Detecting defects on a sample

Abstract

Methods and systems are provided for detecting defects on a sample. A system calculates different candidate reference images from different combinations of images of the sample produced by an inspection subsystem and combines different portions of the candidate reference images without modification to thereby produce a final reference image. The final reference image is then used for defect detection, which may be a single or dual detection. Embodiments are particularly useful for defect detection in regions of a sample containing only indistinguishable, repeating device patterns, such as cell regions, but can also be used for inspection of other types of regions.

Description

Technical Field

[0001] The present invention generally relates to methods and systems for detecting defects on a sample in semiconductor technology. Background Art

[0002] The following description and examples are not to be considered prior art by virtue of their inclusion in this section.

[0003] Fabricating semiconductor devices such as logic and memory devices typically involves processing a sample such as a semiconductor wafer using several semiconductor manufacturing processes to form various features and multiple levels of the semiconductor device. For example, lithography is a semiconductor manufacturing process that typically involves transferring a pattern to a resist disposed on a semiconductor wafer. Additional examples of semiconductor manufacturing processes include but are not limited to chemical mechanical polishing, etching, deposition, and ion implantation. Multiple semiconductor devices can be fabricated on a semiconductor wafer in a certain arrangement and then separated into individual semiconductor devices.

[0004] In the semiconductor industry, inspection using optical or electron beam imaging is an important technique for debugging semiconductor manufacturing processes, monitoring process variations, and improving production yield. As the scale of modern integrated circuits (ICs) continues to shrink and the complexity of manufacturing processes increases, inspection becomes increasingly difficult.

[0005] In each processing step performed on a semiconductor wafer, the same circuit pattern is printed on each die. Most wafer inspection systems take advantage of this fact and use relatively simple inter-die comparison to detect defects on the wafer. However, the printed circuits in each die may contain many patterned feature regions that are repeated in the x or y direction, such as DRAM, SRAM, or FLASH regions. This type of region is commonly referred to as an "array region" (the remaining regions are referred to as "random" or "logic regions"). To achieve better sensitivity, advanced inspection systems employ different strategies to inspect array regions and random or logic regions.

[0006] Array detection algorithms are designed to achieve relatively high sensitivity for such cell regions by taking advantage of the repeatability of DRAM cell regions. For example, inspection systems configured for array region inspection typically perform inter-cell comparison, where images of different cells in the array region of the same die are subtracted from each other and the differences are examined to find defects. This array inspection strategy can achieve much higher sensitivity in the array region than random inspection (which is typically performed by subtracting the image of one die from the image of another die), because it avoids noise caused by inter-die variations.

[0007] Many currently used array defect detection methods divide the entire region of interest into several units (referred to as cells). All cells are placed in separate groups, and each group is used to generate a single reference cell, which serves as the basis for the final reference image. Currently used array defect detection methods generally assume that, except for some random noise, all cells in each group are identical. The gray levels at the same positions of all cells only fluctuate around the noise-free ground truth value. However, due to highly non-uniform reflectivity from different die regions, such an assumption of uniform gray levels may not always be correct. For example, relatively strong leakage from the cell edge regions can introduce relatively low-frequency gray level variations, which can ultimately lead to relatively low inspection sensitivity. In addition, currently used array detection methods treat each region of interest as a single detection region. By bundling different positions together, it is not possible to achieve a good balance between sensitivity and nuisance control. With a setting where optimal sensitivity can be achieved from the internal region, there may be a large amount of nuisance from the edge / corner regions.

[0008] In a random defect detection algorithm, arbitrating the results of two detections has shown better performance than relying on only a single detection. This dual detection ability will also benefit array defect detection methods in terms of nuisance reduction, but it is not used in currently used array defect detection methods.

[0009] Accordingly, it would be advantageous to develop systems and methods for inspecting samples that do not have one or more of the drawbacks described above. SUMMARY OF THE INVENTION

[0010] The following description of the various embodiments should in no way be construed as limiting the subject matter of the appended claims.

[0011] One embodiment relates to a system configured to detect defects on a sample. The system includes an inspection subsystem configured to generate an image of the sample, the image including a test image and two or more other images corresponding to the test image. The system also includes a computer subsystem configured to compute first and second candidate reference images from different combinations of at least two of the test image and the two or more other images. The computer subsystem is further configured to select at least a portion of the first candidate reference image corresponding to a first portion of the test image and a portion of the second candidate reference image corresponding to a second portion of the test image. Additionally, the computer subsystem is configured to combine the selected portions of the first and second candidate reference images without modifying the selected portions to thereby generate a final reference image. The computer subsystem is further configured to generate a difference image by comparing the test image with the final reference image and to detect a defect in the test image by applying a defect detection method to the difference image. The system may be further configured as described herein.

[0012] Another embodiment relates to a computer-implemented method for detecting defects on a sample. The method includes obtaining an image of the sample generated by an inspection subsystem, the image including a test image and two or more other images corresponding to the test image. The method also includes the computing, selecting, combining, generating, and detecting steps described above. The obtaining, computing, selecting, combining, generating, and detecting are performed by a computer subsystem coupled to the inspection subsystem. Each of the steps of the method described above may be performed as further described herein. Additionally, embodiments of the method described above may include any other steps of any other methods described herein. The method described above may be performed by any one of the systems described herein.

[0013] Another embodiment relates to a non-transitory computer-readable medium storing program instructions executable on a computer system to perform a computer-implemented method for detecting defects on a sample. The computer-implemented method includes the steps of the method described above. The computer-readable medium may be further configured as described herein. The steps of the computer-implemented method may be performed as further described herein. Additionally, the computer-implemented method for which the program instructions are executable may include any other steps of any other methods described herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Further advantages of the present invention will become apparent to those skilled in the art upon a reading of the following detailed description of the preferred embodiments and upon reference to the accompanying drawings, in which:

[0015] Figure 1 and 1a is a schematic side view illustration of an embodiment of a system configured as described herein;

[0016] Figure 2 is a schematic plan view illustration of an embodiment showing an example of a test image, a corresponding reference image, and a difference image generated therefrom using a currently employed inspection method, and a final reference image generated as described herein and a difference image generated therefrom;

[0017] Figure 3 is a schematic plan view illustration of an embodiment showing an example of an image generated for a sample and different embodiments of dividing the image into test cells and two or more other cells adjacent to the test cells;

[0018] Figure 4 is a flowchart illustration of an embodiment showing steps that may be performed to detect defects on a sample;

[0019] Figure 5 is a schematic plan view illustration of an embodiment showing an example of an operation of an image frame including different unit regions on a sample;

[0020] Figure 6 is a schematic plan view illustration of an embodiment showing different candidate reference images, wherein different portions corresponding to different parts of a test image are selected and combined without modification to thereby generate a final reference image;

[0021] Figure 7 is a schematic plan view illustration of an embodiment showing a region of interest divided into different segments and a diagram showing an embodiment of separately detecting defects in each of the different segments; and

[0022] Figure 8 is a block diagram illustration of an embodiment showing a non - transitory computer - readable medium storing program instructions for causing a computer system to perform the computer - implemented methods described herein.

[0023] While the present invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and are herein described in detail. The drawings may not be drawn to scale. However, it is to be understood that the drawings and the detailed description thereof are not intended to limit the invention to the particular form disclosed, but on the contrary, the invention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims. Detailed implementation mode

[0024] Turning now to the drawings, it should be noted that the figures are not drawn to scale. Specifically, the scale of some elements of the figures is greatly exaggerated to emphasize the characteristics of the elements. It should also be noted that the figures are not drawn to the same scale. The same reference numerals have been used to indicate elements that can be similarly configured shown in more than one figure. Unless otherwise stated herein, any of the elements described and shown may include any suitable commercially available elements.

[0025] Generally speaking, the embodiments described herein are configured to detect defects on a sample. The embodiments described herein provide new reference image generation and detection methods that are particularly useful for non-pattern detection regions. The currently used methods for defect detection in non-pattern array (NPA) regions assume that the entire inspection area has a uniform gray level and only wafer / imaging noise causes gray level fluctuations. However, the presence of color variations due to different material and structure reflectivities breaks this assumption. Failure to properly handle color variations may result in additional noise, which is sometimes even greater than the defect of interest (DOI) signal, and reduces the NPA defect detection sensitivity. Based on these observations, the steps further described herein are proposed to improve the defect detection performance at non-pattern regions.

[0026] As used herein, the term "nuisance" (which may sometimes be used interchangeably with "nuisance defect") is generally defined as an event that is detected on a sample but is not actually a real defect on the sample. Nuisances that are not actually defects may be detected as events due to non-defect noise sources on the sample (e.g., particles in metal lines on the sample, signals from the underlying layer or material on the sample, line edge roughness (LER), relatively small critical dimension (CD) variations in patterned features, thickness variations, etc.) and / or due to marginality of the inspection system itself and its configuration for inspection.

[0027] The terms "first" and "second" are used herein only for the convenience of referring to different things, but those terms do not imply any other meaning for the embodiments described herein.

[0028] In some embodiments, the sample is a wafer. The wafer can include any wafer known in semiconductor technology. Although some embodiments may be described herein with respect to one or several wafers, the embodiments are not limited to the samples for which they can be used. For example, the embodiments described herein can be used for samples such as photomasks, flat panels, personal computer (PC) boards, and other semiconductor samples. <C

[0029] Figure 1 An embodiment of a system configured to detect defects on a sample is shown. The system includes an inspection subsystem 100. In Figure 1In the embodiments shown, the inspection subsystem is configured as an optical-based inspection subsystem. However, in other embodiments described herein, the inspection subsystem is configured as an electron beam or charged particle beam-based inspection subsystem. Generally, the inspection subsystems described herein include at least one energy source and a detector. The energy source is configured to generate energy that is directed to the sample. The detector is configured to detect energy from the sample and generate an output in response to the detected energy.

[0030] In an optical-based inspection subsystem, the energy directed to the sample includes light, and the energy detected from the sample includes light. For example, in Figure 1 an embodiment of the system shown, the inspection subsystem includes an illumination subsystem configured to direct light to sample 14. The illumination subsystem includes at least one light source. For example, as Figure 1 shown, the illumination subsystem includes light source 16. The illumination subsystem is configured to direct light to the sample at one or more incident angles that can include one or more tilt angles and / or one or more normal angles. For example, as Figure 1 shown, light from light source 16 is directed through optical element 18 at a tilted incident angle and then through lens 20 to sample 14. The tilted incident angle can include any suitable tilted incident angle that can vary depending on, for example, the characteristics of the sample and the defects to be detected on the sample.

[0031] The illumination subsystem can be configured to direct light to the sample at different incident angles at different times. For example, the inspection subsystem can be configured to change one or more characteristics of one or more elements of the illumination subsystem such that light can be directed to the sample at an incident angle different from Figure 1 the incident angle shown. In one such example, the inspection subsystem can be configured to move light source 16, optical element 18, and lens 20 such that light is directed to the sample at different tilted incident angles or normal (or near-normal) incident angles.

[0032] The inspection subsystem can be configured to direct light to the sample at more than one incident angle at the same time. For example, the illumination subsystem can include more than one illumination channel, one of the illumination channels can include light source 16, optical element 18, and lens 20 as Figure 1 shown, and another of the illumination channels (not shown) can include similar elements that can be configured differently or the same or can include at least one light source and possibly one or more other components (such as components further described herein). If this light is directed to the sample at the same time as other light, then one or more characteristics (e.g., wavelength, polarization, etc.) of the light directed to the sample at different incident angles can be different such that the light generated by illuminating the sample at different incident angles can be distinguished from each other at the (one or more) detectors.

[0033] The illumination subsystem may include only one light source (e.g., the source 16 shown in Figure 1 ) and the light from the light source may be split into different optical paths (e.g., based on wavelength, polarization, etc.) by the (several) optical elements (not shown) of the illumination subsystem. Then, the light in each of the different optical paths may be directed to the sample. Multiple illumination channels may be configured to direct light to the sample at the same time or at different times (e.g., when different illumination channels are used to sequentially illuminate the sample). In another example, the same illumination channel may be configured to direct light with different characteristics to the sample at different times. For example, the optical element 18 may be configured as a spectral filter and may change the properties of the spectral filter in various different ways (e.g., by swapping out one spectral filter with another spectral filter) such that different wavelengths of light may be directed to the sample at different times. The illumination subsystem may have any other suitable configuration known in the art for directing light with different or the same characteristics to the sample at different or the same incident angles, either sequentially or simultaneously.

[0034] The light source 16 may include a broadband plasma (BBP) light source. In this way, the light generated by the light source and directed to the sample may include broadband light. However, the light source may include any other suitable light source, such as a laser. The laser may include any suitable laser known in the art and may be configured to generate light of any suitable wavelength(s) known in the art. Additionally, the laser may be configured to generate monochromatic or near-monochromatic light. For example, the laser may be a narrowband laser. The light source may also include a polychromatic light source that generates light of multiple discrete wavelengths or bands.

[0035] The light from the optical element 18 may be focused onto the sample 14 by the lens 20. Although the lens 20 is shown in Figure 1 as a single refractive optical element, in reality, the lens 20 may include several refractive and / or reflective optical elements that collectively focus the light from the optical element onto the sample. Figure 1 The illumination subsystem shown in and described herein may include any other suitable optical elements (not shown). Examples of such optical elements include, but are not limited to, (several) polarization components, (several) spectral filters, (several) spatial filters, (several) reflective optical elements, (several) apodizers, (several) beam splitters, (several) apertures, and the like, which may include any such suitable optical elements known in the art. Additionally, the system may be configured to change one or more of the elements of the illumination subsystem based on the type of illumination to be used for inspection.

[0036] The inspection subsystem may also include a scanning subsystem configured to change positions on the sample (direct light to and detect light from the positions) and possibly scan light across the sample. For example, the inspection subsystem may include a stage 22 on which the sample 14 is placed during inspection. The scanning subsystem may include any suitable mechanical and / or robotic assembly (which includes the stage 22) configured to move the sample such that light can be directed to different positions on the sample and light can be detected from different positions on the sample. Additionally or alternatively, the inspection subsystem may be configured such that one or more optical elements of the inspection subsystem perform some scanning of light across the sample such that light can be directed to different positions on the sample and light can be detected from different positions on the sample. Light can be scanned across the sample in any suitable manner (e.g., in a serpentine path or in a spiral path).

[0037] The inspection subsystem further includes one or more detection channels. At least one of the (several) detection channels includes a detector configured to detect light from the sample attributable to illumination of the sample by the system and to generate an output in response to the detected light. For example, Figure 1 the inspection subsystem shown in includes two detection channels, one detection channel formed by the light collector 24, the element 26, and the detector 28 and the other detection channel formed by the light collector 30, the element 32, and the detector 34. As Figure 1 shown in, the two detection channels are configured to collect and detect light at different collection angles. In some examples, the two detection channels are configured to detect scattered light, and the detection channels are configured to detect light scattered from the sample at different angles. However, one or more of the detection channels may be configured to detect another type of light from the sample (e.g., reflected light).

[0038] As Figure 1 further shown in, the two detection channels are shown positioned in the plane of the paper and the illumination subsystem is also shown positioned in the plane of the paper. Thus, the two detection channels are positioned (e.g., centered) in the plane of incidence. However, one or more of the detection channels may be positioned outside the plane of incidence. For example, the detection channel formed by the light collector 30, the element 32, and the detector 34 may be configured to collect and detect light scattered out of the plane of incidence. Thus, this detection channel may generally be referred to as a “side” channel, and this side channel may be centered in a plane substantially perpendicular to the plane of incidence.

[0039] The inspection subsystem may include associated with Figure 1Different numbers of detection channels (e.g., only one detection channel or two or more detection channels) are shown. In one such example, the detection channel formed by the light collector 30, the element 32, and the detector 34 can form one side channel as described above, and the inspection subsystem can include additional detection channels (not shown) formed as the other side channel positioned on the opposite side of the incident plane. Thus, the inspection subsystem can include a detection channel that includes the light collector 24, the element 26, and the detector 28 and is centered in the incident plane and configured to collect and detect light at (a) scattering angle(s) that is / are at or near normal to the sample surface. Thus, this detection channel can generally be referred to as the "top" channel, and the inspection subsystem can also include two or more side channels configured as described above. Thus, the inspection subsystem can include at least three channels (i.e., one top channel and two side channels), and each of the at least three channels has its own light collector, each of which is configured to collect light at a different scattering angle from each of the other light collectors.

[0040] As further described above, each of the detection channels included in the inspection subsystem can be configured to detect scattered light. Thus, Figure 1 the inspection subsystem shown can be configured for dark field (DF) inspection of a sample. However, the inspection subsystem can also or alternatively include (a) detection channel(s) configured for bright field (BF) inspection of the sample. In other words, the inspection subsystem can include at least one detection channel configured to detect light specularly reflected from the sample. Thus, the inspection subsystem described herein can be configured for only DF, only BF, or both DF and BF inspection. Although each of the light collectors is shown in Figure 1 as a single refractive optical element, it should be understood that each of the light collectors can include one or more refractive optical elements and / or one or more reflective optical elements.

[0041] One or more detection channels may include any suitable detector known in the art. For example, the detector may include a photomultiplier tube (PMT), a charge-coupled device (CCD), and a time delay integration (TDI) camera. The detector may also include a non-imaging detector or an imaging detector. If the detector is a non-imaging detector, then each of the detectors may be configured to detect certain characteristics (e.g., intensity) of the scattered light, but may not be configured to detect such characteristics that vary according to the position within the imaging plane. Thus, the output generated by each of the detectors included in each of the detection channels may be a signal or data, rather than an image signal or image data. In such examples, a computer subsystem (e.g., the computer subsystem 36 that forms part of the inspection system having the inspection subsystem) may be configured to generate an image of the sample from the non-imaging output of the detector. However, in other examples, the detector may be configured as an imaging detector that is configured to generate an imaging signal or image data. Thus, the inspection subsystem may be configured to generate an image in several ways.

[0042] Note that the present disclosure Figure 1 is provided generally to illustrate the configuration of the inspection subsystem that may be included in the system embodiments described herein. Obviously, the inspection subsystem configuration described herein may be modified to optimize the performance of the inspection subsystem as is typically done when designing a commercial inspection system. Additionally, the systems described herein may be implemented using an existing inspection system (e.g., by adding the functionality described herein to an existing inspection system) such as the 29xx / 39xx series tools commercially available from KLA Corp., Milpitas, Calif. For some such systems, the methods described herein may be provided as optional functionality of the inspection system (e.g., in addition to other functionality of the inspection system). Alternatively, the inspection systems described herein may be designed "from scratch" to provide a brand new inspection system.

[0043] The computer subsystem 36 may be coupled to the detector of the inspection subsystem in any suitable manner (e.g., via one or more transmission media, which may include "wired" and / or "wireless" transmission media) such that the computer subsystem may receive the output generated by the detector. The computer subsystem 36 may be configured to perform several functions using the output of the detector, as further described herein. The computer subsystem coupled to the inspection subsystem may be further configured as described herein.

[0044] A computer subsystem (and other computer subsystems described herein) coupled to an inspection subsystem may also be referred to herein as a (plural) computer system. Each of the (plural) computer subsystems or (plural) systems described herein may take various forms, including a personal computer system, an image computer, a mainframe computer system, a workstation, a network appliance, an Internet appliance, or other devices. In general, the term "computer system" may be broadly defined to cover any device having one or more processors that execute instructions from a memory medium. The (plural) computer subsystems or (plural) systems may also include any suitable processor known in the art (e.g., a parallel processor). Additionally, the (plural) computer subsystems or (plural) systems may include a computer platform with high-speed processing and software, either as a stand-alone tool or a networked tool.

[0045] If the system includes more than one computer subsystem, then the different computer subsystems may be coupled to each other such that images, data, information, instructions, etc. may be transmitted between the computer subsystems. For example, computer subsystem 36 may be coupled to the (plural) computer systems 102 by any suitable transmission medium that may include any suitable wired and / or wireless transmission media known in the art (as shown by the dashed line in Figure 1 . Two or more such computer subsystems may also be effectively coupled by sharing a computer-readable storage medium (not shown).

[0046] Although the inspection subsystem was described above as an optical or light-based inspection subsystem, in another embodiment, the inspection subsystem is configured as an electron beam-based inspection subsystem. In an electron beam type inspection subsystem, the energy directed to the sample includes electrons, and the energy detected from the sample includes electrons. In Figure 1a In one such embodiment shown in, the inspection subsystem includes an electron column 122, and the system includes a computer subsystem 124 coupled to the inspection subsystem. The computer subsystem 124 may be configured as described above. Additionally, this inspection subsystem may be coupled to another or more computer subsystems in the same manner as described above and Figure 1 shown in.

[0047] Also as shown in Figure 1a , the electron column includes an electron beam source 126 configured to generate electrons that are focused onto the sample 128 by one or more elements 130. The electron beam source may include, for example, a cathode source or an emitter tip, and the one or more elements 130 may include, for example, gun lenses, anodes, beam limiting apertures, valves, beam current selection apertures, objective lenses, and scanning subsystems, all of which may include any such suitable elements known in the art.

[0048] Electrons returned from the sample (e.g., secondary electrons) can be focused by one or more elements 132 onto the detector 134. The one or more elements 132 can include, for example, a scanning subsystem, which can be the same scanning subsystem included in the (several) elements 130.

[0049] The electron column can include any other suitable elements known in the art. Additionally, the electron column can be further configured as described in the following patents: U.S. Patent No. 8,664,594, issued April 4, 2014, to Jiang et al.; U.S. Patent No. 8,692,204, issued April 8, 2014, to Kojima et al.; U.S. Patent No. 8,698,093, issued April 15, 2014, to Gubbens et al.; and U.S. Patent No. 8,716,662, issued May 6, 2014, to MacDonald et al., which are hereby incorporated by reference in their entirety as if fully set forth herein.

[0050] Although the electron column is shown in Figure 1a configured such that electrons are directed onto the sample at an oblique incident angle and scattered from the sample at another oblique angle, the electron beam can be directed onto the sample and scattered from the sample at any suitable angle. Additionally, as further described herein, the electron beam inspection subsystem can be configured to produce an output of the sample (e.g., having different illumination angles, collection angles, etc.) using multiple modes. The multiple modes of the electron beam inspection subsystem can differ in any output generation parameter of the inspection subsystem.

[0051] The computer subsystem 124 can be coupled to the detector 134, as described above. The detector can detect electrons returned from the surface of the sample, thereby forming an electron beam image (or other output) of the sample. The electron beam image can include any suitable electron beam image. The computer subsystem 124 can be configured to perform any of the (several) steps described herein. The system including Figure 1a the inspection subsystem shown in can be further configured as described herein.

[0052] Provided herein Figure 1aTo generally illustrate the configuration of an electron beam inspection subsystem that may be included in the embodiments described herein. Similar to the optical inspection subsystem described above, the electron beam inspection subsystem configuration described herein can be modified to optimize the performance of the inspection subsystem as is typically done when designing a commercial inspection system. Additionally, the systems described herein can be implemented using existing inspection systems such as tools that are commercially available from KLA-Tencor (e.g., by adding the functionality described herein to an existing inspection system). For some such systems, the methods described herein can be provided as optional functionality of the system (e.g., in addition to other functionality of the system). Alternatively, the systems described herein can be designed "from scratch" to provide a brand new system.

[0053] Although the inspection subsystem was described above as an optical or electron beam inspection subsystem, the inspection subsystem can be an ion beam inspection subsystem. This inspection subsystem can be configured as shown in Figure 1a , with the only difference being that the electron beam source can be replaced with any suitable ion beam source known in the art. Additionally, the inspection subsystem can include any other suitable ion beam systems, such as those included in commercially available focused ion beam (FIB) systems, helium ion microscopy (HIM) systems, and secondary ion mass spectrometry (SIMS) systems.

[0054] The inspection subsystems described herein can be configured to generate outputs of a sample, such as images, in a variety of modes. Generally, a "mode" is defined by parameter values of the inspection subsystem used to generate the output and / or image of the sample (or the output that generates an image of the sample). Thus, modes can differ in terms of the value of at least one of the parameters of the inspection subsystem (other than the location on the sample where the output is generated). For example, in an optical subsystem, different modes can illuminate with light of (a) different wavelengths. As further described herein, modes can differ in terms of (a) illumination wavelength (e.g., by using different light sources, different spectral filters, etc. for different modes). In another example, different modes can use different illumination channels of the optical subsystem. For example, as described above, an optical subsystem can include more than one illumination channel. Thus, different illumination channels can be used for different modes. Modes can also or alternatively differ in terms of one or more collection / detection parameters of the optical subsystem. Modes can differ in terms of any one or more changeable parameters of the inspection subsystem (e.g., (a) illumination polarization, (a) angle, (a) wavelength, etc., (a) detection polarization, (a) angle, (a) wavelength, etc.). The inspection subsystem can be configured to scan the sample in different modes in the same scan or different scans (e.g., depending on the ability to simultaneously scan the sample using multiple modes).

[0055] In a similar manner, the output generated by the electron beam subsystem may include outputs (e.g., images) generated by the electron beam subsystem using two or more different values of the parameters of the electron beam subsystem. Multiple modes of the electron beam subsystem may be defined by the parameter values of the electron beam subsystem used to generate the output and / or image of the sample. Thus, the modes may differ in terms of the value of at least one of the electron beam parameters of the electron beam subsystem. For example, different modes may use different incident angles for illumination.

[0056] As described above, the inspection subsystem is configured to scan energy (e.g., light, electrons, etc.) over a physical version of the sample, thereby generating an output of the physical version of the sample. In this way, the inspection subsystem may be configured as an “actual” subsystem rather than a “virtual” subsystem. However, the storage medium (not shown) and Figure 1 the (several) computer subsystems 102 shown therein may be configured as “virtual” systems. Specifically, the storage medium and the (several) computer subsystems may be configured as a “virtual” inspection system as described in the following co-owned U.S. patents: U.S. Patent No. 8,126,255, issued to Bhaskar et al. on February 28, 2012, and U.S. Patent No. 9,222,895, issued to Duffy et al. on December 29, 2015, the two patents being incorporated herein by reference in their entirety as if fully set forth. The embodiments described herein may be further configured as described in these patents.

[0057] Non-pattern array (NPA) reference generation is an important part of array inspection. The NPA inspection method is used to perform array inspection on highly repetitive but indistinguishable pattern regions. Thus, for NPA inspection, there are no patterns in the optical (or other) image. One problem that affects the sensitivity of NPA inspection is color variation (CV). CV may manifest in the relatively low-frequency gray-scale variations of the input NPA image, especially near the corner or edge regions. CV poses a significant challenge to NPA inspection. A relatively high number of nuisances is another significant challenge for NPA inspection. The embodiments described herein provide new methods for addressing the above two challenges by means of a new reference generation method for handling CV and a new detection strategy for suppressing nuisances.

[0058] Figure 2Describe these challenges and how the embodiments described herein are designed to overcome the challenges. In this figure, image 200 is a test image generated by an optical inspection subsystem (such as the optical inspection subsystem described herein) for a sample. Reference image 202 is a reference image generated by the currently used method for NPA defect detection, and a difference image 204 is generated by subtracting reference image 202 from test image 200. As can be seen by comparing reference image 202 with test image 200, reference image 202 is much quieter near the left and right edges of the cell region in the reference image (the brighter regions in the reference image) than near the same edges in the test image. These differences between the test and reference images can occur when, for example, the test image contains noise in some regions such as near the edges and the reference image generation fails to account for this noise.

[0059] These differences between the reference image and the test image appear as darker regions in the difference image located near the left and right edges of the cell region in the difference image. Then, when a defect detection method is applied to this difference image, these noise differences between the reference and test images can be misdetected as defects. If the defect detection threshold is increased to avoid the detection of this noise, then the DOI present in the test image may not be detected. For several obvious reasons, both of these situations are disadvantageous.

[0060] In contrast, reference image 206 illustrates an embodiment of a reference image that can be generated by the embodiments described herein. A difference image 208 is generated by subtracting reference image 206 from test image 200. As can be seen by comparing reference image 206 with test image 200, reference image 206 and test image 200 contain similar noise near the left and right edges of the cell region in the test image (the brighter regions). These similarities in the noise characteristics become possible because the reference generation described herein enables the noise in the test image to be replicated much better in the reference image, which thereby results in a much quieter difference image, as can be seen by comparing difference image 208 with difference image 204. Therefore, compared with difference image 204, defect detection performed on difference image 204 and difference image 208 using the same parameters (e.g., threshold) will not detect as many nuisances in difference image 208. Additionally, due to the relatively quieter nature of difference image 208, a more sensitive inspection can be performed on difference image 208 without detecting an excessive number of nuisances, while also detecting more DOIs, including those with relatively weak signals.

[0061] One embodiment of a system configured to detect defects on a sample includes an inspection subsystem configured to generate an image of the sample, and the inspection subsystem can include any of the inspection subsystems described herein. The image includes, but is not limited to, a test image and two or more other images corresponding to the test image. The test image and the two or more other images can be generated in an area of the sample that contains only indistinguishable, repeating device patterns. In other words, the images used in the embodiments described herein do not contain images of device patterns formed in the area of the sample where the images are generated. In one such example, the test image 200 is an image of two unit areas on the sample that contain patterned features formed therein, but none of the patterned features are distinguishable in this image. Although the embodiments described herein are particularly suitable and advantageous for NPA inspection, the embodiments can also be used for inspections performed using images in which the patterned features are distinguishable.

[0062] The terms "image" and "image frame" can be used interchangeably herein. Generally, an "image frame" is defined as a set of pixels in an image that are processed together for purposes such as defect detection. Thus, the size of an image or image frame can vary depending on certain characteristics of the inspection subsystem or computer subsystem. In some inspection use cases, an image can actually consist of multiple image frames, but the embodiments described herein are not inherently limited in terms of the size of the images that can be processed.

[0063] As used herein, the term "job" is defined as a number of image frames that are processed together by a computer subsystem to detect defects on a sample. Generally, not all of the images generated during an inspection process can be processed together (e.g., even if possible, it is usually not time or cost efficient). Therefore, all the images are divided into jobs of image frames that can be processed together in a cheaper and more timely manner for defect detection.

[0064] Obtaining an image can include generating an image using an inspection subsystem configured as described herein. This image acquisition can be accomplished when the computer subsystem and the inspection subsystem are coupled in a single tool and possibly when performing the defect detection described herein on the tool and / or when generating the image. In other examples, the computer subsystem can obtain an image from another method, system, or storage medium. For example, the computer subsystem and the inspection subsystem can be coupled or not coupled to a single tool, and the inspection subsystem, the computer subsystem, or another computer subsystem can store the image generated by the inspection subsystem. Then, the computer subsystem can obtain the image from the storage medium in which the image is stored. This image acquisition can be accomplished when the computer subsystem performs the steps described herein outside of the tool and / or after all (or at least some) of the images have been generated. Each of the elements described above can be configured as further described and shown herein.

[0065] In one embodiment, a computer subsystem is configured to divide an image corresponding to a region of interest on a sample into a test image corresponding to a test cell and two or more other images corresponding to two or more other cells adjacent to the test cell. Each region of interest can be divided into a number of cells, as in the currently used NPA defect detection method. However, instead of placing all cells into some non-overlapping groups as in the currently used methods and systems, the computer subsystem can form a group containing several adjacent cells for each cell. Grouping the cells in this way can provide a CV compensation method for NPA reference generation. By reorganizing the cell grouping, a clearer difference image can be achieved.

[0066] Figure 3 An example of an image illustrating a region of interest and different embodiments of dividing the region of interest image into a test image of a test cell and other images corresponding to other cells adjacent to the test cell is described. Specifically, image 300 is a hypothetical image of a region of interest. Although this image is only a single grayscale value image and thus does not truly represent an image that might be produced by an inspection subsystem configured as described herein, it accurately represents an NPA image because none of the patterned device features formed in the region of interest on the sample are distinguishable in the image. There may be multiple instances of the same region of interest on the sample, and image 300 represents the image of only one such instance.

[0067] Each region of interest can be divided into several cells, for example, only along the horizontal or vertical direction. As Figure 3 shown, if the region of interest is divided along the horizontal direction, the test cell image can be image portion 302 and the other cell images adjacent to the test cell can be image portions 304, 306, 308, and 310. If the region of interest is divided along the vertical direction, the test cell image can be image portion 312 and the other cell images adjacent to the test cell can be image portions 314, 316, 318, and 320. In this way, adjacent cells are the cells to the left and / or right of a target cell (if the region of interest is divided along the horizontal direction) or above and / or below the target cell (if the region of interest is divided along the vertical direction). The computer subsystem can form a group of adjacent cells for each test cell target. In other words, if an inspection is performed on the test cell corresponding to image portion 304, a different group of image portions than the group of image portions described above can be selected as the cells adjacent to that test cell.

[0068] Therefore, as Figure 3As shown, the units adjacent to the test unit need not be adjacent to the test unit and may be separated from the test unit by one or more other units. However, the units adjacent to the test unit may be selected as the adjacent units. Additionally, the units adjacent to the test unit may include units on both sides of the test unit, but this is not required. Further, although the number of units adjacent to the test unit is shown as four units in Figure 3 the number of adjacent units can vary widely and depends on several factors such as the degree of change in color variation between units, the actual amount of image data that can be processed simultaneously by the computer subsystem, etc. Different numbers of adjacent units can also be selected for the units in the horizontal and vertical directions. The same region of interest image can also be divided in the horizontal and vertical directions, and the resulting test units can be processed separately. For example, reference generation and defect detection can be performed for test unit 302 and its adjacent units 304, 306, 308, and 310, and reference generation and defect detection can also be performed separately for test unit 312 and its adjacent units 314, 316, 318, and 320.

[0069] In some embodiments, the computer subsystem is configured to apply CV compensation to the test image and two or more other images before calculating the first and second candidate reference images as further described herein. For example, the first step can be to add CV compensation to the reference generation process. One CV compensation can be applied to all the units in a group before creating the reference units. For example, CV compensation can be applied to units 304, 306, 308, and 310 before using them to generate the reference unit image. After this step, the CV is inherently integrated into the reference image that is ultimately subtracted from the test image, which thereby results in a clearer final difference image. In this way, the embodiments described herein provide a CV compensation method for NPA reference generation. By reorganizing the unit grouping and compensating for CV, a clearer difference image is achieved.

[0070] CV can cause each unit to have a different gray level, which would limit the unit-based noise removal ability unless compensated. A particularly suitable way to perform CV compensation is to assume the gray level in a unit as the variable x i . Then, several unknown parameters p n can be used to create a special function f(x i |p1, p2,...). This function can be applied to each pixel position. The unknown parameters can be obtained by minimizing the objective function between the target unit and the function operation ||x t - f(x i|p1, p2,...)| Calculate the parameters of each unit based on the differences between subsequent units. Applying this function to the corresponding unit will compensate for the CV. In this way, the computer subsystem can fit the gray-scale changes within each unit. The basic fitting function can be (for example) piecewise linear. After the fitting is completed, the fitting result f(x i |p1, p2,...) becomes the local reference image within each unit. Since the fitted reference image is substantially very consistent with the test image, the CV embedded in the original test image is removed or significantly reduced.

[0071] As mentioned above, the CV is integrated into the reference unit image. One of the reasons is the CV compensation described above. Specifically, the purpose of generating a better reference described herein is to obtain a clearer difference image. In the past, CV compensation could not be performed or was not performed, which means that the CV would remain in the difference image and could make the difference image too noisy for defect detection. However, by performing CV compensation on the image used to generate the reference image, integrating the CV compensation into any reference image generated therefrom, and then using the reference image to generate a clearer difference image. In other words, the CV is inherently disposed of during the reference image generation process. In this way, the CV in the reference image will compensate for the CV in the test image. The CV compensation can be inherently integrated into the reference generation for all reference images generated as described herein.

[0072] The embodiments described herein generate multiple candidate reference images for each test image and mix them together as further described herein to further improve the CV handling ability. The computer subsystem is configured to calculate first and second candidate reference images from different combinations of at least two of the test image and two or more other images. For example, as Figure 4 shown in step 400 of, the computer subsystem can calculate the candidate reference image. Calculating any of the candidate reference images described herein can be performed by combining two or more images using (for example) linear combination, median calculation, average calculation, another currently used method for generating a calculated reference (CR), etc. In this way, calculating the candidate reference image from two or more images can generate an image with characteristics different from one or more of the input images. In other words, the pixels in the calculated candidate reference image can have different image characteristics, such as gray-scale values, from the corresponding pixels in each of the images used to generate the calculated candidate reference image. Therefore, the candidate reference image generation is fundamentally different from the final reference image generation further described herein.

[0073] In most instances, two or more candidate reference images can be calculated for each test image. In other words, a candidate reference image may not be used for more than one test image in a job. Instead, after grouping images as further described herein, for each test image within each job, the computer subsystem will create at least two candidate reference images. However, in other instances, candidate reference images can be reused for more than one test image.

[0074] In one embodiment, a first of different combinations of at least two of the test image and two or more other images used to calculate a first candidate reference image includes a portion of an image generated across the sample only in the x direction. In another such embodiment, a second of different combinations of at least two of the test image and two or more other images used to calculate a second candidate reference image includes an additional portion of an image generated across the sample only in the y direction. Currently used NPA defect detection methods always generate references in a row-independent manner (i.e., using images generated along the x direction to generate references). The embodiments described herein can also employ the same strategy to form one candidate reference image. Additionally, a similar operation can be applied in a column-independent manner to generate another candidate reference image (i.e., using images generated along the y direction to generate the candidate reference image). Generating two such candidate reference images provides several significant advantages further described herein.

[0075] In a further embodiment, one of different combinations of at least two of the test image and two or more other images used to calculate the first and second candidate reference images includes all images generated across two or more die on the sample in a job of images. For example, an additional reference can be generated by considering all die. "All die" in this context represents all die included in a job, which is the smallest unit of image processing. A job can include all die in at most one row of die or at least three die. Figure 5 Illustrate an example of a job that includes multiple unit regions for inspection. Specifically, job 500 includes unit regions 502, 504, 506, and 508, and each unit region can include different die on the sample. In this embodiment, candidate reference images can be generated from all images of all unit regions included in job 500.

[0076] In this way, some candidate reference images (e.g., row-independent and column-independent references) can be generated from a single die, while one or more other candidate reference images (e.g., candidate reference images generated from "all dies") can be generated from two or more die images. Generally, in the embodiments described herein, at least two candidate reference images are generated such that a final candidate reference image can be created from those two images. However, in many use cases, it will be advantageous to generate all three types of candidate reference images described herein (i.e., row-independent, column-independent, and multi-die candidate reference images). Two or more of these candidate reference images can be used to generate the final reference image. In an example of performing dual detection, one or more of these candidate reference images can also be used as a second or additional final reference image. Such embodiments are further described herein.

[0077] In currently used NPA defect detection, reference generation occurs within each individual frame. In other words, the reference generation for frame 3 does not require any input from frame 2 or frame 4. In contrast, in the embodiments described herein, for generating the reference image for frame 3, both frame 3 and all other frames can be used. It can be advantageous to use these images to generate candidate reference images, for example, when the unit corner region has a unique gray value relative to the rest of the unit area. Such a unique gray value may make it difficult to find a good matching reference image for each corner within each unit area. However, it may be significantly easier to find a good match from other corners of other unit areas. By using all other frames in the job, better candidate reference images can be created and used to create a better final reference image.

[0078] The computer subsystem is also configured to at least select a portion of a first candidate reference image corresponding to a first portion of the test image and a portion of a second candidate reference image corresponding to a second portion of the test image. For example, as shown in step 402 of Figure 4 , the computer subsystem can be configured to select different portions of candidate reference images corresponding to different portions of the test image. As further described herein, the embodiments are configured to provide better reference images, which can then be used to generate better difference images, thereby achieving less nuisance detection and / or more sensitive inspection. To do so, a better reference will then be a reference that better matches the test image, particularly with respect to the noise in the test image. For example, if the noise in the reference image substantially matches the noise in the test image, then when the reference image is subtracted from the test image, that noise in the test image will be canceled out, thereby preventing it from being detected as a nuisance and / or interfering with the detection of DOI. Thus, the goal of the selection step is to find the best matching reference image from a set of candidate reference images on a per-test-image-part basis, which can be performed in several ways further described herein.

[0079] In one embodiment, a selection is made that includes identifying which portions of first and second candidate reference images best match different portions of a test image. Figure 6 An embodiment is shown of how this selection may be performed. In this embodiment, test image 600 is an example of a test image that may be generated for a region of interest on a sample. As with Figure 2 the test image shown, this test image includes a central portion 606 that is relatively more quiescent compared to portions 602 near the left edge and portions 604 near the right edge. These portions are relatively noisier compared to the central portion due to, for example, color variations that may occur near the edges of the region of interest image. As further described above, if the reference images used for defect detection with the test image do not contain noise similar to the test image, then that will cause problems with defect detection. Thus, a new way of generating the reference images described herein is created.

[0080] The selection step may include dividing the test image into an array of blocks, as shown in image 608. Each of the blocks may have the same predetermined characteristics, such as size, and is only used to divide the image into smaller portions that can be evaluated on an individual basis, as further described herein. The number of blocks into which the image is divided may vary (potentially greatly) and may be determined based on several factors, such as the initial image size and the rate at which noise varies across the image (when it varies relatively rapidly, smaller blocks may be more suitable for capturing different values of noise on the same or similar scale at which it changes).

[0081] As mentioned above, the portions of the candidate reference images that best match different portions of the test image may be identified by a computer subsystem, which means that even when each test image portion is processed individually, the correspondence between the test image portion and the reference image portion may be relaxed. In other words, the portions of the candidate reference images that best match a portion of the test image may not necessarily be limited to those portions of the candidate reference images that have the same in-image location as the test image portion.

[0082] To illustrate this concept, consider two example candidate reference images 614 and 616, which can be any of the candidate reference images generated in any of the ways described herein. Each of these candidate reference images can be partitioned into blocks in the same way as the test image, as shown by the black lines superimposed on these images. The identification step can start with a test image block in the upper row of column 610 of the block. This test image block can be compared with each block in candidate reference image 614 to determine which blocks in this candidate reference image best match the test image block. The block of the candidate reference image that best matches the test image block can be saved, and then the same comparison step can be performed using the blocks of candidate reference image 616 to determine if there are even better matching blocks in that candidate reference image. Then, the best matching block in candidate reference image 614 or 616 can be identified and saved to be included in the final reference image at the location of the test image block.

[0083] In this way, for each test image block, all blocks in all (or at least some) candidate reference images can be considered to find the best matching image block. Then, this same process can be performed for the next block in the test image. In some examples, the same block in one of the candidate reference images can be identified as the best match for more than one test image block. By enabling any block in any of the candidate reference images to be used as the best matching block for a test image block, a better reference image may be created than the case where only blocks at the same location within the same image are considered. This method can provide a better final reference image when each of the test image and the candidate reference images exhibits noise with substantially different spatial or other characteristics.

[0084] In another embodiment, a selection is made to include identifying which portions of the first and second candidate reference images best match the corresponding portions of the test image. In this way, different from the embodiment described above, for each test image block, only the candidate reference image blocks with the same in-image location can be considered. This embodiment may be faster than the embodiment described above because the number of blocks considered for any one test image block will be limited to the number of candidate reference images that have been generated. However, different from the embodiment described above, this embodiment will be less flexible or exhaustive in the blocks considered and may therefore be more suitable for examples where at least one of the known or expected candidate reference images contains noise with spatial and possibly other characteristics similar to the test image.

[0085] To illustrate this concept, the image 608, which is the test image 600 divided into blocks, can be used again together with the candidate reference images 614 and 616. The identification step can start with the test image block in the top row of column 610 of the block. This block can be compared with the image block in the top row of column 618 in the candidate reference image 614 and the image block in the top row of column 626 in the candidate reference image 616 to determine which of these blocks best matches the test image block. If no other candidate reference images have been generated for this test image, then no other candidate reference image blocks may be considered for this test image block. As shown by images 608, 614, and 616, the image block in the top row of column 618 of the candidate reference image 614 is a much better match for the test image block in the top row of column 610 than the image block at the same position in the candidate reference image 616. Then, the best matching candidate reference image block identified by this step can be saved to generate the final reference image, as further described herein.

[0086] Next, the identification step can proceed to the test block in the second row from the top in column 610 of the block. This block can be compared with the image block in the second row from the top in column 618 in the candidate reference image 614 and the image block in the second row from the top in column 626 of the candidate reference image 616 to determine which of these two blocks best matches the test image block. Also, if no other candidate reference images have been generated for this test image, then no other candidate reference image blocks may be considered for this test image block. As shown by images 608, 614, and 616, the image block in the second row from the top of column 618 of the candidate reference image 614 is a better match for the test image block in the second row from the top of column 610 than the image block at the same position in the candidate reference image 616. The best matching candidate reference image block identified by this step can also be saved to generate the final reference image, as further described herein. Then, the identification step can be continued for all the remaining test blocks in image 608.

[0087] Regardless of which way for Figure 6Perform a comparison on the images shown in the figure. Different candidate reference images all contain noise different from the test image. Specifically, the noise at the left and right edges of the candidate reference image 614 (shown by the darker parts of the image) extends further into the candidate reference image than the noise at the left and right edges of the test image (also shown by the darker parts of the test image). Specifically, the columns 618, 620, 622, and 624 of the blocks in the candidate reference image 614 contain noise, while only the columns 610 and 612 of the blocks in the test image 608 contain noise. In contrast, the candidate reference image 616 contains almost no noise at the left and right edges of this image (as shown by the absence of any relatively dark regions in the candidate reference image). Specifically, the columns 626 and 628 of the candidate reference image 616 contain no or almost no noise, while the corresponding columns 610 and 612 of the image 608 clearly contain noise. Therefore, if any of the candidate reference images is used with the test image for defect detection, then any of the reference images may result in a large number of nuisance detections or other problems in defect detection. However, by selectively identifying the parts of the candidate reference images to be included in the final reference image based on the degree of matching between the candidate reference images and the test image, a final reference image 634 that clearly contains noise and better matches the test image 600 can be produced.

[0088] The computer subsystem is further configured to combine the selected parts of the first and second candidate reference images without modifying the selected parts of the first and second candidate reference images to thereby produce a final reference image. As Figure 4 shown in step 404 of, the computer subsystem can combine the selected parts to produce a first final reference image, which can be the only final reference image in the case of a single detection or one of the two final reference images in the case of a double detection. For example, the final reference image 634 shown in can be produced by combining the selected parts of those images without modifying the selected parts of the first candidate reference image 614 and the second candidate reference image 616. Figure 6 shown in.

[0089] Specifically, as Figure 6 shown in, for the image blocks in column 610 of the image 608, the image blocks in column 618 of the candidate reference image 614 match much better than the image blocks in column 626 of the candidate reference image 616. Therefore, the image blocks in column 636 of the final reference image 634 can be the image blocks in column 618. In a similar manner, for the image blocks in column 612 of the image 608, the image blocks in column 620 of the candidate reference image 614 match much better than the image blocks in column 628 of the candidate reference image 616. Thus, the image blocks in column 642 of the final reference image 634 can be the image blocks in column 620.

[0090] In contrast, the image blocks in columns 622 and 624 of the candidate reference image 614 do not match the corresponding columns of the image blocks in the image 608 very well. However, the image blocks in columns 630 and 632 of the candidate reference image 616 are much better matches for the corresponding columns of the image blocks in the image 608. Therefore, these image blocks can be included in the final reference image 634 as columns 638 and 640, respectively. The remaining four central columns of the blocks in the image 608 can match very well with the four central columns of the blocks in either of the two candidate reference images. Therefore, the four central columns of the blocks in either of the candidate reference images can be used as the four central columns of the blocks in the final reference image 634.

[0091] In this way, mixing multiple candidate reference images together to create a final reference image can be block-based. The image is cut into relatively small blocks. For each block, the test image is compared with each individual reference image. The reference image block that best matches the test image block is selected to be included in the final reference image.

[0092] As further described above, the combining step is performed without modifying the selected portions of the first and second candidate reference images to thereby produce the final reference image. Therefore, this reference image generation is significantly different from other computational reference (CR) generation methods. Specifically, the final reference image generation described herein can be considered a "pick and place" operation, where once different portions of the candidate reference images are identified and selected as described above, they are placed in the positions of the portions of the test image for which they are selected. Then, the selected portions of the different candidate reference images can be "stitched together" to some extent to form the final reference image. However, these steps are performed without modifying the images themselves. In other words, when actually combining the images, the selected image portions are not modified.

[0093] In contrast, in currently used CR generation methods, combining two or more images to produce a CR involves combining the images such that the resulting image data is different from the image data of the original images. For example, a linear combination operation can be used to produce a CR and thus one or more pixels in the resulting CR can be different from the same one or more pixels in all of the images used to produce the CR. This modification of at least some of the pixels of the image is actually the main point of currently used CR methods. Specifically, the goal of currently used CR methods is usually to produce a reference image that is as quiet as possible such that it is essentially a "defect-free" image.

[0094] That goal is fundamentally different from the goal of the combining step described herein, which is to produce a final reference image that is as similar as possible to the test image in terms of noise. This is shown by Figure 6The test image 600 in [description] is described, where the test image 600 contains relatively significant noise near the left and right edges of the image as further described herein. If a significantly noise-free reference image such as the candidate reference image 616 is used to generate the difference image of the test image, then the noise at the left and right edges of the test image 600 can be detected as defects. In contrast, the final reference image 634 can be generated by the embodiments described herein and can contain relatively significant noise similar to that of the test image with the same or substantially the same characteristics. In this way, this final reference image is by no means noise-free, but when subtracted from the test image to generate the difference image, it will result in a substantially silent difference image, which can be used to detect DOIs containing DOIs with relatively low signals without detecting many nuisances. Therefore, the embodiments described herein generate a final reference image that is fundamentally different from the currently generated and used CRs, which provides a significant advantage for the embodiments described herein.

[0095] Instead of mixing together multiple candidate reference images to generate different final reference images as described above, the comparison steps described above can be performed on a per-test-image-part basis to determine which of the candidate reference images best matches the test image. In this way, after multiple candidate reference images have been generated for each test image, the best reference image can be selected from the multiple reference images. For example, in some examples different from the example shown in [description], there may be a situation where one of the candidate reference images generated in one of the various ways described herein has noise characteristics that are substantially similar to those of the test image. Specifically, if the image 634 is the third candidate reference image instead of the final reference image as described above, then clearly the image 634 matches the noise characteristics of the test image 600 much better than both of the other candidate reference images 614 and 616. Therefore, it may not be necessary to combine different parts of different candidate reference images, because the suitable final candidate reference image already exists in the candidate reference image 634. Figure 6 In some examples different from the example shown in [description], there may be a situation where one of the candidate reference images generated in one of the various ways described herein has noise characteristics that are substantially similar to those of the test image. Specifically, if the image 634 is the third candidate reference image instead of the final reference image as described above, then clearly the image 634 matches the noise characteristics of the test image 600 much better than both of the other candidate reference images 614 and 616. Therefore, it may not be necessary to combine different parts of different candidate reference images, because the suitable final candidate reference image already exists in the candidate reference image 634.

[0096] In some embodiments, the computer subsystem is configured to compute third candidate reference images from different combinations and select at least a portion of the third candidate reference image corresponding to a third portion of the test image, and the combining step includes combining the selected portions of the first, second, and third candidate reference images without modifying the selected portions of the first, second, and third candidate reference images to thereby produce a final reference image. For example, one final reference image can be constructed from three candidate reference images of each test image. One of the references can be produced in a row-independent manner, another of the references can be produced in a column-independent manner, and the third of the references can be produced by using more than one die, thereby imparting a three-dimensional (3D) nature to the resulting final reference image. In this way, the embodiments described herein can be referred to as 3DNPA. The three candidate reference image generation and merging methods described herein provide advantages for the embodiments described herein. For example, by generating three candidate reference images in a row-independent, column-independent, and multi-die manner and then merging all three reference images, better noise reduction can be obtained. In addition to using a greater number of candidate reference images, these steps can be performed as described above.

[0097] In one embodiment, following the combining step, the computer subsystem is configured to apply color variation compensation to the result of the combination. For example, even though combining different portions of different candidate reference images into the final reference image itself does not involve modifying the selected portions of the different candidate reference images, one or more additional image processing steps can be performed on the final reference image before using it for defect detection. As described above, CV compensation can be applied to the candidate reference images before the steps performed to produce the final reference image, which means that the resulting final reference image will inherently have the same CV compensation. Therefore, it may not be necessary to perform additional CV steps on the final reference image and it can be evaluated on a case-by-case basis whether to perform this step. In any case, any CV compensation applied to the resulting final reference image can be performed as further described above.

[0098] The computer subsystem is further configured to produce a difference image by comparing the test image with the final reference image. For example, as Figure 4 shown in step 408 of, the computer subsystem can be configured to produce difference image 1 by subtracting final reference image 1 from the test image. Producing the difference image can be performed in any suitable manner known in the art. Specifically, subtracting the final reference image from the test image can be performed in any suitable manner.

[0099] The computer subsystem is further configured to detect defects in the test image by applying a defect detection method to the difference image. For example, as Figure 4As shown in step 412, the computer subsystem can be configured to perform detection using the difference image 1 as input. In the simplest possible implementation, detecting a defect in step 412 can include comparing the difference image pixels or signals with a threshold and determining that any pixel or signal having a value above the threshold is a defect and any pixel or signal not having a value above the threshold is not a defect. However, many more complex defect detection methods have been developed in the art and such methods can be used in step 412. In other words, the difference image generated by the embodiments described herein can be used for defect detection in the same manner as any other difference image is used for defect detection. Additionally, the defect detection method can include any defect detection method known in the art, such as the MCAT defect detection method, which is a defect detection algorithm used by some inspection systems commercially available from KLA-Tencor or another suitable commercially available defect detection method and / or algorithm. In the case of a single detection method, the output of this step can be the final defect result 418.

[0100] The embodiments described herein can be configured for both single detection and dual detection modes. The two detection modes can be based on multiple reference generation performed as described herein. In the single detection mode, only one final reference image is required and it is used to generate a single difference image for each test image. Then, defect detection is performed using the single difference image. More than one final reference image can be generated for any one test image and used for dual detection. For example, the embodiments described herein implement dual detection of NPA defects, which is not supported in currently used NPA defect detection methods. To achieve dual detection, two final reference images can be generated from the candidate reference images instead of one final reference image.

[0101] Several embodiments can be configured and used for this defect detection. In one embodiment, the computer subsystem is configured to repeatedly select and combine to thereby generate additional final reference images. For example, using the steps described herein, two equally clear but different reference images can be constructed based on three candidate reference images. In this way, the embodiments described herein advantageously provide a dual reference generation method for nuisance reduction.

[0102] In one such embodiment, Figure 4 the step 402 shown can be performed twice, once to select different portions of the candidate reference images for the first final reference image and once to select other different portions of the candidate reference images for the second final reference image. The two selection steps can be performed in one or more of the ways described above. Additionally, the selection steps performed for each final reference image can be performed in the same or different ways. As Figure 4As shown in step 406, the computer subsystem may combine the selected portions to generate a second final reference image. Specifically, the selection step may generate a first set of selected portions for generating a first final reference image and a second set of selected portions for generating a second final reference image. The first and second sets of selected portions should include at least some different portions of any portion of the test image; otherwise, the second final reference image is redundant. The first set of selected portions may be used to generate the first final reference image in step 404, and the second set of selected portions may be used to generate the second final reference image in step 406. The combining step may be performed in other ways, as further described herein.

[0103] In another embodiment, the computer subsystem is configured to select one of the first and second candidate reference images as an additional final reference image. For example, one final reference image may be generated in the manner described, while the other final reference image may be only one of the candidate reference images, rather than generating two final reference images that are combinations of different portions from different candidate reference images. The best candidate reference image to be used as the second final reference image may be selected as further described herein.

[0104] In a further embodiment, the computer subsystem is configured to select an additional candidate reference image as an additional final reference image. For example, the additional candidate reference image may be available and / or specifically generated for consideration as the second final reference image. In one such instance, the computer subsystem may be configured to acquire an image of an adjacent region of interest in the same die or an adjacent die. In this way, this image may not be a CR because it may not be generated by combining two or more images. Instead, this image may be an image generated only by the inspection subsystem, possibly using one or more image processing steps (such as CV compensation, high-pass filtering, and the like) performed on it. Then, this image may be compared with the test image and / or one or more of the other candidate reference images to determine which one or both of the images are most suitable for defect detection. In any case, multiple candidate reference images may be generated for each test image, and the computer subsystem may select the best two reference images instead of the best one reference image for defect detection.

[0105] In another such instance, the two final reference images may result from a selection process performed on three generated candidate reference images. Specifically, the best two candidate reference images may be selected for dual detection. Special attention may be paid to ensure that both of the selected reference images have good quality. For example, in the cell area (e.g., Figure 5At the corner regions of the cell regions shown in, the best reference image typically comes from the third reference image (e.g., the inter-die operation type reference). The second-best reference image may have a much degraded quality at the corner regions (i.e., it may be very different from the same region in the test image). In such cases, the computer subsystem can force the two selected reference images for dual detection to be of the same type, which means they come from the best multi-die candidate reference images. In this way, the embodiments described herein can ensure that the dual detection has a relatively high sensitivity.

[0106] The generation of the dual final reference images can activate dual detection in NPA defect detection. In any of the above embodiments, the computer subsystem can be configured to generate an additional difference image by comparing the test image with an additional final reference image and to detect defects in the test image by applying a defect detection method to the additional difference image. Regardless of which additional final reference image is generated and selected for use in the embodiments described herein, the computer subsystem can also generate a difference image 2 by subtracting the final reference image 2 from the test image (as shown in step 410 of Figure 4 and performing detection (as shown in using the difference image 2 as an input in step 414). These steps can be performed as further described herein.

[0107] The defect detection method determines that there is a defect at the location in the test image only when the defect detection method detects a defect at the corresponding locations in the difference image and the additional difference image. For example, as shown in step 416 of Figure 4 , the defects detected by both detection steps 412 and 414 can be combined into a set of defects detected in the two comparisons. Then, the set of combined defects can be analyzed to generate Figure 4 the final defect result 418 shown in, which contains only the defects detected at the corresponding locations in the first and second difference images. Specifically, the computer subsystem can compare the defects detected in the difference images, and any commonality among the defects will be designated as a real defect, and any differences between the defects will be designated as a nuisance or a ghost. The computer subsystem can compare the results of different defect detection steps performed using different difference images in any suitable way such that the results at the corresponding locations in different difference images can be compared to thereby determine whether the detected defects are real defects or nuisances or ghosts. In this way, the defect detection method is preferably configured such that a defect is reported only when a defect is detected at the same location or the same pixel in two difference images. Thus, the embodiments described herein can perform dual detection that will help suppress nuisance detection. In other words, the CR dual detection can minimize the impact of the only nuisance or ghost in the reference image on the defect detection result through arbitration.

[0108] The embodiments described herein also advantageously provide a new fragment-based detection strategy for suppressing nuisance detection. In one embodiment, a computer subsystem is configured to divide an image of a region of interest corresponding to a cell region on a sample into test images corresponding to test cells and two or more other images corresponding to two or more other cells adjacent to the test cells, and a defect detection method divides the region of interest into different fragments of the cell region and separately detects defects in each of the different fragments. In this way, the embodiments described herein can add image segmentation to group different locations into separate detection processes. The computer subsystem can divide the entire image into different regions and can determine the size of each region based on nuisance differences. During the detection phase, each region can have independent parameters to accommodate different nuisance control requirements.

[0109] In one embodiment, the different fragments include a first fragment for the top and bottom edges of the cell region, a second fragment for the left and right edges of the cell region, a third fragment for the center of the cell region, and a fourth fragment for the corners of the cell region. Figure 7 FIG. shows one such embodiment of fragment-based detection for improved nuisance reduction. In this embodiment, the entire image of the region of interest 700 is divided into different fragments. The fragments include a center fragment 700a, left / right edge fragments 700b, top / bottom edge fragments 700c, and corner fragments 700d. Each instance of this region of interest on the sample can be divided into these four fragments.

[0110] Each of the steps described herein can be performed either jointly or independently for each of the fragments. For example, for a test image corresponding to the entire region of interest, a final reference image can be generated as described herein, and a difference image can be generated for the entire region of interest by subtracting the final reference image from the test image. Then, the difference image can be divided into corresponding parts for each of the different fragments, and defect detection can be performed separately for each of the parts of the difference image.

[0111] In one such instance, in Figure 7 the graph 702 shown in FIG., different noise clouds can be generated separately and independently for each different fragment in the region of interest. Specifically, a noise cloud 702a can be generated only for the center fragment 700a, a noise cloud 702b can be generated only for the left / right edge fragments 700b, a noise cloud 702c can be generated only for the top / bottom edge fragments 700c, and a noise cloud 702d can be generated only for the corner fragments 700d. Then, defect detection can be performed separately for each noise cloud. During the detection phase, each region can have independent parameters to accommodate different nuisance control requirements. In this way, individual parameters can be used to form separate noise clouds for each fragment to achieve an optimal balance between DOI detection and nuisance suppression.

[0112] Multiple noise cloud graphs can be generated for each test image, such as the graph shown in Figure 7 The noise cloud shown in graph 702 is only for a single test image, and similar graphs can be generated for other test images. Additionally, the noise cloud shown in graph 702 is based on the difference image generated for one test image. A separate noise cloud graph can be generated for another difference image generated for the same test image (such as in double detection, where multiple difference images are generated for a single test image).

[0113] Any of the computer subsystems described herein can generate test results, which can include the results of any of the steps described herein. The test results can include information about the detected defects, such as the defect ID, location, size, detection score, information about the defect classification (such as the class label or ID, etc.) of the bounding box of the detected defect, or any such suitable information known in the art. The results of the defects can be generated by the computer subsystem in any suitable manner. The results of the defects can have any suitable form or format, such as a standard file type. The computer subsystem can generate and store the results such that the results can be used by the computer subsystem and / or another system or method to perform one or more functions on the sample or another sample of the same type.

[0114] The computer subsystem can be configured to store the information about the detected defects in any suitable computer-readable storage medium. The information can be stored together with any of the results described herein and can be stored in any manner known in the art. The storage medium can include any of the storage media described herein or any other suitable storage medium known in the art. After the information is stored, the information can be accessed in the storage medium and used by any of the method or system embodiments described herein, formatted for display to the user, used by another software module, method, or system, etc.

[0115] The results and information generated by performing an inspection on a sample can be used in various ways through the embodiments described herein and / or other systems and methods. Such functions include, but are not limited to, changing a process that has been or will be performed on the inspected sample or another sample in a feedback or feedforward manner, such as a manufacturing process or step. For example, a computer subsystem can be configured to determine one or more changes to a process that has been or will be performed on a sample inspected as described herein based on the detected defect(s). The process change can include any suitable change to one or more parameters of the process. The computer subsystem preferably determines those changes such that defects on other samples on which a corrective process is performed can be reduced or prevented, defects on the sample can be corrected or eliminated in another process performed on the sample, defects can be compensated for in another process performed on the sample, and so on. The computer subsystem can determine such changes in any suitable manner known in the art.

[0116] Those changes can then be sent to a semiconductor manufacturing system (not shown) or a storage medium (not shown) that has access to the computer subsystem and the semiconductor manufacturing system. The semiconductor manufacturing system can be or can not be part of the system embodiments described herein. For example, the computer subsystem and / or inspection subsystem described herein can be coupled to a semiconductor manufacturing system via, for example, one or more common elements such as a housing, a power supply, a sample handling device or mechanism, and the like. The semiconductor manufacturing system can include any semiconductor manufacturing system known in the art, such as a lithography tool, an etching tool, a chemical mechanical polishing (CMP) tool, a deposition tool, and the like.

[0117] The embodiments described herein have several advantages over other methods and systems for detecting defects on a sample. For example, the embodiments described herein provide better reference generation to reduce inspection noise in the presence of color variations and provide new detection strategies to further suppress nuisances. Additionally, the embodiments described herein provide new reference generation methods and detection strategies at non-pattern regions, which provide better differential image generation with less and smaller noise and thus higher sensitivity and stronger nuisance suppression ability in the presence of color variations.

[0118] The advantages described above are achieved by several important new features of the embodiments described herein. For example, the embodiments described herein enable the new possibility of generating two or even three candidate reference images simultaneously for any test image. This ability is new for both indistinguishable device patterned features and distinguishable features. For distinguishable repeating unit regions, different from the embodiments described herein, currently used inspection systems and methods typically can generate reference images only along the repeating direction. The multi-candidate reference image mixing process described herein can also be useful for inspecting indistinguishable patterned regions as well as other types of regions. Additionally, the embodiments described herein support single detection using one best final reference image and dual detection using two best reference images generated in any of the ways described herein. The embodiments described herein are also not specific to any one defect detection method and can be used to generate inputs for any suitable defect detection method known in the art.

[0119] Each of the embodiments of the system described above can be combined together into a single embodiment. In other words, unless otherwise stated herein, no system embodiment is mutually exclusive with any other system embodiment.

[0120] Another embodiment relates to a computer-implemented method for detecting defects on a sample. The method includes obtaining an image of the sample generated by an inspection subsystem, the image including a test image and two or more other images corresponding to the test image. The method also includes calculating first and second reference images from different combinations of at least two of the test image and the two or more other images (as in step 400 shown in Figure 4 ). The method also includes at least selecting a portion of a first candidate reference image corresponding to a first portion of the test image and a portion of a second candidate reference image corresponding to a second portion of the test image (as in step 402 of Figure 4 ). Additionally, the method includes combining the selected portions of the first and second candidate reference images without modifying the selected portions of the first and second candidate reference images to thereby generate a final reference image (as in step 404 of Figure 4 ). The method also includes generating a difference image by comparing the test image with the final reference image (as in step 408 of Figure 4 ). Additionally, the method includes detecting a defect in the test image by applying a defect detection method to the difference image (as in step 412 shown in Figure 4 ). The obtaining, calculating, selecting, combining, generating, and detecting steps are performed by a computer subsystem coupled to an inspection subsystem configurable according to any of the embodiments described herein.

[0121] Each of the steps of the method may be performed as further described herein. The method may also include any other steps that may be performed by the inspection subsystem and / or computer subsystem described herein. Additionally, the methods described above may be performed by any of the system embodiments described herein.

[0122] Additional embodiments relate to a non-transitory computer-readable medium that stores program instructions executable on a computer system to perform a computer-implemented method for detecting defects on a sample. Figure 8 One such embodiment is shown in. Specifically, as Figure 8 shown, the non-transitory computer-readable medium 800 includes program instructions 802 executable on a computer system 804. The computer-implemented method may include any steps of any of the methods described herein.

[0123] The program instructions 802 for implementing the methods described herein may be stored on the computer-readable medium 800. The computer-readable medium may be a storage medium, such as a magnetic disk or optical disk, magnetic tape, or any other suitable non-transitory computer-readable medium known in the art.

[0124] The program instructions may be implemented in any of a variety of ways, including process-based techniques, component-based techniques, and / or object-oriented techniques, etc. For example, the program instructions may be implemented using ActiveX controls, C++ objects, JavaBeans, Microsoft Foundation Classes ("MFC"), SSE (Streaming SIMD Extensions), or other techniques or methodologies as desired.

[0125] The computer system 804 may be configured according to any of the embodiments described herein.

[0126] In view of this description, further modifications and alternative embodiments of various aspects of the invention will be apparent to those skilled in the art. For example, methods and systems for detecting defects on a sample are provided. Accordingly, this description should be construed as merely illustrative and for the purpose of teaching those skilled in the art the general manner of practicing the invention. It is to be understood that the forms of the invention shown and described herein are to be taken as presently preferred embodiments. Elements and materials may be substituted for those illustrated and described herein, parts and processes may be reversed, and certain features of the invention may be utilized independently, all of which will be apparent to those skilled in the art after benefiting from the description of the invention. Changes may be made to the elements described herein without departing from the spirit and scope of the invention as described in the appended claims.

Claims

1. A system configured to detect defects on a sample, comprising: An inspection subsystem configured to generate an image of the sample, the image including a test image and two or more other images corresponding to the test image; And A computer subsystem configured to: Calculate first and second candidate reference images from different combinations of at least two of the test image and the two or more other images; Select at least a portion of the first candidate reference image corresponding to a first portion of the test image and a portion of the second candidate reference image corresponding to a second portion of the test image; Combine the selected portions of the first and second candidate reference images without modifying the selected portions of the first and second candidate reference images to thereby generate a final reference image; Generate a difference image by comparing the test image with the final reference image; And Detect a defect in the test image by applying a defect detection method to the difference image.

2. The system according to claim 1, wherein the test image and the two or more other images are generated in an area of the sample that includes only indistinguishable, repeating device patterns.

3. The system according to claim 1, wherein the computer subsystem is further configured to divide the image corresponding to the area of interest on the sample into the test image corresponding to a test cell and the two or more other images corresponding to two or more other cells adjacent to the test cell.

4. The system according to claim 1, wherein prior to the calculation, the computer subsystem is further configured to apply color variation compensation to the test image and the two or more other images.

5. The system according to claim 1, wherein after the combination, the computer subsystem is further configured to apply color variation compensation to the result of the combination.

6. The system according to claim 1, wherein the first of the different combinations of at least two of the test image and the two or more other images used to calculate the first candidate reference image includes a portion of the image generated only in the x direction across the sample.

7. The system according to claim 6, wherein the second of the different combinations of at least two of the test image and the two or more other images used to calculate the second candidate reference image includes an additional portion of the image generated only in the y direction across the sample.

8. The system according to claim 1, wherein one of the different combinations of at least two of the test image and the two or more other images used to calculate one of the first and second candidate reference images includes all of the images generated on two or more die on the sample during the operation of the image.

9. The system according to claim 1, wherein the computer subsystem is further configured to compute a third candidate reference image from the different combinations and at least select a portion of the third candidate reference image corresponding to a third portion of the test image, and wherein the combination includes combining the selected portions of the first, second, and third candidate reference images without modifying the selected portions of the first, second, and third candidate reference images to thereby produce the final reference image.

10. The system according to claim 1, wherein the selection includes identifying which portions of the first and second candidate reference images best match different portions of the test image.

11. The system according to claim 1, wherein the selection includes identifying which portions of the first and second candidate reference images best match corresponding portions of the test image.

12. The system according to claim 1, wherein the computer subsystem is further configured to repeat the selection and combination to thereby produce additional final reference images, generate additional difference images by comparing the test image with the additional final reference images, and detect defects in the test image by applying the defect detection method to the additional difference images, and wherein the defect detection method determines that a defect exists at a location in the test image only if the defect detection method detects a defect at corresponding locations in the difference image and the additional difference image.

13. The system according to claim 1, wherein the computer subsystem is further configured to select one of the first and second candidate reference images as an additional final reference image, generate an additional difference image by comparing the test image with the additional final reference image, and detect defects in the test image by applying the defect detection method to the additional difference image, and wherein the defect detection method determines that a defect exists at a location in the test image only if the defect detection method detects a defect at corresponding locations in the difference image and the additional difference image.

14. The system according to claim 1, wherein the computer subsystem is further configured to select an additional candidate reference image as an additional final reference image, generate an additional difference image by comparing the test image with the additional final reference image, and detect defects in the test image by applying the defect detection method to the additional difference image, and wherein the defect detection method determines that a defect exists at a location in the test image only if the defect detection method detects a defect at corresponding locations in the difference image and the additional difference image.

15. The system according to claim 1, wherein the computer subsystem is further configured to divide the image of the region of interest corresponding to the unit region on the sample into the test image corresponding to the test unit and the two or more other images corresponding to two or more other units adjacent to the test unit, and wherein the defect detection method divides the region of interest into different segments of the unit region and separately detects defects in each of the different segments.

16. The system according to claim 15, wherein the different segments include a first segment for the top and bottom edges of the unit region, a second segment for the left and right edges of the unit region, a third segment for the center of the unit region, and a fourth segment for the corners of the unit region.

17. The system according to claim 1, wherein the inspection subsystem is further configured as an inspection subsystem based on light.

18. The system according to claim 1, wherein the inspection subsystem is further configured as an inspection subsystem based on an electron beam.

19. A non-transitory computer-readable medium storing program instructions executable on a computer system to perform a computer-implemented method for detecting defects on a sample, wherein the computer-implemented method comprises: obtaining an image of the sample generated by an inspection subsystem, the image including a test image and two or more other images corresponding to the test image; calculating first and second candidate reference images from different combinations of at least two of the test image and the two or more other images; selecting at least a portion of the first candidate reference image corresponding to a first portion of the test image and a portion of the second candidate reference image corresponding to a second portion of the test image; combining the selected portions of the first and second candidate reference images without modifying the selected portions of the first and second candidate reference images to thereby generate a final reference image; generating a difference image by comparing the test image with the final reference image; and detecting defects in the test image by applying a defect detection method to the difference image.

20. A computer-implemented method for detecting defects on a sample, comprising: obtaining an image of the sample generated by an inspection subsystem, the image including a test image and two or more other images corresponding to the test image; calculating first and second candidate reference images from different combinations of at least two of the test image and the two or more other images; selecting at least a portion of the first candidate reference image corresponding to a first portion of the test image and a portion of the second candidate reference image corresponding to a second portion of the test image; combining the selected portions of the first and second candidate reference images without modifying the selected portions of the first and second candidate reference images to thereby generate a final reference image; generating a difference image by comparing the test image with the final reference image; and Defects in the test image are detected by applying a defect detection method to the difference image, wherein the acquisition, calculation, selection, combination, generation, and detection are performed by a computer subsystem coupled to the inspection subsystem.

Images