Inspecting noisy patterned features
The SDI/SID approach for RDL inspection addresses noise and misalignment issues by using design-based masks and spatial continuity/relative intensity metrics, enhancing defect detection accuracy in RDL layers.
Patent Information
- Application Number
- JP2023507276
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-08-02
- Filing Date
- 2021-07-21
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2041-07-21
AI Technical Summary
Traditional inspection methods for detecting defects in redistribution layers (RDL) on semiconductor wafers are hindered by non-consistent sub-die shifts and metal surface roughness, which introduce significant noise, making it difficult to detect line open defects using intensity difference-based techniques.
A system and method that utilizes single die inspection (SDI) and single image detection (SID) to identify patterned features on RDL layers, employing design-based masks and spatial continuity or relative intensity metrics to detect defects without relying on die-to-die comparisons, effectively addressing noise and misalignment issues.
The method enhances the detection of complete and partial line open defects in RDL layers by reducing noise interference and improving sensitivity, allowing for more accurate defect identification and reducing false positives.
Smart Images

Figure 0007590545000001 
Figure 0007590545000002 
Figure 0007590545000003
Abstract
Description
[Technical field]
[0001] The present invention generally relates to methods and systems configured to detect defects on a specimen, such as defects on a redistribution layer (RDL) on a back-end packaging wafer. [Background technology]
[0002] The following descriptions and examples are not admitted to be prior art by virtue of their inclusion in this section.
[0003] In the fabrication of semiconductor devices, such as logic and memory devices, a substrate, such as a semiconductor wafer, is typically processed using a number of manufacturing processes to form the various features and levels of the semiconductor devices. For example, in a semiconductor manufacturing process known as lithography, a pattern is transferred from a reticle onto a photoresist disposed on the semiconductor wafer. Further examples of semiconductor manufacturing processes include, but are not limited to, chemical mechanical polishing (CMP), etching, deposition, and ion implantation. After the semiconductor devices are formed in an array on a semiconductor wafer, they may be separated into individual semiconductor devices.
[0004] Inspection processes are used at various steps during the semiconductor manufacturing process to detect wafer-side defects, helping to increase yields and therefore profits in the manufacturing process. Inspection has always been an important part of semiconductor device manufacturing. However, as the dimensions of semiconductor devices shrink, even smaller defects can cause failures in those devices, making inspection more important than ever for the successful manufacture of acceptable semiconductor devices.
[0005] Since the early days of the integrated circuit (IC) industry, wafer inspection has been primarily based on adjacent die-to-die image intensity differences, since lithography typically prints adjacent dies that are "almost identical". As technology advances, other techniques, called "golden reference die" or "standard reference die", have been developed for inspecting dies with common die defects caused by mask defects. Whether performing die-to-die inspection or die-to-golden die inspection, the common assumption is that each die on the wafer is nearly "identical". These traditional inspection methods have been very effective in lithography-based wafer manufacturing processes.
[0006] In modern IC manufacturing, a new packaging process is applied; a final die unit is assembled by connecting two or more good sub-dies that have been diced and tested via redistribution layer (RDL) lines, as shown in Figure 2. Specifically, a final die unit 200 is assembled with sub-dies 202 and 204 that are connected and coupled by multiple RDL lines 206. In this way, a new die is created from the sub-dies in a novel back-end advanced wafer-level packaging (aWLP) process.
[0007] Because the sub-dies are mechanically placed on the mold and the RDL lines are placed on the sub-dies, there may be inconsistent sub-die shifts between adjacent die units. This type of inconsistency may add a significant amount of noise to the inspection results when performing traditional die-to-die difference calculations, even when the RDL lines in adjacent die units are perfectly aligned. For example, as shown in FIG. 3, adjacent assembled die units may exhibit inconsistent shifts below the RDL lines. Specifically, FIG. 3 shows two final die units 300 and 302. The final die unit 300 is formed by sub-dies 304 and 306 connected by an RDL line 308, and the final die unit 302 is formed by sub-dies 310 and 312 connected by an RDL line 314. As shown more clearly in the exploded view 316 of the final die unit 300 and the exploded view 318 of the final die unit 302, in separate final die units, a non-consistent shift may appear between the sub-dies below the RDL line.
[0008] Moreover, the RDL line is mainly made of metal (aluminum or copper). The defects of interest (DOI) in the RDL line include open line defects (line breaks) and partial open line defects (also called "mouse bite" defects), as shown in FIG. 4 and FIG. 5. Specifically, FIG. 4 shows a test image 400, a reference image 402, and a difference image 404 of a complete open line defect in an RDL line. As can be clearly seen in the circle portion 406 of the test image 400, the open line defect is evident in the test image. However, as shown in the difference image 404 (generated by subtracting the reference image 402 from the test image 400), the open line defect cannot be detected in the circle portion 408 corresponding to the circle portion 406 of the test image 400, and the portion of the difference image corresponding to the open line defect cannot be distinguished from other portions of the difference image, such as the portion shown in the square portion 410. Therefore, the difference image 404 cannot be used to successfully detect the open line defect.
[0009] 5 shows a reference image 500, a test image 502, and a difference image 504 for another line open defect on an RDL line. As can be seen from the test image 502, a partial line open defect located in a circle portion 510 of the test image is evident in the test image. However, as shown in the difference image 504 (generated by subtracting the reference image 500 from the test image 502), the partial line open defect cannot be detected in a circle portion 514 corresponding to the circle portion 510 in the test image. In addition, the portion of the difference image corresponding to the partial line open defect cannot be distinguished from other portions of the difference image, such as shown in a square portion 512 corresponding to a non-defect portion 506 of the reference image and a non-defect portion 508 of the test image. Therefore, the difference image 504 cannot be used to detect the partial line open defect visible in the test image 502.
[0010] Thus, it is clear that there are obvious open line or partial open line defects in the test images in both Figures 4 and 5. However, the surface roughness of the metal RDL lines makes intensity difference based inspection extremely difficult even though real defect signals are present in the test images; because the open line or partial open line signals in the difference images are completely buried in the noise caused by the random roughness in the corresponding lines in the test and reference images. [Prior art documents] [Patent documents]
[0011] [Patent Document 1] U.S. Patent No. 7,570,796 [Patent Document 2] U.S. Patent No. 7,676,077 [Patent Document 3] U.S. Patent No. 7,782,452 [Patent Document 4] U.S. Patent No. 8,664,594 [Patent Document 5] U.S. Patent No. 8,692,204 [Patent Document 6] U.S. Patent No. 8,698,093 [Patent Document 7] U.S. Patent No. 8,716,662 [Patent Document 8] U.S. Patent No. 8,126,255 [Patent Document 9] U.S. Patent No. 9,222,895 [Patent Document 10] U.S. Pat. No. 10,607,119 Summary of the Invention [Problem to be solved by the invention]
[0012] Therefore, it would be advantageous to develop a system and method for detecting defects, such as full and / or partial line open defects in RDL lines on a specimen, such as a back-end aWLP specimen, that does not have one or more of the drawbacks discussed above. [Means for solving the problem]
[0013] The following description of embodiments is not to be construed in any way as limiting the subject matter of the appended claims.
[0014] One embodiment relates to a system configured to detect defects on a specimen. The system includes an inspection subsystem configured to generate a set of images of the specimen. The system also includes one or more computer subsystems configured to detect defect candidates on the specimen. During the defect candidate detection, a patterned feature is identified in a test image included in the set of images generated for the specimen. During the defect candidate detection, for at least one pixel in the test image that is within the patterned feature, a difference is determined between a characteristic of the at least one pixel and a characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel. Additionally, during the defect candidate detection, a defect candidate is detected at the at least one pixel based on the determined difference. The system may be further configured as described herein.
[0015] Another embodiment relates to a computer-implemented method for detecting defects on a specimen. The method includes identifying a patterned feature in a test image included in a set of images generated of the specimen by an inspection subsystem. The method also includes, for at least one pixel in the test image that is within the patterned feature, determining a difference between a characteristic of the at least one pixel and a characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel. Additionally, the method includes detecting a defect candidate at the at least one pixel based on the determined difference. The identification, determination, and detection are performed by one or more computer subsystems coupled to the inspection subsystem.
[0016] Each of the steps of the above-described method may be performed as further described herein. In addition, the above-described method embodiments may include any other step(s) of any other method(s) described herein. Furthermore, the above-described methods may be performed by any of the systems described herein.
[0017] Another embodiment relates to a non-transitory computer-readable medium having stored thereon program instructions executable on one or more computer systems to execute a computer-implemented method for detecting defects on a specimen, the computer-implemented method including 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 described herein. In addition, the computer-implemented method on which the program instructions may be executed may include any other step(s) of any other method(s) described herein.
[0018] Further advantages of the present invention will become apparent to those skilled in the art in light of the following detailed description of the preferred embodiments and by reference to the accompanying drawings, in which: [Brief description of the drawings]
[0019] [Figure 1] FIG. 1 is a schematic diagram illustrating a side view of an embodiment of a system configured as described herein. [Figure 1a] FIG. 2 is a schematic diagram illustrating a side view of another system embodiment configured as described herein. [Diagram 2] FIG. 2 is a schematic illustrating a top view of an example die constructed from multiple sub-dies and having redistribution layer (RDL) lines formed thereon. [Diagram 3] FIG. 13 is a schematic diagram illustrating a top view of an example die comprised of multiple sub-dies exhibiting inconsistent shifting between the sub-dies below the RDL lines formed thereon. [Figure 4] FIG. 13 shows examples of test, reference and difference images for a line open defect in an RDL line. [Diagram 5] FIG. 13 shows example test, reference and difference images for another line open defect in an RDL line. [Figure 6]FIG. 2 is a schematic diagram illustrating a plan view of an example of portions having different orientations and comprising different patterned features. [Figure 7] FIG. 1 shows an image in which examples of various types of RDL intra-line defects are displayed. [Figure 8] FIG. 2 is a schematic diagram illustrating the plan view of an example patterned feature in which a difference can be determined within a given window overlaid thereon as described herein and the difference can be used to detect defect candidates within the patterned feature. [Figure 9] 1A-1C are diagrams illustrating example test images of defects that may be detected as described herein. [Figure 10] 1A-1C are example test images of different defects that can be detected as described herein. [Figure 11] 1A-1C show example test images of various patterned features and embodiments of predefined windows overlaid thereon whose properties vary depending on the orientation of the different patterned features. [Figure 12] 1 is a flowchart illustrating embodiments of steps that may be performed by embodiments described herein to detect defects on a specimen. [Figure 13] 1 is a flowchart illustrating an embodiment of steps that may be performed according to another embodiment described herein to detect defects on a specimen. [Figure 14] 1 is a flowchart illustrating an embodiment of steps that may be performed according to another embodiment described herein to detect defects on a specimen. [Figure 15] FIG. 1 is a block diagram illustrating one embodiment of a non-transitory computer-readable medium having stored thereon program instructions for causing one or more computer systems to perform the computer-implemented methods described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0020] Because the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and described in detail herein. The drawings may not be to scale. It is to be understood, however, that the drawings and detailed description thereof are not intended to limit the invention to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope of the invention as defined by the appended claims.
[0021] First, with respect to the drawings, it is noted that the drawings are not drawn to scale. In particular, the scale of some of the components in the drawings has been significantly exaggerated to emphasize the characteristics of those components. It is also noted that the drawings are not drawn to scale. Elements that are shown in more than one drawing and that may be similarly configured are designated with the same reference numerals. Unless otherwise noted in this application, all of the elements described and illustrated include any suitable and commercially available elements.
[0022] The words "design" and "design data", as used herein, generally refer to the physical design (layout) of an IC and the data derived from that physical design through complex simulations or simple geometric and Boolean operations. Such physical design may be in a data structure, such as a Graphical Data Stream (GDS) file, any other standard machine-readable file, any other file known and suitable in the art, a design database, etc. A GDSII file is one type of file used to represent that design layout data. Other examples of such files are GL1 files and OASIS files, as well as proprietary file formats, such as RDF data specific to KLA Corp., Milpitas, Calif., USA. The design may include any other design data or design data proxies, such as those described in commonly assigned U.S. Patent No. 6,233,633 issued to Zafar et al. on August 4, 2009, and U.S. Patent No. 6,233,633 issued to Kulkarni et al. on March 9, 2010, both of which are incorporated by reference herein as if fully set forth herein. In addition, the design data may include standard cell library data, integrated layout data, design data for one or more layers, derivatives of the design data, and full or partial chip design data.
[0023] "Design" and "design data" as used herein also refer to information and data generated by a semiconductor device designer in a design process and therefore available to the embodiments described herein well in advance of printing the design onto any tangible wafer. The "design" or "physical design" may be the design that would ideally be formed on the wafer.
[0024] A "nuisance," as the terminology is used herein, is a defect that is not of interest to the user and / or an event that is detected by inspection but is not actually a defect. Nuisances that are detected as events (also called "defect candidates") but are not actually defects are detected due to non-defect noise sources on the specimen (e.g., line edge roughness (LER), relatively small critical dimension (CD) variations in patterned features, thickness variations, metal grain noise, etc.) and / or marginality in the inspection subsystem itself, i.e., the configuration used for inspection. Thus, in general, the goal of inspection is to prevent nuisances on a specimen, e.g., a wafer, from being detected.
[0025] The term "defect of interest (DOI)" as used herein is defined as a real defect that is detected on a specimen and is actually present on the specimen. Typically, such DOIs are of interest to a user because the user is interested in how many and what types of real defects are present on the specimen being inspected. In some contexts, the term "DOI" may also be used to refer to a subset of all real defects on a specimen that includes only those real defects of interest to the user. For example, there may be multiple types of DOIs on any given specimen, with one or more of these types receiving more attention than the other one or more types. However, in the context of the embodiments described herein, the term "DOI" is used to refer to any real defect on a specimen.
[0026] One embodiment relates to a system configured to detect defects on a specimen. In one embodiment, the specimen is a wafer, which may include any wafer known in the semiconductor art. Although some embodiments may be described herein with reference to one or more wafers, the embodiments are not limited according to the specimens with which they may be used. For example, specimens with which the embodiments described herein may be used include reticles, flat panels, personal computer (PC) boards, and other semiconductor specimens.
[0027] In some embodiments, the specimen has a metallic upper layer. According to one such embodiment, the upper layer can be an aluminum layer. The aluminum layer can include any suitable aluminum layer known in the art. The upper layer of the specimen can include a back-end-of-line (BEOL) layer, such as any BEOL layer known in the art, including those described herein. According to further embodiments, the upper layer of the specimen can be a redistribution layer (RDL), such as any suitable configuration known in the art.
[0028] The upper layers of the specimen may also have metal lines. For example, metal lines may be provided in the BEOL and RDL layers to form various elements of devices formed on the specimen. Such metal lines may have a significant amount of "grain" noise, as detailed herein. "Grain" is the term used herein to refer to dislocations in the crystal structure of the metal (e.g., aluminum or copper). The embodiments described herein are configured to detect defects on such layers despite the grain noise.
[0029] In certain embodiments, the specimen is a post-dicing specimen. A "post-dicing" specimen can be broadly defined as a wafer or other substrate, particularly one on which a number of devices have been formed (e.g., in the form of separate dies or dice) and which has been further separated from one another according to one of a variety of techniques. A "post-dicing" specimen can also be a specimen that has been separated into a number of dies or dice but has not yet been subjected to a packaging process.
[0030] In some embodiments, the specimen also includes a high-noise layer. A "high-noise" layer, as the term is defined herein, is generally a layer whose noise is a dominant obstacle to inspection of that layer. For example, while inspection of an individual wafer layer by any inspection tool will show more or less noise relative to other layers (and techniques to handle the detection of such noise must generally be used for inspection of the individual wafer layer), the main obstacle to successful inspection of the wafer layer is often the extremely small size of the defects that must be detected. In contrast, the embodiments described herein are particularly suited to the detection of relatively large ("macro") defects having a size of about 200 nm or more. Thus, the main obstacle in such inspection is not necessarily the size of the defects that must be detected (since many inspection tool configurations are capable of detecting such large defects on most layers). Instead, the layers described herein will generally exhibit "high noise" levels in the images generated for those layers, which may make even such large size defects difficult, if not impossible to detect, although the embodiments described herein are designed to handle such noise levels through the defect detection described herein.
[0031] The embodiments described herein are designed to be particularly effective for detecting such defects for a variety of reasons. For example, in traditional front-end lithographic wafer manufacturing processes, wafer inspection is usually performed by adjacent die-to-die image comparison, due to the valid assumption that the images originating from adjacent dies are identical. With reasonable pre-processing steps, such as adjacent die image alignment and gray level uniformity scaling through histogram remapping, anomalies and thus potential defects usually cause differences above the threshold.
[0032] In a new advanced wafer level package (aWLP) technology, sub-dies (which have been tested to be good) can be joined together via RDL lines to create an organized die unit. An example of such an organized die unit is shown in FIG. 2, detailed above. As shown in FIG. 3, detailed above, the RDL lines and the sub-dies below them can have inconsistent relative spatial shifts. Even if adjacent die-to-die alignment is performed based on the RDL lines, the sub-die structures below those RDL lines can be misaligned. Such misalignment creates misalignment noise from the structures below the RDL lines, making the traditional die-to-die difference-based approach very challenging.
[0033] In addition, intensity difference based detection of RDL line open defects can be more challenging because the RDL lines are typically metallic (copper) resulting in surface roughness induced random optical features in the RDL line images, and the open line defect signal is completely buried in the surface roughness induced noise. In one such example, the inspection can detect significant noise due to RDL intra-metal line noise, such as that caused by excess metal grains. In another such example, the inspection can detect significant noise due to RDL inter-metal layer noise caused by transparent dielectric polymers above or below the RDL layer.
[0034] In contrast, the embodiments described herein provide inspection of the RDL lines without being hindered by the severe die-to-die differential noise. In addition, the embodiments described herein can perform test image only (i.e., non-die-to-die) inspection of the RDL lines. Thus, the embodiments described herein provide single die inspection (SDI) or single image detection (SID) systems and methods that can detect these defects and eliminate the die-to-die noise source in the inspection.
[0035] FIG. 1 illustrates an embodiment of a system configured to detect defects on a specimen. The system includes an inspection subsystem 10 and one or more computer subsystems, such as computer subsystem 36 and computer subsystem(s) 102. The inspection subsystem 10 is configured to generate output, such as images, related to the specimen. In some embodiments, the inspection subsystem is configured as an optical subsystem. For example, the inspection subsystem in the embodiment of FIG. 1 is configured to scan or direct light onto a tangible specimen while detecting light from the specimen to generate output related to the specimen. The inspection subsystem may also be configured to perform such scanning (or directing) and detection in a number of different modes.
[0036] In one embodiment, the inspection subsystem is configured for macro inspection. The system described herein may be referred to as a macro inspection tool. Macro inspection tools are particularly suited for inspection of relatively noisy BEOL layers, such as RDL and post-dicing applications, where defects are detected in the presence of significant noise, such as grains, on the metal line side. A macro inspection tool is defined herein as a system that has a spatial resolution of about 200 nm to about 2.0 μm or greater, not necessarily diffraction limited. This spatial resolution means that the smallest defect that can be detected by the system has a dimension of greater than about 200 nm, which is significantly larger than the smallest defect that can be detected by the most advanced inspection tools available on the market today, thus meriting the name "macro" inspector. Such systems tend to utilize longer wavelengths of light (e.g., about 500 nm to about 700 nm) than those available in the most advanced inspection tools available on the market today. Such systems can be used when the DOI has a relatively large size and potentially when a throughput of 100 wafers per hour (wph) or more is required (wafer throughput here refers to the number of 300 mm wafers inspected per hour).
[0037] In the embodiment of the system shown in FIG. 1, the inspection subsystem 10 includes an illumination subsystem configured to direct light to the specimen 14. The illumination subsystem includes at least one light source. For example, the illumination subsystem shown in FIG. 1 includes a light source 16. The illumination subsystem can be configured to direct light to the specimen at one or more angles of incidence, including one or more oblique angles and / or one or more perpendicular angles. For example, in the example shown in FIG. 1, light from the light source 16 is directed through an optical element 18 and then through a lens 20 to the specimen 14 at an oblique angle of incidence. The oblique angle of incidence can be any suitable oblique angle of incidence and can vary according to the characteristics of the specimen.
[0038] The inspection subsystem can also be configured to direct light to the specimen at different angles of incidence at different times. For example, the inspection subsystem can be configured to direct light to the specimen at angles of incidence different than that shown in FIGURE 1 by modifying one or more characteristics of one or more components of the illumination subsystem. According to 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 specimen at a different oblique or orthogonal (or near-orthogonal) angle of incidence.
[0039] The inspection subsystem may also be configured to simultaneously direct light to the sample at multiple angles of incidence. For example, the illumination subsystem may have multiple illumination channels, one of which may include light source 16, optical elements 18, and lens 20 as shown in FIG. 1, and another (not shown) may include similar elements, such as those configured differently or similarly, or may include at least one light source and potentially one or more other components, such as those described herein. When the light is directed to the sample simultaneously with other light, one or more characteristics (e.g., wavelength, polarization, etc.) of the light directed to the sample at different angles of incidence may be different so that the light resulting from the illumination of the sample at the different angles of incidence may be distinguishable from one another at the detector(s).
[0040] In some embodiments, the illumination subsystem may include only a single light source (e.g., light source 16 shown in FIG. 1 ), and the light from the light source may be split into separate optical paths (e.g., based on wavelength, polarization, etc.) by one or more optical elements (not shown) in the illumination subsystem. Each of the separate optical paths may then be directed to the sample. Multiple illumination channels may be configured to direct light to the sample simultaneously or at different times (e.g., when sequentially illuminating the sample using separate illumination channels). In some embodiments, the same illumination channel may be configured to direct light to the sample with different characteristics at different times. For example, in some embodiments, optical element 18 may be configured as a spectral filter, and the characteristics of the spectral filter may be changed in various ways (e.g., by changing the spectral filter) to direct different wavelengths of light to the sample at different times. The illumination subsystem may direct light having different or the same characteristics to the sample at different or the same angles of incidence, either sequentially or simultaneously, and may have any other configuration known and suitable in the art.
[0041] According to an embodiment, the light source 16 may comprise a broadband plasma (BBP) light source, such that the light generated by the light source and directed to the sample includes broadband light. However, the light source may comprise any other suitable light source, such as a laser. The laser may be any suitable laser known in the art and may be configured to generate light at any suitable wavelength or wavelengths known in the art. Additionally, the laser may be configured to generate monochromatic or near monochromatic light, such that the laser is a narrowband laser. The light source may also comprise a polychromatic light source that emits light at multiple discrete wavelengths or wavebands.
[0042] Light from optical element 18 may be focused onto sample 14 by lens 20. Although lens 20 is shown in FIG. 1 as a single refractive optical element, in practice lens 20 may comprise multiple refractive and / or reflective optical elements that cooperate to focus light from the optical element onto the sample. The illumination subsystem shown in FIG. 1 and described herein may include any other suitable optical elements (not shown). Examples of such optical elements include, but are not limited to, deflector(s), spectral filter(s), spatial filter(s), reflective optical element(s), apodizer(s), beam splitter(s), aperture(s), and any other suitable type of optical element known in the art. Additionally, the inspection subsystem may be configured to vary one or more of the components of the illumination subsystem based on the type of illumination used for inspection.
[0043] The inspection subsystem may also include a scanning subsystem configured to scan light over the sample. For example, the inspection subsystem may include a stage 22 on which the sample 14 is positioned during inspection. The scanning subsystem may include any suitable mechanical and / or robotic assembly (including the stage 22) configured to move the sample and scan light over the sample. Additionally or alternatively, the inspection subsystem may be configured such that one or more optical elements included in the inspection subsystem perform some type of optical scanning of the sample. The scanning of light over the sample may be performed in any suitable manner, for example along a serpentine or spiral path.
[0044] The inspection subsystem further includes one or more detection channels. At least one of the one or more detection channels includes a detector configured to detect light resulting from illumination of the sample and generate an output in response to the detected light. For example, the inspection subsystem shown in FIG. 1 includes two detection channels, one formed by collector 24, element 26, and detector 28, and the other formed by collector 30, element 32, and detector 34. As shown in FIG. 1, the two detection channels are configured to collect and detect light at different collection angles. In some examples, both 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 can be configured to detect other types of light from the sample (e.g., specular light).
[0045] As detailed in FIG. 1, both detection channels are in the plane of the paper, and the illumination subsystem is also in the plane of the paper. That is, in this embodiment, both detection channels are in the plane of incidence (e.g., are centered therein). However, one or more of the detection channels may be located outside the plane of incidence. For example, the detection channel formed by collector 30, element 32, and detector 34 may be configured to collect and detect light scattered outside the plane of incidence. In such a case, the detection channel may be referred to as a "side" channel, and may be centered in a plane that is substantially perpendicular to the plane of incidence.
[0046] Although the embodiment shown in FIG. 1 has two detection channels, the inspection subsystem may have other numbers of detection channels (e.g., a single detection channel or more than one detection channel). In one such example, the detection channel formed by collector 30, element 32, and detector 34 may form a side channel as described above, and the inspection subsystem may include an additional detection channel (not shown) formed as another side channel on the opposite side of the incidence face. The inspection subsystem may then include a detection channel having collector 24, element 26, and detector 28, located at the center of the incidence face, configured to collect and detect light at one or more scattering angles normal to or near the sample surface. This detection channel may thus be referred to as a "top" channel, and the inspection subsystem may also include more than one side channel configured as described above. In this way, the inspection subsystem can be configured to have at least three channels (i.e., one top channel and two side channels), each of the at least three channels having its own collector, with each collector configured to collect light at a different scattering angle than any of the other collectors.
[0047] As detailed above, each of the detection channels in the inspection subsystem can be configured to detect scattered light. Thus, the inspection subsystem depicted in FIG. 1 can be configured for dark-field (DF) inspection of the specimen. However, the inspection subsystem can also or alternatively have detection channel(s) configured for bright-field (BF) inspection of the specimen. In other words, the inspection subsystem can have at least one detection channel configured to detect light specularly reflected from the specimen. Thus, the inspection subsystem described herein can be configured for DF imaging alone, BF imaging alone, or both. Although each collector is depicted in FIG. 1 as a single refractive optical element, each collector can also have one or more refractive optical elements and / or one or more reflective optical elements.
[0048] The detection channel or channels may include any suitable detector known in the art. For example, the detectors may include photomultiplier tubes (PMTs), charge-coupled devices (CCDs), time-delay integration (TDI) cameras, or any other suitable detector known in the art. The detectors may include non-imaging and imaging detectors. If the detectors are non-imaging detectors, each detector may be configured to detect a particular characteristic of the scattered light, such as intensity, but not as a function of position in the imaging plane. In this way, the output generated by each detector in each detection channel of the inspection subsystem may be a signal or data, but not an image signal or image data. In this type of example, a computer subsystem, such as computer subsystem 36, may be configured to generate an image of the sample based on the non-imaging output of the detector. However, in another example, the detector may be configured as an imaging detector configured to generate an image signal or image data. Thus, the inspection subsystem may be configured to generate an image in a variety of ways.
[0049] It is noted that FIG. 1 is provided herein to generally depict inspection subsystem configurations that may be incorporated into the system embodiments described herein or that may generate output used by the system embodiments described herein. As is apparent, the inspection subsystem configurations described herein may be modified to optimize the performance of the inspection subsystem, as is commonly done when designing commercial inspection systems. In addition, the systems described herein may be implemented using existing systems (e.g., by adding the described functionality to an existing system), such as the Altair™ series of tools commercially available from KLA. For such systems, the described embodiments may be provided as optional features of the system (e.g., in addition to other features of the system). Alternatively, the described inspection subsystem may be designed “from scratch” to provide an entirely new inspection subsystem. The inspection subsystem may further be configured as described in U.S. Patent Application Publication No. 2010 / 0133934, issued to Mehanian et al. on Aug. 24, 2010, which is incorporated herein by reference and is considered fully disclosed herein.
[0050] The computer subsystem 36 may be coupled in any suitable manner (e.g., via one or more transmission media, including, for example, "wired" and / or "wireless" transmission media) to the detectors of the inspection subsystem so that the computer subsystem can receive output generated by the detectors during scanning of the sample. The computer subsystem 36 may be configured to use the detector outputs to perform a number of functions detailed herein.
[0051] The computer subsystem shown in FIG. 1 (as well as other computer subsystems described herein) may be referred to herein as computer system(s). Each of the computer subsystem(s) or system(s) described herein may take a variety of forms, including a personal computer system, an image computer, a mainframe computer system, a workstation, a network appliance, an Internet appliance, or other device. In general, the term "computer system" may be broadly defined to include any device having one or more processors that execute instructions obtained from a storage medium. The computer subsystem(s) or system(s) may include any suitable processor known in the art, such as a parallel processor. In addition, the computer subsystem(s) or system(s) may include a computer platform with high speed processing and software, whether a stand-alone or networked tool.
[0052] When there are multiple computer subsystems in a system, the separate computer subsystems may be coupled together to allow images, data, information, instructions, etc. to be sent between the computer subsystems. For example, computer subsystem 36 may be coupled to computer subsystem(s) 102, as shown by the dashed lines in Figure 1, by any suitable transmission medium, including any suitable wired and / or wireless transmission medium known in the art. Two or more of such computer subsystems may also be substantially coupled by a shared computer-readable storage medium (not shown).
[0053] Although the inspection subsystem has been described above as being an optical or light-based inspection subsystem, in some embodiments, the inspection subsystem is configured as an electron beam subsystem. One such embodiment is shown in FIG. 1a, where the inspection subsystem includes an electron column 122 coupled to a computer subsystem 124. As also shown in FIG. 1a, the electron column includes an electron beam source 126 configured to generate electrons that are focused onto a specimen 128 by one or more elements 130. The electron beam source may include, for example, a cathode source or emitter tip, and the one or more elements 130 may include, for example, a gun lens, an anode, a beam limiting aperture, a gate valve, a beam current selection aperture, an objective lens, and a scanning subsystem, any of which may include any such element known and suitable in the art.
[0054] Electrons returning from the sample (e.g., secondary electrons) may be focused by one or more elements 132 onto a detector 134. The one or more elements 132 may include, for example, a scanning subsystem, which may be the same as the scanning subsystem included in element(s) 130.
[0055] Any other suitable elements known in the art may be included within the electron column. In addition, the electron column may be further configured as described in U.S. Patent No. 6,399,433 issued to Jiang et al. on April 4, 2014, U.S. Patent No. 6,399,433 issued to Kojima et al. on April 8, 2014, U.S. Patent No. 6,399,433 issued to Gubbens et al. on April 15, 2014, and U.S. Patent No. 6,399,433 issued to MacDonald et al. on May 6, 2014, each of which is incorporated by reference herein as if fully set forth herein.
[0056] Although the electron column is shown in FIG. 1a as being configured such that electrons are directed at the sample at one oblique angle of incidence and scattered from the sample at another oblique angle, the angle at which the electron beam is directed toward and scattered from the sample can be any suitable angle. In addition, the electron beam-based subsystem can be configured to generate images of the sample using multiple modes (e.g., with different illumination angles, collection angles, etc.) as described herein. The multiple modes of the electron beam-based subsystem can differ in any imaging parameters of the inspection subsystem.
[0057] The computer subsystem 124 may be coupled to a detector 134 as described above. The detector may detect electrons returning from the surface of the specimen to form an electron beam image of the specimen, which may include any suitable electron beam image. The computer subsystem 124 may be configured to use the output generated by the detector 134 to perform one or more of the functions described herein with respect to the specimen. The computer subsystem 124 may also be configured to perform any additional step(s) described herein. The system having the inspection subsystem shown in FIG. 1a may be further configured as described herein.
[0058] It is noted that FIG. 1a is provided herein to provide a general depiction of an e-beam based subsystem configuration that may be incorporated into the embodiments described herein. As with the optical subsystems described above, the e-beam 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 existing systems (e.g., by adding the described functionality to an existing system), such as the eSxxx series and eDR-xxxx series tools commercially available from KLA. For certain such systems, the described embodiments may be provided as an optional feature of the system (e.g., in addition to other features of the system). Alternatively, the described systems may be designed "from scratch" to provide an entirely new system.
[0059] Although the inspection subsystem has been described above as being an optical or electron beam subsystem, the inspection subsystem can also be an ion beam based subsystem. Such an inspection subsystem can be configured as shown in FIG. 1a, except that the electron beam source would be replaced with any suitable ion beam source known in the art. In addition, the inspection subsystem can be any other suitable ion beam based subsystem, such as those incorporated into commercially available focused ion beam (FIB) systems, helium ion microscopy (HIM) systems, and secondary ion mass spectrometry (SIMS) systems.
[0060] The inspection subsystem noted above is configured to generate an output related to the tangible specimen by scanning energy (e.g., light, electrons, etc.) over the tangible specimen. In this way, the inspection subsystem can be configured as a "real" subsystem rather than a "virtual" subsystem. However, the storage medium (not shown) and computer subsystem(s) 102 shown in FIG. 1 can also be configured as a "virtual" inspection subsystem configured to generate an image of the specimen. In particular, the storage medium and computer subsystem(s) can be configured as a "virtual" inspection system and used as an inspection subsystem in the embodiments described herein, as described in U.S. Patent No. 6,399,433 issued to Bhaskar et al. on February 28, 2012, and U.S. Patent No. 6,399,433 issued to Duffy et al. on December 29, 2015, both of which are commonly assigned and incorporated herein by reference and are considered fully set forth herein. The embodiments described herein can be further configured as described in both patent documents.
[0061] As detailed above, an inspection subsystem may be configured to generate output associated with the specimen in a number of different modes. In general, a "mode" may be defined by the value of a parameter of the inspection subsystem used to generate output associated with the specimen (e.g., an image). Thus, different values of at least one parameter of the inspection subsystem (other than the location on the specimen at which the output is generated) may define different modes. For example, an optical inspection subsystem may provide illumination using at least one different wavelength of light in different modes. The modes may have different illumination wavelengths as described herein (e.g., by using different light sources, different spectral filters, etc., in different modes). According to another example, the different modes may use different illumination channels of the inspection subsystem. For example, as noted above, multiple illumination channels may be provided within the inspection subsystem. In that case, different illumination channels may be used for different modes. Additionally or alternatively, the modes may differ in one or more acquisition / detection parameters of the inspection subsystem. The inspection subsystem may also be configured to scan the specimen in separate modes, either in the same scan or in separate scans, eg, in light of the ability to simultaneously scan the specimen using multiple modes.
[0062] As mentioned above, the traditional pure die-to-die intensity difference based defect detection algorithm is not effective for detecting open line defects on aWLP samples for the reasons shown in Figures 2-5. The main obstacle is due to excessive nuisance caused by random metal line surface roughness and / or inconsistent spatial shift, i.e., although the difference based signal can be detected, the gray level intensity and geometric size of the noise in the difference signal will be larger than that of the actual open line defect. To deal with the open line defects, the embodiments described herein are specialized inspection methods and systems for detecting open line defects on RDL lines, e.g., in a single die based inspection scheme.
[0063] Although the embodiments described herein are suitable for detecting open line defects in RDL lines, and certain embodiments are described herein in connection with such defects and specimens, the embodiments described herein are not limited to such defects and specimens. For example, the embodiments described herein may be used for inspection of defects in other relatively noisy patterned features and / or where the defects and patterned features exhibit substantially separate characteristics in an inspection image.
[0064] The computer subsystem(s) is / are configured to detect defect candidates on the specimen. A "defect candidate" is defined in the present terminology as any event detected on the specimen that is not filtered out as a nuisance. For example, after defect detection, the defect candidates will include all events detected on the specimen, both real defects (or DOIs) and nuisances. After nuisance filtering, the defect candidates will include all detected defect candidates that were not eliminated by nuisance filtering. Whether the defect candidates after nuisance filtering include only DOIs depends on how well the nuisance filtering distinguishes between DOIs and nuisances. After all nuisance filtering is performed, the remaining defect candidates may be qualified as detected defects, but one or more of the qualified defects may actually be nuisances.
[0065] Candidate defects are detected by identifying patterned features in a test image of the generated images of the specimen. The test image may be any test, target, or frame image generated by the inspection subsystem. The test image may have any suitable size, such as a relatively small patch image, and the size of the patch image may vary depending on the configuration of the inspection subsystem and / or computer processing of the output generated by the inspection subsystem.
[0066] In some embodiments, the patterned features include lines on the RDL layer, and in some embodiments, the patterned features are formed of metal. For example, as described herein, a specimen for which defects may be detected by the embodiments described herein may have RDL lines formed of metal. The computer subsystem(s) may then be configured to identify the RDL lines in the test image, which may then be inspected for open line defects, as described herein. Similarly, the computer subsystem(s) may be configured to identify and inspect other patterned features formed of metal.
[0067] In certain embodiments, when identifying a patterned feature, a mask is applied to the test image to separate pixels in the test image that are within the patterned feature from all other pixels in the test image. In other words, when identifying a patterned feature in the test image, pixels in the test image that correspond to the patterned feature can be identified. One particularly suitable way in which this can be done is by applying a mask to the test image. The mask may have clearly defined and qualified areas that correspond to the patterned features and areas that do not. For example, the mask may have transparent areas that correspond to the identified patterned features and opaque areas that do not correspond to the patterned features (including patterned features not of interest, underlying or non-patterned areas, etc.). In general, the mask may have any suitable configuration known in the art. When the mask is applied to the test image, the mask may be overlaid on the test image and the test image may be aligned with respect to the mask. Such registration and alignment may be performed according to any suitable technique known in the art.
[0068] Although the identification step is described herein in the context of a patterned feature, in practice the identification step may identify multiple patterned features of interest within the same test image, whether they have the same characteristics or different characteristics (e.g., multiple RDL lines having the same characteristics, multiple RDL lines at least some of which have different shapes, orientations, sizes, etc.). In other words, the identification step is described herein in the context of a patterned feature for the purpose of brevity and clarity of understanding of the present invention. However, the identification step may be performed simultaneously or sequentially with respect to multiple patterned features. Similarly, although the present invention is described herein in the context of a single test image for clarity and brevity, the steps described herein may be performed sequentially or simultaneously with respect to multiple test images. A test image set may include multiple test images generated at the same in-die location within one or more dies on the specimen or at different in-die locations within one or more dies on the specimen; this means that the test images do not have to include the same patterned features as each other. The test images may also include any of the test images described above that are generated using one or more modes of the inspection subsystem.
[0069] In one such embodiment, one or more computer subsystems are configured to generate a mask based on the design of the specimen. In this way, the design information can be used to generate the RDL line mask according to the embodiments described herein. For example, in the Kronos™ 1080 wafer level packaging inspection system commercially available from KLA, the FlexPoint™ algorithm architecture (which is a feature of adaptive inspection with the introduction of the Kronos™ 1080 product from the Altair™ production line) performs median intensity based segmentation to separate the RDL lines and the spaces between them. FlexPoint™ significantly improves the sensitivity-nuance tradeoff of the Altair™ inspection tool, especially in the space areas where the image intensity is normally relatively quiet, making RDL line-to-line bridging easier to detect. However, the inventors have also found that the LER of the RDLs can make it difficult to generate a clean line mask. Some of the 89xx tools commercially available from KLA will incorporate features that allow users to generate FlexPoint™ masks (for RDL lines / spaces) based on design information (e.g., GDS files) that can also be provided by the user. Generating such masks is advantageously much faster, cleaner, and less susceptible to RDL LER than mask generation based on sample images. With design-based FlexPoint™ masks, RDL line opens as well as RDL space bridging can be detected by the embodiments described herein.FlexPoint™ mask generation using design information may be performed as described in commonly assigned U.S. patent application Ser. No. 16 / 901,102, by Zhang et al., filed Jun. 15, 2020, which is incorporated by reference and is considered fully set forth herein, and this feature may be available as a DesignWise® feature on the 89xx tools commercially available from KLA. The embodiments described herein may be further configured as described in that patent application.
[0070] With regard to handling line open defect detection in RDL lines, the embodiments described herein benefit from the relatively quick and easy generation of design-based RDL line space masks. A direct advantage of using design-based FlexPoint™ mask generation is that the embodiments can perform patch-to-mask alignment (PMA), a FlexPoint™ intra-feature component for accurately placing the mask on the image, without relying on pure total image intensity. Thus, ambiguity or confusion from the RDL to previous layer shift (as shown in FIG. 3) will not affect the alignment process.
[0071] In one such embodiment, the mask separates pixels within portions of the patterned feature having different orientations into separate regions, and the determining and detecting steps described herein are performed separately for the separate regions. For example, RDL line portions having different orientations can be easily assigned separate region mask IDs, and thus can be subject to separate inspection algorithms and associated parameter sets. The design-based RDL line / space mask can be overlaid on a median reference die (MRD) image, such as one generated according to any suitable manner known in the art. Different colors or other suitable indicia can be used to represent the separate region IDs. Pixels lacking a colored mask or other indicia can be designated as non-inspection regions, i.e., regions of no interest or areas of no interest.
[0072] As shown in FIG. 6, the design information can easily separate the continuous RDL lines into separate regions with different orientations, such as horizontal, vertical, 45 degree diagonal, and 135 degree diagonal, in addition to the spaces between the RDL lines. Specifically, FIG. 6 shows examples of separate patterned features 600, 602, 604, and 606 with the portions of the patterned features with different orientations separated into separate regions. Patterned feature 600 has horizontal portions 608, 612, and 616, a 45 degree diagonal portion 610, and a 135 degree diagonal portion 614. Patterned feature 602 has horizontal portions 618, 622, and 626, a 45 degree diagonal portion 620, and a 135 degree diagonal portion 624. Patterned feature 604 has vertically oriented portions 628, 632, and 636, a 135 degree diagonal portion 630, and a 45 degree diagonal portion 634. Patterned feature 606 has vertically oriented portions 638, 642, and 646, a 135 degree diagonal portion 640, and a 45 degree diagonal portion 644.
[0073] Portions of a patterned feature(s) having the same orientation can be sorted into the same region, and portions of a patterned feature(s) having different orientations can be sorted into different regions. According to one such example, pixels in portions 608, 612, and 616 of patterned feature 600 can be sorted into a first region, pixels in portion 610 of patterned feature 600 can be sorted into a second region, and pixels in portion 614 of patterned feature 600 can be sorted into a third region. Pixels in portions of the other patterned features shown in FIG. 6 can be sorted in the same manner, with regions being generated separately for each patterned feature.
[0074] Pixels in portions of a plurality of patterned features having the same orientation may be sorted into the same region. For example, pixels in portions 608, 612, and 616 of patterned feature 600 and portions 618, 622, and 626 of patterned feature 602 may be sorted into a first region. Pixels in portion 610 of patterned feature 600, portion 620 of patterned feature 602, portion 634 of patterned feature 604, and portion 644 of patterned feature 606 may be sorted into a second region. Pixels in portion 614 of patterned feature 600, portion 624 of patterned feature 602, portion 630 of patterned feature 604, and portion 640 of patterned feature 606 may be sorted into a third region. Pixels within portions 628, 632, and 636 of patterned feature 604 and within portions 638, 642, and 646 of patterned feature 606 can be sorted into a fourth region.
[0075] Different Region IDs can be defined for different line segments throughout the RDL line and can be represented by different fill patterns as shown in FIG. 6 or in other ways, such as different mask colors. The masks described herein can be easily generated and used in one or more computer subsystems with the DesignWise® features using appropriate RDL extraction rule scripts. The setup of the masks can be performed during recipe setup.
[0076] For at least one pixel in the test image and located in the patterned feature, a difference between a characteristic of the at least one pixel and a characteristic of other pixels in the test image and located in a predetermined window of the at least one pixel is determined during defect candidate detection, and a defect candidate is detected at the at least one pixel based on the determined difference. In these steps, structural continuity / discontinuity in the RDL line / space and noise statistics specific to the RDL line / space are accounted for and utilized to detect defects, such as open line defects. The goal of these steps can be to detect open line defects with the aid of the RDL line mask as described above. As described above, the difference intensity provided by die-to-die or other test-to-reference comparisons will introduce a lot of noise due to the roughness of the RDL lines. However, if one looks at the test image itself, it is clear that the open line defect signal is there. For example, the test image in FIG. 7 clearly shows an open line defect. Specifically, the test image 700 shows a relatively small open line defect 702. Test image 704 shows a faded line defect 706 and a relatively large missing line defect 708, both on a diagonal segment of the line and on another sideways segment of the line, that are not visible in reference image 710. Line roughness is also very apparent in these images.
[0077] In the embodiments described herein, the structural continuity / discontinuity property is used to detect open line defects. For RDL line structures, there are two observations that are relevant to the steps described herein. The first observation is for pixels in the open line defect area that "belong" to spaces, which usually have a lower and / or uniform intensity than the lines in BF mode. The second observation is for pixels near the edges of the RDL lines, whose intensity usually has a dip, i.e., lower than the intensity of the pixels in the line, but higher than the intensity of the pixels in the space.
[0078] As mentioned above, in some embodiments, the mask separates portions of the patterned features having different orientations into different regions, and in such embodiments, the determining and detecting steps detailed herein are performed separately for the different regions. For example, by utilizing the convenience of a design-based FlexPoint™ mask (based on native inspection pixels), the embodiments described herein can easily identify whether a pixel in a frame belongs to a horizontal RDL line, a vertical RDL line, a 135 degree RDL line, a 45 degree RDL line, a space between RDL lines, a space between RDL lines with different orientations, a don't care area, etc., with an accuracy of ±1 pixel. The steps described herein can then be performed separately for the different regions of lines and spaces with different orientations, potentially with one or more different parameters, such as different predefined window sizes, different thresholds, different sensitivities, etc. For example, vertical regions of the lines can be inspected with a first predefined window size and shape and a first threshold or sensitivity, and horizontal regions of the lines can be inspected with a second predefined window size and shape and a second threshold or sensitivity. Other parameters of the steps described herein can be varied as well. Parameters can be varied for other patterned structures and orientations described herein.
[0079] In one such embodiment, one or more computer subsystems are configured to independently determine one or more characteristics of a predefined window used to perform separate determination steps for different regions. For example, a predefined window size used for spatial discontinuity discrimination can be defined in an inspection recipe for different RDL line segments. Each type of line segment can have its own specific window size definition. These window size definitions can be determined during the recipe setup phase.
[0080] 11 shows different predefined window size definitions for various RDL line segments. The predefined windows are indicated in the images by the white outlined areas. For example, image 1100 shows the predefined window size definition for a horizontal line segment. Image 1102 shows the predefined window size definition for a vertical line segment. Image 1104 shows the predefined window size definition for a 135 degree diagonal RDL line segment, and image 1106 shows the predefined window size definition for a 45 degree diagonal RDL line segment.
[0081] In one embodiment, in determining the difference, a range of a characteristic of the at least one pixel and other pixels in a predefined window, i.e., a maximum-minimum difference of the characteristic of the at least one pixel and other pixels in the predefined window, is determined, and a threshold is applied to the range in detecting defect candidates. Two embodiments of a method for single die-based inspection are described herein. The method uses a spatial continuity metric based on the spatial range window (i.e., the predefined window). Given the two observations above, for each pixel on the RDL line, a spatial window centered on a pixel on the RDL line segment can be devised and the range (maximum-minimum intensity pixel difference) can be calculated.
[0082] FIG. 8 shows how the technique successfully detects defects on RDL lines. Specifically, FIG. 8 shows an example 800 of using range as an indicator to detect structural discontinuities due to open line defects in horizontal RDL lines, and an example 810 of using range as an indicator to detect structural discontinuities due to open line defects in diagonal RDL lines. The image in example 800 is of a line 802 and a space 804 and two predefined windows 806 and 808 that overlie different portions of one of the lines. The predefined windows 806 and 808 are two instances of the exact same predefined window, centered on different pixels in one of the lines as shown. The image in example 810 is of a line 812 and a space 814 and two predefined windows 816 and 818 that overlie one of the lines. As above, predetermined windows 816 and 818 are just two instances of the same predetermined window, centered on different pixels within one of the lines as shown. The range values determined for the pixels at the centers of predetermined windows 806 and 816 will be relatively low because the pixels within those predetermined windows are both in relatively low intensity, quiet space. In contrast, the range values determined for predetermined windows 808 and 818 will be relatively high. For example, based on the inspection subsystem light level settings for the inspection image, the ranges may be the maximum of a bright line structure and a dim blank background.
[0083] As shown in FIG. 8, the predetermined window allows the embodiments described herein to detect RDL discontinuities with relatively low information processing costs. Of course, the sensitivity with which such RDL line open defects are detected depends on the size of the window. If the window size is too large, some rather small line opens may be missed. If it is too small, nuisance may sneak in. However, this detection method itself does not necessarily aim to achieve a high enough capture rate with zero or very low nuisance rate. Rather, this detection technique can be used as a "cheap and fair" RDL line open detection method. By performing other steps described herein, the detected defect candidates can be classified into nuisance and defects.
[0084] In some such embodiments, one or more computer subsystems are also configured to determine whether a defect candidate is a defect based on one or more characteristics of the defect candidate. For example, such detection methods may be combined with some morphological post-processing to filter out relatively small line open nuisances and / or assign a final defect classification through iDO™, a defect classification module available on some commercially available tools from KLA, and / or with deep learning (DL) based binning to filter out excess nuisances and retain real line open DOIs. Such morphological post-processing and defect classification and binning may be performed as described herein.
[0085] In one such embodiment, defect candidates include full opens in patterned features. For example, the first technique described above is relatively simple and fast, and is less susceptible to nuisances than other methods and systems currently used for RDL line inspection, but its performance is good for full line opens but not necessarily for partial line opens.
[0086] The second approach is based on the difference of pixel intensity relative to the average intensity in the patterned feature, potentially in the same corresponding RDL line segment in each die frame (also referred to herein simply as "single die difference vs. average" or "SDDiftMean"). For example, in another embodiment, in determining the difference, the difference is determined between a characteristic of at least one pixel and an average of that characteristic for the at least one pixel and other pixels within a given window. The average or the difference between the pixel characteristic and the average may be determined in any suitable manner.
[0087] In such an embodiment, the entire patterned feature in the test image is enclosed within the predetermined window. For example, unlike the first approach, the entire patterned feature in the test image can be located within the predetermined window within which the difference is determined. In particular, the average of a property, e.g., intensity, of all pixels in the identified patterned feature can be determined, and then a difference between the property of any one pixel in the patterned feature and the average can be determined and used for the detection step described herein. Such an embodiment can be preferred when the entire patterned feature in the test image has a single orientation, such as the lines shown in example 800 of FIG. 8.
[0088] In some such embodiments, the window encompasses the entire patterned feature in the test image and one or more other patterned features in the test image, where the entire patterned feature has the same type and orientation as the patterned feature. For example, if the test image has multiple patterned features of the same type and orientation, such as the lines shown in example 800 of FIG. 8, the average of the properties of all the pixels in all the patterned features in the test image can be determined, and the difference between the property of any one pixel in the patterned features and the average can be determined and used for the detection steps described herein. Determining the average over multiple (or all) patterned features of the same type and orientation in the test image can provide advantages related to the detection steps described herein, such as making the average more stable and less sensitive to noise in the patterned features. Additionally, by determining the average based solely on the patterned feature or features in the test image (and not on any other images generated for the sample, such as other test or reference images), the effect of non-defective within-sample or across-sample variations on the average can be advantageously reduced.
[0089] The predetermined window may be similarly varied for different portions of the patterned features that are oriented differently and are separated into different pixel regions. For example, one average may be determined based on all pixels in horizontal segments of RDL lines in the test image, another average may be determined based on all pixels in vertical segments of RDL lines in the test image, and so on. The average used to determine the difference for any pixel may then be selected based on the segment of the RDL line in which the pixel is located. In other words, if a pixel in the test image is located in a vertical segment of an RDL line, then the average determined based on all (or at least some of) the vertical segments of the RDL lines in the test image may be selected and used to determine the difference for that pixel.
[0090] In a further such embodiment, the defect candidates are detected by determining the mean and standard deviation of the characteristics of the at least one pixel and other pixels in the window, dividing the difference by the standard deviation to determine a signal strength, and applying a threshold to the signal strength. The idea is relatively simple: for each group of line segments (e.g., landscape, portrait, 135° diagonal and 45° diagonal RDL segments) in each frame, the mean intensity and the standard deviation of the intensity are calculated. The test is performed by applying the following equation to the corresponding group of line segments: Signal intensity = (pixel intensity - average intensity) / standard deviation As described above, the relative intensity of each pixel is simply calculated.
[0091] If the signal strength is greater than a predefined threshold (polarity specific, with different thresholds for dark and light defects), the pixel is abnormal. The method is conceptually simple, but can be relatively computationally expensive. For example, the method could include two frame pixel visit passes: a first pass to collect noise statistics (e.g., mean and standard deviation) in each region or group, and a second pass for relative intensity calculation and anomaly detection. The mean and standard deviation could also be determined in other ways according to any suitable procedure known in the art.
[0092] In one such embodiment, the one or more computer subsystems are further configured to determine the mean and standard deviation based on the generated median reference image for the test images. For example, corresponding parameters of the range of mean intensity in the space segments may be automatically calculated using the RDL space segment mask and the underlying MRD image. The MRD image may be generated according to any suitable manner known in the art. This step may be performed during recipe setup.
[0093] An optional step that may be performed in light of throughput optimization may be to generate a mask based on the RDL line segment design polygons (i.e., align the design of those RDL polygons to the RDL line structures in the MRD image) and then calculate the mean and standard deviation of pixel intensities for the horizontal, vertical, 45 degree diagonal, and 135 degree diagonal RDL line segments, and perform similar calculations for those pixels in the space segments between the RDL line segments. These steps may be performed during recipe setup.
[0094] In some such embodiments, one or more computer subsystems are configured to determine the mean and standard deviation from the test images, for example, from a single test image, providing the same advantages as described above. In this manner, true single die image (SDI) defect detection can be performed according to the embodiments described herein.
[0095] In certain embodiments, a patterned feature may be identified by applying a mask to a test image to separate pixels within the patterned feature in the test image from all other pixels in the test image, and the results of application of the mask may be pre-processed to identify one or more pixels beyond one or more edges of the patterned feature and to exclude the identified one or more pixels from the set of pixels in the patterned feature. In certain embodiments, a patterned feature may be identified by applying a mask to a test image to separate pixels within the patterned feature in the test image from all other pixels in the test image, and the results of application of the mask may be post-processed to identify one or more pixels beyond one or more edges of the patterned feature and to exclude the identified one or more pixels from the set of pixels in the patterned feature. According to these embodiments, any marginality in the steps can be addressed by applying a mask to the test image as detailed above and applying either pre-processing or post-processing (or both) to the result of applying the mask. For example, due to system limitations, most systems will have a mask placement accuracy of approximately ±1.5 pixels. As can be seen in FIG. 9, detailed below, at the zoom level, extra space pixels may appear in the RDL line segments. These extra space pixels will result in excessive nuisance detections. Therefore, a more robust pre-processing or post-processing would be advantageous to mitigate such nuisance detections. There are two ways that these extra space pixels can be handled: (1) pre-processing the RDL line segment mask to remove the space pixels in the line segments, or (2) post-processing to remove the excess space pixels with lower intensity. Either approach involves computations that are more expensive than the first approach described in this application. These pre- and post-processing steps may be otherwise performed according to any suitable manner known in the art.
[0096] Specifically, FIG. 9 shows two test images 900 and 908 generated for a sample. There are two RDL lines 902 and 904 in the test image 900, and an open line defect 906 in the RDL line 904. There are also two RDL lines 910 and 912 in the test image 908, but neither of them has an open line defect. However, by determining the relative signal strengths of these images as described herein, both the open line DOI and nuance at the RDL line edge 914 in the test image 900 are detected. In other words, the RDL line open detection by the SDDiftMean method not only effectively detects open line defects and partial open line defects, but also detects excessive nuance for pixels at the RDL line edge. Therefore, the above pre- and post-processing steps can be an important nuance reduction technique for the embodiments described herein.
[0097] In further embodiments, defect candidates include full or partial opens in patterned features. For example, as shown in FIG. 10, the SDDiftMean method can detect both full line open defects and relatively large partial line opens. In particular, the SDDiftMean method can detect both full line open defects, such as the line open defect 906 shown in the test image 900 of FIG. 9, and partial line open defects, such as those shown in the circled portion 1002 of the test image 1000 shown in FIG. 10. In this manner, partial line opens can be detected by determining relative signal strengths using the SDDiftMean method.
[0098] In certain embodiments, one or more computer subsystems are configured to determine whether a defect candidate is a defect by performing a morphological operation on the results of the detection step. For example, in the embodiments described herein, a morphological operation can be used to eliminate a relatively small number of candidate pixel cliques. In both of the single-die-based RDL line open candidate detection methods described herein, inevitably, neither method performs DOI-limited capture. Both methods provide a better DOI capture rate vs. nuisance rate relationship than traditional die-to-die difference-based inspection methods, which typically overwhelm the inspection with a large number of nuisance detections, making it impossible to detect real DOI. The novel single-die-based method, whether spatial discontinuity-based or relative signal intensity-based, provides a fairly good capture rate with a fairly low nuisance rate. However, further reduction of nuisance is desirable for production-useful wafer inspection. One observation about such nuisances is that they are usually sporadic and diffuse, and have relatively small pixel count cliques. Therefore, by performing morphological operations, it is possible to filter out such relatively small sized nuisances. Such morphological operation based nuisance filtering can reduce the nuisances by a significant amount at the post-processing stage of the inspection. The analysis of defect candidates in relation to their morphological characteristics and the use of such morphological characteristics to separate potential nuisances from potential DOIs can also be performed in any suitable manner.
[0099] In an additional embodiment, the one or more computer subsystems are configured to determine whether a defect candidate is a defect by binning the defect candidate based on the test image. In an embodiment, the one or more computer subsystems are configured to determine whether a defect candidate is a defect by applying a deep learning (DL) based classifier to the test image. For example, in the embodiments, the iDO™ based binning or DL based binning can be used to further filter out nuissance. According to one such example, even if some structure with a relatively large spatial dimension exists, it can be identified as a nuissance by its appearance, such as RDL line edge roughness vs. partial line open relationship or line bending. The iDO™ in-line binning can be used to filter out potential nuissance at each parallel processing unit level. If the DOI and the nuissance can be visually distinguished, the DL binning using the defect candidate patch may provide a better nuissance reduction result than the iDO™ classifier. DL-based binning may be performed as described in U.S. Patent Application Publication No. 2010 / 0133633, issued to He et al. on March 31, 2020, which is incorporated by reference herein in its entirety. DL-based nuisance filtering may be performed as described in U.S. Patent Application No. 16 / 663,283, filed on October 24, 2019 by Huang et al., which is incorporated by reference herein in its entirety. The embodiments described herein may be further configured as described in those patents and patent applications.
[0100] In some embodiments, the determination and detection steps are performed without a reference image. In other words, the detection of defect candidates can be by single image defect detection, instead of two die images as in die-to-die inspection, or test and reference images as in die-to-reference inspection. For example, metal grains are major nuisances in back-end wafer inspection. Metal grains are randomly distributed across a specimen, e.g., a wafer (e.g., across all dies on a wafer), so the metal grains cannot be cancelled out by die-to-die subtraction. Specifically, the difference image generated by die-to-die subtraction for a metal layer contains grain noise that cannot be cancelled out by subtraction and can be detected as an event. In fact, die-to-die subtraction based techniques may amplify noise from metal grains, which may result in failure to detect DOI. Therefore, currently used die-to-die defect detection and binning techniques do not work well for such specimens. In that regard, the embodiments described herein can be configured to detect and isolate defects in any die without using any reference die, whether it is in an adjacent die or a standard reference die. According to the embodiments described herein, a single die defect patch image can be used to detect and separate DOI candidates from true DOI metal grain nuisances through DL or as otherwise described herein.
[0101] FIG. 12 illustrates one embodiment of steps that may be performed to detect defects as described herein. These steps may be performed in two separate phases: a setup phase 1200 and a run-time phase 1202. In the setup phase, one or more computer subsystems may perform a rendering and placement step 1206 using design data 1204, e.g., a GDS file, as input in any manner known and suitable in the art. In the setup phase, one or more computer subsystems may also perform a fiducial generation step 1208, e.g., generating MRDs or other types of fiducials for specimen inspection. Results of the fiducial generation step and the rendering and placement step may be used by the computer subsystem(s) to perform a segment mask generation step 1210, e.g., to generate one of the segment masks described herein. The computer subsystem(s) may also perform an anchor site search step 1212 using the output of the fiducial generation step 1208 to select anchor sites according to the previously generated fiducials. These anchor sites can then be used in the inspection process to align the images produced by the inspection subsystem to a common reference.
[0102] When the inspection is to be performed, the recipe parameters may be sent to the individual CPU cores performing the inspection. Additional recipe parameters for different RDL line / space segments may be sent to the inspection computer during this run-time phase along with the mask used for the inspection. During the run-time phase, one or more computer subsystems may use the output of the anchor site search step to perform a patch to mask alignment step 1214 to align the images generated by the inspection subsystem to the mask so that patterned features in the images can be identified by a place mask on scanned image step 1216 as described herein. Thereafter, as shown in step 1218, the one or more computer subsystems may perform segment-based die-to-die (DtD) differential intensity-based inspection or single die (SD) inspection as described herein. For example, in the runtime phase, the segment-based DtD / SD target intensity-based inspection step can perform traditional die-to-die inspection using a segment mask, or it can perform a novel segment mask-guided single-die inspection using one of two techniques detailed in this application: (1) spatial windowing based on structural continuity, and / or (2) the difference between pixel intensity and its average intensity within the same segment, normalized, for example, by the standard deviation of the segment.
[0103] FIG. 13 illustrates one embodiment of steps that may be performed by embodiments described herein for "spatial continuity" based single die RDL line open detection. As shown in FIG. 13, these steps are performed in two phases: a detection process phase 1300 and a post-processing phase 1302. One or more computer subsystems perform a pre-processing step 1304, such as any of the pre-processing steps described herein. After the pre-processing step is completed, the detection process can begin with a determination of whether a pixel loop is complete, as shown in step 1306. Performing a "pixel loop" means going through (visiting) each individual pixel in a job frame, patch image, test image, etc. If the pixel loop is complete, the one or more computer subsystems proceed to a post-processing phase.
[0104] If it is determined that the pixel loop has not been completed, the one or more computer subsystems determine whether the pixel is on an RDL line n, as shown in step 1308. If the pixel is not on an RDL line, the one or more computer subsystems repeat step 1306. If the pixel is determined to be on an RDL line, the computer subsystem(s) determine whether the pixel's intensity is within a range (n), as shown in step 1310. An in-range pixel intensity can be defined as an intensity that is greater than the minimum value of the range setting and less than the maximum value of the range setting. If the pixel is not within range (n), the one or more computer subsystems return to step 1306. If the pixel intensity is determined to be within range (n), the one or more computer subsystems determine whether the single-die pixel window range is less than a threshold, as shown in step 1312. A single-die pixel window range that is less than a threshold means that the neighboring pixel maximum intensity minus the neighboring pixel minimum intensity is less than a threshold setting. If the single-die pixel window range is not less than the threshold, the one or more computer subsystems return to step 1306. If the single-die pixel window range is determined to be less than the threshold, the pixel is marked as defective in the one or more computer subsystems, as shown in step 1314. After step 1314, the one or more computer subsystems return to step 1306.
[0105] If the one or more computer subsystems determine in step 1306 that they have completed the pixel loop, then the one or more computer subsystems generate defect blobs in step 1316 of the post-processing phase 1302. The one or more computer subsystems then determine whether the defect blob loop is complete in step 1318. If it is determined that the defect blob loop is not complete, then the one or more computer subsystems determine whether the number of defective pixels in the blob is greater than a threshold, as shown in step 1320. If the number of defective pixels in the blob is not greater than the threshold, then the one or more computer subsystems return to step 1318. If the number of defective pixels in the blob is greater than the threshold, then the one or more computer subsystems generate defects, as shown in step 1322. After the one or more computer subsystems generate the defects, the one or more computer subsystems return to step 1318. When the one or more computer subsystems determine that the defect blob loop is complete, the one or more computer subsystems perform a post-processing step 1324, such as any of the post-processing steps described herein, and then the one or more computer subsystems perform a result return step 1326, in which the results of any of the steps performed by the computer subsystem(s) are output as an inspection results file or other suitable file.
[0106] FIG. 14 illustrates one embodiment of steps that may be performed by embodiments described herein for "normalized relative intensity vs. average" based single die RDL line open detection. As shown in FIG. 14, these steps may be performed in two phases: a noise collection process phase 1400 and a detection process phase 1402. Prior to the start of the noise collection process, one or more computer subsystems perform pre-processing 1404, such as any of the pre-processing steps described herein. After the pre-processing step(s) are completed, the one or more computer subsystems begin the noise collection process by determining whether a first pixel loop has been completed, as shown in step 1406. If a first pixel loop has not been completed, the one or more computer subsystems determine whether the pixel is on an RDL line n or space m, as shown in step 1408. If it is determined that the pixel is not on an RDL line n or space m, the one or more computer subsystems return to step 1406. If the pixel is determined to be on an RDL line n or space m, the one or more computer subsystems collect the mean and standard deviation (std) of the intensities for the segments of the line (n) and / or space (m), as shown in step 1410. After the one or more computer subsystems perform step 1410, the one or more computer subsystems return to step 1406.
[0107] If the one or more computer subsystems determine in step 1406 that the first pixel loop has been completed, the one or more computer subsystems begin the detection process 1402 by determining whether a second pixel loop has been completed in step 1412. If the one or more computer subsystems determine that the second pixel loop has not been completed, the one or more computer subsystems determine whether the pixel is on an RDL line n or space m as shown in step 1414. If the pixel is not on an RDL line n or space m, the one or more computer subsystems return to step 1412. If the pixel is on an RDL line n or space m, the one or more computer subsystems calculate the signal intensity of the pixel as (pixel intensity-mean[n]) / std[n] as shown in step 1416. After the one or more computer subsystems calculate the signal strength of the pixel, the one or more computer subsystems determine whether the signal strength of the pixel is greater than a threshold, as shown in step 1418. If the signal strength of the pixel is not greater than the threshold, the one or more computer subsystems return to step 1412. If the pixel signal strength threshold is greater, the computer subsystem(s) mark the pixel as defective, as shown in step 1420.
[0108] When the one or more computer subsystems determine in step 1412 that the second pixel loop is complete, the one or more computer subsystems begin a post-processing phase 1422. The post-processing phase performed in this embodiment may be the same as the post-processing phase shown in FIG.
[0109] The steps of the embodiments described and illustrated in Figures 12-14 above may be performed by any of the system embodiments described herein. In addition, the steps of the embodiments described and illustrated in the figures may be performed as detailed herein and may be combined with any other steps described herein.
[0110] There are some differences and similarities between the first and second approaches described herein that can be used to select the best of them for any particular application. For example, the first pixel range-based approach requires only one detection pass visiting each pixel in the frame, making it computationally less expensive than the second approach, which may include two passes visiting the frame pixels, one for collecting noise statistics (mean and standard deviation of the intensity in each segment) and one for inspection. In addition, the first pixel range-based approach detects full line open defects at a relatively low cost, while the second average-based approach can detect RDL partial line open defects in addition to full line open defects. To reduce the cost of the second average-based approach, embodiments may also calculate the mean and standard deviation for MRD or other suitable criteria in the setup phase as described herein. In this way, the embodiments may avoid performing two passes per pixel, but such criteria may be less tolerant of wafer-scale local process variations on different dies. The post-processing phase of both approaches may be the same. The post-processing step(s) in both approaches may include any of the post-processing step(s) described herein, such as iDO™-based nuance reduction and DL-binning-based nuance reduction.
[0111] The described embodiments have several advantages over other systems and methods currently used to detect defects on RDL lines and other noisy patterned features. For example, the described embodiments allow single-die intensity-based features to be utilized for (partial) line open defect detection, thereby overcoming the problem of excessive noise in die-to-die intensity difference due to the relatively rough metal surface of RDL lines. In addition, the use of design-based masks allows the embodiments to avoid inconsistent shifts between RDL lines and the sub-die structures below them. The described embodiments also help improve the sensitivity of RDL line open capture by existing back-end inspection tools for aWLP applications.
[0112] The described embodiments differ in several important ways from other systems and methods currently used to detect defects on RDL lines and other noisy patterned features, thereby achieving the above-mentioned advantages. For example, the described embodiments allow single-die intensity-based inspection to avoid the excessive noise of die-to-die difference-based inspection. In addition, the described embodiments provide a new spatial continuity-based metric for RDL line open detection. The described embodiments also provide a new relative intensity-based metric for RDL line open detection. Furthermore, the described embodiments can be configured to use design polygon information, thus making it simple and easy to separate different segments of RDL lines (e.g., horizontal, vertical, 135° diagonal, 45° diagonal) into different regions, which is especially important for situations where traditional intensity-based mask generation is not possible. According to the embodiments described herein, DL-based binning can also be used to filter out nuances that evade intensity-based detection techniques, which can aid in better nuance-DOI discrimination.
[0113] The one or more computer subsystems may be configured to generate results related to the detected defect candidates and / or the detected defects. The results related to the defect candidates and / or defects may include any of the results described herein, such as information such as the location of bounding boxes for the detected defects, detection scores, information about defect classifications such as class labels or IDs, or any suitable information of that nature known in the art. The results related to the defect candidates and / or defects may be generated by the computer subsystem(s) in any suitable manner. The results related to the defect candidates and / or defects may be in any suitable form or format, such as a standard file type. The results may be generated and stored by the computer subsystem(s) and / or may be used by the computer subsystem(s) and / or other systems or methods to perform one or more functions on the sample or other samples of the same type. Such functions include, but are not limited to, those that provide alerts about processes, e.g., manufacturing processes or steps, performed in a feedback manner on the sample, those that provide alerts about processes, e.g., manufacturing processes or steps, performed in a feedforward manner on the sample, and the like.
[0114] Each of the embodiments of each of the systems described above may be combined together into a single, unitary embodiment.
[0115] Another embodiment includes a computer-implemented method for detecting defects on a specimen. The method includes identifying a patterned feature in a test image of a set of images generated by an inspection subsystem of the specimen. The method also includes determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and a characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel. The method also includes detecting a defect candidate at the at least one pixel based on the determined difference. The identifying, determining, and detecting steps are performed by one or more computer subsystems coupled to the inspection subsystem.
[0116] Each step of the method may be performed as detailed herein. The method may also include any other step(s) performable by the system, computer subsystem(s) and / or inspection subsystem described herein. The inspection subsystem and one or more computer subsystems may be configured according to any of the embodiments described herein, such as inspection subsystem 10 and computer subsystem(s) 36 and 102, respectively. In addition, the above-described method may be performed by any of the system embodiments described herein.
[0117] An additional embodiment relates to a non-transitory computer readable medium having stored thereon program instructions that may be executed on one or more computer systems to perform a computer-implemented method for detecting defects on a specimen. One such embodiment is illustrated in FIG. 15. Specifically, as illustrated in FIG. 15, non-transitory computer readable medium 1500 contains program instructions 1502 that are executable on computer system(s) 1504. The computer-implemented method may include any step(s) of any method(s) described herein.
[0118] Program instructions 1502 implementing methods, such as those described herein, can be stored on a computer readable medium 1500. The computer readable medium can be a storage medium, such as a magnetic or optical disk, magnetic tape, or any other suitable non-transitory computer readable medium known in the art.
[0119] The program instructions may be implemented in any of a number of ways, including procedure-based techniques, component-based techniques, and / or object-oriented techniques, among others. For example, the program instructions may be implemented using ActiveX controls, C++ objects, JavaBeans, Microsoft Foundation Classes ("MFC"), Streaming SIMD Extensions (SSE), or any other technology or methodology, as desired.
[0120] The computer system(s) 1504 may be configured according to any of the embodiments described herein.
[0121] Further modifications and alternative embodiments of various aspects of the invention will be apparent to those skilled in the art in view of this specification. For example, a method and system for detecting defects on a specimen is provided. This specification is therefore to be construed as being illustrative only and is intended to teach those skilled in the art the general practice of the present invention. As will be understood, the forms of the invention shown and described herein are to be taken as preferred embodiments. Elements and materials may be substituted for those depicted and described herein, parts and processes may be reversed, and certain features of the invention may be utilized independently, all of which would be apparent to a person skilled in the art in light of the description of the invention. Modifications may be made to the elements described herein without departing from the spirit and scope of the invention as set forth in the following claims.
Claims
1. 1. A system configured to detect defects on a specimen, comprising: an inspection subsystem configured to generate images of the specimen; one or more computer subsystems configured to detect defect candidates on the specimen; and detecting the defect candidates comprises: identifying a patterned feature in a test image included in the set of images generated for the sample; determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and said characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel; and detecting a defect candidate at the at least one pixel based on the determined difference; Including, The system, wherein the identifying includes separating pixels in the test image that are within the patterned feature from all other pixels in the test image by applying a mask to the test image.
2. The system of claim 1 , wherein the patterned features include lines on a redistribution layer.
3. 10. The system of claim 1, wherein the patterned features are formed of a metal.
4. 10. The system of claim 1, wherein said determining and said detecting are performed without a reference image. A system that is.
5. 10. The system of claim 1 , wherein the one or more computer subsystems are further configured to generate the mask based on a design of the specimen.
6. 2. The system of claim 1, wherein the mask classifies portions of the patterned features having different orientations into separate regions, and the determining and detecting are performed separately for the separate regions.
7. 7. The system of claim 6, wherein the one or more computer subsystems are further configured to independently determine one or more characteristics of the predetermined window for each of the determining performed separately for the separate regions.
8. A system configured to detect defects on a specimen, comprising: an inspection subsystem configured to generate images of the specimen; one or more computer subsystems configured to detect defect candidates on the specimen; and detecting the defect candidates comprises: identifying a patterned feature in a test image included in the set of images generated for the sample; determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and said characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel; and detecting a defect candidate at the at least one pixel based on the determined difference; Including, The system, wherein determining the difference includes determining a range for the characteristic of the at least one pixel and the other pixels within the specified window, the range being the difference between a maximum value and a minimum value of the characteristic of the at least one pixel and the other pixels within the specified window, and detecting the defect candidates includes applying a threshold to the range.
9. 10. The system of claim 8, wherein the defect candidates include a complete open in the patterned feature.
10. 9. The system of claim 8, wherein the one or more computer subsystems are further configured to determine whether the defect candidate is a defect based on one or more characteristics of the defect candidate.
11. A system configured to detect defects on a specimen, comprising: an inspection subsystem configured to generate images of the specimen; one or more computer subsystems configured to detect defect candidates on the specimen; and detecting the defect candidates comprises: identifying a patterned feature in a test image included in the set of images generated for the sample; determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and said characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel; and detecting a defect candidate at the at least one pixel based on the determined difference; Including, determining the difference includes determining a difference between the characteristic of the at least one pixel and an average of the characteristics of the at least one pixel and the other pixels within the predetermined window; The system wherein the predetermined window encompasses the entire patterned feature in the test image.
12. 12. The system of claim 11, wherein the predetermined window encompasses the entirety of the patterned feature in the test image and the entirety of one or more other patterned features in the test image, the entirety of the other one or more patterned features having the same type and orientation as the patterned feature.
13. 12. The system of claim 11, wherein detecting the defect candidates includes determining a standard deviation based on the average and the characteristics of the at least one pixel and the other pixels within the predetermined window, determining a signal strength by dividing the difference by the standard deviation, and applying a threshold to the signal strength.
14. 14. The system of claim 13, wherein the one or more computer subsystems are further configured to determine the mean and the standard deviation based on a median reference image generated for the test image.
15. 14. The system of claim 13, wherein the one or more computer subsystems are further configured to determine the mean and the standard deviation based on the test images.
16. 12. The system of claim 11 , wherein the identifying includes: applying a mask to the test image to separate pixels in the test image that are within the patterned feature from all other pixels in the test image; and pre-processing a result of applying the mask to identify one or more pixels that are beyond one or more edges of the patterned feature and excluding the identified one or more pixels from pixels within the patterned feature.
17. 12. The system of claim 11 , wherein the identifying includes applying a mask to the test image to separate pixels in the test image that are within the patterned feature from all other pixels in the test image, and post-processing a result of the application of the mask to identify one or more pixels that are beyond one or more edges of the patterned feature and excluding the identified one or more pixels from pixels within the patterned feature.
18. 12. The system of claim 11, wherein the defect candidates include full or partial opens in the patterned features.
19. 2. The system of claim 1, wherein the one or more computer subsystems are further configured to determine whether the defect candidate is a defect by performing a morphological operation on a result of the detection.
20. 2. The system of claim 1, wherein the one or more computer subsystems are further configured to determine whether the defect candidates are defects by binning the defect candidates based on the test images.
21. 2. The system of claim 1, wherein the one or more computer subsystems are further configured to determine whether the defect candidates are defects by applying a deep learning based classifier to the test image.
22. 10. The system of claim 1, wherein the inspection subsystem is further configured for macro inspection.
23. 10. The system of claim 1, wherein the inspection subsystem is further configured as an electron beam subsystem.
24. 10. The system of claim 1, wherein the inspection subsystem is further configured as an optical subsystem.
25. 10. The system of claim 1, wherein the specimen is a wafer.
26. 1. A non-transitory computer readable medium having stored thereon program instructions executable on one or more computer systems to perform a computer-implemented method for detecting defects on a specimen, the computer-implemented method comprising: identifying a patterned feature in a test image included in the set of images generated by the inspection subsystem of the specimen; determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and said characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel; and detecting a defect candidate at the at least one pixel based on the determined difference; Including, A non-transitory computer-readable medium, wherein the identifying includes separating pixels in the test image that are within the patterned feature from all other pixels in the test image by applying a mask to the test image.
27. 1. A computer-implemented method for detecting defects on a specimen, comprising: identifying a patterned feature in a test image included in the set of images generated by the inspection subsystem of the specimen; determining, for at least one pixel in the test image that is within the patterned feature, a difference between a characteristic of the at least one pixel and said characteristic of other pixels in the test image that are within a predetermined window of the at least one pixel; and detecting a defect candidate at the at least one pixel based on the determined difference; Including, A computer-implemented method, wherein the identifying includes separating pixels in the test image that are within the patterned feature from all other pixels in the test image by applying a mask to the test image, and the identifying, determining, and detecting are performed by one or more computer subsystems coupled to the inspection subsystem.
Citation Information
Patent Citations
Wiring pattern inspection apparatus
JP2000046745A
Pattern inspection device and pattern inspection method
JP2006098151A
Defect inspection method and apparatus thereof
JP2012063209A
Pattern inspection device and pattern inspection method
JP2017216392A
Unified neural network for defect detection and classification
US10607119B2
