Multi-view Chip Analysis

Abstract

A method for detecting defects on a sample is disclosed herein. The method includes the following steps: obtaining scan data of regions of the sample in multiple perspectives; and performing an integrated analysis of the obtained scan data. The integrated analysis includes: calculating a cross-perspective covariance based on the obtained scan data, and / or estimating a cross-perspective covariance; and determining the presence of a defect in the region while taking into account the cross-perspective covariance.

Description

Technical Field

[0001] The present disclosure generally relates to wafer analysis. Background Art

[0002] As design rules shrink, there is a corresponding need for wafer analysis tools to detect ever-smaller defects. Previously, defect detection was mainly limited by laser power and detector noise. Currently, existing wafer analysis tools are mainly limited by wafer noise caused by diffuse reflection from the wafer surface: surface irregularities on the wafer composed of the roughness of the etched pattern typically appear as bright spots (speckles) in the scanned image. These bright spots can be very similar to the "fingerprints" (characteristic markings) of defects. Therefore, there is a need for improved techniques for differentiating defects from wafer noise. Summary of the Invention

[0003] Aspects of the present disclosure according to some embodiments relate to methods and systems for wafer analysis. More specifically, but not exclusively, aspects of the present disclosure according to some embodiments relate to methods and systems for multi-perspective wafer analysis, in which measurement data from multiple perspectives is subjected to integrated analysis.

[0004] Thus, according to one aspect of some embodiments, a method for detecting defects on a sample (e.g., a wafer or an optical mask) is provided. The method includes the following steps:

[0005] - Obtain scan data of a first region (e.g., on the surface) of the sample in multiple perspectives.

[0006] - Perform integrated analysis on the obtained scan data. The integrated analysis includes:

[0007] ■ Calculate cross-perspective covariance (i.e., covariance between different perspectives) based on the obtained scan data, and / or estimate cross-perspective covariance.

[0008] ■ Determine the presence of defects in the first region taking into account the cross-perspective covariance.

[0009] According to some embodiments of the method, the sample is a patterned wafer.

[0010] According to some embodiments of the method, the sample is a bare wafer.

[0011] According to some embodiments of the method, the multiple perspectives include two or more of the following items: one or more incident angles of one or more illumination beams, one or more collection angles of one or more collected beams, at least one intensity of one or more illumination beams, at least one intensity of one or more collected beams, and compatible combinations of the above items.

[0012] In some embodiments of the method, the method is optical, scanning electron microscopy-based, and / or atomic force microscopy-based.

[0013] In some embodiments of the method, the method is optical, and the multiple perspectives include two or more of the following items: one or more illumination angles, intensity of the illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focus offsets of the illumination beam, one or more collection angles, intensity of the collected radiation, collection polarization, phase of one or more of the collected beams, brightfield channel, grayfield channel, Fourier filtering of the returned light, and sensing type selected from intensity, phase, or polarization, and compatible combinations of the above items.

[0014] In some embodiments of the method, the integrated analysis includes:

[0015] - For each of a plurality of sub-regions of the first region, generating a difference value in each of the multiple perspectives based on the acquired scan data and the corresponding reference data of the first region in each of the multiple perspectives. (That is, generating a set of difference values, where each difference value in the set corresponds to a different perspective.)

[0016] - Determining whether each of the plurality of sub-regions is defective based at least on the difference values corresponding to the sub-region and the sub-regions adjacent to the sub-region and the noise values (i.e., a set of noise values) corresponding to the sub-region and the adjacent sub-regions. The noise values include the corresponding covariance from the cross-perspective covariance.

[0017] In some embodiments of the method, the method further includes the step of: generating a difference image of the first region in each of the multiple perspectives based on the acquired scan data and the reference data. The difference value corresponding to each sub-region from the plurality of sub-regions is derived from and / or characterizes the sub-image of the difference image corresponding to the sub-region. (Such that given N difference images, N sub-images (i.e., a set of N sub-images) correspond to each sub-region. More specifically, N sub-images (one sub-image for each of the N difference images) and N corresponding difference values correspond to each sub-region.)

[0018] In some embodiments of the method, the noise value is calculated based at least on the difference value.

[0019] In some embodiments of the method, the step of determining whether each of the plurality of sub-regions is defective includes:

[0020] - Generate a covariance matrix that includes noise values corresponding to a sub-region and sub-regions adjacent to the sub-region.

[0021] - Multiply a first vector that includes difference values corresponding to a sub-region and adjacent sub-regions by the inverse of the covariance matrix to obtain a second vector.

[0022] - Calculate a scalar product of the second vector and a third vector, where the components of the third vector include values characterizing one or more defects.

[0023] - If the scalar product is greater than a predetermined threshold, label (designate) the sub-region as defective.

[0024] According to some embodiments of the method, at least one of the plurality of sub-regions has a size corresponding to a single (image) pixel.

[0025] According to some embodiments of the method, cross-view covariance is estimated based at least on scan data obtained during a preliminary scan of a sample, where regions of the sample (e.g., regions on a surface) are sampled during the preliminary scan. Each sampled region represents a group of regions of the sample, where at least one of the sampled regions represents a first region.

[0026] According to some embodiments of the method, the method further includes the steps of: when the presence of a defect is determined, determining whether the defect is a defect of interest, and optionally, when the defect is determined to be of interest, classifying the defect.

[0027] According to some embodiments of the method, the method is repeated for each of a plurality of additional regions so as to scan a larger region of the sample (e.g., a larger region on the surface of the sample) formed by the first region and the additional regions.

[0028] According to one aspect of some embodiments, there is provided a computerized system for obtaining and analyzing multi-view scan data of a sample (e.g., a wafer or an optical mask). The computerized system is configured to implement the above method.

[0029] According to one aspect of some embodiments, there is provided a non-transitory computer-readable storage medium storing instructions that cause a computerized analysis system (e.g., a wafer analysis system) to implement the above method.

[0030] According to one aspect of some embodiments, there is provided a computerized system for obtaining and analyzing multi-view scan data of a sample. The system includes:

[0031] - A scanning device configured to scan regions of the sample (e.g., regions on a surface) in multiple viewing angles.

[0032] - A scan data analysis module (including one or more processors and memory components), the scan data analysis module being configured to perform an integrated analysis of the scan data obtained in a scan, wherein the integrated analysis includes:

[0033] ■ Calculating a cross - perspective covariance based on the obtained scan data, and / or estimating a cross - perspective covariance.

[0034] ■ Determining the presence of defects in a region, taking into account the cross - perspective covariance.

[0035] According to some embodiments of the system, the system is configured to analyze scan data of a patterned wafer.

[0036] According to some embodiments of the system, the system is configured to analyze scan data of a bare wafer.

[0037] According to some embodiments of the system, the multiple perspectives include two or more of the following items: one or more incident angles of one or more illumination beams, one or more collection angles of one or more collected beams, at least one intensity of one or more illumination beams, and at least one intensity of one or more collected beams.

[0038] According to some embodiments of the system, the scanning device includes an optically - based imager.

[0039] According to some embodiments of the system, the scanning device includes a scanning electron microscope.

[0040] According to some embodiments of the system, the scanning device includes an atomic force microscope.

[0041] According to some embodiments of the system, the multiple perspectives include two or more of the following items: one or more illumination angles, intensity of illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focus offsets of an illumination beam, one or more collection angles, intensity of collected radiation, collection polarization, phase of one or more collected beams, bright - field channel, gray - field channel, Fourier filtering of the returned light, and a sensing type selected from intensity, phase, or polarization, as well as compatible combinations of the above items.

[0042] According to some embodiments of the system, the integrated analysis includes:

[0043] - For each of a plurality of sub - regions of a first region, generating a difference value in each of the multiple perspectives based on the obtained scan data of the first region and corresponding reference data in each of the multiple perspectives.

[0044] - Determine whether each of the plurality of sub-regions is defective based at least on a difference value corresponding to the sub-region and a sub-region adjacent to the sub-region, and a noise value corresponding to the sub-region and an adjacent sub-region. The noise value includes a corresponding covariance from cross-view covariance.

[0045] According to some embodiments of the system, the scan data analysis module is further configured to: generate a difference image of a first region in each of the multiple perspectives based on the acquired scan data and reference data, wherein the difference value corresponding to each sub-region from the plurality of sub-regions is derived from and / or characterizes a sub-image corresponding to the sub-region of the difference image.

[0046] According to some embodiments of the system, the scan data analysis module is configured to calculate a noise value based at least on the difference value.

[0047] According to some embodiments of the system, the step of determining whether each of the plurality of sub-regions is defective includes:

[0048] - Generate a covariance matrix that includes noise values corresponding to the sub-region and a sub-region adjacent to the sub-region.

[0049] - Multiply a first vector including difference values corresponding to the sub-region and an adjacent sub-region by the inverse of the covariance matrix to obtain a second vector.

[0050] - Calculate the scalar product of the second vector and a third vector, the components of the third vector including values characterizing one or more defects.

[0051] - If the scalar product is greater than a predetermined threshold, mark the sub-region as defective.

[0052] According to some embodiments of the system, at least one of the plurality of sub-regions has a size corresponding to a single (image) pixel.

[0053] According to some embodiments of the system, the scan data analysis module is configured to estimate cross-view covariance based at least on scan data obtained during a preliminary scan of a sample, wherein an area of the sample (e.g., an area on a surface) is sampled during the preliminary scan. Each sampled area represents a group of areas of the sample, wherein at least one of the sampled areas represents the first region.

[0054] According to some embodiments of the system, the scan data analysis module is further configured to: after determining the presence of a defect, further determine whether the defect is a defect of interest, and optionally, classify the defect when the defect is determined to be of interest.

[0055] According to some embodiments of the system, the system is further configured to repeat the scanning and integrated analysis for each of a plurality of additional regions so as to scan a larger region of the sample (e.g., a larger region on a surface) formed by the first region and the additional regions.

[0056] According to one aspect of some embodiments, there is provided a non-transitory computer-readable storage medium storing instructions that cause a computerized analysis system (e.g., a wafer analysis system) to perform the following operations:

[0057] - Scan a region (e.g., a region on a surface) of a sample (e.g., a wafer or an optical mask) from multiple perspectives.

[0058] - Perform an integrated analysis of the scan data obtained in the scan, the integrated analysis including:

[0059] ■ Calculate cross-perspective covariance and / or estimate cross-perspective covariance based on the obtained scan data.

[0060] ■ Determine the presence of defects in the region taking into account the cross-perspective covariance.

[0061] According to some embodiments of the storage medium, the sample is a patterned wafer.

[0062] According to some embodiments of the storage medium, the sample is a bare wafer.

[0063] According to some embodiments of the storage medium, the multiple perspectives include two or more of the following items: one or more incident angles of one or more illumination beams, one or more collection angles of one or more collected beams, at least one intensity of one or more illumination beams, and at least one intensity of one or more collected beams.

[0064] According to some embodiments of the storage medium, the computerized analysis system is optical-based.

[0065] According to some embodiments of the storage medium, the scanning of the computerized analysis system is electron microscopy-based or atomic force microscopy-based.

[0066] According to some embodiments of the storage medium, the multiple perspectives include two or more of the following items: one or more illumination angles, intensity of illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focus offsets of an illumination beam, one or more collection angles, intensity of collected radiation, collection polarization, phase of one or more collected beams, bright field channel, gray field channel, Fourier filtering of the returned light, and a sensing type selected from intensity, phase, or polarization, and compatible combinations of the above items.

[0067] According to some embodiments of the storage medium, the integrated analysis includes:

[0068] - For each of a plurality of sub-regions of a first region, generating a difference value in each of a plurality of perspectives based on the acquired scan data and corresponding reference data of the first region in each of the plurality of perspectives.

[0069] - Determining whether each of the plurality of sub-regions is defective based at least on the difference values corresponding to the sub-region and a sub-region adjacent to the sub-region and the noise values corresponding to the sub-region and an adjacent sub-region. The noise values include corresponding covariances from cross-perspective covariance.

[0070] According to some embodiments of the storage medium, the stored instructions cause the scan data analysis module of the computerized system to perform the following operations: generating a difference image of a first region in each of a plurality of perspectives based on the acquired scan data and reference data, wherein the difference value corresponding to each sub-region from a plurality of sub-regions is derived from and / or characterizes a sub-image of the difference image corresponding to the sub-region.

[0071] According to some embodiments of the storage medium, the stored instructions cause the scan data analysis module to calculate the noise values based at least on the difference values.

[0072] According to some embodiments of the storage medium, the step of determining whether each of the plurality of sub-regions is defective includes:

[0073] - Generating a covariance matrix that includes the noise values corresponding to the sub-region and a sub-region adjacent to the sub-region.

[0074] - Multiplying a first vector that includes the difference values corresponding to the sub-region and an adjacent sub-region by the inverse of the covariance matrix to obtain a second vector.

[0075] - Calculating a scalar product of the second vector and a third vector, the components of the third vector including values characterizing one or more defects.

[0076] - If the scalar product is greater than a predetermined threshold, marking the sub-region as defective.

[0077] According to some embodiments of the storage medium, at least one of the plurality of sub-regions has a size corresponding to a single (image) pixel.

[0078] According to some embodiments of the storage medium, the stored instructions cause the scan data analysis module to perform the following operations: estimate the cross-view covariance based at least on scan data obtained during a preliminary scan of a sample, sampling regions (e.g., regions on a surface) of the sample during the preliminary scan. Each sampled region represents a group of regions of the sample, where at least one of the sampled regions represents a first region.

[0079] Certain embodiments of the present disclosure may include some, all, or none of the above advantages. Those skilled in the art can readily appreciate one or more other technical advantages based on the drawings, descriptions, and claims included herein. Moreover, while specific advantages are listed above, various embodiments may also include all, some, or none of the listed advantages.

[0080] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. In case of conflict, the patent specification, including definitions, will control. As used herein, the indefinite articles "a" and "an" mean "at least one" or "one or more" unless the context clearly dictates otherwise.

[0081] Unless otherwise specifically recited, it will be apparent from the present disclosure that, according to some embodiments, terms such as "processing," "computing," "determining," "estimating," "evaluating," "measuring," etc. can refer to actions and / or processes of a computer or computing system, or similar electronic computing device, that manipulate and / or transform data represented as physical (e.g., electronic) quantities within the registers and / or memories of the computing system into other data similarly represented as physical quantities within the memories, registers, or other such information storage devices, information transmission devices, or information display devices of the computing system.

[0082] Embodiments of the present disclosure may include apparatuses for performing the operations herein. The apparatuses may be specially constructed for the desired purposes or may include one or more general-purpose computers selectively activated or reconfigured by a computer program stored in the computer. Such computer programs may be stored in a computer-readable storage medium, such as but not limited to any type of disk including floppy disks, optical disks, CD-ROMs, magneto-optical disks, read-only memory (ROM), random access memory (RAM), electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic or optical cards, or any type of medium suitable for storing electronic instructions and capable of being coupled to a computer system bus.

[0083] The processes and displays presented herein are not inherently related to any particular computer or other device. Various general-purpose systems may be used with the programs according to the teachings herein, or it may prove expedient to construct more specialized devices to perform the desired method(s). From the following description, the desired structures of various such systems will be apparent. Additionally, embodiments of the present disclosure are not described with reference to any particular programming language. It will be understood that various programming languages may be used to implement the teachings of the present disclosure as described herein.

[0084] Aspects of the present disclosure may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The disclosed embodiments may also be practiced in a distributed computing environment where tasks are performed by remote processing devices linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Some embodiments of the present disclosure are described herein with reference to the accompanying drawings. The description and the drawings together enable those of ordinary skill in the art to understand how some embodiments may be implemented. The drawings are for illustrative purposes only and show only the necessary structural details for a basic understanding of the embodiments of the present disclosure, and are not intended to show the structural details in more detail. For clarity, some of the objects depicted in the drawings are not drawn to scale. Moreover, two different objects in the same drawing may be drawn to different scales. In particular, the scale of some objects may be greatly exaggerated compared to other objects in the same drawing. Figure 1 In the drawings:

[0086] is a flowchart of a method for multi-view wafer analysis according to some embodiments;

[0087] Figure 1 is a flowchart of operations for integrated analysis of multi-view scan data according to some specific embodiments of the method;

[0088] Figure 2 is according to Figure 1 is a flowchart of sub-operations for identifying (detecting) defects in a scanned area of a wafer according to some specific embodiments of the operations;

[0089] Figure 3 is according to Figure 2 is a flowchart of sub-operations for identifying (detecting) defects in a scanned area of a wafer according to some specific embodiments of the operations;

[0090] Figures 4A - 4G presents according to some embodiments including in Figure 3The algebraic representation used in the calculations of the sub-operations of;

[0091] Figure 5A and Figure 5B Presents two different ways of enumerating sub-images according to some embodiments;

[0092] Figure 6 Presents a block diagram of a computerized system for obtaining and analyzing multi-view scan data of a wafer (which is also depicted);

[0093] Figure 7A Schematically depicts a computerized system for obtaining and analyzing multi-view scan data of a wafer (which is also depicted), and the depicted computerized system is Figure 6 A specific embodiment of the computerized system of;

[0094] Figure 7B Schematically depicts light rays mirror-reflecting off Figure 7A the wafer of;

[0095] Figure 8 Schematically depicts a computerized system for obtaining and analyzing multi-view scan data of a wafer (which is also depicted), and the depicted computerized system is Figure 6 A specific embodiment of the computerized system of;

[0096] Figure 9 Schematically depicts a computerized system for obtaining and analyzing multi-view scan data of a wafer (which is also depicted), and the depicted computerized system is Figure 6 A specific embodiment of the computerized system of; and

[0097] Figures 10A - 10C Presents simulation results showing the efficacy of the Figure 1 method of; Detailed Description

[0098] The principles, uses, and implementations of the teachings herein can be better understood with reference to the accompanying specification and drawings. After reading the description and drawings presented herein, those skilled in the art will be able to implement the teachings herein without undue effort or experimentation. In the drawings, like reference numerals always refer to like parts.

[0099] In the specification and claims of this application, the words "comprising" and "having" and their forms are not limited to the members in the lists that may be associated with those words.

[0100] As used herein, the term "about" may be used to specify a value of a quantity or parameter (e.g., the length of a component) within a continuous range of values near (and including) a given (stated) value. According to some embodiments, "about" may specify a value of a parameter as being between 80% and 120% of a given value. For example, the statement "the length of the component is equal to about 1 m" is equivalent to the statement "the length of the component is between 0.8 m and 1.2 m". According to some embodiments, "about" may specify a value of a parameter as being between 90% and 110% of a given value. According to some embodiments, "about" may specify a value of a parameter as being between 95% and 105% of a given value.

[0101] As used herein, according to some embodiments, the terms "substantially" and "about" may be interchangeable.

[0102] Referring to the accompanying drawings, in a flowchart, optional operations may appear within boxes delineated by dashed lines.

[0103] As used herein, the term "multi-perspective wafer analysis" is used to refer to wafer analysis that employs scan data from multiple perspectives. For example, the different perspectives may differ from one another in terms of polarization, collection pupil segment, phase information, focus offset, etc. The additional information provided by multiple perspectives can be used, in particular, to more efficiently cope with wafer noise. Scan data from several perspectives may produce predictable or self-learnable patterns that can be distinguished from wafer noise, thus leading to an improved defect detection rate.

[0104] As used herein, according to some embodiments, the terms "identify" and "detect" and their derivatives, when employed with reference to, for example, defects on a wafer, may be used interchangeably.

[0105] As used herein, according to some embodiments, the term "sample" may refer to a wafer or an optical mask. The wafer may be patterned or bare.

[0106] Method

[0107] According to one aspect of some embodiments, a computer-implemented method for wafer analysis is provided, in which scan data from multiple perspectives is subjected to integrated analysis (as defined and explained below). Figure 1 A flowchart presenting such a method (method 100) according to some embodiments is shown.

[0108] According to some embodiments, method 100 includes operation 110, in which scan data of multiple perspectives of a region (area) of a wafer is obtained. More specifically, in operation 110, multiple images (e.g., image frames) of multiple perspectives of a scanned region of the wafer (e.g., a segment corresponding to an image frame) can be obtained. As detailed below, the multiple images can be obtained using a scanning device configured to scan the wafer from multiple perspectives. In particular, the scanning device can include an imager (imaging module or unit) configured to irradiate (e.g., illuminate) a region of the wafer and collect radiation from the region. According to some embodiments, the imager can be optically based (configured to illuminate a region of the wafer with electromagnetic radiation such as visible light and / or ultraviolet (UV) radiation). According to some embodiments, the UV radiation can be or can include deep UV radiation and / or extreme UV radiation. According to some embodiments, the imager can be configured to irradiate a region of the wafer with one or more charged particle beams (e.g., electron beams).

[0109] According to some embodiments, the imager can be configured to allow simultaneous irradiation of the wafer with multiple radiation beams, thereby facilitating simultaneous scanning of multiple regions of the wafer.

[0110] Generally speaking, the perspectives can be classified into two groups: radiation channel perspectives and collection channel perspectives. Broadly, the radiation channel determines one or more physical properties of the irradiation beam incident on the wafer, such as the trajectory line of the beam, the shape of the beam, and / or the polarization of the beam (when the beam is a light beam). In contrast, the collection channel includes the sensing type (intensity, polarization, phase) and a "filter", which in this context broadly refers to a mechanism (e.g., segmented pupil, Fourier filter, polarization beam splitter) configured to allow selective collection (and sensing) of components of the radiation returning from the wafer characterized by certain physical properties, such as the return (reflection, scattering) angle, intensity, and polarization (when the radiation is electromagnetic radiation).

[0111] In some embodiments where the imager is optical-based, the multiple viewing angles can include two or more of the following items: one or more focus offsets of (a plurality of) illumination angles (i.e., (a plurality of) incident angles of illumination radiation), illumination intensity (as determined by the amplitude of the illumination radiation), illumination polarization (i.e., the polarization of the illumination radiation), illumination wavefront (the shape of the wavefront of the illumination radiation when monochromatic), illumination spectrum (i.e., the spectrum of the illumination radiation), and illumination beam (which can be slightly defocused), (a plurality of) collection angles (thereby allowing selective sensing of light returning at a certain angle or range of angles), intensity of the collected radiation (thereby allowing selective sensing of light returning at a certain intensity or range of intensities), collection polarization, phase of the (a plurality of) collected beams (when the (a plurality of) illumination beams are monochromatic), bright field channel, gray field channel (which can be further subdivided into darkfield and "pure" gray field), Fourier filtering of the returned light, sensing type (e.g., amplitude, phase, and / or polarization), and compatible combinations of the items listed above.

[0112] In particular, it is to be understood that a viewing angle can be characterized by more than one item from the above list. That is, a combination of items from the above list. For example, a viewing angle can be characterized by the angle at which an incident light beam strikes the wafer surface (i.e., the illumination angle) and the polarization of the incident light beam (i.e., the illumination polarization). As another example, a viewing angle can be characterized by the collection angle and the collection phase (i.e., the phase of the collected beam). Further, it is to be understood that a viewing angle can combine characteristics from both the illumination channel and the collection channel. For example, a viewing angle can be characterized by the illumination polarization and the collection polarization. As another example, a viewing angle can be characterized by the illumination angle and polarization and the collection intensity and phase.

[0113] Thus, the acquired images can differ from each other by at least one parameter selected from the list of viewing angles specified above.

[0114] As used herein, in some embodiments, with reference to a list that includes a sublist (which includes a plurality of items (e.g., elements or claim limitations)) and at least one item not in the sublist, words such as "two or more of..." and "at least two of..." can refer only to two elements of the sublist, one element of the sublist and one listed element not in the sublist, two elements not in the sublist, and so on. For example, in some embodiments where at least one illumination spectrum includes two illumination spectra, the multiple viewing angles can consist of or include these two illumination spectra.

[0115] More generally, in some embodiments, the reflected and / or scattered light can undergo Fourier filtering before being detected. Fourier filtering can be used to increase the number of viewing angles and the amount of information obtainable therefrom. In some embodiments, the multiple viewing angles can include slightly defocused illumination.

[0116] According to some embodiments, for example when the illumination source is a laser, the illumination spectrum can be narrow. According to some embodiments, for example when the illumination source is an incoherent source such as a lamp, the illumination spectrum can be wide. According to some embodiments, at least one illumination spectrum includes a plurality of illumination spectra. Each illumination spectrum of the plurality of illumination spectra can be narrow - and optionally coherent (e.g., when the illumination light is a coherent laser) - or wide.

[0117] According to some embodiments, multi-view scan data can be obtained from a bright-field channel (i.e., bright-field reflected light) and / or a gray-field channel (i.e., gray-field scattered light). As used herein, according to some embodiments, the term "gray-field scattered light" is used broadly to refer to non-bright-field reflected light. In particular, according to some embodiments, the term "gray-field scattered light" can also be used to refer to dark-field scattered light.

[0118] According to some embodiments, images corresponding to different views can be obtained simultaneously or substantially simultaneously. According to some embodiments, images corresponding to different views can be obtained sequentially. According to some embodiments, some images corresponding to different views can be obtained simultaneously or substantially simultaneously, while some images corresponding to other views can be obtained at an earlier or later time.

[0119] According to some embodiments, the imager used to obtain scan data in operation 110 can include a plurality of detectors. For example, a first detector can be configured to detect the intensity of the returned beam, while a second detector can be configured to detect the polarization of the returned beam.

[0120] According to some embodiments in which all views are obtained simultaneously, each detector can be assigned (designated) to a different view. Alternatively, according to some embodiments in which all views are obtained sequentially, a single detector can be employed. According to some embodiments in which some of the views are obtained simultaneously and some of the views are obtained sequentially, at least some of the detectors can be assigned to subsets of multiple views, each subset including at least two of the views.

[0121] According to some embodiments, a segmented pupil can be employed to separate the returned radiation beam arriving at the pupil according to the reflection angle or scattering angle of a sub-beam of the returned radiation beam from the wafer. Different detectors can be assigned to detect the radiation from different pupil segments (one detector per pupil segment), such that each pupil segment constitutes a different collection channel corresponding to a different collection angle (and a different view). (The detectors can be positioned in a conjugate plane of the pupil plane, and the segmented pupil can be positioned in the pupil plane.)

[0122] According to some embodiments, method 100 includes operation 120, in which the scan data obtained in operation 110 undergoes integrated analysis to identify (detect) defects in the scanned area. As used herein, the term "integrated analysis" employed with respect to the analysis of multi-view scan data (i.e., scan data from at least two different views) refers to an analysis that utilizes scan data from multiple views in order to obtain an improved defect detection rate. According to some embodiments, the integrated analysis can take into account cross-view covariance, i.e., the covariance between at least some of the different views.

[0123] Optionally, according to some embodiments, method 100 may further include operation 125, in which it is determined whether the identified defect (i.e., the defect identified in operation 125) is of interest (or a nuisance). According to some such embodiments, defects determined to be of interest may be further classified. That is, operation 125 may determine the type of distortion that caused the defect. Some distortions may be specific to certain types of components (semiconductor devices) fabricated on the wafer, such as chips or other components, e.g., transistors. The classification may be based on measured or derived characteristics of the identified defects in the multiple views.

[0124] According to some embodiments, method 100 further includes operation 130, in which operations 110 and 120 (and optional operation 125) may be repeated with respect to additional regions of the wafer (e.g., with respect to other die segments). In particular, the additional regions may constitute one or more predefined larger regions of the wafer to be scanned (e.g., one or more dies). According to some embodiments, operations 110 and 120 (and optional operation 125) may be repeated until the entire wafer has been scanned.

[0125] According to some embodiments in which method 100 includes both operations 125 and 130, the order of operations 125 and 130 may be reversed.

[0126] Figure 2 A flowchart of operation 220 is presented, which is a specific embodiment of operation 120. According to some embodiments, operation 220 may include:

[0127] - Sub-operation 220a, in which a set of difference images of the scanned area is generated based on the acquired images (i.e., the multiple images acquired in operation 110) and corresponding reference data.

[0128] Each difference image in the set of difference images corresponds to one of the views (from the multiple views). Each difference image may be generated using one or more of the acquired images corresponding to the view and the reference data of the scanned area corresponding to the view.

[0129] Sub-operation 220b, in which, for each one of the multiple sub-images (e.g., pixels) of each difference image in the set, a (multiple) difference value (also referred to as a “(multiple) attribute”) is calculated. Sub-images corresponding to the same sub-region of the wafer in the scanned wafer region define a respective set of sub-images such that each sub-image in the set of sub-images can correspond to a different viewing angle (from multiple viewing angles). (In particular, in sub-operation 220b, for each sub-image in the set of sub-images (which corresponds to the same sub-region of the wafer), a respective difference value can be calculated, thereby resulting in a set of difference values corresponding to the sub-region of the wafer (and the set of sub-images).)

[0130] - Sub-operation 220c, in which each one of the multiple sub-regions of the wafer corresponding to the multiple sub-images of sub-operation 220b can be determined to be defective (or not defective) based at least on the set of difference values corresponding to the sub-region of the wafer and a respective (corresponding) set of noise values.

[0131] As used herein, according to some embodiments, a sub-region (e.g., having a size corresponding to a pixel or a small group of pixels) is referred to as “defective” when it includes a defect or a part of a defect.

[0132] As used herein, according to some embodiments, when the sub-image is a pixel, the term “(multiple) difference value” with respect to the sub-image and the “(multiple) pixel value” with respect to the same sub-image can be used interchangeably.

[0133] According to some embodiments, the reference data can include a reference image that has been obtained, for example, when scanning a wafer or when scanning wafers manufactured to have the same design, or a reference image generated based on design data of the wafer such as CAD data.

[0134] As used herein, the term "difference image" should be understood in a broad sense and can refer to any image obtained by combining at least two images (e.g., a first image (e.g., an image of a scanned area of a wafer or an image obtained from multiple images of the scanned area) and a second image (e.g., a reference image derived from reference data related to the scanned area)). The combination of the two images can involve any manipulation of the two images that results in at least one "difference image" that can reveal changes (differences) between the two images or, more generally, can distinguish (discriminate) between the two images (when differences exist). In particular, it is to be understood that the term "combination" with respect to two images can be used more broadly than subtracting one image from another and encompasses other mathematical operations that can be performed in addition to or in place of subtraction. Further, it is to be understood that one or both of the two images can be individually manipulated (i.e., pre-processed) before combining the two images to obtain a difference image. For example, the first image can be registered relative to the second image.

[0135] As used herein, the term "reference data" should be interpreted broadly to cover any data that indicates the physical design of a (patterned) wafer and / or data derived from the physical design (e.g., derived by simulation). According to some embodiments, "reference data" can include or consist of "design data" of the wafer (such as CAD data in various formats).

[0136] Additionally or alternatively, "reference data" can include or consist of data obtained by, for example, fully or partially scanning the wafer during recipe setup or even during runtime. For example, the scan of one die or multiple dice with the same architecture during runtime can be used as reference data for another die with the same architecture. Further, a first wafer manufactured to a certain design can be scanned during recipe setup, and the obtained scan data can be processed to generate reference data or additional reference data for a wafer of the same design (the same design as the first wafer) for subsequent manufacturing. Such "self-generated" reference data is necessary when design data is not available but can also be beneficial even when design data is available.

[0137] More generally, it is to be understood that the term "difference image" can refer to any set of derived values obtained by jointly manipulating two sets of values: a first set of values (obtained during a scan) and a second set of values (reference values obtained from reference data), such that each derived value in the set corresponds to a sub-region (e.g., a pixel) of the scanned region on the wafer. The joint manipulation can involve any mathematical operation on the two sets of values such that the (resulting) set of derived values can reveal a difference (if any) between the two sets of values or more generally can distinguish between the two sets of values. (The mathematical operation can include or can exclude subtraction.) In particular, the joint manipulation is not limited to the manipulation of corresponding value pairs. That is, each (difference) value in the set of difference values can be produced by the joint manipulation of multiple values in the first set and multiple values in the second set.

[0138] According to some embodiments, the set of difference values associated with a sub-image can also include scan data associated with adjacent sub-images or data generated based on scan data associated with adjacent sub-images. For example, according to some embodiments in which each sub-image is a pixel, the set of pixel values (e.g., intensity values) corresponding to a pixel can also include the pixel values of adjacent pixels. As used herein, according to some embodiments, two sub-images in a (given image, e.g., a difference image) can be referred to as "neighbors" when they are "closest neighbors". That is, the two sub-images are adjacent to each other in the sense that there are no other sub-images between them. According to some embodiments, two pixels can be referred to as "neighbors" not only when they are closest neighbors, but also when they are separated from each other by at most one pixel, at most two pixels, at most three pixels, at most five pixels, or even at most ten pixels. Each possibility corresponds to a different embodiment.

[0139] According to some embodiments, the set of difference values associated (corresponding) with a first sub-image also includes scan data associated with adjacent sub-images such that the first sub-image is centered relative to the adjacent sub-images.

[0140] In some embodiments in which the sub-images are pixels, the set of difference values associated with a first pixel also includes scan data associated with adjacent pixels, such that the first pixel and the adjacent pixels form a block of m×n pixels, where 3≤m≤11 and 3≤n≤11. Larger values of n and m are possible and may be desirable, for example when the size of a defect or the correlation length of noise is large. According to some such embodiments, the first pixel may be located at the center of the block. In particular, when the size of a suspected defect is larger than the first pixel (i.e., when the first pixel may only include (in the depicted sense) a part of the suspected defect), n and m may be selected such that the block (formed by the first pixel and the adjacent first pixels) completely depicts the suspected defect.

[0141] According to some embodiments, sub-operation 220c may include calculating a set of noise values. According to some embodiments, the set of noise values may be calculated based on the set of difference values corresponding to the sub-region.

[0142] According to some embodiments, method 100 may include a preliminary scan operation in which the wafer is partially scanned. More specifically, the wafer may be "sampled" in the sense of scanning a sample region of the wafer. Each region in the sample (i.e., each region from the sampled region) represents a wafer region characterized by a certain architecture, component type(s), etc.

[0143] According to some embodiments, in order to reduce the computational load and speed up wafer analysis, certain computational operations may be performed only on the preliminary scan data. For example, one or more dies from a group of dies fabricated with the same design may be sampled (in the preliminary scan operation). The scan data obtained from corresponding regions within the sampled dies may be used later (e.g., in sub-operation 220c) for corresponding regions of the non-sampled dies. In particular, according to some such embodiments, a set of noise values corresponding to the sampled region may be calculated and stored in a memory (i.e., before operation 110). The set of noise values may later be used in sub-operation 220c as part of the determination of whether the scanned region includes a defect.

[0144] According to some embodiments, operation 120 may additionally include a sub-operation in which images (of the same region) obtained at different times (in particular, times that differ by more than a typical time scale affecting the wafer analysis system (used to inspect the wafer) and / or high-frequency physical effects of the wafer) and associated with different perspectives are registered with each other. For example, before sub-operation 220a (i.e., according to Figure 2In embodiments implementing operation 120), this “viewpoint-to-viewpoint” registration can be implemented. In addition to standard die-to-die registration and / or cell-to-cell registration, viewpoint-to-viewpoint registration can also be implemented. According to some embodiments, for example, where different images associated with different viewpoints are offset from each other by a sub-pixel, a pixel, or even up to ten pixels, an alignment protocol can be employed. This can advantageously avoid the need to apply a registration protocol, which is relatively more cumbersome.

[0145] According to some embodiments, prior to sub-operation 220a, images in different viewpoints of the same scan region can be registered with each other. The registration can be implemented using scan data obtained from a common channel (which does not change when switching between viewpoints). According to some such embodiments, the multi-view scan data is obtained from a bright-field channel, while the gray-field channel is used to register the images with each other. Alternatively, according to some embodiments, the multi-view scan data is obtained from the gray-field channel, while the bright-field channel is used to register the images with each other. (In addition to standard die-to-die registration and / or cell-to-cell registration, “viewpoint-to-viewpoint” registration can also be implemented.) According to some embodiments, at least two viewpoints are always acquired at a time, with one viewpoint being common to all the acquired viewpoints.

[0146] Figure 3 A flowchart of sub-operation 320c, which is a specific embodiment of sub-operation 220c, is presented. According to some embodiments, sub-operation 320c can include the calculation of a covariance matrix (which constitutes a set of noise values). According to some embodiments, the calculation of the covariance matrix can be based on the corresponding set of difference values calculated in sub-operation 220b and / or scan data obtained during a preliminary scan of the wafer. The terms in the off-diagonal blocks of the covariance matrix include cross-viewpoint covariances (cross-viewpoint covariances between sub-images corresponding to different (adjacent) sub-regions and cross-viewpoint covariances between sub-images corresponding to the same sub-region). According to some such embodiments, determining whether a sub-region includes a defect (or a part of a defect) in sub-operation 220c can include:

[0147] - Sub-operation 320c1, multiplying a first vector v (whose components include the difference values in the set of difference values corresponding to the sub-region) by the inverse of the corresponding covariance matrix C to obtain a second vector u. (Note that the set of difference values corresponding to the sub-region also includes difference values related to adjacent sub-regions.)

[0148] - Sub - operation 320c2, which takes the scalar product of a second vector u and a third vector k (e.g., a predetermined kernel corresponding to a sub - region). The components of the third vector k can characterize the signature of a particular type of (multiple) defect that would appear in the difference image obtained (ideally) with substantially no wafer noise - the sub - region is suspected of at least partially including the particular type of (multiple) defect.

[0149] - Sub - operation 320c3, in which it is checked whether the scalar product exceeds a predetermined threshold B, and if so, the sub - region is marked as including a defect (or a part of a defect).

[0150] Figures 4A - 4G Present the algebraic representations used in the calculations involved in the Figure 3 sub - operations according to some embodiments. Figure 4A The first vector v for the case where the number of sub - images is n and the number of viewpoints is m is shown in. v thus includes n×m components. (Note that each of the vectors v, u, and k is defined as a column vector.) Each component of v can be labeled by a pair of indices i and j, where the index i = 1, 2, …, n represents the sub - images (e.g., pixels), and the index j = 1, 2, …, m represents the viewpoints. Thus, as Figure 4A defined in, the first n components of v (i.e., v 11 、v 12 、…、v 1n ) represent the difference values of (n sub - images) in the first viewpoint. Similarly, the components n + 1 to 2n of v (i.e., v 21 、v 22 、…、v 2n ) represent the difference values in the second viewpoint, and so on. The vector v is thus "composed" of m Figure 4B n - component vectors v j shown in. Each of v j corresponds to a different viewpoint (which is labeled by the index j).

[0151] Also refer to Figure 5A and Figure 5B , Figure 5A shows the possible ways of enumerating the pixels (more generally, sub - images) in the case where the number of pixels p i (i = 1, 2, …, 9) under consideration is nine, and thus shows the order of the terms in v (as well as C and k). In addition to the central pixel p 5 (which is the pixel to be analyzed), eight of its closest pixels are also shown. The set of pixel values (of the central pixel) includes not only the value related to the central pixel but also the values related to the eight surrounding pixels.

[0152] Figure 5B shows the case where the number of pixels p j (j = 1, 2, …, 5) is five, enumerating the possible ways of pixels (more generally sub - images), and thus showing the order of the terms in v (and C and k). In addition to the central pixel p 1 , four pixels closest to it are also shown. The set of pixel values (of the central pixel) includes not only the values related to the central pixel but also the values related to the four closest neighboring pixels.

[0153] Figure 4C shows the covariance matrix C. For the choice of the arrangement of the components within the first vector v above (i.e., as defined in Figure 4A and Figure 4B ), C takes a structure in which C is “composed” of m × m smaller matrices C ab (a = 1, 2, …, m; b = 1, 2, …, m) such that each of C ab is the covariance matrix of an n × n matrix. Each of the m C aa (a = 1, 2, …, m) corresponds respectively to the a - th perspective and is “associated” between different sub - images corresponding to the same (i.e., the a - th) perspective. Each of the “off - diagonal” matrices C a,b≠a (i.e., when b ≠ a) is “associated” between sub - images in different perspectives (i.e., the a - th and the b - th perspectives). C ab is shown in Figure 4D .

[0154] For the same case (i.e., where the number of sub - images is n and the number of perspectives is m), the third vector k is shown in Figure 4E . Similar to the first vector v, the third vector k is “composed” of m Figure 4F n - component vectors k j shown in j . Each of k

[0155] The second vector u (obtained in sub-operation 320c1) is the matrix product of the (one-dimensional matrix) v and the inverse of C. In sub-operation 320c3, it is checked whether k·u > B. Note that the value of the threshold B can depend on a predetermined kernel (i.e., on the characteristics of the defect(s) that the sub-region is suspected of including or partially including). The value of the threshold B can also vary from one sub-region to an adjacent sub-region, depending on the geometry of the corresponding pattern on the sub-region. This can be the case even when the sub-region and the adjacent sub-region each correspond to a pixel in size and each include a corresponding part of the same defect. A defect can typically have an area of at least about 10 nm × 10 nm and can affect the signal obtained from an area measured about 100 nm × 100 nm around the defect (i.e., corresponding to at least about 3 × 3 pixels in the case where a pixel on the wafer corresponds to an area of about 10 nm × 10 nm). The threshold B can be selected such that the percentage of false positives (i.e., the case where a defect-free sub-region of the wafer is incorrectly determined to be defective) does not exceed a predefined (threshold) ratio.

[0156] According to some embodiments, to accelerate the calculation, some of the non-diagonal terms or non-diagonal blocks of the covariance matrix are not calculated (e.g., some of the matrix C a,b≠a ). (If no non-diagonal blocks are calculated, the calculations involved are equivalent to calculating m smaller covariance matrices (e.g., Figure 4G as shown for the case of m = 3). Each of the m smaller covariance matrices corresponds to one of the perspectives, where m is the number of perspectives.)

[0157] As described above, the third vector k (i.e., the predetermined kernel) characterizes the signature of a defect or a family of defects (i.e., similar defects) in the absence (or substantially absence) of wafer noise and can be obtained by applying a matched filter to the signature of the defect or the family of defects in the presence of wafer noise in order to maximize the signal-to-noise ratio. According to some embodiments, the third vector k characterizes the signature of a specific type of defect that the sub-region is suspected of including (or partially including).

[0158] According to some embodiments, the predetermined kernel can be derived based on one or more of the following items: (i) experimental measurements performed on a wafer region known to include one or more defects; (ii) computer simulations of light scattering from defects; (iii) physical models describing the behavior of defects; and (iv) machine learning algorithms designed to provide an optimized kernel.

[0159] According to some embodiments, it may be known that some perspective pairs exhibit weaker correlations than other perspective pairs (e.g., based on scan data obtained during an initial scan). According to some such embodiments, in sub-operation 320c, items in blocks corresponding to perspective pairs known to exhibit weaker correlations are not computed to speed up the analysis.

[0160] According to some embodiments, in addition to covariance, higher moments of the joint probability distribution (which are related to the measured values obtained by the imager in operation 110) may be considered as part of the determination in sub-operation 220c as to whether a sub-region includes (or partially includes) a defect. For example, according to some embodiments, skewness and / or kurtosis may be considered.

[0161] While some of the above embodiments involve using optical scanning to implement method 100, as already mentioned, according to some embodiments, method 100 may also be implemented using a scanning electron microscope (SEM). According to some such embodiments, multiple perspectives include two or more of at least one intensity of the (multiple) illuminating electron beam(s) (e-beam), at least one intensity of the (multiple) returned electron beam(s), at least one spin of the (multiple) illuminating electron beam(s), at least one spin of the (multiple) returned electron beam(s), one or more incident angles of the (multiple) illuminating electron beam(s), and one or more collection angles of the (multiple) returned electron beam(s).

[0162] According to some alternative embodiments, method 100 may be implemented using an atomic force microscope (AFM). According to some such embodiments, multiple perspectives may include different types of AFM tips, different tapping modes, and / or applying the AFM at different resonant frequencies.

[0163] According to some embodiments where the image resolution (i.e., pixel size) provided by the imager may be higher than required (e.g., when implementing method 100 using SEM or AFM), or to accelerate wafer analysis, difference value pairs corresponding to pixels within a sub-image of the difference image may be averaged to obtain a single ("coarse-grained") difference value corresponding to the sub-image. In such embodiments, the set of difference values corresponding to a sub-region may include the average difference value related to the sub-image of the sub-region and the average difference value related to the sub-image of an adjacent sub-region in each of the multiple perspectives. (Each average difference value is obtained by averaging the difference values related to the pixels constituting the corresponding sub-image). Then, a covariance matrix may be computed based on the average difference values, potentially allowing a significant reduction in the computational load.

[0164] System

[0165] According to one aspect of some embodiments, a computerized system for obtaining and analyzing multi-perspective scan data of a wafer is provided.Figure 6 is a block diagram of such a computerized system (computerized system 600) according to some embodiments. System 600 includes a scanning device 602 and a scan data analysis module 604.

[0166] The scanning device 602 is configured to scan a wafer in each of multiple perspectives (such as the multiple perspectives listed above in the method subsection). According to some embodiments, scan data related to two or more of the multiple perspectives can be obtained simultaneously or substantially simultaneously. Additionally or alternatively, according to some embodiments, the scanning device 602 can be configured to scan the wafer, one perspective at a time (from the multiple perspectives). That is, the scanning device 602 can be configured to switch between perspectives.

[0167] The scan data analysis module 604 is configured to (i) receive the multi-perspective scan data obtained by the scanning device 602, and (ii) perform an integrated analysis of the multi-perspective scan data, as further detailed below.

[0168] According to some embodiments, the scanning device 602 includes a platform 612, a controller 614, an imager 616 (imaging device), and an optical device 618. The scanning device 602 is delineated by a double-dotted line box to indicate that components therein (such as the platform 612 and the imager 616) can be separated from each other, for example, in the sense of not being included in a common housing.

[0169] The platform 612 is configured to hold a sample to be inspected, such as a wafer 620 (or an optical mask). The wafer 620 can be patterned, but those skilled in the art will understand that method 100 can also be used to detect defects in a bare wafer. According to some embodiments, the platform 612 can be movable, as described below. The imager 616 can include one or more light emitters (such as visible light and / or ultraviolet light sources) configured to illuminate the wafer 620. Further, the imager 616 can include one or more light detectors. In particular, the imager 616 can apply collection techniques, including bright field collection, gray field collection, and so on. The optical device 618 can include optical filters (such as spatial filters, polarization filters, Fourier filters), beam splitters (such as polarization beam splitters), mirrors, lenses, prisms, gratings, deflectors, reflectors, apertures, etc., which are configured to allow the acquisition of scan data related to multiple perspectives. According to some embodiments, the optical device 618 can be configured to allow the scanning device 602 to switch between different perspectives. For example, the optical device 618 can include a polarization filter and / or a beam splitter configured to set the polarization of the emitted (illuminating) light and / or select the polarization of the collected (returned) light.

[0170] More specifically, according to some embodiments, the optical device 618 can include any arrangement of optical components configured to perform the following operations: determining (setting) one or more optical properties (such as shape, divergence, polarization) of a radiation beam from a radiation source of the imager 616 and the trajectory line of the incident radiation beam. According to some embodiments, the optical device 618 can further include any arrangement of optical components configured to perform the following operations: selecting (e.g., by filtering) one or more optical properties of one or more returned radiation beams before detecting the one or more returned radiation beams (e.g., beams specularly reflected by the wafer 620 or diffusely scattered from the wafer 620), and selecting the trajectory line followed by the one or more returned beams when the one or more returned beams return from the wafer 620. According to some embodiments, the optical device 618 can further include optical components configured to direct one or more returned radiation beams toward a detector of the imager 616.

[0171] The controller 614 can be functionally associated with the platform 612, the imager 616, the optical device 618, and the scan data analysis module 604. More specifically, the controller 614 is configured to control and synchronize the operations and functions of the modules and components listed above during the scanning of the wafer. For example, the platform 612 is configured to support the sample to be inspected (such as the wafer 620) and mechanically translate the sample to be inspected along a trajectory line set by the controller 614, and the controller 614 also controls the imager 616.

[0172] The scan data analysis module 604 includes computer hardware (one or more processors, such as image and / or graphics processing units, and volatile and non-volatile memory components; not shown). The computer hardware is configured to analyze the multi-view scan data received from the imager 616 of the area on the wafer 620 substantially as described above in the method subsection to determine the presence of defects.

[0173] The scan data analysis module 604 can further include an analog-to-digital (signal) converter (ADC) and a frame grabber (not shown). The ADC can be configured to receive an analog image signal from the imager 616. Each analog image signal can correspond to a different view from multiple views. The ADC can be further configured to convert the analog image signal into a digital image signal and transmit the digital image signal to the frame grabber. The frame grabber can be configured to obtain a digital image (block image or image frame) of the scanned area on the scanned wafer (such as the wafer 620) from the digital image signal. Each digital image can be in one of the multiple views. The frame grabber can be further configured to transmit the digital image to one or more of the processor and / or the memory components.

[0174] More specifically, the scan data analysis module 604 can be configured to:

[0175] - Generate a set of difference values in each of multiple perspectives based on the scan data of the scanned area received from imager 616 and corresponding reference data that may be stored in the (one or more) memory components. Substantially as described above in the method subsection in Figure 2 the description, each set of difference values corresponds to a sub-region (e.g., "pixel") of the scanned area.

[0176] - Substantially as described above in the method subsection in Figure 2 the description and in Figure 3 the description according to some embodiments of system 600, for each sub-region, determine whether the sub-region is defective based at least on the corresponding set of difference values and the corresponding set of noise values.

[0177] According to some embodiments, substantially as described above in the method subsection in Figure 2 the description and in Figure 3 the description according to some embodiments of system 600, the scan data analysis module 604 may be configured to: for each set of difference values and based at least on the set of difference values, generate a corresponding set of noise values. According to some embodiments, the generation of the set of noise values may be based at least on the scan data obtained in the (one or more) preliminary scans of the wafer, where a representative area of the wafer is scanned.

[0178] According to some embodiments, the determination of whether a sub-region is defective may be implemented taking into account the type of the (one or more) defects that the sub-region is suspected to include or partially include. In particular, the determination may involve calculating a covariance matrix and may further include a calculation involving a predetermined kernel and a corresponding threshold, the predetermined kernel characterizing the features of the (one or more) suspected defect types in the case of substantially no wafer noise.

[0179] According to some alternative embodiments not depicted in the drawings, a computerized system for obtaining and analyzing multi-perspective scan data of a wafer is provided. The system may be similar to system 600, but differs at least in that it uses (one or more) electron beams instead of electromagnetic radiation to irradiate the wafer. In such embodiments, the imager of the system may include a scanning electron microscope.

[0180] According to some alternative embodiments not depicted in the drawings, a computerized system for obtaining and analyzing multi-perspective scan data of a wafer is provided. The system may be similar to system 600, but differs at least in that it uses an atomic force microscope instead of an optically-based imager.

[0181] Figure 7AA computerized system 700 is schematically depicted. The computerized system 700 is a specific embodiment of the system 600. The system 700 includes a radiation source 722 and a plurality of detectors 724. The radiation source 722 and the plurality of detectors 724 together constitute an imager (or form part of an imager), which is a specific embodiment of the imager 616 of the system 600. The system 700 further includes a scan data analysis module 704, which is a specific embodiment of the scan data analysis module 604 of the system 600. The system 700 further includes a beam splitter 732 and an objective lens 734. The beam splitter 732 and the objective lens 734 together constitute an optical device (or form part of an optical device), which is a specific embodiment of the optical device 618 of the system 600. Also shown is a platform 712 (which is a specific embodiment of the platform 612 of the system 600) and a wafer 720 disposed on the platform 712.

[0182] Also indicated is the optical axis O of the objective lens 734. The optical axis O extends parallel to the z-axis.

[0183] In operation, light is emitted by the radiation source 722. The light is directed towards the beam splitter 732, and some of the light is transmitted through the beam splitter 732. The transmitted light is focused by the objective lens 734 onto the wafer 720 to form an illumination point S on the wafer 720. The returned light (which undergoes specular reflection from the wafer 720) is directed back towards the objective lens 734 and refracted by the objective lens 734 towards the beam splitter 732. A portion of the returned light (which has been refracted by the objective lens 734) is reflected by the beam splitter 732 towards the detector 724.

[0184] For ease of illustration, the trace lines of a pair of light rays are indicated. More specifically, a first light ray L 1 and a second light ray L 2 indicate the light rays emitted by the radiation source 722. A third light ray L 3 and a fourth light ray L 4 indicate the (returned) light rays that travel towards the detector 724 after being reflected by the beam splitter 732 (after having been scattered from the wafer 712 and refracted by the objective lens 734). The third light ray L 3 constitutes a portion of the first light ray L 1 that remains after passing through the beam splitter 732 and then being reflected by the beam splitter 732. The fourth light ray L 4 constitutes a portion of the second light ray L 2 that remains after passing through the beam splitter 732 and then being reflected by the beam splitter 732.

[0185] Also indicated is a segmented pupil 740 (segmented aperture, which also forms part of the optical device). The segmented pupil 740 may be positioned in the pupil plane, and the detector 724 may be positioned in a plane conjugate to the pupil plane. The segmented pupil 740 is divided into a plurality of pupil segments (or sub-apertures). The segmentation of the pupil allows the returned beam reaching the pupil (e.g., the light beam reflected from the wafer) to be separated into sub-beams, which is done according to the respective return angles of each of the sub-beams, such that each pupil segment will correspond to a different viewing angle. That is, each of the viewing angles generated by the segmented pupil 740 corresponds to a different collection angle.

[0186] As a non-limiting example, in Figure 7A it is shown that the segmented pupil 740 is divided into nine pupil segments 740a to 740i, which are arranged in a square array, and the detector 724 includes nine corresponding detectors 724a to 724i. The system 700 is configured such that the light reaching each of the pupil segments (which originates from the radiation source 722 and has undergone specular reflection from the wafer 720) continues from the pupil segment towards the corresponding detector from the detector 724. That is, the light passing through the first pupil segment 740a is sensed by the first detector 724a, the light passing through the second pupil segment 740b is sensed by the second detector 724b, and so on. Thus, each of the detectors 724 is configured to sense light returning at different angles.

[0187] According to some embodiments, the optical device may further include an optical guiding mechanism (not shown) for guiding the light passing through each of the pupil segments. The optical guiding mechanism may be configured to ensure that the light passing through the pupil segments is guided to the corresponding (target) detector (from the detector 724) without "leaking" to other detectors.

[0188] According to some embodiments, and as Figure 7A depicted in, the optical device may be configured such that the light reaching the objective lens 734 directly from the radiation source 722 arrives there as a collimated beam. The wafer 720 may be positioned or substantially positioned at the focal plane of the objective lens 734 such that the light rays incident on the wafer 720 form an illumination point S on the wafer 720, and the illumination point S may be as small as about 100 nanometers.

[0189] The different light rays that have been refracted by the objective lens 734 from the collimated beam may be incident on the wafer 720 at different angles. The refracted portion of the first light ray L 1 is incident on the wafer 720 at a first incident angle θ 1 (i.e., the angle formed by the refracted portion and the optical axis O), and the refracted portion of the second light ray L 2 is incident on the wafer 720 at a second incident angle θ 2is incident on the wafer 720. For the sake of illustration, it is assumed that θ 2 is equal to θ 1 , such that the trajectory line followed by the refracted portion of the second light ray L 2 from the objective lens 734 to the wafer 720 is retraced by the refracted portion of the first light ray L 1 after being reflected from the wafer 720. Similarly, the trajectory line followed by the refracted portion of the first light ray L 1 from the objective lens 734 to the wafer 720 is retraced by the refracted portion of the second light ray L 2 after being reflected from the wafer 720.

[0190] Thus, when there is no ambiguity in the context, θ 2 can be used to refer to the reflection (return) angle at which the refracted portion of the first light ray L 1 leaves the wafer 720, rather than the incident angle of the refracted portion of the second light ray L 2 on the wafer 720. Similarly, when there is no ambiguity in the context, θ 1 can be used to refer to the reflection (return) angle at which the refracted portion of the second light ray L 2 leaves the wafer 720, rather than the incident angle of the refracted portion of the first light ray L 1 on the wafer 720.

[0191] Note that not only the incident angle can be relevant to multi - perspective wafer analysis, but the azimuth angle can also be relevant to multi - perspective wafer analysis, especially when the wafer 720 is patterned (due to one or more asymmetries introduced by the pattern relative to the wafer surface). That is, the angle formed by the "projection" of the incident light ray on the wafer surface with the x - axis of the orthogonal coordinate system that parameterizes the lateral dimension of the wafer surface. In Figure 7B , the incident angle (or polar angle) θ i of the light ray L i incident on the wafer 720 and the first azimuth angle are also indicated. The reflection angle (or polar angle) θ r of the light ray L r reflected from the wafer 720 is also indicated, where θ i = θ

[0192] Each of the detectors 724 is positioned to detect light rays that have impinged on the wafer 720 at a polar angle θ (or more precisely, a continuous range of polar angles centered on θ) and an azimuth angle (or more precisely, a continuous range of azimuth angles centered on ).

[0193] According to some embodiments, system 700 may further include an infrastructure for sensing light (e.g., a suitably positioned detector) that has diffusely scattered from wafer 720 (specifically, light rays outside the light cone generated by objective 734). According to some embodiments, system 700 may be configured to use an image generated from sensed gray-field scattered light as an additional perspective(s) and / or reference image for perspective-to-perspective registration.

[0194] Substantially as described with respect to scan data analysis module 604 of system 600, scan data analysis module 704 is configured to receive scan data from detector 724 and determine whether the scanned region includes one or more defects based on the scan data. The scan data from each of detectors 724a through 724i may be used separately to generate difference images I 1 to I 9 , each difference image being at a different perspective.

[0195] Since in Figure 7A the segmented pupil 740 is depicted as including nine pupil segments and the detectors from detector 724 correspond to each of these pupil segments, the number of perspectives is nine. Thus, the set of difference values associated with the first "pixel" on the wafer (i.e., a sub-region of the size corresponding to an image pixel) includes 9×(N + 1) elements (difference values), where N is the number of neighboring pixels considered. That is, N is the number of neighboring pixels whose difference values are included in the set of difference values associated with the first pixel. For example, when the number of neighboring pixels is eight (substantially as depicted in Figure 5A ), the set of difference values includes 81 elements. (The predetermined kernel also includes 81 elements.) Then, the covariance matrix is an 81×81 matrix.

[0196] Although in Figure 7A the pupil segments are depicted as having equal shape and size, it is to be understood that in general, the shapes and / or sizes of the different pupil segments of segmented pupil 740 may differ from each other. In particular, according to some embodiments, the different pupil segments may differ in area (i.e., the lateral dimension of the pupil segment parallel to the zx plane) and in the corresponding longitudinal extent of the pupil segment (e.g., the y coordinate of the entrance and / or exit of the pupil segment may vary from one pupil segment to another).

[0197] Figure 8FIG. 800 schematically depicts a computerized system 800, which is a specific embodiment of system 600. System 800 is similar to system 700, but differs in that it includes optical components that separate the light returning from the wafer into different polarizations, thereby allowing the number of viewing angles to be doubled. More specifically, system 800 includes a radiation source 822, a first plurality of detectors 824, and a second plurality of detectors 826, which together constitute an imager (or form part of an imager), and the imager is a specific embodiment of imager 616 of system 600. System 800 further includes a scan data analysis module 804, which is a specific embodiment of scan data analysis module 604 of system 600. System 800 further includes a first beam splitter 832, an objective lens 834, a second beam splitter 836, a first segmented pupil 840, and a second segmented pupil 850, which together constitute an optical device (or form part of an optical device), and the optical device is a specific embodiment of optical device 618 of system 600. The second beam splitter 836 is a polarization beam splitter. A platform 812 (which is a specific embodiment of platform 612 of system 600) and a wafer 820 disposed on the platform 812 are also shown.

[0198] According to some embodiments, the radiation source 822 may be similar to the radiation source 722, and each of the plurality of detectors 824 and 826 may be similar to the plurality of detectors 724. The first beam splitter 832 and the objective lens 834 may be similar to the beam splitter 732 and the objective lens 734, and each of the segmented pupils 840 and 850 may be similar to the segmented pupil 740.

[0199] Substantially as described above with respect to system 700, in operation, a portion of the light beam emitted by the radiation source 822 is transmitted through the first beam splitter 832, focused by the objective lens 834 (to form an illumination point S' on the wafer 820), returned by the wafer 820, refocused by the objective lens 834 again, and reflected from the first beam splitter 832. The portion of the returned light beam reflected from the first beam splitter 832 travels towards the second beam splitter 836 and is split by the second beam splitter 836 into two light beams of different polarizations (e.g., s-polarized light and p-polarized light): a first polarized light beam and a second polarized light beam. The first polarized light beam travels towards the first segmented pupil 840 and the first plurality of detectors 824, and the second polarized light beam travels towards the second segmented pupil 850 and the second plurality of detectors 826 (such that each combination of pupil segment and polarization is assigned a detector).

[0200] The arrows indicating the trajectory lines of the light rays emitted by the radiation source 822 are not numbered.

[0201] Basically as described with respect to the scan data analysis module 604 of system 600, the scan data analysis module 804 is configured to receive scan data from detectors 824 and 826 and determine whether the scanned area includes one or more defects based on the scan data. The scan data from each of the first detectors 824a through 824i can be used separately to generate difference images J 1 through J 9 , each difference image at a different perspective. The scan data from each of the second detectors 826a through 826i can be used separately to generate difference images J 10 through J 18 , each difference image at a different perspective (and at a polarization different from that of difference images J 1 through J 9 ). Thus, two difference images in two different perspectives can be obtained from each pair of the polar and azimuth angles characterizing the light rays returned from the wafer 820: a first difference image corresponding to a first polarization and a second polarization corresponding to a second polarization.

[0202] Since in Figure 8 , each of the segmented pupils 840 and 850 is depicted as including nine pupil segments, and the detectors from detectors 824 and 826 respectively correspond to each of the pupil segments, the number of perspectives is eighteen. Thus, the set of difference values associated with the first "pixel" on the wafer includes 18×(N'+1) elements (difference values), where N' is the number of neighboring pixels considered. For example, when the number of neighboring pixels is eight, the set of difference values includes 162 elements. (The predetermined kernel also includes 162 elements). Then, the covariance matrix is a 162×162 matrix.

[0203] Figure 9A computerized system 900 is schematically depicted. The computerized system 900 is a specific embodiment of the system 600. The system 900 includes a radiation source 922, a first detector 924, a second detector 926, and a third detector 928, which together constitute an imager (or form part of an imager), and the imager is a specific embodiment of the imager 616 of the system 600. The system 900 further includes a scan data analysis module 904, and the scan data analysis module 904 is a specific embodiment of the scan data analysis module 604 of the system 600. The system 900 further includes a first beam splitter 932, an objective lens 934, a second beam splitter 936, a third beam splitter 938, a first polarizer 942, and a second polarizer 944, which together constitute an optical device (or form part of an optical device), and the optical device is a specific embodiment of the optical device 618 of the system 600. The (unsegmented) pupils in front of each of the detectors 924, 926, and 928 are not shown. A platform 912 (which is a specific embodiment of the platform 612 of the system 600) and a wafer 920 placed on the platform 912 are also shown.

[0204] The first polarizer 942 is positioned in front of the second detector 926, and the second polarizer 944 is positioned in front of the third detector 928. The first polarizer 942 is configured to filter out light of a first polarization, and the second polarizer 944 is configured to filter out light of a second polarization, and the second polarization is different from the first polarization.

[0205] Substantially as described above with respect to the system 700, in operation, a portion of the light beam emitted by the radiation source 922 is transmitted through the first beam splitter 932, focused by the objective lens 934 (to form an illumination point S” on the wafer 920), returned by the wafer 920, focused again by the objective lens 934, and reflected from the first beam splitter 932. The portion of the returned beam reflected from the first beam splitter 932 travels towards the second beam splitter 936 and is split by the second beam splitter 936 into a first returned sub-beam and a second returned sub-beam. The first returned sub-beam constitutes the portion of the returned beam transmitted through the second beam splitter 936. The second returned sub-beam constitutes the portion of the returned beam reflected by the second beam splitter 936.

[0206] The first returned sub-beam travels towards the first detector 924 and is sensed by the first detector 924. The second returned sub-beam travels towards the third beam splitter 938 and is split by the third beam splitter 938 into a transmitted portion and a reflected portion. The transmitted portion travels towards the first polarizer 942, and the reflected portion travels towards the second polarizer 944. The polarizers 942 and 944 can be aligned at different angles such that each of the second detector 926 and the third detector 928 senses light of a different polarization. Thus, the detectors 924, 926, and 928 can be configured to provide readings sufficient to fully characterize the polarization of the returned beam (which is reflected from the wafer 920).

[0207] The arrow indicating the trace line of the light emitted by the radiation source 922 is not numbered.

[0208] Substantially as described with respect to the scan data analysis module 604 of system 600, the scan data analysis module 904 is configured to receive scan data from detectors 924, 926, and 928 and determine whether the scanned area includes one or more defects based on the scan data. The scan data from each of detectors 924, 926, and 928 can be used separately to generate difference images K 1 , K 2 and K 3 , each difference image being at a different perspective.

[0209] Since the pupil of system 900 (not shown) is not segmented, unlike the pupils of systems 700 and 800, the number of perspectives is three (one perspective for each detector). Thus, the set of difference values associated with the first "pixel" on the wafer includes 3×(N″ + 1) elements (difference values), where N″ is the number of adjacent pixels considered. For example, when the number of adjacent pixels is eight, the set of difference values includes 27 elements. (The predetermined kernel also includes 27 elements). Then, the covariance matrix is a 27×27 matrix.

[0210] Note that according to some embodiments, a single polarization beam splitter can be used instead of the combination of the third beam splitter 938 with the first polarizer 942 and the second polarizer 944.

[0211] Simulation results

[0212] This subsection describes simulation results that demonstrate the efficacy of the above-described methods (e.g., method 100) and systems. Figure 10A Multi-perspective scan data obtained by a simulation computerized system (such as system 700) is presented. The multi-perspective scan data includes nine images (enumerated by Roman numerals I through IX) of a square region of a (simulated) wafer, each image at a different perspective. The region is considered to be uniform except for a distortion (possibly introduced by a dust particle) in the center of the region (i.e., at the central pixel). The scale of the region is set to 1 μm 2 . Each of images I through IX is an intensity image corresponding to a different collection angle (which can be obtained by way of a segmented pupil such as segmented pupil 740). An intensity scale ranging from black to white is also indicated, where black corresponds to zero (I = 0) or the minimum intensity, and white corresponds to the maximum intensity reading (I = I max ) or an intensity reading above that.

[0213] In each of Images I through IX, the intensity of the central pixel typically varies from one pixel to the next and on average appears neither brighter nor darker than the surrounding pixels. In other words, in any of the images, the defect is not apparent to the naked eye, even when the images are viewed side by side.

[0214] As explained above, a pixel can be determined to be defective when the quantity s ij is greater than a corresponding threshold, where s ij = k ij ·((C ij ) -1 v ij ), and the indices i and j label the pixel (i and j represent the row and column of the pixel, respectively). Here, v ij is the first vector corresponding to the (i,j)th pixel, C ij is the covariance matrix corresponding to the (i,j)th pixel, and k ij is the third vector (kernel) corresponding to the (i,j)th pixel.

[0215] Figure 10B is a graphical representation of s ij corresponding to the simulation region when cross-view covariance is not taken into account. This effectively amounts to setting the off-diagonal blocks of C ij to zero. s ij is arranged in a square array according to the values assumed by i and j. (Note that since the simulation region is intended to be uniform according to its "bare design", the threshold B for all pixels can be taken to be the same, and no additional information is obtained by subtracting B from s ij .) Also indicated is the scale s = k·(C -1 v), which ranges from black to white, with black corresponding to s = s min and white corresponding to s = s max .

[0216] Figure 10C is a graphical representation of s ij corresponding to the simulation region when cross-view covariance is taken into account (i.e., all components of C ij are calculated).

[0217] Figure 10B A dashed circle D is drawn around the central pixel (which corresponds to the defective pixel), where the central pixel is indicated by the arrow d. Figure 10C A dashed circle D' is drawn around the central pixel, where the central pixel is indicated by the arrow d'. It can be readily seen that the central pixel appears much brighter in Figure 10C than in Figure 10B , i.e., the defect signal is inFigure 10C is much stronger than that in Figure 10B which demonstrates the improved defect detection ability of the disclosed method. Considering cross - perspective covariance increases the signal - to - noise ratio from ≈0.7 to ≈2.2.

[0218] According to one aspect of some embodiments, a method for obtaining information about an area of a sample (e.g., a wafer) is provided. The method includes the following steps:

[0219] - Obtaining a plurality of images of the area by an imager. The plurality of images may differ from each other in at least one parameter selected from the following items: illumination spectrum, collection spectrum, illumination polarization, collection polarization, illumination angle, collection angle, and sensing type (e.g., intensity, phase, polarization). The step of obtaining a plurality of images includes illuminating (irradiating) the area and collecting radiation from the area. The area includes a plurality of area pixels (i.e., the area includes a plurality of sub - areas, each of the sub - areas having a size corresponding to a pixel).

[0220] - Receiving or generating a plurality of reference images.

[0221] - Generating a plurality of difference images by an image processor (e.g., a scan data analysis module), the difference images representing the differences between the plurality of images and the plurality of reference images.

[0222] - Calculating a set of area pixel attributes (a set of pixel values) for each area pixel in the plurality of area pixels (i.e., for each pixel in the area). The calculation is performed based on the pixels of the plurality of difference images.

[0223] - Calculating a set of noise attributes based on the sets of area pixel attributes of the plurality of area pixels (i.e., based on the sets of pixel values corresponding to each of the plurality of area pixels). Note that the covariance matrix (and its inverse) is a set of numbers characterizing the statistical properties of the noise. Those statistical properties are generally referred to as “attributes”. Using the covariance matrix as the statistical property is a particular non - limiting embodiment.

[0224] - Determining whether an area pixel represents a defect for each area pixel based on the relationship between the set of noise attributes and the set of area pixel attributes of the pixel.

[0225] According to some embodiments of the method, the step of determining whether an area pixel represents a defect is also performed in response to a set of attributes of an actual defect.

[0226] According to some embodiments of the method, the step of determining whether an area pixel represents a defect is also performed in response to a set of attributes of an estimated defect.

[0227] According to some embodiments of the method, the method includes the following step: calculating a set of noise attributes by calculating a covariance matrix.

[0228] According to some embodiments of the method, the step of calculating the covariance matrix includes: calculating a set of covariance values for each regional pixel, where the set of covariance values represents the covariance between different attributes of the set of regional pixel attributes of the regional pixel (i.e., between different perspectives), and calculating a given covariance matrix based on the multiple sets of covariance values of multiple regional pixels.

[0229] According to some embodiments of the method, after calculating the covariance matrix, the inverse of the covariance matrix is used for further calculations. The inverse of the covariance matrix is multiplied by the set of attributes representing the defect of interest (rather than noise).

[0230] According to some embodiments of the method, the method further includes the following steps: for each regional pixel, determining whether the regional pixel represents a defect by comparing the product of the multiplication between the following items with a threshold (e.g., threshold B): (i) the set of attributes of the regional pixel (e.g., the first vector v), (ii) the inverse of the covariance matrix corresponding to the noise affecting the set of attributes of the regional pixel (e.g., matrix C -1 ), and (iii) the set of attributes of the defect of interest.

[0231] According to some embodiments of the method, the set of pixel attributes of the regional pixel includes data on the regional pixel and the adjacent regional pixels of the regional pixel.

[0232] According to some embodiments of the method (e.g., as shown in Figure 7A and Figure 8 ), the imager includes multiple detectors for generating multiple images, and the method further includes the following steps: allocating different detectors to detect radiation from different pupil segments among multiple pupil segments (of a segmented pupil).

[0233] According to some embodiments of the method, different pupil segments among the multiple pupil segments exceed four pupil segments.

[0234] According to some embodiments of the method (e.g., as shown in Figure 8 ), the imager includes multiple detectors for generating multiple images, and the method further includes the following steps: allocating different detectors to detect radiation from different combinations of (a) polarization and (b) different pupil segments among the multiple pupil segments.

[0235] According to some embodiments of the method, the method includes the following steps: obtaining multiple images at the same time point.

[0236] According to some embodiments of the method, the method includes the following steps: obtaining multiple images at different time points.

[0237] According to some embodiments of the method, the method further comprises the step of: classifying the defects.

[0238] According to some embodiments of the method, the method further comprises the step of: determining whether the defect is a defect of interest or not a defect of interest.

[0239] According to one aspect of some embodiments, there is provided a computerized system for obtaining information about a region of a sample (e.g., a region on a wafer). The system includes an imager, the imager including optics and an image processor. The imager is configured to obtain a plurality of images of the region. The plurality of images may differ from each other in at least one parameter selected from the following items: illumination spectrum, collection spectrum, illumination polarization, collection polarization, illumination angle, and collection angle. The step of obtaining the plurality of images includes illuminating the region and collecting radiation from the region. The region includes a plurality of region pixels. The computerized system is configured to receive or generate a plurality of reference images. The image processor is configured to:

[0240] - Generate a plurality of difference images representing the differences between the plurality of images and the plurality of reference images.

[0241] - Calculate a set of region pixel attributes for each of the plurality of region pixels. The set of region pixel attributes is calculated based on the pixels of the plurality of difference images.

[0242] - Calculate a set of noise attributes based on the sets of region pixel attributes of the plurality of region pixels.

[0243] - For each region pixel, determine whether the region pixel represents a defect based on the relationship between the set of noise attributes and the set of region pixel attributes of the pixel.

[0244] According to one aspect of some embodiments, there is provided a non - transitory computer - readable medium storing instructions that cause a computerized system to perform the following operations:

[0245] - Obtain a plurality of images of a region of an object (e.g., a region on a wafer) by the imager of the computerized system (as described above). The plurality of images differ from each other in at least one parameter selected from the following items: illumination spectrum, collection spectrum, illumination polarization, collection polarization, illumination angle, collection angle, and sensing type. The step of obtaining the plurality of images includes illuminating the region and collecting radiation from the region. The region includes a plurality of region pixels.

[0246] - Receive or generate a plurality of reference images.

[0247] - Generate, by the image processor of the computerized system, a plurality of difference images that represent the differences between the plurality of images and the plurality of reference images.

[0248] - Calculate a set of region pixel attributes for each of a plurality of region pixels, wherein the calculation steps are performed based on pixels of a plurality of difference images.

[0249] - Calculate a set of noise attributes based on the sets of region pixel attributes of the plurality of region pixels.

[0250] - For each region pixel, determine whether the region pixel represents a defect based on the relationship between the set of noise attributes of the pixel and the set of region pixel attributes.

[0251] Although the present disclosure focuses on the scanning and inspection of wafers, those skilled in the art will understand that the disclosed methods and systems can also be applied to detecting irregularities in optical masks used in wafer manufacturing ("mask inspection").

[0252] As used herein, according to some embodiments, the terms "collection channel" and "detection channel" may be used interchangeably. According to some embodiments, the symbols "V data", "Cov", and "V defect" may be used to indicate the first vector v, the covariance matrix C, and the third vector k, respectively.

[0253] As used herein, according to some embodiments, the term "group" may refer not only to a plurality of elements (e.g., components, features), but also to a single element. In the latter case, the group may be referred to as a "single-member group".

[0254] It should be understood that certain features of the present disclosure that are described in the context of separate embodiments for clarity may also be provided in combination in a single embodiment. Conversely, the various features of the present disclosure that are described in the context of a single embodiment for brevity may also be provided separately or in any suitable sub-combination or in a suitable manner in any other described embodiment of the present disclosure. The features described in the context of an embodiment are not considered to be essential features of the described embodiment unless specifically so specified.

[0255] Although the operations of the methods according to some embodiments may be described in a specific sequence, the methods of the present disclosure may also include some or all of the described operations implemented in a different order. The methods of the present disclosure may include some or all of the described operations. No particular operation in the disclosed methods is considered to be an essential operation of the method unless specifically so specified.

[0256] Although the present disclosure is described in connection with its specific embodiments, it will be apparent that many alternatives, modifications, and variations will be apparent to those skilled in the art. Accordingly, the present disclosure encompasses all such alternatives, modifications, and variations that fall within the scope of the appended claims. It is to be understood that the present disclosure is not necessarily limited in its application to the details of construction and arrangement of components and / or methods set forth herein. Other embodiments may be practiced and the embodiments may be implemented in various ways.

[0257] The language and terminology used herein are for the purpose of description and should not be regarded as limiting. The citation or identification of any reference in this application should not be construed as an admission that such reference is available as prior art to the present disclosure. Section headings are used herein to facilitate understanding of the specification and should not be construed as necessarily limiting.

Claims

1. A method for detecting defects on a sample, the method comprising the steps of: obtaining scan data of a first region of the sample in multiple perspectives; and performing an integrated analysis of the obtained scan data, the integrated analysis including: calculating a cross-perspective covariance based on the obtained scan data; and determining the presence of a defect in the first region, taking into account the cross-perspective covariance, wherein determining the presence of the defect in the first region further includes: for each of a plurality of sub-regions of the first region, generating a difference value in each of the multiple perspectives based on the obtained scan data of the first region in each of the multiple perspectives and corresponding reference data; and determining whether each of the plurality of sub-regions is defective based at least on the difference value corresponding to the sub-region and a sub-region adjacent to the sub-region and a noise value corresponding to the sub-region and the adjacent sub-region, the noise value including a corresponding covariance from the cross-perspective covariance, wherein the step of determining whether each of the plurality of sub-regions is defective includes: generating a covariance matrix C, the covariance matrix C including the noise value corresponding to the sub-region and a sub-region adjacent to the sub-region; multiplying a first vector v including the difference value corresponding to the sub-region and the adjacent sub-region by the inverse of the covariance matrix C to obtain a second vector; calculating a scalar product of the second vector and a third vector, the components of the third vector including values characterizing a defect; and if the scalar product is greater than a predetermined threshold, marking the sub-region as defective, wherein: n is the number of the plurality of sub-regions, m is the number of the plurality of perspectives, the first vector v includes n×m components, v ij represents the value of the j-th sub-region in the i-th perspective, i = 1, 2, …, n, j = 1, 2, …, m, the covariance matrix C is composed of m×m matrices C ab such that each of the matrices C ab is the covariance matrix of an n×n matrix, a = 1, 2, …, m, b = 1, 2, …, m, C a,b＝a (a = 1, 2, …, m) correlates between different sub-regions in the a-th perspective respectively, C a,b≠a correlates between sub-regions in the a-th perspective and the b-th perspective respectively.

2. The method according to claim 1, wherein the sample is a patterned wafer.

3. The method according to claim 1, wherein the multiple perspectives include two or more of the following items: one or more incident angles of one or more illumination beams, one or more collection angles of one or more collected beams, at least one intensity of the one or more illumination beams, and at least one intensity of the one or more collected beams, and compatible combinations of the above items.

4. The method according to claim 1, wherein the method is optical, and wherein the multiple perspectives include two or more of the following items: one or more illumination angles, intensity of illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focal offsets of an illumination beam, one or more collection angles, intensity of collected radiation, collection polarization, phase of one or more collected beams, bright field channel, gray field channel, Fourier filtering of returned light, and a sensing type selected from intensity, phase, or polarization, and compatible combinations of the above items.

5. The method according to claim 1, further comprising the steps of: Generating a difference image of the first region in each of the multiple perspectives based on the acquired scan data and the corresponding reference data, and wherein the difference value corresponding to each sub-region from the plurality of sub-regions is derived from and / or characterizes the sub-image of the difference image corresponding to the sub-region.

6. The method according to claim 1, wherein the noise value is calculated based at least on the difference value.

7. The method according to claim 1, wherein at least one of the plurality of sub-regions has a size corresponding to a single pixel.

8. The method according to claim 1, wherein the cross-perspective covariance is calculated based at least on the scan data obtained during a preliminary scan of the sample, in which regions of the sample are sampled, each sampled region representing a group of regions of the sample, wherein at least one of the sampled regions represents the first region.

9. The method according to claim 1, further comprising the steps of: When the presence of a defect is determined, determining whether the defect is a defect of interest.

10. The method according to claim 9, further comprising the steps of: When the defect is determined to be of interest, classifying the defect.

11. The method according to claim 1, characterized in that repeating for each of a plurality of additional regions so as to scan a larger region of the sample formed by the first region and the additional regions.

12. A computerized system for acquiring and analyzing multi-perspective scan data of a sample, the system comprising: A scanning device configured to scan regions of a sample in multiple perspectives; and A scan data analysis module configured to perform an integrated analysis of the scan data obtained in the scan, the integrated analysis including: Calculating a cross-perspective covariance based on the acquired scan data; and Determining the presence of a defect in the region, taking into account the cross-perspective covariance, wherein determining the presence of the defect in the region further includes: For each of a plurality of sub-regions of the region, generating a difference value in each of the multiple perspectives based on the acquired scan data and the corresponding reference data of the region in each of the multiple perspectives; and Determining whether each of the plurality of sub-regions is defective based at least on the difference value corresponding to the sub-region and sub-regions adjacent to the sub-region and a noise value corresponding to the sub-region and the adjacent sub-regions, the noise value including the corresponding covariance from the cross-perspective covariance, wherein the step of determining whether each of the plurality of sub-regions is defective includes: Generating a covariance matrix C, the covariance matrix C including the noise values corresponding to the sub-region and the sub-regions adjacent to the sub-region; Multiplying a first vector v including the difference values corresponding to the sub-region and the adjacent sub-regions by the inverse of the covariance matrix C to obtain a second vector; Calculate the scalar product of the second vector and the third vector k, where the components of the third vector k include values characterizing the defect; and If the scalar product is greater than a predetermined threshold, label the sub-region as defective, wherein: n is the number of the multiple sub-regions, m is the number of the multiple perspectives, the first vector v includes n×m components, v ij represents the value of the j-th sub-region in the i-th perspective, i = 1, 2, …, n, j = 1, 2, …, m, the covariance matrix C consists of m×m matrices C ab such that each of the matrices C ab is the covariance matrix of an n×n matrix, a = 1, 2, …, m, b = 1, 2, …, m, C a,b＝a (a = 1, 2, …, m) correlates between different sub-regions in the a-th perspective respectively, and C a,b≠a correlates between sub-regions in the a-th perspective and the b-th perspective respectively.

13. The system according to claim 12, wherein the scanning device includes an optically based imager, and wherein the multiple viewing angles include two or more of the following items: one or more illumination angles, intensity of illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focus offsets of the illumination beam, one or more collection angles, intensity of the collected radiation, collection polarization, phase of one or more of the collected beams, bright field channel, gray field channel, Fourier filtering of the returned light, and a sensing type selected from intensity, phase, or polarization, and compatible combinations of the above items.

14. The system according to claim 12, wherein the scan data analysis module is further configured to: generate a difference image of the region in each of the multiple viewing angles based on the obtained scan data and the corresponding reference data, and wherein the difference value corresponding to each sub-region from the plurality of sub-regions is derived from and / or characterizes the sub-image corresponding to the sub-region of the difference image.

15. A non-transitory computer-readable storage medium storing instructions that cause a sample analysis system to perform the following operations: Scan a region of a sample in multiple viewing angles ； and Perform an integrated analysis of the scan data obtained in the scan, the integrated analysis including: Calculate the cross-view covariance based on the obtained scan data; and Determine the presence of a defect in the region taking into account the cross-view covariance, wherein determining the presence of the defect in the region further includes: For each of a plurality of sub-regions of the region, generate a difference value in each of the multiple viewing angles based on the obtained scan data and the corresponding reference data of the region in each of the multiple viewing angles; and Determine whether each of the plurality of sub-regions is defective based at least on the difference value corresponding to the sub-region and the sub-regions adjacent to the sub-region and the noise value corresponding to the sub-region and the adjacent sub-regions, the noise value including the corresponding covariance from the cross-view covariance, wherein the step of determining whether each of the plurality of sub-regions is defective includes: Generate a covariance matrix C, the covariance matrix C including the noise values corresponding to the sub-region and the sub-regions adjacent to the sub-region; Multiply a first vector v including the difference values corresponding to the sub-region and the adjacent sub-regions by the inverse of the covariance matrix C to obtain a second vector; Calculate the scalar product of the second vector and the third vector k, where the components of the third vector k include values characterizing the defect; and If the scalar product is greater than a predetermined threshold, label the sub-region as defective, wherein: n is the number of the multiple sub-regions, m is the number of the multiple perspectives, the first vector v includes n×m components, v ij represents the value of the j-th sub-region in the i-th perspective, i = 1, 2, …, n, j = 1, 2, …, m, the covariance matrix C is composed of m×m matrices C ab such that each of the matrices C ab is the covariance matrix of an n×n matrix, a = 1, 2, …, m, b = 1, 2, …, m, C a,b＝a (a = 1, 2, …, m) correlates between different sub-regions in the a-th perspective respectively, and C a,b≠a correlates between the sub-regions in the a-th perspective and the b-th perspective respectively.

16. The storage medium according to claim 15, wherein the multiple perspectives include two or more of the following items: one or more illumination angles, the intensity of the illumination radiation, illumination polarization, illumination wavefront, illumination spectrum, one or more focus offsets of the illumination beam, one or more collection angles, the intensity of the collected radiation, collection polarization, the phase of one or more collected beams, bright field channels, gray field channels, Fourier filtering of the returned light, and a sensing type selected from intensity, phase, or polarization, and compatible combinations of the above items.

Images