positions in the FOV. Therefore, unique filters are needed to individually correct each reference patch to address one or more of the above-mentioned noises and aberrations.

[0112] In some embodiments, an optimization method can be used to obtain a filter to minimize the difference between the inspection patch and the corresponding corrected reference patch (when the filter is applied to the corresponding reference patch). For example, the objective loss function of the optimization process can be expressed as: min{∑ i,j h*ref-ins}, where h represents the filter, ref represents the pixel values of the reference patch, and ins represents the pixel values of the inspection patch. For example, the filter can have a fixed size (e.g., 5*5, or 7*7), and when the filter is applied to the reference patch, it can be sequentially convolved with the corresponding part of the reference patch to obtain the corrected reference patch. Such a calculated filter can correct the above-mentioned variations caused by one or more of the following: registration residuals, intensity gain and offset, defocus, and FOV distortion, etc.

[0113] In some embodiments, when the desired correction can be considered a linear transformation, least squares (LS) optimization can be used to calculate the filter. An exemplary implementation of this optimization is described below.

[0114] For a given image I and a reference image R, search for a filter f such that f = argmin ||R * f - I||2. f is the filter that minimizes the difference between the reference patch and the image patch. The optimization algorithm finds the 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 LS optimization is calculated for the filter variables. The output of the algorithm includes the estimated optimal filter f and the corrected image R_corrected. By applying an unknown filter f (e.g., of size 3×3) to the image R and comparing it with the image I, a set of approximate linear equations can be generated. This set of linear equations is transformed into matrix notation x = Ab, where:

[0115] b = [I 22 , I 32 , I 42 , …, I 32 , I 33 , I 34 ,..] T

[0116] x = [f 11 , f 21 , f 31 , …, f 31 , f 32 , f 33 T

[0117]

[0118] The discrete 2D convolution can be represented as matrix multiplication, and the problem is transformed into the following LS square problem: And the LS solution is given by: (where b is the vector of known observations, A is the known matrix, and x is the vector of unknown parameters).

[0119] ​In some other embodiments, in cases where the required correction can be considered a non - linear transformation, an iterative optimization method, for example, can be used to calculate the filter. For example, the optimization method can be implemented as a first - order (gradient - based) line search optimization scheme. A first rough fit with the initial value of the filter can be calculated and used as the initial starting point for the optimization. In each iteration, the gradient can be calculated at the starting point, and the direction of descent (e.g., the anti - gradient direction) can be identified along which the loss function will be sufficiently reduced. A step size is calculated, which determines the distance to move along this direction. When moving along the direction of descent with the determined step size, a new starting point is derived and used to start the next iteration. The iteration can be repeated until convergence (i.e., the error of the loss function reaches a minimum). The filter with the optimized value that can minimize the loss function becomes the optimal filter for correcting the reference patch.

[0120] In some cases, optionally, the filter can additionally include one or more image masks. For example, one mask can be used to avoid invalid image information, such as defective pixels (e.g., burnt pixels) in the image patch. As another example, a weight mask can be used for the purpose of applying different weights to different regions of the image patch. For example, the weight mask can be configured such that the weights of important regions in the image patch are increased, while the weights of less important regions can be reduced. Examples of regions that may require specific weights include (but are not limited to): suspicious defect pixels, regions around suspicious defects (excluding the defect itself), peripheral pixels, edges within the image (usually having high gradients). Such masks can be used alone or in combination.

[0121] Optionally, in certain cases, the size of the filter can be adjusted according to the noise characteristics (e.g., noise level) of the reference image.

[0122] According to certain embodiments, as Figure 3 shown, for each reference patch, one or more noises can be selected (302) for correction (e.g., different types / sources of noise), for example, selected (302) from the group including: registration residuals (including rigid registration residuals and / or elastic registration residuals, etc.), intensity gain and offset, defocus, and / or field - of - view (FOV) distortion. A filter including a set of filter components can be specifically calculated (304) (using any suitable optimization method) to correct the corresponding noise of each reference patch. As described above, each filter component can be linear or non - linear and can address one or more of the above - mentioned noise types, optionally with one or more image masks. The calculated filter can be applied to (306) the reference patch to obtain the corrected reference patch.

[0123] Once the corresponding filters are computed for each reference fill block, a set of filters corresponding to a set of reference fill blocks is obtained, and a set of corrected reference fill blocks can be created (e.g., by applying the set of filters to the corresponding reference fill blocks). A set of corrected reference fill blocks can be combined (214) to obtain a composite reference fill block. The check block can be compared with the composite reference fill block to obtain a differential block.

[0124] For example, a differential block can be generated based on the difference between the pixel values of the check block and the pixel values of the composite reference fill block. In some cases, one or more difference normalization factors can be used to further normalize the differential block (for at least some of the pixels in the block). For example, the difference normalization factors can be determined based on the behavior of the normal population of pixel values in the differential block and / or the composite reference fill block. For example, to reduce the shot noise effect, the differential block can be normalized according to the gray level of the corresponding pixel values in the composite reference fill block (e.g., pixels with higher gray level values are generally noisier than pixels with lower gray level values, so different normalization factors can be assigned to different pixels). In some cases, the pixel values in the differential block can be normalized to the ratio between the corresponding original pixel values of the differential block and the corresponding difference normalization factors.

[0125] In Figure 6 the example of

[0126] Once the above process of blocks 208 to 214 is performed for each check image and the differential blocks corresponding to the check blocks are obtained, a plurality of differential blocks corresponding to a plurality of check images (for a given DCI) are obtained. In Figure 6 the example of

[0127] the above process of blocks 208 to 214 is repeated for each of the three check images 603, 604, and 605 for the shown DCI, and three differential blocks corresponding to the three check blocks 610 are obtained.

[0128] As Figure 4Illustratively, in some embodiments, the rank may be calculated (402) by calculating a score for each of a plurality of differential patches based on the highest pixel value in the differential patch, thereby generating a plurality of scores corresponding to the plurality of differential patches, and averaging (404) the plurality of scores to obtain the rank. For example, the score for each differential patch may be the highest pixel value in the patch, which indicates the strength of the suspected defect signal. As another example, the score may be derived as an average (e.g., mean, weighted mean, or average value) of the pixel values in the patch.

[0129] Figure 8 An example of a plurality of differential patches according to certain embodiments of the presently disclosed subject is shown. As shown, for example, by comparing each inspection patch with a corresponding composite reference patch (which is generated based on six reference patches), three differential patches 802, 804, and 806 corresponding to Figure 6 three inspection patches 610 in are generated. As described above, the three differential patches are further normalized. As shown, a score for each differential patch is calculated based on the highest pixel value in the patch, thereby generating three scores. The rank can be obtained by averaging the three scores.

[0130] In some embodiments, generation (206) of a plurality of differential patches, calculation of a rank, and application of a detection threshold (216) may be performed for each DCI in a DCI list to determine whether the DCI is a DOI. An updated defect map corresponding to the inspection area may be provided, the updated defect map including one or more DOIs detected by the determination. In Figure 6 the example of, assuming there are M DCIs in the set of DCIs selected during the process described with reference to frame 204 (although Figure 6 only one DCI is shown therein), the determination as described above may be made for each of the M DCIs, and N of the M DCIs may be determined to be DOIs. An updated defect map may be generated, the updated defect map including N DOIs corresponding to the inspection area.

[0131] The process described with reference to Figure 2 may be repeated for one or more additional inspection areas on the mask. In some embodiments, an area of interest (ROI) to be inspected on the mask may be predefined, and one or more inspection areas within the ROI may be inspected using the above-described process. In some cases, the ROI may be defined as the entire mask, while in some other cases, the ROI may be defined as a part of the mask.

[0132] It should be noted that although the mask inspection process described with reference to Figure 2 is performed using as Figure 6illustrated by way of example of the multi-die mask shown, but this is in no way intended to limit the present disclosure in any way. It should be understood that the methods and systems presented can be similarly applied to single-die masks. Figure 7 Examples of a single-die mask, inspection regions of the single-die mask, a plurality of inspection images, and reference images are shown in accordance with certain embodiments of the presently disclosed subject matter.

[0133] As shown, the single-die mask 700 has a mask field consisting of a single die. For a given inspection region 702 in the single die, three inspection images 703, 704, and 705 are sequentially acquired. Similar to that described above with reference to Figure 6 these three images are captured at a specific step size such that the FOVs of the three images overlap, for example, by one-third of the FOV of an image. Thus, the inspection region 702 is acquired three times in the three images (e.g., the inspection region 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).

[0134] For the inspection region 702, three reference regions 706, 708, and 709 from the same die that share the same design pattern as the inspection region can be used as a reference for comparison. For each reference region, three reference images can be similarly acquired, where the FOVs overlap by one-third of the FOV of an image. Thus, a set of nine reference images is obtained for each of the inspection images 703, 704, and 705.

[0135] The reference regions in the single-die mask can be identified in various ways. The design data of the die (or its (multiple) parts) can include various design patterns having specific geometries and arrangements. The design patterns can be defined as being composed of one or more structural elements, each structural element having a geometry with a contour (e.g., one or more polygons).

[0136] In some embodiments, the design data of the single-die mask can be received, and a plurality of design groups can be retrieved, each of the plurality of design groups corresponding to one or more die regions having the same design pattern. Thus, the regions in the die corresponding to the same design pattern can be identified. It should be noted that design patterns are considered to be "the same" when the design patterns are the same or when the design patterns are highly correlated or similar to each other. Various similarity metrics and algorithms can be applied to match and cluster similar design patterns, and the present disclosure should not be construed as being limited by any particular metric used to derive the design groups. The clustering of the design groups (i.e., the partitioning from the CAD data into a plurality of design groups) can be performed in advance or by the PMC 102 as a preparatory step of this inspection process.

[0137] Once in as Figure 7In the single-die scenario shown, by identifying a reference region and obtaining a reference image, defect map generation, differential patch generation, and grade calculation and DOI determination can be performed similarly to the multi-die scenario described above with reference to Figure 2 and Figure 6 .

[0138] According to some embodiments, the output generated by the inspection process described with reference to Figure 2 may include one or more of the following: a determined DOI, an updated defect map for the inspection region, and / or a composite reference patch for each inspection patch. The output may be used by the mask inspection tool 120 and / or one or more inspection modules included in the mask inspection system 100 to further inspect the mask, such as, for example, additional defect detection, defect review, defect classification, metrology-related operations (e.g., CD measurement), and / or any other inspection operation. For example, the composite reference patch may be used for defect detection related to EPD.

[0139] It should be noted that the mask applicable to the currently disclosed inspection process can be any type of mask, including but not limited to memory masks and / or logic masks, and / or Arf masks and / or EUV masks, etc. The present disclosure is not limited to a specific type or function of the mask to be inspected.

[0140] According to some embodiments, as described above with reference to Figure 2 , Figure 3 and Figure 4 , the mask inspection process can be included as part of an inspection recipe that can be used by the system 101 and / or the inspection tool 120 for in-line mask inspection at runtime. Accordingly, the presently disclosed subject matter also includes systems and methods for generating an inspection recipe during the recipe setup phase, where the recipe includes the steps (and their various embodiments) described with reference to Figure 2 , Figure 3 and Figure 4 . It should be noted that the term "inspection recipe" should be broadly interpreted to cover any recipe that can be used by an inspection tool to perform operations related to any type of mask inspection (including the embodiments described above).

[0141] It should be noted that the examples shown in the present disclosure, such as, for example, mask inspection tool architectures and configurations, mask types and / or layouts, illustrated image patches, specific noise types / sources and filters, and the optimization methods described above, are for illustrative purposes only and should not be considered to limit the present disclosure in any way. Other suitable examples / embodiments may be used in addition to or in place of the above.

[0142] One of the advantages of certain embodiments of the mask inspection process as described herein is the ability to detect defects of interest (DOI) on a mask before mass-producing wafers in a semiconductor foundry, with improved detection sensitivity, thus achieving a higher DOI capture rate while suppressing the FAR.

[0143] This is achieved at least by the ability to acquire overlapping images to increase the number of references, calculate specific filters customized for corresponding reference patches to effectively remove various noises in each reference patch, and combine multiple corrected reference patches into a composite reference patch that further eliminates noise, thereby obtaining a differential patch with improved SNR.

[0144] In addition, the above process is repeated for multiple overlapping inspection images that capture the same inspection area, and when making a decision on the presence of a DOI (e.g., generating a rank based on the scores of multiple differential patches), multiple results (i.e., multiple differential patches) are taken into account, which can further remove false positives and effectively improve detection sensitivity without adjusting the detection threshold. Other advantages are that processing small image patches instead of the complete inspection and reference images allows for more accurate reference correction, thus enabling higher-quality reference creation.

[0145] One of the advantages of certain embodiments of the mask inspection process as described herein is to apply multiple reference inspection processes as described above to a single-die mask, thereby enabling the detection of DOI on the single-die mask with increased detection sensitivity.

[0146] It should be understood that the present disclosure is not limited to its application to the details contained in the description or shown in the drawings herein.

[0147] It should also be understood that the system according to the present disclosure can be implemented at least in part on a suitably programmed computer. Similarly, the present disclosure contemplates a computer program that can be read by a computer for performing the method of the present disclosure. The present disclosure further contemplates a non-transitory computer-readable memory that tangibly embodies an instruction program executable by a computer for performing the method of the present disclosure.

[0148] The present disclosure can have other embodiments and can be practiced and carried out in various ways. Therefore, it should be understood that the expressions and terms used herein are for descriptive purposes and should not be regarded as restrictive. Thus, those skilled in the art will understand that the concepts on which the present disclosure is based can easily be used as a basis for designing other structures, methods, and systems to achieve several purposes of the presently disclosed subject matter.

[0149] Those skilled in the art will readily understand that various modifications and alterations can be applied to the embodiments of the present disclosure as described above without departing from the scope of the present disclosure defined and limited by the appended claims.

Claims

1. A computerized system for inspecting a mask that can be used to fabricate a semiconductor sample, the system including a processing and memory circuitry (PMC), the processing and memory circuitry (PMC) configured to: For an inspection area of the mask, obtain a plurality of inspection images having a plurality of fields of view (FOVs) that at least overlap with the inspection area, and for each inspection image, obtain a set of reference images, the set of reference images including a plurality of reference images for each corresponding reference area in one or more corresponding reference areas; Generate a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more candidate defects located in the inspection area of the corresponding inspection image, and align the one or more candidate defects of the corresponding inspection image to generate a list of defects of interest (DCI); For at least one given DCI in the list, generate a plurality of differential patches, wherein the PMC is configured to generate differential patches corresponding to each of the plurality of inspection images by: Extract image patches around the location of the given DCI from the inspection image and the set of reference images respectively, thereby generating an inspection patch and a set of reference patches; For each reference patch, calculate a filter that is optimized to minimize the difference between the inspection patch and a corrected reference patch obtained using the filter, thereby generating a set of filters and a set of corrected reference patches corresponding to the set of reference patches; And Combine the set of corrected reference patches to obtain a composite reference patch, and compare the inspection patch with the composite reference patch to obtain the differential patch; Calculate a score based on the plurality of differential patches, and apply a detection threshold to the score to determine whether the given DCI is a defect of interest (DOI).

2. The computerized system according to claim 1, wherein the mask is a multi-die mask, the inspection area is located in an inspection die on the mask, and the one or more reference areas are respectively from one or more reference dies of the inspection die on the mask.

3. The computerized system according to claim 1, wherein the mask is a single-die mask, and the inspection area and the one or more reference areas are from a single die on the mask and share the same design pattern.

4. The computerized system according to any one of the preceding claims, wherein the PMC is configured to register the inspection patch with each reference patch from the set respectively to correct a corresponding offset between the inspection patch and each reference patch before calculating the filter.

5. The computerized system according to any one of the preceding claims, wherein the filter is calculated to correct at least one of the following noises of the reference patch: registration residual, intensity gain and offset, defocus, or field of view (FOV) distortion.

6. The computerized system according to any one of the preceding claims, wherein the filter includes a set of filter components for correcting the corresponding noise of the reference patch.

7. The computerized system according to any one of the preceding claims, wherein the filter is calculated using least squares optimization.

8. The computerized system according to any one of the preceding claims, wherein the grade is calculated by: calculating a score for each of the plurality of differential compensation blocks based on the highest pixel value in the differential compensation block, thereby generating a plurality of scores corresponding to the plurality of differential compensation blocks, and averaging the plurality of scores to obtain the grade.

9. The computerized system according to any one of the preceding claims, wherein the PMC is further configured to perform the generating of the plurality of differential compensation blocks, calculating the grade, and applying a detection threshold for each DCI in the DCI list to determine whether the DCI is a DOI, and providing an updated defect map corresponding to the inspection area and including one or more DOIs detected by the determination.

10. The computerized system according to any one of the preceding claims, wherein the PMC is further configured to repeat the following for one or more additional inspection areas on the mask: the obtaining of the plurality of inspection images, the generating of the plurality of defect maps, the aligning of the one or more candidate defects, the generating of the plurality of differential compensation blocks, the calculating of the grade, and the applying of the detection threshold.

11. The computerized system according to any one of the preceding claims, wherein the plurality of inspection images are sequentially acquired by a lithographic inspection tool having a predefined step size, the lithographic inspection tool being configured to simulate the optical configuration of a lithography tool that can be used to fabricate the semiconductor sample.

12. The computerized system according to claim 11, further comprising the lithographic inspection tool.

13. The computerized system according to any one of claims 1 to 10, wherein the plurality of inspection images are obtained by: sequentially acquiring a plurality of images using a non-lithographic inspection tool having a predetermined step size, and performing a simulation on the plurality of images to simulate the optical configuration of a lithography tool that can be used to fabricate the semiconductor sample, thereby generating the plurality of inspection images.

14. The computerized system according to any one of the preceding claims, wherein the list of candidate defects of interest (DCI) includes one or more candidate defects common in at least a majority of the plurality of inspection images.

15. A computerized method for inspecting a mask that can be used to fabricate a semiconductor sample, the method being performed by a processing and memory circuitry (PMC) and the method comprising: For an inspection area of the mask, obtaining a plurality of inspection images having a plurality of fields of view (FOV) that at least overlap the inspection area, and for each inspection image, obtaining a set of reference images, the set of reference images including a plurality of reference images for each corresponding reference area in one or more corresponding reference areas; Generate a plurality of defect maps corresponding to the plurality of inspection images, each defect map including one or more candidate defects located in the inspection area of the corresponding inspection image, and align the one or more candidate defects of the corresponding inspection image to generate a list of candidate defects of interest (DCI); For at least one given DCI in the list, generate a plurality of differential patches, including generating differential patches corresponding to each of the plurality of inspection images by: Extract image patches around the location of the given DCI from the inspection image and the set of reference images respectively, thereby generating an inspection patch and a set of reference patches; For each reference patch, calculate a filter that is optimized to minimize the difference between the inspection patch and the corrected reference patch obtained using the filter, thereby generating a set of filters and a set of corrected reference patches corresponding to the set of reference patches; and Combine the set of corrected reference patches to obtain a composite reference patch, and compare the inspection patch with the composite reference patch to obtain the differential patch; Calculate a score based on the plurality of differential patches, and apply a detection threshold to the score to determine whether the given DCI is a defect of interest (DOI).

16. The computerized method according to claim 15, wherein the mask is a multi-die mask, the inspection area is located in an inspection die on the mask, and the one or more reference areas are respectively from one or more reference dies of the inspection area on the mask.

17. The computerized method according to claim 15, wherein the mask is a single-die mask, and the inspection area and the one or more reference areas are from a single die on the mask and share the same design pattern.

18. The computerized method according to any one of claims 15 to 17, further comprising registering the inspection patch with each reference patch from the group before calculating the filter to correct the corresponding offset between the inspection patch and each reference patch.

19. The computerized method according to any one of claims 15 to 18, wherein the filter is calculated to correct at least one of the following noises of the reference patch: registration residual, intensity gain and offset, defocus, or field of view (FOV) distortion.

20. The computerized method according to any one of claims 15 to 19, wherein the filter includes a set of filter components for correcting the corresponding noise of the reference patch.

21. The computerized method according to any one of claims 15 to 20, wherein the filter is calculated using least squares optimization.

22. The computerized method according to any one of claims 15 to 21, wherein the score is calculated by: calculating a score for each of the plurality of differential patches based on the highest pixel value in the differential patches, thereby generating a plurality of scores corresponding to the plurality of differential patches, and averaging the plurality of scores to obtain the score.

23. The computerized method according to any one of claims 15 to 22, further comprising performing said generating of a plurality of differential fill blocks, calculating a rank, and applying a detection threshold for each DCI in said DCI list to determine whether the DCI is a DOI, and providing an updated defect map corresponding to said inspection area and including one or more DOIs detected by said determination.

24. The computerized method according to any one of claims 15 to 23, further comprising repeating, for one or more additional inspection areas on said mask: said obtaining of a plurality of inspection images, said generating of a plurality of defect maps, said aligning of said one or more candidate defects, said generating of a plurality of differential fill blocks, said calculating of a rank, and said applying of a detection threshold.

25. The computerized method according to any one of claims 15 to 24, wherein said plurality of inspection images are sequentially acquired by a lithographic inspection tool having a predefined step size, the lithographic inspection tool being configured to simulate the optical configuration of a lithographic tool that can be used to fabricate said semiconductor sample.

26. The computerized method according to any one of claims 15 to 24, wherein said plurality of inspection images are obtained by: sequentially acquiring a plurality of images using a non-lithographic inspection tool having a predetermined step size, and performing a simulation on said plurality of images to simulate the optical configuration of a lithographic tool that can be used to fabricate said semiconductor sample, thereby generating said plurality of inspection images.

27. The computerized method according to any one of claims 15 to 26, wherein said list of candidate defects of interest (DCI) includes one or more candidate defects common to at least a majority of said plurality of inspection images.

28. A non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium tangibly embodying an instruction program, the instruction program when executed by a computer causes the computer to perform the computerized method according to any one of claims 15 to 27.

Images