Data processing method, charged particle evaluation method and system

Abstract

A data processing method for image data obtained by scanning a charged particle beam across a sample; the method comprises the following steps: classifying pixels of the image data into foreground pixels and background pixels to generate a foreground pixel map; and encoding the foreground pixel map into encoded data using a sparse matrix encoding technique.

Description

Cross-references to related applications

[0001] This application claims priority to EP application 23188908.0, filed on August 1, 2023, and EP application 24169155.9, filed on April 9, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The embodiments provided herein generally relate to data processing methods, charged particle evaluation systems, and methods for operating charged particle evaluation systems. Background Technology

[0003] In the manufacture of semiconductor integrated circuit (IC) chips, undesirable pattern defects inevitably occur on the substrate (i.e., wafer) or mask during the fabrication process due to factors such as optical effects and random particles, thereby reducing yield. Therefore, monitoring the extent of these undesirable pattern defects is a crucial process in IC chip manufacturing. More generally, the evaluation (e.g., inspection and / or measurement) of the surface of the substrate or other objects / materials before, during, and / or after surface fabrication steps is an important process.

[0004] Pattern inspection devices with at least one beam of charged particles have been used to inspect objects (which may be referred to as samples), for example, to detect pattern defects. These devices typically employ electron microscopy techniques, such as scanning electron microscopy (SEM). In some SEMs, a primary electron beam of relatively high-energy electrons targets a final deceleration step to land on the sample with a relatively low landing energy. The electron beam is focused onto the sample as a probe spot. The interaction between the material structure at the probe spot and the landing electrons from the electron beam causes signal electrons, such as secondary electrons, backscattered electrons, or Auger electrons, to be emitted from the surface. Signal electrons can be emitted from the material structure of the sample. By scanning the primary electron beam as a probe spot across the sample surface, signal electrons can be emitted across the sample surface. By collecting these emitted signal electrons from the sample surface, the pattern inspection device can obtain an image representing the properties of the material structure of the sample surface. Evaluation systems exist with detectors (with a single detector or an array of many detector elements) and / or with a single beam or tens, hundreds, or thousands of beams.

[0005] When such evaluation devices (such as pattern inspection devices for detecting defects) are used to evaluate samples at high throughput, for example, image data is generated at very high rates through many detector elements. Processing this data requires significant computing power, data storage, and / or communication bandwidth. Summary of the Invention

[0006] The purpose of this disclosure is to provide embodiments that reduce the requirements for computing power, data storage, and / or communication bandwidth in processing image data from evaluation devices, such as pattern inspection devices.

[0007] According to a first aspect of the present invention, a data processing method for image data obtained by scanning a charged particle beam across a sample is provided; the method includes: The image data pixels are classified into foreground pixels and background pixels to generate a foreground pixel map; and The foreground pixel image is encoded into coded data using sparse matrix coding techniques.

[0008] According to a second aspect of the present invention, an evaluation method is provided, comprising: Guide the charged particle beam onto the sample; A digital detection signal that generates pixel values ​​in response to signal particles from a sample; Scanning samples and charged particle beams; and Use the above method to process digital detection signals.

[0009] According to a third aspect of the present invention, a data encoding apparatus is provided for encoding image data obtained by scanning a charged particle beam across a sample, the apparatus comprising: A classifier is used to classify pixels in image data into foreground pixels and background pixels to generate a foreground pixel map; and An encoder is used to encode a foreground pixel image into encoded data using sparse matrix coding techniques.

[0010] According to a fourth aspect of the present invention, a data comparison apparatus for detecting candidate defects in a sample image is provided, the apparatus comprising: An interface for receiving encoded sample data representing a foreground pixel map encoded using sparse matrix coding techniques; and A comparator is used to compare coded sample data with coded reference data, which represents a foreground pixel map derived from a reference image and encoded using sparse matrix coding techniques.

[0011] According to a fifth aspect of the present invention, a method for processing detection data from the sample surface of a detector element of a detector array is provided, the method comprising: Up-scanning charged particle beams across the sample region The signal particles are detected using one or more detector elements of a charged particle detector array to obtain detection data; Pixels that receive detection data from the data stream from the detector element; In the data stream, the pixels of the detected data are classified into foreground pixels and background pixels to generate a foreground pixel dataset representing the scanned sample area.

[0012] According to a sixth aspect of the present invention, a method for processing detection data from the sample surface of a detector element of a detector array is provided, the method comprising: Scanning charged particle beams across the sample region; Detecting signal particles using one or more detector elements of a charged particle detector array to obtain detection data; pixels receiving detection data from a data stream from the detector elements; In the data stream, the pixels of the detected data are classified into two or more different categories to generate a pixel dataset of the scanned sample region representing pixels of at least one category. Attached Figure Description

[0013] The above and other aspects of this disclosure will become more apparent from the description of exemplary embodiments in conjunction with the accompanying drawings.

[0014] Figure 1 This is a schematic diagram illustrating an exemplary charged particle beam inspection system.

[0015] Figure 2 It is illustrated as Figure 1 A schematic diagram of an exemplary multi-beam charged particle evaluation apparatus, which is part of an exemplary charged particle beam inspection system.

[0016] Figure 3 This is a schematic diagram illustrating an exemplary arrangement with a detector module and a data path for transmitting data from the detector module to a data processing device, such as an external charged particle evaluation device.

[0017] Figure 4 The illustration shows the classification process of an embodiment.

[0018] Figure 5 The illustration shows a comparative process of an embodiment.

[0019] Figure 6 This is a schematic diagram of a data processing system according to an embodiment.

[0020] The schematic diagrams and views illustrate the components described below. However, the components depicted in the accompanying drawings are not drawn to scale. Detailed Implementation

[0021] Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings, wherein, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The implementations set forth in the following description of the exemplary embodiments do not represent all implementations consistent with the invention. Rather, they are merely examples of apparatuses and methods consistent with the aspects of the invention listed in the appended claims.

[0022] The increased computing power of electronic devices can be achieved by significantly increasing the packaging density of circuit components such as transistors, capacitors, and diodes on IC chips, which reduces the physical size of the device. This is achieved by increasing resolution, enabling the manufacture of smaller structures. For example, the IC chip for a smartphone (this is the size of the thumbnail and was available in 2019 or earlier) could include more than 2 billion transistors, each less than 1 / 1000th the size of a human hair. Therefore, it is not surprising that semiconductor IC manufacturing is a complex and time-consuming process with hundreds of individual steps. Even an error in one step can significantly impact the operation of the final product. The goal of a manufacturing process is to increase the overall yield of the process. For example, to achieve a 75% yield in a 50-step process (where one step can indicate the number of layers formed on the wafer), the yield of each individual step must be greater than 99.4%. If the yield of each individual step is only 95%, the total process yield will be as low as 7%.

[0023] While high process yields are required in IC chip manufacturing facilities, maintaining high substrate (i.e., wafer) throughput (defined as the number of substrates processed per hour) is also essential. The presence of defects can impact both high process yields and high substrate throughput. This is especially true if reviewing defects requires operator intervention. Therefore, detecting and identifying micron- and nanometer-scale defects at high throughput using inspection equipment such as scanning electron microscopy ('SEM') is important for maintaining high yields and low costs.

[0024] SEM embodiments include scanning devices and detector devices. The scanning device includes an illumination device comprising an electron source for generating primary electrons (or charged particles) and a projection device for scanning a sample, such as a substrate, using one or more focused primary electron beams (or multiple beams). Multiple beams can be arranged as a multi-beam system. The multiple beams can have a multi-beam arrangement (it can be in the form of a grid) that can be guided towards the sample along a multi-beam path. At least the illumination device or system and the projection device or system together can be referred to as an electron optics system or apparatus. Primary electrons interact with the sample and generate secondary electrons. The detection device captures these secondary electrons from the sample as it is scanned, allowing the SEM to create an image of the scanned area of ​​the sample. For high-throughput inspection, some inspection devices use multiple focused primary electron beams, i.e., multi-beam systems. Component beams within the multi-beam system can be referred to as sub-beams or beam waves. Multiple beams can scan different portions of the sample simultaneously. Therefore, multi-beam inspection devices can inspect samples at much higher speeds than single-beam inspection devices.

[0025] The following describes the known implementation methods of multi-beam inspection devices.

[0026] Although the description and figures relate to electron optical systems, it is to be understood that the embodiments are not intended to limit this disclosure to specific charged particles. Therefore, references to electrons throughout this document can be considered more generally as references to charged particles, which are not necessarily electrons.

[0027] Now for reference Figure 1 , Figure 1 This is a schematic diagram illustrating an exemplary charged particle beam evaluation system 100 (or simply evaluation system), which may also be referred to as a charged particle beam inspection system or simply an inspection system or a charged particle measurement system or simply a measurement system. Figure 1 The charged particle beam evaluation system 100 includes a main chamber 10, a loading and locking chamber 20, an electron beam system 40, an equipment front-end module (EFEM) 30, and a controller 50. The electron beam system 40 is located within the main chamber 10.

[0028] EFEM 30 includes a first loading port 30a and a second loading port 30b. EFEM 30 may include multiple additional loading ports. The first loading port 30a and the second loading port 30b may, for example, house a front-opening substrate transfer box (FOUP) containing a substrate (e.g., a semiconductor substrate or a substrate made of (a variety of) other materials) or a sample (substrate, wafer, and sample are collectively referred to as “sample”). One or more robotic arms (not shown) in EFEM 30 transport the sample to the loading locking chamber 20.

[0029] Loading lock chamber 20 is used to remove gas from the area surrounding the sample. This creates a vacuum with a local pressure lower than the ambient pressure. Loading lock chamber 20 can be connected to a loading lock vacuum pump system (not shown), which removes gas particles from loading lock chamber 20. Operation of the loading lock vacuum pump system allows the loading lock chamber to reach a first pressure below atmospheric pressure. After reaching the first pressure, one or more robotic arms (not shown) transport the sample from loading lock chamber 20 to main chamber 10. Main chamber 10 is connected to a main chamber vacuum pump system (not shown). The main chamber vacuum pump system removes gas particles from main chamber 10, causing the pressure around the sample to reach a second pressure below the first pressure. After reaching the second pressure, the sample is transported to an electron beam system, through which the sample can be examined. Electron beam system 40 may include a multi-beam electron optics unit (which may be referred to as an electron optics array or (occasionally) an electron optics device).

[0030] The controller 50 is electronically connected to the electron beam system 40. The controller 50 may be a processor (such as a computer) configured to control the charged particle beam evaluation device 100. The controller 50 may also include a processing circuit system (which may be referred to as a processing circuit system) configured to perform various signal, data, and image processing functions. Although the controller 50 is... Figure 1 The structure is shown external to the main chamber 10, the loading and locking chamber 20, and the EFEM 30; however, it should be understood that the controller 50 may be part of the structure. The controller 50 may be located in one of the components of the charged particle beam inspection apparatus, or it may be distributed across at least two components. While this disclosure provides an example of a main chamber 10 housing the electron beam system, it should be noted that the broadest aspects of this disclosure are not limited to the chamber housing the electron beam system. Rather, it should be understood that the foregoing principles can also be applied to other equipment and other arrangement of devices operating under a second pressure. For example, the controller may be a distributed controller having one or more portions under the second pressure, and optionally one or more portions under the first pressure.

[0031] Now for reference Figure 2 , Figure 2 This is a schematic diagram illustrating an exemplary electron beam system 40, including a multi-beam electron optical system 41 (or electron optical device), which is Figure 1This is part of an exemplary charged particle beam evaluation apparatus 100. The electron beam system 40 includes an electron source 201 and a projection device 230. The electron beam system 40 also includes an actuation platform 209 and a sample holder 207. The electron source 201 and the projection device 230 may be collectively referred to as an electron optical system 41 or an electron optical array. The sample holder 207 is supported by the actuation platform 209 to hold a sample 208 (e.g., a substrate or mask) for inspection. The multi-beam electron optical system 41 also includes a detector 240 (e.g., an electron detection device). It should be noted that while an electron beam system including a multi-beam electron optical system has been described, this is only one embodiment of the invention. The invention disclosed herein can be applied to electron beam systems having a single-beam electron optical system, for example, for projecting a single beam of electrons toward a sample.

[0032] The electron source 201 may include a cathode (not shown) and an extractor or anode (not shown). During operation, the electron source 201 is configured to emit electrons from the cathode as primary electrons. The primary electrons are extracted or accelerated by the extractor and / or anode to form a primary electron beam 202.

[0033] Projection device 230 is configured to convert primary electron beam 202 into multiple sub-beams 211, 212, 213 and guide each sub-beam onto sample 208. Although three sub-beams are illustrated for simplicity, there may be tens, hundreds, thousands (e.g., approximately ten thousand), tens of thousands, or hundreds of thousands of sub-beams. A greater number of sub-beams may exist in charged particle beam system 40, which includes multiple charged particle optical systems 41; each charged particle optical system can project many sub-beams toward the sample, such as thousands of beams. Sub-beams may be referred to as beam waves. Charged particle beam system 40, which includes multiple charged particle optical systems 41, may be referred to as a multi-column system.

[0034] Controller 50 can be connected to Figure 1 The charged particle beam evaluation apparatus 100 comprises various components, such as an electron source 201, a detector 240, a projection device 230, and an actuation platform 209 (e.g., a motorized platform). A controller 50 can perform various image and signal processing functions. The controller 50 can also generate various control signals to manage the operation of the charged particle beam inspection apparatus, including a multi-beam charged particle device.

[0035] Electron optics 230 can be configured to focus sub-beams 211, 212, and 213 onto sample 208 for inspection, and can form probe spots 221, 222, and 223 (three probe spots for this example, one for each sub-beam) on the surface of sample 208. Projection device 230 can be configured to deflect primary sub-beams 211, 212, and 213 to scan probe spots 221, 222, and 223 across segments of the surface of sample 208. In response to primary sub-beams 211, 212, and 213 incident on probe spots 221, 222, and 223 on sample 208, electrons (which may be referred to as signal particles) comprising secondary electrons and backscattered electrons are generated from sample 208. The energy of the secondary electrons is typically less than or equal to 50 eV. Practical secondary electrons can have energies less than 5 eV, but any energy below 50 eV is typically handled at the secondary electron level. The electron energies of backscattered electrons are typically between 0 eV and the landing energies of the primary sub-beams 211, 212, and 213. Since electrons with detected energies less than 50 eV are generally considered secondary electrons, a certain percentage of actual backscattered electrons will be counted as secondary electrons.

[0036] Detector 240 is configured to detect signal particles, such as secondary electrons and / or backscattered electrons, and generate corresponding signals, which are sent to signal processing system 280 for preprocessing, such as analog-to-digital conversion. (Signal processing device 280 may be distributed across multiple different devices, which may be collectively referred to as the signal processing system.) Detector 240 may be incorporated into projection device 230. Further details and alternative arrangements of detector modules, sensors, and detector arrays positioned proximate to, upstream of, downstream of, or otherwise integrated into the objective lens can be found in EP application 20216890.2 and PCT application PCT / EP2021 / 068548, which are incorporated herein by reference, at least if they disclose details of detector modules, sensors, detector arrays, and similar elements.

[0037] The detector can be provided with multiple parts, more specifically multiple detection parts (such as detector elements). A detector comprising multiple parts can be associated with one of sub-beams 211, 212, 213. Thus, multiple parts of a detector 240 can be configured to detect signal particles emitted from sample 208 with respect to one of the primary beams (otherwise, they could be referred to as sub-beams 211, 212, 213). In other words, a detector comprising multiple parts can be associated with one of the apertures in at least one electrode of the objective lens assembly. The multiple parts can be segments arranged radially and / or at an angle. More specifically, a detector comprising multiple parts can be arranged around a single aperture, providing an example of such a detector.

[0038] As mentioned, the detection signal from the detector module is used to generate an image. Utilizing multiple detection sections, the detection signal includes components from different detection signals, which can be processed as a dataset or as a detection image. A signal conduit (or data path) may exist between the detector and an external device (such as controller 50 or another processor). The signal conduit can be used for data streaming. Signal processing device 280 may be included within the signal conduit. Many different detection signals can be sent through the signal conduit, which may correspond to the data signal of each detection section. The data signals together constitute a data stream. Considering that there may be many sub-beams, or even thousands of sub-beams from different charged particle optical systems, in a charged particle beam system 40 including multiple charged particle optical systems 41, the data load generated by the detector may be large. Note: In embodiments with a single-beam charged particle system, the detector may include multiple detector sections, each outputting a corresponding detection signal. Further, in embodiments, the charged particle beam system 40 may include multiple single-beam charged particle optical systems.

[0039] An objective lens may be an array of objectives and may include multiple planar electrodes or plates comprising apertures for respective paths of the multiple beams. Each plate may extend across the multiple beams. An objective lens (e.g., as an array of objectives) may include at least two electrodes that can be connected to and controlled by a corresponding potential. Additional plates may be present to control additional degrees of freedom separately. A detector may be a plate that is part of, associated with, or connected to an objective lens having an aperture for each path of the multiple beams. The detector may be located above, below, or inside the objective lens (or particularly the array of objectives).

[0040] A scanning deflector can be associated with or even integrated into an objective lens, for example, as a scanning deflector array. A scanning deflector array can be referred to as a deflector array, and can be integrated into an objective lens array, for example. In an arrangement, the scanning deflector can be positioned upstream of the objective lens. In an arrangement where the path of the primary beam is collimated upstream of the objective lens, the scanning deflector can be positioned upstream of the objective lens. In an arrangement where multiple beams are generated from a collimated primary beam from a source through a beam-limiting aperture array of the objective lens or a beam-limiting aperture array associated with the objective lens, the scanning deflector can be a macro-scanning deflector positioned upstream of the objective lens for operation with the collimated primary beam. Other electro-optical arrangements including one or more elements described herein are conceivable. Such a scanning deflector can be controlled by a controller to deflect one of the multiple beams along an axis in the sample plane or two primary axes on the sample surface, for example, in the sample plane (which may be orthogonal to each other).

[0041] It should be noted that embodiments of scanning deflectors located close to the sample (e.g., integrated into or near the objective lens) may have a limited scanning deflection range. However, scanning deflectors located close to the sample can be precisely controlled and have a fast response relative to other types of scanning actuators, such as actuation platforms.

[0042] Controller 50 can control actuation platform 209 to move sample 208 during inspection. Controller 50 can enable actuation platform 209 to move sample 208 in at least one direction during sample inspection, preferably continuously, for example, at a constant speed, which can be referred to as a scan. The speed of the actuation platform can be referred to as the rate of movement. Controller 50 can control the movement of actuation platform 209 such that it varies the speed of movement of sample 208 relative to the multi-beam path depending on various parameters. Controller 50 can control the deflection of the scan deflector such that the multi-beam path moves relative to the platform, thereby moving across the sample surface. Controller 50 can vary the beam deflection of the scan deflector depending on various parameters, thereby changing the scan of the beams on the sample surface. For example, controller 50 can control the platform speed (including its direction) and / or the scan deflector depending on, for example, the characteristics of the inspection scanning elements and steps in the scanning process, as disclosed in EPA 21171877.0, filed May 3, 2021. This application is incorporated herein by reference as long as it discloses at least the combined stepping and scanning strategy of the platform and scan deflector. Therefore, the movement rate may include stepping frequency and / or platform scan rate at different times.

[0043] To obtain a two-dimensional image of the sample, the probe beam (or each one) is scanned across the sample surface in a two-dimensional grating pattern. This involves movement of the beam in two directions: a primary scan direction and a secondary scan direction, which have different orientations, such as orthogonal ones. The primary scan direction can be referred to as the fast scan direction. For example, a scan deflector can be controlled to actuate the primary beam (e.g., multiple beams) in the fast direction. The secondary scan direction can be referred to as the slow scan direction. For example, a platform can be controlled to actuate the sample relative to the path of the primary beam; a scan deflector can be controlled to actuate the path of the beam on the sample surface in the slow direction; or the platform and scan deflector can be actuated to achieve scanning in the slow direction. The platform is preferably used only in the slow direction because the inertia due to its large mass makes achieving scanning movement (e.g., acceleration) in the fast direction more challenging than alternatives such as scan deflectors.

[0044] Known multi-beam systems (such as the electron beam system 40 and charged particle beam evaluation apparatus 100 described above) are disclosed in US2020 / 118784, US202002 / 03116, US2019 / 0259564 and WO2021078352, which are last incorporated herein by reference, as long as they are electron beam systems and / or charged particle beam evaluation apparatuses.

[0045] The electron beam system 40 may include a projection component to adjust the accumulated charge on the sample 208 by irradiating the sample 208.

[0046] Data output from the charged particle evaluation device (e.g., electron beam system 40) is expected to be automatically processed by a data processing device to detect defects in the sample being evaluated. The data processing device or at least a part thereof may be part of controller 50, part of another computer in the wafer fab, or elsewhere integrated into the charged particle evaluation device.

[0047] In the charged particle inspection system, the electron-optical system 41, or each electron-optical system 41, is located within the main (vacuum) chamber 10. Therefore, it is necessary to transmit data from the electron-optical system(s) to devices or other components outside the vacuum chamber, such as the controller 50. The raw data generated by the electron-optical system(s) can be transmitted out of the vacuum chamber via a signal conduit. Considering the amount of data, the data can be transmitted optically, for example, using a signal conduit with optical capabilities, such as an optical fiber. An optical transceiver is used to convert the electrical signals representing the data into optical signals. The optical transceiver is located near the detector module 240 of the electron-optical system 41. The optical transceiver is configured to convert the electrical signals output from the detector module 240 into optical signals for transmission along the optical fiber. The optical fiber may be capable of transmitting multiple channels simultaneously (e.g., using different wavelengths). Therefore, the detection signal from each individual electrode of the detector module is converted into an appropriate number of data streams. Multiple optical fibers can be used, whether single-channel or multi-channel.

[0048] The signal conduit, or each signal conduit (e.g., optical fiber), passes through the wall of the main chamber 10 via a vacuum feedthrough 11 (in use, the main chamber 10 is under vacuum). A suitable vacuum feedthrough is described in US2018 / 0182514 A1, which is incorporated herein by reference at least to the extent that it relates to the feedthrough device. The signal conduit (e.g., optical fiber) is connected to the data processing device 50, so that the data processing device 50 can be located outside the vacuum for easy access and to avoid the need to increase the size of the vacuum chamber to accommodate the data processing device. In single-row and multi-row systems, multiple optical transceivers and multiple optical fibers per row can be used, if convenient.

[0049] exist Figure 3 The diagram illustrates an arrangement involving multiple optical fibers (or other communication channels). Each optical fiber 242 is connected to a subunit (or detector unit) 241 of the detector module 240 and transmits the signal generated by the corresponding subunit 241 to a data processing device 50 outside the main chamber 10. Each subunit 241 may include one or more individual electrodes of the detector module 240. It will be seen that a large number of optical fibers 242 are required in a multi-row system capable of imaging most of the sample 208 at once, therefore the vacuum feedthrough 11 must be large. Besides the space required, a large vacuum feedthrough or multiple smaller feedthroughs may be difficult to seal.

[0050] Various methods can be employed to detect defects in images generated by charged particle evaluation devices. A common approach is to compare an image of a portion of a sample (referred to herein as a sample image) with one or more reference images. In practice, data points representing the data stream of the sample image are compared with data points of the reference images retrieved from memory or delivered as parallel data streams. For simplicity, this process can be referred to below as image comparison; the data points can be called pixels. As mentioned above, high-throughput evaluation devices, such as multi-beam or multi-column evaluation devices (i.e., generating many simultaneously detected signals), generate data at very high rates, making pixel-by-pixel comparisons require very high-bandwidth data communication paths to obtain data from the evaluation device and / or significant computational resources to perform the comparisons. A two-stage defect detection process has been proposed, in which an initial comparison is performed on the sample data to locate candidate defects, and then a “clip” (an image data region surrounding the candidate defect) is sent to an external data processing system for more detailed analysis. However, even this process requires significant computational resources to perform the initial comparison in a high-throughput evaluation device.

[0051] More generally, in charged particle evaluation apparatuses, particularly those using multiple charged particle beams, it is typically desirable to perform some initial data processing on the large volume of data generated. For example, it may be desirable to reduce noise, compress and / or encode the data to reduce bandwidth and / or computational load. This initial data processing can be performed entirely or partially within the charged particle evaluation apparatus (e.g., within the vacuum chamber of the charged particle evaluation apparatus, or even within the detector module of the charged particle evaluation apparatus). This processing can be inline, i.e., optionally applied unbuffered to the detection signal, such as in an external processor or data path outside the charged particle evaluation system, such as a remote data rack, from the detector module (e.g., from the detector or even the detector element) to other places within the charged particle apparatus.

[0052] The following is for reference. Figures 4 to 6 The method for evaluating samples according to the embodiments is described. At a higher level, a two-stage process is performed.

[0053] exist Figure 4 In the first stage described, a classification process is performed on the sample image output by the detector in the charged particle evaluation device at the local level of detection. The sample image represents a portion of the sample being evaluated. The classification process identifies regions of different categories in the dataset (e.g., as in the sample image), such as foreground and background regions of the sample image, and generates a pixel map of one category, such as a foreground pixel map. More generally, the classification process can generate a sample dataset of one category, such as a foreground pixel dataset. The sample dataset can have a structure that does not directly correspond to a 'graph'; where the graph directly provides information about the relative positions of the foreground regions.

[0054] Then, the foreground pixel map can undergo a first step to create coded data, which is a compressed representation of the sample image. This first step can be either an encoding or compression step. It can be considered a data encoding step. The coded data can correspond to features in the data. This encoding can be considered feature encoding. The first step can be implemented using sparse matrix coding techniques. The coded data is an efficient compressed representation of the sample image. That is, a high compression ratio is applied to the data to make it a compressed representation of the sample image. The reference image also undergoes the same classification and encoding process, which is not necessarily detector-local, nor necessarily performed simultaneously with the generation of the sample image.

[0055] The first step is as follows: Figure 4 The diagram depicts a multi-valued sample image SI and a reference image RI, transformed by a classifier CL into a two-level sample image SM and a two-level reference image RM, respectively, as compressed representations. Note that a level can be referred to as a 'category'; the classification may depend on factors other than the magnitude of the region parameter or other factors instead. Each two-level image represents a foreground and background region (e.g., pixels) and can be referred to as a binary image, with each pixel having one bit. Note that in the reference... Figure 4 The use of two levels in the illustrated and described embodiments is exemplary. As many levels (or categories) as needed can be used, such as three or more, five or more, and / or in the range of two to ten levels. The number of levels can correspond to different types of features in the encoded data that need to be readily apparent (i.e., represented by corresponding pixels in such different regions). In the case of a graph with three or more levels, foreground pixels can typically be of two or more categories; that is, pixels can be classified as background pixels and at least two different categories of foreground pixels (or foreground pixel categories).

[0056] Additionally or alternatively, there may be more than one background category of pixels. Each of the multiple different levels may correspond to a feature of a category, and each category may be represented by a corresponding dataset (e.g., a graph). These categories may have the same hierarchical structure, such that these categories are not classified as background pixels and foreground pixels. Alternatively, each category may have a hierarchical structure, such that each category has the same hierarchical structure as foreground pixels and / or background pixels. In addition to foreground or background pixels with different categories, there may be other categories, such as mid-ground pixels. Therefore, in an arrangement, there may be one or more categories with the same (or higher) hierarchical structure as foreground pixels and / or background pixels, such as in more mid-ground pixels. Pixels of different categories may be classified into a category within a higher-level hierarchical category. A class of pixels within a higher-level hierarchical category may be referred to as a subclass.

[0057] Figure 4The first-stage processing described herein can be considered a form of feature encoding, where different types of features are converted into data structures with a lower density than the original detection signal. Therefore, this processing can occur without applying any defect detection processing. The encoded data provides the data for later identification of the detection. This differs from another process that processes the detection signal to initially identify candidate defects, in which only data related to the candidate defects (e.g., image clips) needs to be sent from the charged particle evaluation device for further analysis. The first-stage process aims to reduce the amount of data that needs to be sent and / or processed to identify defects.

[0058] In any case, by applying the present invention to the embodiments, the amount of data output from the charged particle evaluation device is greatly reduced by reducing the data structure, and it is expected that even if it cannot be eliminated, the throughput limitation of the data transmission rate can be reduced.

[0059] In the second stage, such as Figure 5 As depicted, the comparator CP compares the compressed representations SM and RM of a sample image and a reference image to identify defects in the sample image. In embodiments, the comparison process may include subtracting the compressed representations of the sample and reference images to generate a differential image DI. However, any suitable comparison algorithm can be used. The comparison algorithm may involve operations such as subtraction, XOR, summation, and / or stitching. If pixels of multiple classes are used, the comparison algorithm can compare a dataset associated with one class of sample images with the corresponding datasets of the reference images(s). If pixels of multiple classes are used, the comparison algorithm can compare each dataset of the sample images with multiple (e.g., all) datasets of the reference images(s). The differential image DI may be filtered and / or thresholded by a filter FI to eliminate false positives and obtain a candidate defect file CDF.

[0060] More specifically, the first-stage classification step simply categorizes pixels into an initial number of categories. In this example, for simplicity, two categories are referenced: foreground pixels and background pixels. While two categories are referenced, specifically foreground and background, it should be understood that this step can be applied to any number of selected categories (such as categories with a higher hierarchical structure), foreground and background. Therefore, the classification step can reduce the data output to a binary bitmap as a foreground pixel map, which in itself represents a significant reduction in data volume.

[0061] The classification of the selected categories (e.g., foreground and background) may depend on the nature of the pattern being examined. For example, foreground pixels might be pixels located at the positions of lines or contact holes, while other pixels are background pixels. Typically, foreground pixels can represent the selected type of topography, surface material, or any other characteristic of the sample being evaluated that can be distinguished in the detection signal. For example, one type of foreground pixel could be a so-called "soft defect," i.e., a portion of the substrate where the surface parameters (e.g., height) represented by the detector intensity signal fall between the expected values ​​of the feature and the background. Soft defects can represent features that are not formed correctly. An example of a soft defect is a "bridging defect," where an undesired structure forms between two desired features.

[0062] It's important to understand that in some cases, a foreground pixel can be a pixel with a high detector intensity signal, but in others, depending on the type of feature considered as foreground, a foreground pixel will be a pixel with a low detector intensity signal. Therefore, when using a threshold to classify pixels as foreground or background, in different evaluation schemes, a foreground pixel can be a pixel with a detector signal below the threshold, or it can be a pixel with a detector signal above the threshold.

[0063] In some cases, foreground pixels can have several different categories, which may be referred to as subclasses. For example, more than one threshold may be used to identify different categories of foreground pixels. In an embodiment, foreground pixels of different categories are combined into a single multi-value map as a foreground pixel map (or more generally, a foreground pixel dataset). In an embodiment, a separate binary map is created for pixels of each category, such that the foreground pixel map comprises multiple binary maps. Whether one multi-value map or several binary maps is preferred may depend on the number of maps and the number of foreground pixels in each map, as well as the encoding and comparison techniques used. Desiredly, the number of foreground categories is small, for example, less than 15 (per pixel). Figure 4 (bits), expected to be 7 or fewer (per pixel) Figure 3 (bits), expected to be 3 or fewer (per pixel) Figure 2 (bits). Conveniently, foreground pixels can be represented as binary "1" and background pixels can be represented as binary "0", but the opposite signs are also useful.

[0064] Encoding the foreground pixel map can further significantly reduce the amount of data used to represent it. This allows for managing larger detection signal data loads without increasing the processing power and memory requirements for managing detection data in systems with many individual detection elements, each generating a different detection signal. Desiredly, lossless encoding techniques are preferred. Suitable techniques include: Key dictionary Nested Lists Coordinate list Compressed sparse lines Compressed sparse columns Run-length encoding

[0065] Typically, the ratio of foreground pixels to total pixels is low before classification. For example, the number of foreground pixels may be less than the square root of the total number of pixels. Therefore, a small number of foreground pixels may be selected, resulting in a low ratio of foreground pixels to total pixels. A sparse matrix or system is a matrix or system that can utilize the percentage and / or distribution of zero elements. In this application, the proportion of data in which defects are detected is expected to be very small. In some use cases, almost no defects, or even none at all, may be detected. Therefore, it is expected that only a small number of foreground pixels (only a small percentage of the total number of pixels) or only a small number of pixels from another category will be defective. For example, the ratio of defective pixels may be as low as about 1 / 10. 8 (One in a hundred million).

[0066] Sparse matrix coding is particularly advantageous when the proportion of foreground pixels is low, and can provide a very high compression level in the compressed representation of foreground pixel maps. Specifically, if foreground pixels of only one class or multiple classes are encoded as separate binary bitmaps, sparse matrix coding can be further reduced by encoding only the positions of the foreground pixels, without needing to encode the value of each pixel. The encoding technique can be selected or adapted based on the pattern to be evaluated and defined in the evaluation scheme. For example, for a pattern of contact holes, foreground pixels may be expected to exist in clusters of predictable shapes, so only the positions of the clusters and their deviations from the predicted shapes need to be encoded. Patterns with predominantly rectangular features can be efficiently encoded by sending the relative corners of a rectangular shape or the positions of a corner and its dimension.

[0067] Classifying detector signals into foreground and background pixels can be performed using a comparator that compares the detector signal pixel-by-pixel to a threshold. The threshold can be predetermined, for example, as part of an evaluation scheme for a specific device pattern. Multiple thresholds can be employed when using multiple foreground categories. In some cases, classification may consider additional factors beyond the detection signal value corresponding to a single pixel. For example, neighboring pixel values ​​may also be considered.

[0068] For example, denoising techniques can be applied to the foreground pixel map before or during the encoding process. Denoising techniques can eliminate isolated foreground pixels. In most applications, the features of interest are larger than a single pixel, so isolated pixels are more likely to be noise than features. Reducing (if not eliminating) noise before or during the encoding process can further reduce the amount of data to be transmitted.

[0069] The classification and encoding steps are suitable for implementation using dedicated hardware, such as FPGAs or ASICs. This dedicated hardware may be more efficient and economical than programmed general-purpose computing devices (e.g., CPUs). Some parts of the process can be performed by simple electronic circuitry that can be placed in the signal processing substrate of the detector module. This electronic circuitry may include analog elements, such as variable threshold circuitry. The processor of the dedicated hardware can be used as intended; while some functionality of the general-purpose computing device (e.g., CPU) may not be needed for the intended processing and may be unused and redundant. Dedicated hardware such as FPGAs or ASICs can be integrated into the detector circuitry system or can be placed in the main (vacuum) chamber near the electron optics column. This detector circuitry system can be located within the detector. This processing, implemented by the circuitry system, can occur within the detector. In this way, the amount of data to be sent outside the main chamber can be significantly reduced, for example, by one or more orders of magnitude. This significant reduction in data volume results in a corresponding significant reduction in the number of signal conduits and the size of the vacuum feedthrough. Conversely, the data reduction can increase the number of detector elements (and optionally, the number of data signals from the detector and / or signal conduits, such as before the feedthrough) without increasing the number of feedthroughs or the effective size. The classification and encoding steps can be performed in parallel using numerous dedicated circuits within the image data, enabling classification and encoding to be performed at high rates, such as matching the detector's data generation rate. This reduces or eliminates the need for data buffering. However, as discussed below, in some cases, some short-term buffering of the detector signal may be desirable.

[0070] It's important to understand that the classification and coding of reference images can be performed offline prior to the evaluation process, thus there are no particular constraints on the timing and method of classification and coding. The source of the reference images is also not particularly limited. For example, reference images can be obtained from design data, such as in the form of GDSII data of the pattern of the device being evaluated; this evaluation can be referred to as a die-to-database evaluation. Alternatively, reference images can be obtained by scanning a reference sample, such as one generated as: the average of sample areas of the same pattern previously scanned, or, for example, a manually determined ideal image without any defects (e.g., a gold image). If reference images are obtained by scanning, the scanning can be performed at a higher fidelity (e.g., at a slower speed), or the scan can be obtained by averaging scans of multiple references.

[0071] In various embodiments, another scan image (or two other scan images) scanned under the same conditions as the image being evaluated can be used as a reference; this can be referred to as die-to-die evaluation. When using such an image, or virtually any scan image, it is desirable to use two reference images derived from the two different scan images to determine the presence of defects; this defect detection process can be referred to as die-to-die inspection. The benefit of this process is that any possible defects identified by comparing the sample image with one of the reference scan images can be verified as defects. It is possible that the reference image may contain defects. Comparing the sample image with the first reference image determines a differential image and a set of possible defects. This differential image can be compared with a second differential image obtained by comparing the sample image with a second reference image, or with a reference differential image determined by comparing two reference images. Any differences present in the two differential images can indicate a possible defect that is not a defect but appears to be present in one of the reference images. Similarly, possible defects shown to be present in the differential image and the reference differential image indicate a defect present in one of the reference images but not in the sample image. Using this invention, comparison between a scanned image and two reference images can be achieved by classifying and encoding a dataset representing a sample image and two reference images.

[0072] Refer again Figure 5 The second stage, comparing the encoded sample image with the encoded reference image, may include simply subtracting the encoded foreground pixel map of the sample image from the foreground pixel map of the reference image (hereinafter referred to as the reference foreground pixel map) (or vice versa), or performing an XOR operation (which can be considered a multiplication operation). This operation (e.g., subtraction) removes foreground pixels present in both images to generate a differential foreground pixel map containing only foreground pixels present in one image and absent in the other. These residual foreground pixels (e.g., having multiple categories when using multiple foreground or other pixel categories) can represent candidate defects. Actual defects may include, for example, features missing in the test image (e.g.,... Figure 5 The dashed circles in the image indicate unprinted contact holes, missing, misaligned, or incorrect features, additional features, particles adhered to the sample, bridging between features, and soft defects. The operation selected for comparing the sample image with the reference image may depend on the type of defect being looked for and the number of pixel categories used.

[0073] In some cases, it may be desirable to perform filtering or thresholding on the differential foreground pixel map before identifying candidate defects. For example, such filtering can eliminate defects smaller than a predetermined size. In cases where there are multiple categories of foreground pixels, filtering can select one or more categories. Post-comparison filtering may be more necessary if denoising is not performed during or before the classification and encoding steps.

[0074] Before comparing coded sample and coded reference images, it may be desirable to perform an alignment process to determine the global offset between the sample and reference foreground pixel maps. In some cases, misalignment may exist between the sample and reference foreground pixel maps, with only a few pixels (e.g., less than 5) in each direction, which can be detected by known techniques. Such small offsets can be easily accommodated in the comparison step. As an alternative, multiple comparisons can be performed using different offsets, and the result with the smallest difference is considered to represent a correct alignment. In some cases, a large offset may exist that is unknown with high precision; in such cases, two-stage alignment can be performed. Route alignment can be performed based on the offset estimate, followed by fine alignment based on known image analysis techniques.

[0075] It's important to understand that, depending on the encoding nature of the sample and reference foreground pixel images, it may not be necessary to convert them back to a full matrix representation or bitmap. For example, if the sample and reference foreground pixel images are encoded as an ordered list of foreground pixel locations, a direct comparison of the list can be performed very easily. In other encoding techniques, such as run-length encoding, pixel values ​​can be dynamically determined for comparison and do not need to be stored.

[0076] Now refer to Figure 6 The description includes a data processing system comprising signal processing device 280 and data processing device 50. Signal processing device 280 is an example of a first data processing device, and data processing device 50 is an example of a second data processing device. Detector signals output by detector 240 are received by signal processing device 280. In embodiments where the electron optical system has multiple electron beams, the detector signals may be separate signals for each beam. In such an arrangement, a separate signal processing device 280 may be provided for each electron beam. (In embodiments, signal processing device 280 may include multiple signal processing elements assigned to different detector signals, for example, derived from different electron beams. The different signal processing elements may have the characteristics of the signal processing devices described elsewhere herein.) Desiredly, the detector signals are multi-valued signals representing the electron intensity returned from the sample in response to the electron beam or each electron beam.

[0077] Detector signals are supplied in parallel to one or more classifiers 2801a to 2801c. Each classifier determines that a given pixel is a foreground pixel, or, in the case of multiple classifiers, determines a specific subclass of the foreground pixel. In an embodiment, each classifier 2801a to 2801c may be a comparator that determines whether the density of the detector signal corresponding to a given pixel is higher or lower than a given threshold, or within a given range of values. The threshold or range of values ​​may be determined in an evaluation scheme for the specific pattern to be evaluated. Alternatively, classifiers 2801a to 2801c may be arranged in series or cascade. Although three classifiers are shown, it is understood that more or fewer classifiers may be provided depending on the number of pixel categories or subclasses to be identified; this may be a small number of categories, such as two categories, or up to ten or fifteen categories. Each classifier in classifiers 2801a to 2801c outputs a corresponding foreground pixel map 2802a to 2802c. As discussed above, each foreground pixel map in foreground pixel maps 2802a to 2802c may be a binary bitmap. Alternatively, the outputs of classifiers 2801a to 2801c can be combined into a single multi-valued foreground pixel map.

[0078] Then, a noise reduction process is performed on the foreground pixel maps 2802a to 2802c, or each foreground pixel map, using a noise filter 2803. The noise filter 2803 may include an erosion kernel 2804, optionally followed by a dilation kernel 2805. The erosion kernel 2804 reduces the size of the foreground pixel region within the foreground pixel map by a predetermined amount, such as one pixel. The dilation kernel 2805 increases the size of the foreground pixel region in the foreground pixel map, for example, by the same amount as the reduction applied by the erosion kernel 2804. The effect of the erosion and dilation operations is to eliminate small regions of the foreground pixels, such as isolated foreground pixels. As discussed above, isolated foreground pixels or small regions of foreground pixels are more likely to represent noise than actual features on the sample.

[0079] Then, encoder 2806 encodes the foreground pixel image or each foreground pixel image using, for example, the sparse matrix encoding technique described above. The encoded foreground pixel image is then sent to data processing device 50, for example, via a vacuum feedthrough.

[0080] The data processing device 50 includes an interface 501 for receiving an encoded foreground pixel image and performing error correction, etc. A reference interface 503 receives a reference foreground pixel image 600, which is expected to be encoded in the same format as the sample foreground pixel image received by interface 501. A candidate defect detector 502 compares the sample foreground pixel image with the reference foreground pixel image to identify the location of any candidate defects in the sample. The location of any candidate defect is transmitted to the output module 504.

[0081] Output module 504 receives the results output by candidate defect detector 502 and prepares them for output to users, other wafer fab systems, or for further processing within the inspection system. The output can be any of several different forms. In the simplest option, the output may simply be an indication of whether the sample has or does not have a defect. However, in some use cases, almost all samples will have at least one potential defect, and more detailed information is desired, for example, as a dataset, such as images (e.g., a dataset that can be rendered as an image). Therefore, the output may include, for example, a defect location map (or more generally, data indicating the defect and its location in a reference frame, such as an image or sample surface), a deviation image, and / or a clip of pixel data. The clip may be a cropped or partial image and / or a reference image. Alternatively or additionally, the output may include information about the severity of a possible defect, represented by the magnitude of the difference between the sample image and the reference image, such as how the sample image deviates from the reference image. The clip of pixel data may be obtained from a buffer 2807 in the signal processing device, which temporarily stores image data derived from the detector signal. The clip can correspond to a portion of an image and can be derived from the data of a sample image, and optionally from a reference image corresponding to that portion of the sample image.

[0082] The output module 504 can also filter potential defects, for example by outputting only defect locations where the difference between the sample image and the reference is greater than a threshold or where the pixel density indicating the difference is higher than a threshold. Another possibility is to output only a predetermined number of the most severe defect sites indicated by the difference magnitude. This can be achieved by storing the defect sites in a buffer, and when the buffer is full, overwriting the lowest magnitude defects if a higher magnitude defect is detected.

[0083] Defect information can be output in any suitable format, such as a list, dataset, image, or graph. Ideally, output module 504 can output clips, i.e., images of sample areas where potential defects are detected. This allows for further examination of the potential defects to determine if they are real and severe enough to affect the operation of devices formed or present on the sample. The remainder of the source image (i.e., portions not saved as clips) can be discarded to save on data storage and transmission requirements. Therefore, a dataset can be a collection of clips; images can include collections of clips.

[0084] The first data processing device (e.g., signal processing device 280) performing the classification and encoding steps may be located remotely from the detector array, for example, within a vacuum chamber. The first data processing device may be part of the electro-optical components of an electro-optical array; such components may include detectors (such as a detector array) and objectives (such as an objective lens array). The detector may be located in the path of the beam grating, and the signal processing device may be located remotely from the path of the beam grating. The first data processing device may be separate from the detector but electrically connected to it. The signal processing device is used to process data, not just signals including data directly from the detector. The signal processing device may be used to sort buffered or stored data files representing previously received detection signals as data. In this context, the signal processing device may be referred to as a data processing device. Although the described embodiments relate to the inline application of the invention, which may be desirable in systems generating large amounts of data, the invention can be applied after receiving and storing detection signals from a portion of the sample surface (such as an image). However, in such an application, the amount of data processed may be reduced, although some or all of the initial reduction in data load and the reduced memory for initial storage of detection signals from the charged particle system 41 may not be realized.

[0085] The first data processing device may include one or more processors, including a general-purpose CPU, a graphics processing unit (GPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or other specialized processors. GPUs or other specialized processors may be particularly efficient when performing classification using machine learning processes. In the case of multiple processors, specific processors may be assigned to process signals received from a specific electron beam or specific electro-optical device 41. The processors may be provided near the corresponding electro-optical device to which they are assigned to process their signals.

[0086] In one embodiment, at least a portion (if not all) of the data processing device 50 is located away from the electron optical array, such as away from the electron optical components, and is preferably located outside the vacuum chamber including the electron optical array. At least a portion of the data processing device can communicate signalingly with the detector via a feedthrough in the chamber wall. At least a portion of the data processing device 50 (e.g., in an arrangement within the data processing device 50) can be provided in a server rack away from the evaluation equipment, such as in the ground floor of a wafer fab (sometimes referred to as a sub-fab). However, the processing can be performed at least partially within the detector; therefore, the processing can be distributed between the data processing device 50 and the detector.

[0087] In embodiments of the invention, data sets associated with information having spatial relationships (e.g., related to the output of an electro-optical device, a reference image, or a candidate defect) are processed. In many cases, it is convenient to include the data set in a data structure, referred to herein as a graph, where the position of a value in the data structure is related to the position of the corresponding property measurement in, for example, the sample being evaluated. A 2D bitmap is an example of such a data structure. However, embodiments of the invention are not limited to this data structure. In alternative or more generally described representations, data values ​​are accompanied by or associated with positional information. That is, a data set may include information relating to the position of a value relative to one or more other values ​​in the data set. This information may be included in a data structure. Unless the advantages of a particular data structure are described, the term "data set" should not be construed as implying any specific data structure. More compact data structure forms, where values ​​are accompanied by positional information, may also be used.

[0088] A data stream is a series of data values ​​(also referred to as data points) sent in a predetermined order. The order of data values ​​in a data stream may be related to their position in an image or data graph. The order of data values ​​in a data stream may correspond to the order in which points on a sample are scanned by an electron beam. When performing processing steps on a data stream, these steps can be performed serially, i.e., on each data value (or a set of consecutive data values) in the order presented in the data stream. Ideally, the processing should not reorder the data stream, i.e., provide output values ​​in the same order as the incoming data stream. Ideally, the processing in the data stream should be performed in real time, i.e., at the same rate as the data values ​​are sent in the data stream.

[0089] References to up and down, upward and downward, above and below, etc., should be understood as referring to directions parallel (usually but not always vertical) to the inverse and downstream of the electron beam or multiple beams impacting sample 208. Therefore, references to the inverse and forward beams are intended to refer to directions about the beam path that are independent of any current gravitational field.

[0090] The embodiments described herein can take the form of a series of aperture arrays or electro-optical elements arranged in an array along a beam or multi-beam path. Such electro-optical elements can be electrostatic. In embodiments, all electro-optical elements (e.g., the last electro-optical element in a sub-beam path from the beam-limiting aperture array to the sample) can be electrostatic and / or can be in the form of an aperture array or a plate array. In some arrangements, one or more electro-optical elements are fabricated as microelectromechanical systems (MEMS) (i.e., using MEMS fabrication techniques). Electro-optical elements can have both magnetic and electrostatic components. For example, a compound array lens can have a macromagnetic lens that covers the multi-beam path, with upper and lower electrodes within the magnetic lens and arranged along the multi-beam path. The electrodes can be an aperture array for the beam paths of the multi-beams. Electrodes can be present above, below, or between the electrodes to control and optimize the electromagnetic field of the compound lens array.

[0091] Although the embodiments described herein are multi-beam electro-optical systems with electroelectro-optical devices, such as electrode plates with apertures for operating on different beams in a multi-beam configuration, the evaluation system may have a multi-beam electro-optical system including magnetic components. For example, the electro-optical system may include at least one of the following: a macromagnetic converging lens that can be configured in a non-rotational setting; an electro-optical array stack for fine manipulation of the beams in a multi-beam configuration; a beam-limiting aperture array for generating and / or shaping multiple beams in a multi-beam configuration, which may be a plate upstream of the converging lens and / or a plate downstream between the converging lens and the electro-optical array stack; a macromagnetic objective for projecting multiple beams onto a sample and may have electrostatic elements; a Wien filter for allowing the primary beam to be directed toward the sample and directing signal particles from the sample to a detector that may be located in a secondary column; and a detector in the secondary column. In another arrangement of the evaluation system, the electro-optical system may be designed to project a single beam toward the sample. This single-beam design can be similar to a multi-beam electro-optics system with a magnetic converging lens; in a variant, the detector is located in the main column (i.e., without a secondary column or Wien filter), for example facing the sample position and / or located between the converging lens and the objective lens.

[0092] The terms “sub-beam” and “beamwave” are used interchangeably herein and are both understood to encompass any radiation beam derived from a parent radiation beam by dividing or splitting it. The term “manipulator” is used to encompass any element that affects the path of a sub-beam or beamwave, such as a lens or deflector. Reference to an element aligned along a beam path or sub-beam path is understood to indicate that the corresponding element is positioned along the beam path or sub-beam path. Reference to an optical device is understood to refer to an electro-optical device.

[0093] The evaluation tools or systems according to this disclosure may include means for qualitative evaluation of a sample (e.g., pass / fail), means for quantitative measurement of a sample (e.g., dimensional features), or means for generating graphical images of the sample. Examples of evaluation tools or systems are inspection tools (e.g., for identifying defects), examination tools (e.g., for classifying defects), and measurement tools, or tools capable of performing any combination of evaluation functionality associated with inspection tools, examination tools, or measurement tools (e.g., subway inspection tools).

[0094] Reference to a system of components or elements that can controllably manipulate a charged particle beam in a certain manner includes configuring a controller, control system, or control unit to control the components to manipulate the charged particle beam in the described manner, and optionally using other controllers or devices (e.g., a voltage supply) to control the components to manipulate the charged particle beam in this manner. For example, under the control of a controller, control system, or control unit, a voltage supply may be electrically connected to one or more components to apply a potential to the components, such as electrodes applied to control lens arrays and objective lens arrays. The actuation of components, such as a platform, may be controlled to actuate another component (e.g., a beam path) and thus move relative to that component using one or more controllers, control systems, or control units.

[0095] The method of the present invention can be performed by a computer system comprising one or more computers. The computer used to implement the present invention may include one or more processors, including a general-purpose CPU, a graphics processing unit (GPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or other special-purpose processors. As discussed above, in some cases, a particular type of processor may offer advantages in reducing cost and / or increasing processing speed, and the method of the present invention can be adapted to the use of a particular processor type. Certain steps of the method of the present invention involve parallel computation, which is readily implementable on processors capable of parallel computation, such as GPUs.

[0096] The computer used to implement this invention can be physical or virtual. It can be a server, client, or workstation. Multiple computers used to implement this invention can be distributed and interconnected via a local area network (LAN) or a wide area network (WAN). The results of the method of this invention can be displayed to a user or stored in any suitable storage medium. This invention can be implemented in a non-transitory computer-readable storage medium that stores instructions for performing the method of this invention. This invention can be implemented in a computer system including one or more processors and a memory or storage device storing instructions for performing the method of this invention.

[0097] The functionality provided by a computer and processor (such as a controller or control system or control unit) can be implemented by the computer. Any suitable combination of components can be used to provide the required functionality, including, for example, a CPU, RAM, SSD, motherboard, network connectivity, firmware, software, and / or other components known in the art that allow the execution of the required computational operations. The required computational operations can be defined by one or more computer programs. Such computer programs can take the form of multiple computer programs, which can be distributed across different processors, for example, implemented. One or more computer programs can be provided in the form of a medium storing computer-readable instructions (optionally a non-transient medium). When the computer-readable instructions are read by the computer, the computer executes the required method steps. The computer can consist of independent units or a distributed computing system having multiple different computers interconnected via a network.

[0098] As used herein, the term "image" is intended to refer to any data structure of values, where each value is associated with a sample at a location, and the arrangement of values ​​in the array corresponds to the spatial arrangement of the sampling locations. The term "data graph" can be used to describe such a data structure. Images can comprise a single layer or multiple layers. In the case of a multi-layer image, each layer (also referred to as a channel) represents a different sample at a location. The term "pixel" is intended to refer to a single value in the array, or, in the case of a multi-layer image, a set of values ​​corresponding to a single location. Images can be stored in any convenient format on a computer-readable storage medium.

[0099] Embodiments of the present invention are described in the following numbered clauses: Clause 1. A data processing method for image data obtained by scanning a charged particle beam across samples; the method comprising: classifying pixels of the image data into foreground pixels and background pixels to generate a foreground pixel map; and encoding the foreground pixel map into encoded data using a sparse matrix coding technique. Clause 2. The method according to Clause 1, wherein classifying pixels includes comparing pixel values ​​with a threshold to generate a comparison map, and a foreground pixel map is generated from the comparison map. Clause 3. The method according to Clause 2, wherein comparing a pixel value with a threshold includes comparing a pixel value with multiple thresholds to generate multiple comparison maps, the multiple foreground pixel maps being generated from the multiple comparison maps, and encoding the foreground pixel maps includes encoding each foreground pixel map as encoded data using a sparse matrix coding technique. Clause 4. The method according to Clause 3, wherein the comparison includes comparing pixel values ​​with no more than 10 thresholds to generate no more than 10 comparison maps. Clause 5. The method according to Clause 1, 2, 3 or 4 also includes reducing noise in the foreground pixel map(s). Clause 6. The method according to Clause 5, wherein noise reduction includes applying an erosion kernel to the comparison map to generate an erosion map. Clause 7. The method according to Clause 6, wherein noise reduction also includes applying a dilation kernel to the erosion map to generate a foreground pixel map. Clause 8. The method according to any of the preceding clauses, wherein the sparse matrix coding technique is selected from the group consisting of: key dictionary; nested list; coordinate list; compressed sparse row; compressed sparse column; run-length encoding. Clause 9. The method according to any of the preceding clauses, wherein classification and encoding are performed in a first data processing device, and further includes sending encoded data to a second data processing device located remotely from the first data processing device. Clause 10. The method according to Clause 9, wherein a first data processing device is included in an evaluation apparatus that scans a beam of charged particles across a sample. Clause 11. The method according to Clause 10, wherein the second data processing device is located away from the evaluation device. Clause 12. The method according to Clauses 9, 10 or 11, wherein the first data processing device includes an FPGA, ASIC or GPU. Clause 13. The method according to any of the preceding clauses further includes comparing the encoded data with a reference image to identify candidate defects, wherein the reference image is also encoded using a sparse matrix coding technique. Clause 14. The method according to Clause 13, wherein the reference image is obtained from GDSII data. Clause 15. The method according to Clause 13, wherein the reference image is obtained by scanning a reference. Clause 16. The method of Clause 13, 14 or 15, depending on any one of Clauses 9 to 12, wherein the comparison is performed in a second data processing device. Clause 17. The method according to any one of Clauses 13 to 16 further includes sending the region containing the image data of the candidate defect to the outside of the evaluation device. Clause 18. The method according to any of the preceding clauses, wherein the image data is obtained using a multi-beam charged particle evaluation device. Clause 19. An evaluation method comprising: directing a beam of charged particles onto a sample; generating a digital detection signal of pixel values ​​in response to signal particles from the sample; scanning the sample and the beam of charged particles; and processing the digital detection signal using a method of any one of Clauses 1 to 18. Clause 20. A computer program for data processing, the computer program comprising instructions that, when executed by a data processing system, cause the data processing system to perform a method according to any one of Clauses 1 to 18. Clause 21. A data encoding apparatus for encoding image data obtained by scanning a charged particle beam across a sample, the apparatus comprising: a classifier for classifying pixels of the image data into foreground pixels and background pixels to generate a foreground pixel map; and an encoder for encoding the foreground pixel map into encoded data using a sparse matrix encoding technique. Clause 22. A data encoding device pursuant to Clause 21, wherein a classifier is configured to compare pixel values ​​with a threshold to generate a comparison map, and a foreground pixel map is generated from the comparison map. Clause 23. A data encoding device pursuant to Clause 22, wherein a classifier is configured to compare pixel values ​​with a plurality of thresholds to generate a plurality of comparison maps, the plurality of foreground pixel maps being generated from the plurality of comparison maps, and an encoder is configured to encode each foreground pixel map as encoded data using a sparse matrix coding technique. Clause 24. A data encoding device pursuant to Clause 23, wherein a classifier is configured to compare pixel values ​​with no more than 10 thresholds to generate no more than 10 comparison maps. Clause 25. The data encoding device according to Clauses 21, 22, 23 or 24 further includes a noise filter configured to reduce noise in the foreground pixel map(s). Clause 26. A data encoding device pursuant to Clause 25, wherein a noise filter includes an erosion kernel configured to generate an erosion map from a comparison map. Clause 27. A data encoding device pursuant to Clause 26, wherein the noise filter further includes an expansion kernel configured to generate a foreground pixel map from an erosion map. Clause 28. A data encoding device according to any one of Clauses 21 to 27, wherein the encoder is configured to employ a sparse matrix encoding technique selected from the group consisting of: a key dictionary; a nested list; a coordinate list; compressed sparse rows; compressed sparse columns; and run-length encoding. Clause 29. A data encoding device pursuant to any one of Clauses 21 to 28, wherein the encoder is further configured to transmit encoded data to a data processing device located remotely from the data encoding device. Clause 30. Data encoding device pursuant to Clause 29, the data encoding device is included in the evaluation apparatus that scans a beam of charged particles across the sample. Clause 31. A data encoding device pursuant to any one of Clauses 21 to 30, wherein at least a portion of the classifier and / or encoder is implemented in an FPGA, ASIC, or GPU. Clause 32. A data comparison apparatus for detecting candidate defects in a sample image, the apparatus comprising: an interface for receiving coded sample data representing a foreground pixel map encoded using a sparse matrix coding technique; and a comparator for comparing the coded sample data with coded reference data, the coded reference data representing a foreground pixel map derived from a reference image and encoded using a sparse matrix coding technique. Clause 33. Data comparison device pursuant to Clause 32, wherein the reference image is obtained from GDSII data. Clause 34. Data comparison device pursuant to Clause 32, wherein the reference image is obtained by scanning a reference. Clause 35. The data comparison device according to any one of Clauses 32 to 34 further includes an output module configured to send a region containing image data of candidate defects to an external part of the data comparison device. Clause 36. A data processing system comprising a data encoding device according to any one of Clauses 21 to 31 and a data comparison device according to any one of Clauses 32 to 35. Clause 37. A charged particle evaluation apparatus for scanning a charged particle beam across a sample; the apparatus comprising: a detector unit configured to output a digital detection signal of pixel values ​​in response to signal particles incident from a sample; a scanning unit for relatively scanning the sample and the charged particle beam; and a data encoding device according to any one of Clauses 21 to 31. Clause 38. Charged particle evaluation apparatus pursuant to Clause 37, wherein the data encoding device is at least partially located within a vacuum chamber. Clause 39. A charged particle evaluation system for evaluating at least a portion of a sample, the charged particle evaluation system comprising the charged particle evaluation apparatus of Clause 37 or 38 and a data comparison apparatus according to any one of Clauses 32 to 35. Clause 40. A data structure comprising encoded sample data including a foreground pixel map obtained by relatively scanning a sample and a beam of charged particles, the foreground pixel map being encoded using a sparse matrix coding technique. Clause 41. A method for processing detection data from a sample surface of detector elements of a detector array, the method comprising: scanning a beam of charged particles across a sample region; detecting the obtained signal particles using one or more detector elements of the charged particle detector array to obtain detection data; receiving pixels of the detection data from the detector elements in a data stream; and classifying the pixels of the detection data in the data stream into foreground pixels and background pixels to generate a foreground pixel dataset representing the scanned sample region. Clause 42. The method according to Clause 41, wherein detection data from multiple detector elements can be represented as an image of the sample surface. Clause 43. The method according to Clause 41 or 42 also includes encoding the foreground pixel dataset into encoded data using an encoding operator in the data stream. Clause 44. The method according to Clauses 41, 42 or 43, wherein classification and / or encoding are performed in a detector and / or a first data processing device, and further comprising sending encoded data to a second data processing device remote from the detector or the first data processing device. Clause 45. A method for processing detection data from a sample surface of detector elements of a detector array, the method comprising: scanning a beam of charged particles across a sample region; detecting the resulting signal particles using one or more detector elements of the charged particle detector array to obtain detection data; receiving pixels of the detection data from the detector elements in a data stream; and classifying the pixels of the detection data in the data stream into two or more distinct categories of pixels to generate a pixel dataset representing pixels of at least one category from the scanned sample region. Clause 46. In accordance with the method of Clause 45, two or more categories have an equal hierarchy. Clause 47. The method according to Clause 45, wherein different categories of pixels include foreground pixels and background pixels, and a dataset of foreground pixels is generated. Clause 48. The method according to Clause 45, wherein the pixel category includes one or more other types of pixels, such as mid-field pixels, that have an equal hierarchical structure with foreground and background pixels. Clause 49. The method according to Clause 47 or 48, wherein pixels of multiple categories include foreground pixels and / or pixels of multiple categories include background pixels. Clause 50. The method according to any one of Clauses 45 to 49, wherein there are three or more, or five or more, or ten or more pixels of different categories.

[0100] Although the invention has been described in conjunction with various embodiments, other embodiments of the invention will be apparent to those skilled in the art from the description and practice of the invention disclosed herein. The description and examples are intended to be illustrative only, and the true scope and spirit of the invention are indicated by the following claims.

Claims

1. A data processing method for image data, said image data being obtained by scanning a charged particle beam across a sample, said method comprising: The pixels of the image data are classified into foreground pixels and background pixels to generate a foreground pixel map; as well as The foreground pixel image is encoded into encoded data using sparse matrix coding technology.

2. The method of claim 1, wherein classifying pixels comprises comparing pixel values ​​with a threshold to generate a comparison map, the foreground pixel map being generated from the comparison map.

3. The method of claim 2, wherein comparing pixel values ​​with thresholds includes comparing pixel values ​​with multiple thresholds to generate multiple comparison maps, generating multiple foreground pixel maps from the multiple comparison maps, and encoding the foreground pixel maps includes encoding each foreground pixel map as encoded data using a sparse matrix coding technique.

4. The method of claim 3, wherein the comparison comprises: Pixel values ​​are compared with no more than 10 thresholds to generate no more than 10 comparison images.

5. The method according to claim 1, 2, 3 or 4, further comprising: Reduce noise in the foreground pixel map(s).

6. The method of claim 5, wherein reducing noise comprises: An erosion kernel is applied to the comparison map to generate an erosion map.

7. The method according to any one of the preceding claims, wherein the classification and the encoding are performed in a first data processing device, and further comprising sending the encoded data to a second data processing device remote from the first data processing device.

8. The method of claim 7, wherein the first data processing device is included in an evaluation apparatus that scans the charged particle beam across the sample.

9. The method of claim 8, wherein the second data processing device is located away from the evaluation device.

10. The method according to any one of the preceding claims further includes comparing the encoded data with a reference image to identify candidate defects, wherein the reference image is also encoded using the sparse matrix coding technique.

11. The method of claim 10, wherein the reference image is obtained from GDSII data, or the reference image is obtained by scanning a reference.

12. The method according to claims 10 to 11, further comprising sending the region of the image data containing the candidate defect to the outside of the evaluation device.

13. The method according to any one of the preceding claims, wherein the image data is obtained using a multi-beam charged particle evaluation device.

14. A computer program for data processing, the computer program comprising instructions that, when executed by a data processing system, cause the data processing system to perform the method according to any one of claims 1 to 13.

15. A data encoding device for encoding image data obtained by scanning a charged particle beam across a sample, the device comprising: A classifier is used to classify the pixels of the image data into foreground pixels and background pixels to generate a foreground pixel map; as well as An encoder is used to encode the foreground pixel map into encoded data using sparse matrix coding techniques.

Images