Compensating target position for temperature in semiconductor applications

Abstract

Methods and systems for determining information about a specimen are provided. One system includes an output acquisition subsystem configured to generate an output for the specimen at one or more target locations on the specimen and one or more temperature sensors configured to measure one or more temperatures within the system. The system also includes a deep learning model configured to predict an error in at least one of the one or more target locations based on at least one of the one or more measured temperatures input to the deep learning model by a computer subsystem. The computer subsystem is configured to determine a corrected target location for at least one of the one or more target locations by applying the predicted error to at least one of the one or more target locations.

Description

[Technical Field]

[0001] The present invention relates generally to methods and systems for determining information about a specimen. Particular embodiments relate to compensating a target position on a specimen for temperature in a system for performing a process on the specimen. [Background technology]

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

[0003] The fabrication of semiconductor devices, such as logic and memory devices, typically involves processing a substrate, such as a semiconductor wafer, using a number of semiconductor manufacturing processes to form various features and levels of the semiconductor device. For example, lithography is a semiconductor manufacturing process that involves transferring a pattern from a reticle to a resist that is placed on the semiconductor wafer. Additional examples of semiconductor manufacturing processes include, but are not limited to, chemical mechanical polishing (CMP), etching, deposition, and ion implantation. Multiple semiconductor devices may be fabricated in an array on a single semiconductor wafer, which is then separated into individual semiconductor devices.

[0004] Inspection processes are used at various steps during the semiconductor manufacturing process to detect defects on specimens to drive higher yields and therefore higher profits in the manufacturing process. Inspection has always been an important part of manufacturing semiconductor devices. However, as semiconductor device dimensions decrease, inspection becomes even more important to the successful manufacture of acceptable semiconductor devices because smaller defects can cause device failure.

[0005] Defect review typically involves re-detecting defects detected as defects by the inspection process and generating additional information about the defects at higher resolution using either a high-magnification optical system or a scanning electron microscope (SEM). Thus, defect review is performed at individual locations on the specimen where defects were detected by inspection. The higher resolution data about the defects generated by defect review is more suitable for determining defect attributes, such as profile, roughness, and more precise size information. Defects can typically be more accurately classified into defect types based on the information determined by defect review compared to inspection.

[0006] Metrology processes are also used at various steps during semiconductor manufacturing processes to monitor and control the process. Metrology processes differ from inspection processes in that, unlike inspection processes in which defects are detected on a specimen, metrology processes are used to measure one or more properties of a specimen that cannot be determined using currently used inspection tools. For example, a metrology process may be used to measure one or more properties of a specimen during processing, such as dimensions (e.g., linewidth, thickness, etc.) of features formed on the specimen, so that the performance of the process can be determined from the one or more properties. Additionally, if one or more properties of a specimen are unacceptable (e.g., outside a predetermined range for the property(ies)), the measurements of one or more properties of the specimen may be used to modify one or more parameters of the process so that additional specimens produced by the process have acceptable property(ies).

[0007] A metrology process also differs from a defect review process in that defects detected by inspection are reviewed in the defect review, and the metrology process may be performed at locations where no defects have been detected. In other words, unlike a defect review, the locations on the specimen where the metrology process is performed may be independent of the results of the inspection process performed on the specimen. In particular, the locations where the metrology process is performed may be selected independent of the inspection results. Additionally, because the locations on the specimen where the metrology is performed may be selected independent of the inspection results, the locations where the metrology process is performed may be determined before the inspection process is performed on the specimen, unlike a defect review in which the locations on the specimen where the defect review is to be performed cannot be determined until inspection results for the specimen are generated and available.

[0008] One aspect of the above-described methods and systems that can be challenging is knowing where on the specimen a result, e.g., a measurement, a detected defect, a redetected defect, etc., was generated. For example, the above-described tools and processes are used to determine information about the structure and / or defects on the specimen. Because structures vary across the specimen (as may form functional devices on the specimen), the result of a measurement, inspection, or defect review is typically useless unless one knows exactly where on the specimen it was generated. In the example of metrology, unless the measurement is performed at a known, predetermined location on the specimen, the measurement may fail if the measurement location does not include the portion of the specimen being measured and / or if a measurement of one portion of the specimen is assigned to another portion of the specimen. In the case of inspection, unless the defect location on the specimen is determined substantially accurately, the defect location may be inaccurately determined with respect to the specimen and / or the specimen's design. In either case, an error in the location on the specimen where the result was generated can render the result useless and even adversely affect the manufacturing process if the result is used to alter the manufacturing process.

[0009] After the sample is aligned with the tool, it may be scanned or measured using known alignment target(s) or mark(s) on the sample. Based on information about the sample, such as where a pattern is formed on the sample and / or the location of the target of interest on the sample, the tool may perform a measurement or scan of the target of interest on the sample. Even if the alignment prior to the measurement or scan is successful, there are several factors that can cause the reported position in the results to be inaccurate. Therefore, many processes involving navigation across the sample involve measuring and correcting position errors on the fly. In such an example, the tool's scanning subsystem, which may include a robot and other associated hardware, may move the sample to a location, and then the tool may perform alignment at each location.

[0010] However, such location-by-location alignment methods have several drawbacks. For example, such methods are generally substantially more time consuming. Location-by-location alignment also requires that more images be recorded and image processing be performed at the time of alignment. Additionally, location-by-location alignment methods do not take into account temperature-induced position errors on the tool. [Prior art documents] [Patent documents]

[0011] [Patent Document 1] International Publication No. 2017 / 123561 [Patent Document 2] U.S. Patent Application Publication No. 2016 / 0104600 Summary of the Invention [Problem to be solved by the invention]

[0012] It would therefore be beneficial to develop a system and method for determining information about a sample that does not suffer from one or more of the above drawbacks. [Means for solving the problem]

[0013] The following description of various 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 determine information about a specimen. The system includes an output acquisition subsystem configured to generate an output of the specimen at one or more target locations on the specimen and one or more temperature sensors configured to measure one or more temperatures within the system. The system also includes a computer subsystem configured to acquire one or more measured temperatures from the one or more temperature sensors and one or more components executed by the computer subsystem. The one or more components include a deep learning (DL) model configured to predict an error at at least one of the one or more target locations based on at least one of the one or more measured temperatures input into the DL model by the computer subsystem. The computer subsystem is configured to determine a corrected target location for at least one of the one or more target locations by applying the predicted error to at least one of the one or more target locations. The system may be further configured as described herein.

[0015] Another embodiment relates to a method for determining information about a specimen. The method includes measuring one or more temperatures in a system using one or more temperature sensors. The system includes an output acquisition subsystem configured to generate specimen outputs at one or more target locations on the specimen. The method also includes predicting an error at at least one of the one or more target locations by inputting at least one of the one or more measured temperatures into a DL model included in one or more components executed by a computer system. Additionally, the method includes determining a corrected target location for at least one of the one or more target locations by applying the predicted error to at least one of the one or more target locations. The inputting and determining are performed by the computer system.

[0016] Each step of the method may be performed as further described herein. The method may include any other step(s) of any other method(s) described herein. The method may be performed by any of the systems described herein.

[0017] Another embodiment relates to a non-transitory computer-readable medium storing program instructions executable on a computer system to perform a computer-implemented method for determining information about a sample. The computer-implemented method includes obtaining one or more temperatures measured within the system using one or more temperature sensors. The system is configured as described above, and the computer-implemented method includes the predicting and determining steps of the above method. The computer-readable medium may be further configured as described herein. The steps of the computer-implemented method may be performed as further described herein. Additionally, the computer-implemented method on which the program instructions are executable 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 with the benefit of the following detailed description of the preferred embodiments and by reference to the accompanying drawings. [Brief explanation 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 2] FIG. 1 is a schematic diagram illustrating a side view of an embodiment of a system configured as described herein. [Figure 3] FIG. 1 is a schematic diagram illustrating a side view of an embodiment of a system configured as described herein. [Figure 4] 1 is a flowchart illustrating steps that may be performed by embodiments described herein during training. [Figure 5]1 is a flowchart illustrating steps that may be performed by embodiments described herein during execution. [Figure 6] FIG. 1 is a block diagram illustrating one embodiment of a non-transitory computer-readable medium storing program instructions for causing a computer system to perform the computer-implemented methods described herein. DETAILED DESCRIPTION OF THE INVENTION

[0020] While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof have been shown by way of example in the drawings and are herein described in detail. The drawings may not be to scale. It should be understood, however, that the drawings and the detailed description relating thereto are not intended to limit the invention to the particular forms disclosed, but rather, the invention is intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.

[0021] Turning now to the drawings, it should be noted that the figures are not drawn to scale. In particular, the scale of some of the elements in the figures may be greatly exaggerated to emphasize the characteristics of the elements. It should also be noted that the figures are not drawn to the same scale. Elements shown in multiple figures that may be similarly configured are indicated using the same reference numerals. Unless otherwise stated herein, any of the elements described and illustrated may include any suitable commercially available elements.

[0022] Generally, the embodiments described herein are systems and methods for determining information about a specimen. The information determined by the embodiments described herein includes position correction estimates for applications such as metrology and inspection. The embodiments may be advantageously used to generate position correction information to improve the accuracy of specimen placement in equipment such as metrology and inspection tools. Additionally, the embodiments described herein advantageously use machine learning (ML) domains to estimate position corrections.

[0023] In some embodiments, the specimen is a wafer. Wafers can include any wafer known in the semiconductor arts. Although some embodiments may be described herein with reference to one or more wafers, the embodiments are not limited to specimens that can be used. For example, the embodiments described herein can be used with specimens such as reticles, flat panels, personal computer (PC) boards, and other semiconductor specimens.

[0024] One embodiment of a system configured to determine information about a sample is shown in Figure 1. The system includes an output acquisition subsystem 100. The output acquisition subsystem may include and / or be coupled to a computer subsystem, such as computer subsystem 36 and / or one or more computer systems 102.

[0025] Generally, the output acquisition subsystems described herein include at least an energy source, a detector, and a scanning subsystem. The energy source is configured to generate energy that is directed to the sample by the output acquisition subsystem. The detector is configured to detect energy from the sample and generate an output in response to the detected energy. The scanning subsystem is configured to vary the location on the sample where the energy is directed and where the energy is detected. In one embodiment, as shown in FIG. 1 , the output acquisition subsystem is configured as an optical-based output acquisition subsystem.

[0026] In the light-based output acquisition subsystems described herein, the energy directed to the sample includes light, and the energy detected from the sample includes light. For example, in the system embodiment shown in FIG. 1 , the output acquisition subsystem includes an illumination subsystem configured to direct light toward the sample 14. The illumination subsystem includes at least one light source. For example, as shown in FIG. 1 , the illumination subsystem includes a light source 16. The illumination subsystem is configured to direct light toward the sample at one or more angles of incidence, which may include one or more oblique angles and / or one or more normal angles. For example, as shown in FIG. 1 , light from the light source 16 passes through an optical element 18 and then through a lens 20 and is directed toward the sample 14 at an oblique angle of incidence. The oblique angle of incidence may include any suitable oblique angle of incidence, which may vary depending, for example, on the properties of the sample and the process being performed on the sample.

[0027] The illumination subsystem may be configured to direct light to the sample at different angles of incidence at different times. For example, the output acquisition subsystem may be configured to change one or more properties of one or more elements of the illumination subsystem, such that light may be directed to the sample at a different angle of incidence than that shown in Figure 1. In one such example, the output acquisition subsystem may be configured to move the light source 16, the optical element 18, and the lens 20, such that light is directed to the sample at a different grazing or normal (or near-normal) angle of incidence.

[0028] In some cases, the output acquisition subsystem may be configured to direct light toward the sample at multiple angles of incidence simultaneously. For example, the illumination subsystem may include multiple illumination channels, one of which may include light source 16, optical element 18, and lens 20 as shown in FIG. 1 , and another illumination channel (not shown) may include similar elements that may be configured differently or the same, or may include at least a light source and, optionally, one or more other components as further described herein. When such light is directed toward the sample simultaneously with other light, one or more properties (e.g., wavelength, polarization, etc.) of the light directed toward the sample at different angles of incidence may be different, and the light resulting from illuminating the sample at the different angles of incidence may be distinguished from one another at the detector(s).

[0029] In another example, the illumination subsystem may include only one light source (e.g., light source 16 shown in FIG. 1 ), and light from the light source may be split into different optical paths (e.g., based on wavelength, polarization, etc.) by one or more optical elements (not shown) in the illumination subsystem. The light in each of the different optical paths is then directed toward the sample. Multiple illumination channels may be configured to direct light toward the sample simultaneously or at different times (e.g., when different illumination channels are used to illuminate the sample sequentially). In another example, the same illumination channel may be configured to direct light with different characteristics toward the sample at different times. For example, optical element 18 may be configured as a spectral filter, and the characteristics of the spectral filter may be changed in various different ways (e.g., by replacing one spectral filter with another) so that light of different wavelengths is directed toward the sample at different times. The illumination subsystem may have other suitable configurations known in the art to illuminate the sample with light having different or the same characteristics, at different or the same angles of incidence, sequentially, or simultaneously.

[0030] The light source 16 may include a broadband plasma (BBP) light source. As such, the light generated by the light source and directed toward the sample may include broadband light. However, the light source may include any other suitable light source, for example, any suitable laser known in the art configured to generate light at any suitable wavelength(s). The laser may be configured to generate monochromatic or nearly monochromatic light. As such, the laser may be a narrowband laser. The light source may also include a polychromatic light source that generates light at multiple discrete wavelengths or wavebands.

[0031] Light from optical element 18 may be focused onto sample 14 by lens 20. While lens 20 is shown in FIG. 1 as a single diffractive optical element, in practice, lens 20 may include multiple diffractive and / or reflective optical elements that, in combination, focus light from the optical element to the sample. The illumination subsystem shown in FIG. 1 and described herein may include other suitable optical elements (not shown). Examples of such optical elements include, but are not limited to, polarizing component(s), spectral filter(s), spatial filter(s), reflective optical element(s), apodizer(s), beam splitter(s), aperture(s), etc., which may include any suitable optical element known in the art. Additionally, the system may be configured to vary one or more elements of the illumination subsystem based on the type of illumination used to generate the output.

[0032] The output acquisition subsystem may also include a scanning subsystem, which is configured to vary the location on the sample where the light is directed and detected, possibly causing the light to scan across the sample. For example, the output acquisition subsystem may include a stage 22, on which the sample 14 is positioned during output generation. The scanning subsystem may include any suitable mechanical and / or robotic assembly (including the stage 22) that may be configured to move the sample so that the light may be directed to and detected at various locations on the sample. Additionally or alternatively, the output acquisition subsystem may be configured such that one or more optical elements of the output acquisition subsystem perform a scan of the light across the sample, directing the light to and detecting various locations on the sample. When the light is scanned across the sample, the light may be scanned across the sample in any suitable manner, such as a serpentine path or a spiral path.

[0033] The output acquisition subsystem further includes one or more detection channels. At least one of the detection channels includes a detector configured to detect light from the sample resulting from illumination of the sample by the output acquisition subsystem and generate an output responsive to the detected light. For example, the output acquisition 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 cases, both detection channels are configured to detect scattered light, i.e., the detection channels are configured to detect light scattered from the sample at different angles. However, one or more detection channels may be configured to detect other types of light (e.g., reflected light) from the sample.

[0034] As further shown in FIG. 1 , both detection channels are shown positioned within the plane of the paper, and the illumination subsystem is also shown positioned within the plane of the paper. Thus, in this embodiment, both detection channels are positioned within (e.g., at) the plane of incidence. However, one or more of the detection channels may be positioned outside the plane of incidence. For example, the detection channel formed by collector 30, element 32, and detector 34 may be configured to collect and detect light scattered out of the plane of incidence. Thus, such a detection channel may be generally referred to as a “side” channel, and such a side channel may be centered in a plane that is approximately perpendicular to the plane of incidence.

[0035] Although FIG. 1 illustrates an embodiment of an output acquisition subsystem including two detection channels, the output acquisition subsystem may include a different number of detection channels (e.g., only one detection channel or two or more detection channels). In one such example, the detection channel formed by collector 30, element 32, and detector 34 may form one side channel as described above, and the output acquisition subsystem may include an additional detection channel (not shown) formed as another side channel located on the opposite side of the incidence face. Thus, the output acquisition subsystem may include a detection channel including collector 24, element 26, and detector 28, and the detection channel is centered on the incidence face and configured to collect and detect light at scattering angles normal or near-normal to the sample surface. This detection channel may therefore be commonly referred to as the “top” channel, and the output acquisition subsystem may also include two or more side channels configured as described above. Thus, the output acquisition subsystem may include at least three channels (i.e., one top channel and two side channels), each of the at least three channels having its own collector, each configured to collect light at a different scattering angle than each of the other collectors.

[0036] As further described above, each of the detection channels included in the output acquisition subsystem can be configured to detect scattered light. Thus, the output acquisition subsystem shown in FIG. 1 can be configured for dark-field (DF) imaging of the sample. However, the output acquisition subsystem may also, or alternatively, include detection channel(s) configured for bright-field (BF) imaging of the sample. In other words, the output acquisition subsystem can include at least one detection channel configured to detect light specularly reflected from the sample. Thus, the output acquisition subsystems described herein can be configured for DF-only, BF-only, or both DF and BF imaging. While each of the collectors is shown in FIG. 1 as a single diffractive optical element, each of the collectors can include one or more diffractive optical elements and / or one or more reflective optical elements.

[0037] The one or more detection channels may include any suitable detector known in the art, such as a photomultiplier tube (PMT), a charge-coupled device (CCD), and a time-delay integration (TDI) camera. The detectors may also include non-imaging detectors or 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 may not be configured to detect such characteristic as a function of position in the imaging plane. Thus, the output generated by each detector included in each detection channel of the output acquisition subsystem may be a signal or data, and may not be an image signal or image data. In such cases, a computer subsystem, such as computer subsystem 36, may be configured to generate an image of the sample from the non-imaging output of the detector. However, in other examples, the detectors may be configured as imaging detectors configured to generate an image signal or image data. Thus, the output acquisition subsystem may be configured to generate an image in many ways.

[0038] It should be noted that FIG. 1 is provided herein to generally illustrate an output acquisition subsystem configuration that may be included in the system embodiments described herein. Clearly, the output acquisition subsystem configuration described herein may be modified to optimize the performance of the output acquisition subsystem, as is typically performed when designing a commercially available system. Additionally, the systems described herein may be implemented using existing systems, such as the 28xx / 39xx series of commercially available tools from KLA Corp. of Milpitas, California (e.g., by adding the functionality described herein to the existing system). In some such systems, the methods described herein may be provided as optional features of the system (e.g., in addition to other functionality of the system). Alternatively, the systems described herein may be designed “from scratch” to provide an entirely new system.

[0039] The computer subsystem 36 may be coupled to the detector of the output acquisition subsystem in any suitable manner (e.g., via one or more transmission media, which may include “wired” and / or “wireless” transmission media) so that the computer subsystem can receive the output generated by the detector. The computer subsystem 36 may be configured to perform numerous functions, including steps and functions further described herein, with or without the detector output. Thus, the steps described herein may be performed “on tool” by a computer subsystem coupled to or part of the output acquisition subsystem. Additionally, or alternatively, the computer system(s) 102 may perform one or more of the steps described herein. Thus, one or more of the steps described herein may be performed “off tool” by a computer system not directly coupled to the output acquisition subsystem. The computer subsystem 36 and the computer system(s) 102 may be further configured as described herein.

[0040] Computer subsystem 36 (and other computer subsystems described herein) may also be referred to herein as computer system(s). Each of the computer subsystem(s) or system(s) described herein may take various forms, including a personal computer system, an image computer, a mainframe computer system, a workstation, a network appliance, an Internet appliance, or other device. In general, the term "computer system" may be broadly defined to encompass any device having one or more processors that executes instructions from a memory medium. Computer subsystem(s) or system(s) may also include any suitable processor known in the art, such as a parallel processor. In addition, computer subsystem(s) or system(s) may include a computer platform with high-speed processing and software, either as a stand-alone or networked tool.

[0041] When a system includes multiple computer subsystems, the different computer subsystems may be coupled to one another such that images, data, information, instructions, etc. may be transmitted between the computer subsystems. For example, computer subsystem 36 may be coupled to computer system(s) 102, as indicated by the dashed lines in FIG. 1, by any suitable transmission medium, which may include any suitable wired and / or wireless transmission medium known in the art. Two or more of such computer subsystems may also be operatively coupled by a shared computer-readable storage medium (not shown).

[0042] While the output acquisition subsystem is described above as an optical or light-based output acquisition subsystem, in other embodiments, the output acquisition subsystem is configured as an electron-based output acquisition subsystem. In an electron beam output acquisition subsystem, the energy directed at the specimen includes electrons, and the energy detected from the specimen includes electrons. In one such embodiment, shown in FIG. 2, the output acquisition subsystem includes an electron column 122, and the system includes a computer subsystem 124 coupled to the output acquisition subsystem. The computer subsystem 124 may be configured as described above. Additionally, such an output acquisition subsystem may be coupled to one or more other computer systems in the same manner as described above and shown in FIG. 1.

[0043] 2, the electron column includes an electron beam source 126 configured to generate electrons that are focused onto a sample 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, all of which may include any such suitable elements known in the art.

[0044] Electrons returning from the sample (e.g., secondary electrons) may be focused onto a detector 134 by one or more elements 132. The one or more elements 132 may include, for example, a scanning subsystem, which may be the same scanning subsystem included in element(s) 130.

[0045] The electron column may include any other suitable elements known in the art. Additionally, the electron column may be further configured as described in U.S. Patent No. 8,664,594, issued April 4, 2014 to Jiang et al., U.S. Patent No. 8,692,204, issued April 8, 2014 to Kojima et al., U.S. Patent No. 8,698,093, issued April 15, 2014 to Gubbens et al., and U.S. Patent No. 8,716,662, issued May 6, 2014 to MacDonald et al., which are incorporated by reference as if fully set forth herein.

[0046] Although the electron column is shown in Figure 2 configured so that electrons are directed at the sample at an oblique angle of incidence and scattered from the sample at another oblique angle, the electron beam may be directed at the sample and scattered from it at any suitable angle. Additionally, the electron beam output acquisition subsystem may be configured to use multiple modes to generate output at the sample as described further herein (e.g., with different illumination angles, collection angles, etc.). The multiple modes of the electron beam output acquisition subsystem may differ in any output generation parameter of the output acquisition subsystem.

[0047] The computer subsystem 124 may be coupled to the detector 134 as described above. The detector may detect electrons returning from the surface of the sample, thereby forming an electron beam image (or other output) of the sample. The electron beam image may include any suitable electron beam image. The computer subsystem 124 may be configured to determine information about the sample using the output generated by the detector 134, and may perform as further described herein. The computer subsystem 124 may be configured to perform any additional step(s) described herein. A system including the output acquisition subsystem shown in FIG. 2 may be further configured as described herein.

[0048] It should be noted that FIG. 2 is provided herein to generally illustrate configurations of electron beam power acquisition subsystems that may be included in embodiments described herein. As with the optical subsystems described above, the electron beam subsystem configurations described herein may be modified to optimize the performance of the power acquisition subsystem, as is typically done when designing commercially available systems. In addition, the systems described herein may be implemented using existing systems (e.g., by adding the functionality described herein to an existing system), such as commercially available tools from KLA. In some such systems, the methods described herein may be provided as optional functionality of the system (e.g., in addition to other functionality of the system). Alternatively, the systems described herein may be designed “from scratch” to provide an entirely new system.

[0049] Although the output acquisition subsystem is described above as being an optical or electron beam subsystem, the output acquisition subsystem may also be an ion beam output acquisition subsystem. Such an output acquisition subsystem may be configured as shown in FIG. 2 , except that the electron beam source is replaced with any suitable ion beam source known in the art. Additionally, the output acquisition subsystem may be any other suitable ion beam output acquisition system, such as those included in commercially available focused ion beam (FIB) systems, helium ion microscopy (HIM) systems, and secondary ion mass spectrometry (SIMS) systems.

[0050] FIG. 3 illustrates another embodiment of a system including various optically based output acquisition subsystems. The output acquisition subsystem illustrated in FIG. 3 is described in more detail in U.S. Pat. No. 6,515,746 to Opsal et al., which is incorporated by reference as if fully set forth herein. Some non-essential details of the system presented in this patent have been omitted from the description corresponding to FIG. 3 presented herein. However, it should be understood that the system illustrated in FIG. 3 may be further configured as described in this patent. Additionally, upon reading the description of several embodiments provided herein, it will be apparent that the system illustrated in FIG. 3 has been modified to improve upon the system described in U.S. Pat. No. 6,515,746 to Opsal et al. The modifications include correcting for temperature-induced variations in target position.

[0051] One of the output acquisition subsystems is configured as a broadband reflectance spectrometer. The broadband reflectance spectrometer (BRS) 230 simultaneously probes the sample 226 with multiple wavelengths of light. The BRS 230 uses a lens 232 and includes a broadband spectrometer 234, which can be of any type commonly known and used in the art. The lens 232 may be a transmissive optical element formed from a material such as calcium fluoride (CaF2). Such a lens may be a spherical microscope objective lens with a high aperture number (approximately 0.90 NA) to create a large spread in the angle of incidence on the sample surface, creating a spot size of approximately 1 micron in diameter. Alternatively, the lens 232 may be a reflective optical element. Such a lens may have a lower aperture number (approximately 0.4 NA) and be capable of focusing light to a spot size of approximately 10-15 microns. The spectrometer 234 shown in FIG. 3 includes a lens 236, an aperture 238, a dispersive element 240, and a detector array 242. The lens 236 may be formed of CaF2.

[0052] In operation, probe beam 244 from light source 246 is collimated by lens 245 and directed by mirror 243 through mirror 266 to mirror 286, which directs the light through mirror 248 to lens 232, which then focuses the light onto sample 226. The light source may include any of the light sources described above. Lens 245 may be formed of CaF2.

[0053] Light reflected from the sample's surface passes through lens 232 and is directed by mirror 248 (via mirror 250) to spectrometer 234. Lens 236 focuses the probe beam through aperture 238, which defines a spot within the field of view of the sample surface to be analyzed. A dispersive element 240, such as a diffraction grating, prism, or holographic plate, angularly disperses the beam as a function of wavelength and directs it to individual detector elements contained in detector array 242.

[0054] The different detector elements measure the light intensity of different wavelengths of light contained in the probe beam, preferably simultaneously. Alternatively, detector 242 can be a charge-coupled device ("CCD") camera or a photomultiplier tube with appropriate dispersive or other wavelength-selective optics. Note that a monochromator can be used to measure different wavelengths sequentially (one wavelength at a time) using a single detector element. Furthermore, dispersive element 240 can be configured to disperse light in one direction as a function of wavelength and in an orthogonal direction as a function of angle of incidence relative to the sample surface, allowing simultaneous measurements as a function of both wavelength and angle of incidence. Computer subsystem 252 processes the intensity information measured by detector array 242.

[0055] Broadband spectroscopic ellipsometer (BSE) 254 is also configured to perform measurements on the sample using light. BSE 254 includes a polarizer 256, a focusing mirror 258, a collimating mirror 260, a rotating compensator 262, and an analyzer 264. In some embodiments, BSE 254 can be configured to perform measurements on the sample using light provided by light source 246, light source 283, or another light source (not shown).

[0056] In operation, mirror 266 directs at least a portion of probe beam 244 toward polarizer 256, thereby creating a known polarization state for the probe beam, preferably linearly polarized. Mirror 258 focuses the beam onto the sample surface at an oblique angle, ideally around 70° relative to the normal to the sample surface. Based on well-known principles of ellipsometry, the reflected beam, after interacting with the sample, typically has a mixture of linear and circular polarization, depending on the composition and thickness of the sample's film 268 and substrate 270.

[0057] The reflected beam is collimated by mirror 260, which directs the beam toward rotating compensator 262. Compensator 262 introduces a relative phase delay δ (phase retardation) between pairs of orthogonally polarized beam components. Compensator 262 rotates at an angular velocity c about an axis substantially parallel to the beam propagation direction, preferably by electric motor 272. Analyzer 264, preferably another linear polarizer, mixes the polarization states incident on it. By measuring the light transmitted by analyzer 264, the polarization state of the reflected probe beam can be determined.

[0058] Mirror 250 directs the beam to spectrometer 234, which simultaneously measures the intensity of different wavelengths of light in the reflected probe beam passing through the compensator / analyzer combination. Computer subsystem 252 receives the output of detector 242 and processes the intensity information measured by detector 242 as a function of wavelength and as a function of the azimuthal (rotational) angle about the axis of rotation of compensator 262, solving for the ellipsometric values ​​ψ and Δ as described in U.S. Patent No. 5,877,859 to Aspnes et al., which is incorporated by reference as if fully set forth herein.

[0059] The systems including the broadband reflectance spectrometer and broadband spectroscopic ellipsometer described above may further include additional output acquisition subsystem(s) configured to perform additional measurements of the sample using the light. For example, the system may include an output acquisition subsystem configured as a beam profile ellipsometer, a beam profile reflectometer, another optical subsystem, or a combination thereof.

[0060] Beam profile ellipsometry (BPE) is described in U.S. Pat. No. 5,181,080 to Fanton et al., which is incorporated by reference as if fully set forth herein. BPE 274 includes a laser 283 that generates a probe beam 284. Laser 283 may be a Toshiba solid-state laser diode emitting a linearly polarized 3 mW beam at 673 nm. BPE 274 also includes a quarter-wave plate 276, a polarizer 278, a lens 280, and a quad detector 282. In operation, linearly polarized probe beam 284 is focused onto sample 226 by lens 232. Light reflected from the sample surface passes through lens 232 and mirrors 248, 286, and 288 and is directed to BPE 274 by mirror 290.

[0061] The position of a ray within the reflected probe beam corresponds to a specific angle of incidence relative to the sample's surface. Quarter-wave plate 276 retards the phase of one of the beam's polarization states by 90°. Linear polarizer 278 causes the beam's two polarization states to interfere with each other. To maximize the signal, the axis of polarizer 278 should be oriented at a 45° angle with respect to the fast and slow axes of quarter-wave plate 276. Detector 282 is a quad-cell detector with four radially arranged quadrants, each of which blocks one-quarter of the probe beam and generates a separate output signal proportional to the power of the portion of the probe beam that strikes that quadrant.

[0062] The output signal from each quadrant is sent to computer subsystem 252. By monitoring the change in the polarization state of the beam, ellipsometry information, such as ψ and Δ, can be determined. To determine this information, computer subsystem 252 takes the difference between the sums of the output signals in opposite quadrants, a value that varies linearly with film thickness for very thin films.

[0063] Beam profile reflectometry (BPR) is described in U.S. Patent No. 4,999,014 to Gold et al., which is incorporated by reference as if fully set forth herein. BPR 292 includes a laser 283, a lens 294, a beam splitter 296, and two linear detector arrays 298 and 300 to measure the reflectivity of a sample. In operation, a linearly polarized probe beam 284 is focused onto the sample 226 by lens 232, with various rays within the beam striking the sample surface at various angles of incidence. Light reflected from the sample surface passes through lens 232 and mirrors 248 and 286 and is directed by mirror 288 to BPR 292. The positions of the rays within the reflected probe beam correspond to specific angles of incidence relative to the sample's surface. Lens 294 spatially expands the beam in two dimensions. A beam splitter 296 separates the S and P components of the beam, and detector arrays 298 and 300 are oriented orthogonally to separate information about the S and P polarizations. Beams with large angles of incidence are located near opposite ends of the array. The output from each element of the diode array corresponds to a different angle of incidence. Detector arrays 298 and 300 measure the intensity of the entire reflected probe beam as a function of the angle of incidence relative to the sample surface. Computer subsystem 252 receives the outputs of detector arrays 298 and 300 and derives the thickness and refractive index of thin film layer 268 based on these angle-dependent intensity measurements using various types of modeling algorithms. Optimization routines using an iterative process, such as a least-squares fitting routine, are typically used.

[0064] 3 may also include additional components, such as a detector / camera 302. The detector / camera 302 may be positioned above the mirror 290 and used to observe the reflected beam from the sample 226 for alignment and focusing purposes.

[0065] To calibrate the BPE 274, BPR 292, BRS 230, and BSE 254, the system may include a wavelength-stable calibration reference ellipsometer 304 used in combination with a reference sample (not shown). For calibration purposes, the reference sample ideally consists of a thin oxide layer of thickness d formed on a silicon substrate. In general, however, the sample may be any suitable substrate of known composition, including bare silicon wafers and silicon wafer substrates with one or more thin films formed thereon. The layer thickness d does not need to be known or consistent between periodic calibrations.

[0066] Ellipsometer 304 includes light source 306, polarizer 308, lenses 310 and 312, rotating compensator 314, analyzer 316, and detector 318. Compensator 314 is rotated at an angular velocity ψ about an axis substantially parallel to the direction of propagation of beam 320, preferably by electric motor 322. It should be noted that the compensator can be located either between the sample and the analyzer (as shown in FIG. 3) or between the sample and polarizer 308. It should also be noted that polarizer 308, lenses 310 and 312, compensator 314, and analyzer 316 all have their structures optimized for the particular wavelength of light generated by light source 306 to maximize the accuracy of the ellipsometer.

[0067] Light source 306 generates a quasi-monochromatic probe beam 320 with a known, stable wavelength and stable intensity. This can be done passively, where light source 306 produces a very stable output wavelength that does not change over time (i.e., less than 1% variation). An example of a passively stable light source is a helium-neon laser or other gas discharge laser system. Alternatively, a non-passive system can be used, where the light source includes a light generator (not shown) that generates light having a wavelength that is not precisely known or stable over time, and a monochromator (not shown) that precisely measures the wavelength of the light generated by the light generator. Examples of such light generators include laser diodes or polychromatic light sources used in combination with color filters, such as gratings. In either case, the wavelength of beam 320 is a known constant or is measured by a monochromator and provided to computer subsystem 252, allowing ellipsometer 304 to precisely calibrate optical measurement devices within the system.

[0068] The operation of ellipsometer 304 during calibration is further described in U.S. Patent No. 6,515,746. Briefly, beam 320 enters detector 318, which measures the intensity of the beam as it passes through the compensator / analyzer combination. Computer subsystem 252 processes the intensity information measured by detector 318 to determine the polarization state of the light after interacting with the analyzer, i.e., the ellipsometry parameters of the sample. This information processing involves measuring the beam intensity as a function of the compensator's azimuthal (rotational) angle about its axis of rotation. This measurement of intensity as a function of compensator rotation angle is effectively a measurement of beam 320 intensity as a function of time, since the angular velocity of the compensator is typically known and constant.

[0069] By knowing the composition of the reference sample and the exact wavelength of the light generated by the light source 306, the optical properties of the reference sample, such as the film thickness d, refractive index, and extinction coefficient, can be determined by the ellipsometer 304. Once the film thickness d is determined by the ellipsometer 304, the same sample is probed by other optical measurement devices BPE 274, BPR 292, BRS 230, and BSE 254, which measure various optical parameters of the sample. The computer subsystem 252 then calibrates the process variables used to analyze the results from these optical measurement devices to produce accurate results. In the above calibration technique, all system variables that affect phase and intensity are determined and corrected using phase offsets and reflectance normalization factors described in U.S. Patent No. 6,515,746, making the optical measurements made by these calibrated optical measurement devices absolute.

[0070] The calibration techniques described above are primarily based on calibration using the derived thickness d of a thin film. However, calibration using the ellipsometer 304 can be based on any optical property of a reference sample that is measurable or determinable by the ellipsometer 304 and / or otherwise known, regardless of whether the sample has a single film thereon, multiple films thereon, or no film thereon (a bare sample).

[0071] In some embodiments, the output acquisition subsystems may have at least one common optical component. For example, lens 232 is common to BPE 274, BPR 292, BRS 230, and BSE 254. Similarly, mirrors 243, 266, 286, and 248 are common to BPE 274, BPR 292, BRS 230, and BSE 254. As shown in FIG. 3 , ellipsometer 304 has no optical components common to the other output acquisition subsystems. Such isolation from the other output acquisition subsystems may be appropriate because the ellipsometer is used to calibrate the other output acquisition subsystems.

[0072] Further, as noted above, the output acquisition subsystem may be configured to have multiple modes. Generally, a "mode" is defined by the values ​​of the parameters of the output acquisition subsystem used to generate an output at the specimen. Thus, different modes may differ in values ​​for at least one of the output-generating parameters of the output acquisition subsystem (other than the location on the specimen where the output is generated). For example, for an optical-based output acquisition subsystem, different modes may use different wavelengths of light. The modes may differ in the wavelength of light directed at the specimen (e.g., by using different light sources, different spectral filters, etc. for the different modes), as described further herein. In another embodiment, different modes may use different illumination channels. For example, as noted above, the output acquisition subsystem may include multiple illumination channels. Thus, different illumination channels may be used for different modes.

[0073] The multiple modes may also differ in illumination and / or collection / detection. For example, as further described above, the output acquisition subsystem may include multiple detectors. Thus, one of the detectors may be used for one mode and another detector may be used for another mode. Furthermore, the modes may differ from each other in multiple ways described herein (e.g., different modes may have one or more different illumination parameters and one or more different detection parameters). In addition, the multiple modes may have different viewpoints, i.e., different angles of incidence and / or collection, which is possible as further described above. The output acquisition subsystem may be configured to scan the sample with different modes, either in the same scan or different scans, depending, for example, on the ability to simultaneously scan the sample using multiple modes.

[0074] In some examples, the systems described herein may be configured as inspection systems. However, the systems described herein may also be configured as other types of semiconductor-related quality control-type systems, such as defect review systems and metrology systems. For example, the output acquisition subsystem embodiments described herein and illustrated in FIGS. 1-3 may have one or more parameters modified to provide different output generation capabilities depending on the application in which they are used. In one embodiment, the output acquisition subsystem is configured as an electron beam defect review subsystem. For example, the output acquisition subsystem illustrated in FIG. 2 may be configured to have higher resolution when used for defect review or metrology rather than inspection. In other words, the output acquisition subsystem embodiments illustrated in FIGS. 1-3 illustrate general and various configurations of output acquisition subsystems, which may be adjusted in several ways, as would be apparent to one skilled in the art, to create output acquisition subsystems with various output generation capabilities that are more or less suitable for different applications.

[0075] As described above, the output acquisition subsystem may be configured to direct energy (e.g., light, electrons) at and / or scan energy over a physical version of the sample, thereby generating an actual output of the physical version of the sample. In this manner, the output acquisition subsystem may be configured as a “real” output acquisition system rather than a “virtual” system. However, the storage medium (not shown) and computer system(s) 102 shown in FIG. 1 and / or other computer subsystems shown and described herein may be configured as “virtual” systems. In particular, the storage medium and computer system(s) 102 are not part of the output acquisition subsystem 100 and do not have any functionality for manipulating a physical version of the sample, but may be configured as a virtual inspector that performs functions such as inspection, a virtual metrology system that performs functions such as metrology, a virtual defect review tool that performs functions such as defect review, etc., using the stored detector output. Systems and methods configured as "virtual" systems are described in commonly assigned U.S. Patent No. 8,126,255, issued to Bhaskar et al. on February 28, 2012; U.S. Patent No. 9,222,895, issued to Duffy et al. on December 29, 2015; and U.S. Patent No. 9,816,939, issued to Duffy et al. on November 14, 2017, which are incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in these patents. For example, the computer subsystems described herein may be further configured as described in these patents.

[0076] One embodiment of a system configured to determine information about a specimen includes an output acquisition subsystem configured to generate output for the specimen at one or more target locations on the specimen. The one or more target locations may be predetermined target locations stored in a recipe for a process to be performed on the specimen using the output acquisition subsystem. The one or more target locations may be discrete locations, such as locations used in a move-acquire-measure type metrology process, or locations defined by an area on the specimen that is scanned, such as in an inspection or scanning type process. As such, the target locations indicate where a process should be performed on the specimen; however, as described further herein, even with state-of-the-art specimen positioning methods and systems, a process does not always occur at the intended target location. As described further herein, the embodiments described herein aim to minimize the difference between the intended target location and the actual target location and / or correct the target location at which the process was performed. By predicting temperature-dependent specimen positioning errors, the embodiments described herein may improve the performance and usability of the quality control type processes and tools described herein.

[0077] In one embodiment, the output acquisition subsystem is configured as a metrology subsystem. As mentioned above, the output acquisition subsystems shown in FIGS. 1-3 may be configured as metrology subsystems. As such, the embodiments described herein may be configured as metrology tools. In the field of semiconductor metrology, a metrology tool may include an illumination subsystem that illuminates a target, a collection subsystem that acquires relevant information provided by the illumination subsystem's interaction (or lack thereof) with the target, device, or feature, and a computer subsystem that analyzes the collected information using one or more algorithms. Metrology tools may be used to measure structural and material properties related to various semiconductor manufacturing processes (e.g., material composition, dimensional properties of structures and films, such as film thickness, and / or critical dimensions (CDs) of structures, overlay, etc.). These measurements are used to facilitate process control and / or yield efficiency in the manufacture of semiconductor dies.

[0078] A metrology tool may include one or more hardware configurations that may be used in conjunction with certain embodiments described herein to measure, for example, the various semiconductor structure and material properties discussed above. Examples of such hardware configurations include, but are not limited to, the following:

[0079] 1. Spectroscopic Ellipsometer (SE) 2. SE with multiple lighting angles 3. Measuring Mueller Matrix Elements (e.g., using rotation compensator(s)) 4. Single-wavelength ellipsometer 5. Beam profile ellipsometer (angle resolved ellipsometer) 6. Beam profile reflectometer (angle resolved reflectometer) 7. Broadband Reflectance Spectrometer (Spectroreflectometer) 8. Single-wavelength reflectometer 9. Angle-resolved reflectometer 10. Imaging System 11. Scatterometers (speckle analyzers, etc.)

[0080] The hardware configurations may be separated into separate operational systems, or one or more hardware configurations may be combined into a single tool. An example of combining multiple hardware configurations into a single tool is shown in FIG. 3, which may be further configured as described in U.S. Pat. No. 7,933,026 to Opsal et al., which is incorporated by reference as if fully set forth herein. The systems described herein may be further configured as described in this reference.

[0081] FIG. 3, for example, shows a schematic diagram of an exemplary metrology tool, including a) a broadband SE (i.e., 254), b) an SE (i.e., 304) with a rotating compensator (i.e., 314), c) a beam profile ellipsometer (i.e., 274), d) a beam profile reflectometer (i.e., 292), e) a broadband reflectance spectrometer (i.e., 230), and f) a deep-ultraviolet reflectance spectrometer (i.e., 230). Additionally, such systems typically include numerous optical elements, including specific lenses, collimators, mirrors, quarter-wave plates, polarizers, detectors, cameras, apertures, and / or light sources. The wavelength of the optical system can vary from approximately 120 nm to 3 microns. For non-ellipsometer systems, the collected signal may be polarization-resolved or unpolarized. FIG. 3 also shows a diagram of multiple metrology heads integrated into the same tool. However, in many cases, multiple metrology tools are used to measure a single or multiple metrology targets, as described, for example, in U.S. Patent No. 7,478,019 to Zangooie et al., which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this reference.

[0082] The illumination subsystem of a particular hardware configuration includes one or more light sources. The light sources may generate light having only one wavelength (i.e., monochromatic light), light having multiple discrete wavelengths (i.e., polychromatic light), light having multiple wavelengths (i.e., broadband light), and / or light that sweeps continuously or hops between wavelengths (i.e., tunable or swept light sources). Examples of suitable light sources include, but are not limited to, white light sources, ultraviolet (UV) lasers, arc lamps or electrodeless lamps, laser-sustained plasma (LSP) sources such as those commercially available from Energetiq Technology, Inc. of Woburn, Massachusetts, supercontinuum sources (e.g., broadband laser sources) such as those commercially available from NKT Photonics Inc. of Morganville, New Jersey, or short-wavelength light sources such as x-ray sources, extreme UV sources, or combinations thereof. The light sources may also be configured to provide light of sufficient brightness, in some cases, that is greater than about 1 W / (nm cm). 2 The measurement system may further include high-speed feedback to the light source to stabilize the power and wavelength of the light source. The output of the light source may be transmitted via free-space propagation, or possibly via optical fiber or any type of light guide.

[0083] Metrology tools can be designed to perform many different types of measurements related to semiconductor manufacturing. Certain embodiments described herein can be applied to such measurements. For example, in certain embodiments, a tool can measure characteristics of one or more targets, such as critical dimensions, overlay, sidewall angle, film thickness, and process-related parameters (such as focus and / or dose). A target can include a particular region of interest that is periodic in nature, such as the grid of a memory die. A target can include multiple layers (or films) whose thicknesses can be measured by a metrology tool. A target can include a target design placed (or already present) on a specimen, such as for use in alignment and / or overlay registration operations. A particular target can be located at various locations on a specimen. For example, a target can be located within a scribe line (such as between dies) and / or on the die itself. In certain embodiments, multiple targets are measured (simultaneously or at different times) by the same or multiple metrology tools, as described in U.S. Patent No. 7,478,019 to Zangooie et al. Data from such measurements can be combined. Data from metrology tools may be used in semiconductor manufacturing processes, for example, to feed-forward, feedback, and / or feed-sideway corrections to processes (lithography, etch, etc.), thereby creating a complete process control solution.

[0084] As semiconductor device feature dimensions continue to shrink, smaller metrology targets are often required. Furthermore, measurement accuracy and matching to actual device characteristics increases the need for device-like targets and in-die and on-device measurements. Various metrology implementations have been proposed to achieve this goal. For example, focused-beam ellipsometry, based primarily on reflective optics, is one such implementation, as described in U.S. Pat. No. 5,608,526 to Piwonka-Corle et al., which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent. Apodizers may be used to mitigate the effects of optical diffraction, which broaden the illumination spot beyond the size defined by geometric optics. The use of apodizers is described in U.S. Pat. No. 5,859,424 to Norton, which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent. The use of high-numerical-aperture tools with simultaneous multiple-angle-of-incidence illumination is another method for achieving small target capabilities. This technique is described, for example, in U.S. Patent No. 6,429,943 to Opsal et al., which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent.

[0085] Other example measurements may include measuring the composition of one or more layers of a semiconductor stack or sample, measuring specific defects on (or within) the sample, and measuring the amount of photolithography radiation to which the sample has been exposed. In some cases, the metrology tools and algorithms may be configured to measure non-periodic targets; see, for example, U.S. Patent No. 9,291,554 to Kuznetsov et al., issued March 22, 2016, and U.S. Patent No. 9,915,522 to Jiang et al., issued March 13, 2018, which are incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent.

[0086] Measuring the parameter of interest typically involves several algorithms. For example, the optical interaction between the incident beam and the sample is modeled using an EM (electromagnetic) solver, using algorithms such as RCWA, FEM, the method of moments, the surface integral method, the volume integral method, and FDTD. The target of interest is typically modeled (parameterized) using a geometry engine, or possibly a process modeling engine, or a combination of both. The use of process modeling is described in U.S. Pat. No. 10,769,320 to Kuznetsov et al., issued September 8, 2020, which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent. The geometry engine is implemented, for example, in KLA's AcuShape software product.

[0087] The collected data can be analyzed by many data fitting and optimization methods and technologies, including libraries, fast low-dimensional models, regression, machine learning algorithms such as neural networks and support vector machines (SVM), dimensionality reduction algorithms such as PCA (Principal Component Analysis), ICA (Independent Component Analysis) and LLE (Locally Linear Embedding), sparse representations such as Fourier transforms or wavelet transforms, Kalman filters, algorithms that facilitate matching from the same or different tool types, etc.

[0088] The collected data may also be analyzed by algorithms that do not involve modeling, optimization, and / or fitting, such as those described in U.S. Patent No. 10,591,406 to Bringoltz et al., issued March 17, 2020, which is incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in this patent.

[0089] Computational algorithms are typically optimized for measurement applications using one or more approaches such as computational hardware design and implementation, parallelization, computation distribution, load balancing, multi-service support, dynamic load optimization, etc. Various implementations of the algorithms may be performed in firmware, software, FPGAs, programmable optical components, etc.

[0090] The data analysis and fitting step typically pursues one or more of the following goals:

[0091] 1. Measurement of critical dimensions (CD), sidewall angle (SWA), shape, stress, composition, film, bandgap, electrical properties, focus / dose, overlay, generation of process parameters (resist state, partial pressure, temperature, focus model, etc.), and / or combinations thereof; 2. Modeling and / or design of measurement systems; and 3. Modeling, design, and / or optimization of metrology targets.

[0092] The embodiments described herein that are configured for the field of semiconductor metrology are not limited to the hardware, algorithmic / software implementations and architectures, and use cases summarized above.

[0093] In another embodiment, the output acquisition subsystem is configured as an inspection subsystem. The inspection subsystem may be configured to perform inspection using other energy types, such as light, electrons, or ions. Such an output acquisition subsystem may be configured, for example, as shown in FIGS. 1 and 2. In a system in which the output acquisition subsystem is configured as an inspection subsystem, the computer subsystem may be configured to detect defects on the specimen based on the output generated by the output acquisition subsystem. For example, in perhaps the simplest scenario, the computer subsystem may subtract a reference from the output to generate a difference signal or image and apply a threshold to the difference signal or image. The computer subsystem may determine that any difference signal or image having a value above the threshold is a defect or potential defect, and any difference signal or image having a value below the threshold is not a defect or potential defect. Of course, many defect detection methods and algorithms used in commercially available inspection tools are much more complex than this example, and any such method or algorithm may be applied to the output generated by the output acquisition subsystem configured as an inspection subsystem.

[0094] Similarly, the process may be a defect review process. Unlike an inspection process, a defect review process typically returns to a distinct location on the specimen where a defect was detected. An output acquisition subsystem configured for defect review may generate and use specimen images as described herein to determine one or more attributes of the defects, such as their shape, dimensions, roughness, background pattern information, and / or classification of the defects (e.g., bridging-type defects, missing feature defects, etc.). For defect review applications, the computer subsystem may be configured to determine information about the defects or specimen from output generated by the output acquisition subsystem using any suitable defect review method or algorithm used in any suitable defect review tool.

[0095] In some embodiments, the output acquisition subsystem is an optical-based subsystem. Such an output acquisition subsystem may be further configured as described herein and shown in FIGS. 1 and 3. In other embodiments, the output acquisition subsystem is an electronic-based subsystem. Such an output acquisition subsystem may be further configured as described herein and shown in FIG. 2. Thus, as can be seen from the above, the embodiments described herein may be used with output acquisition subsystems having many different characteristics. In general, the embodiments described herein may be used with any tool that can measure temperature near or within hardware elements responsible for positioning the sample within the tool at runtime and can measure position error when moving across the sample during calibration.

[0096] The system also includes one or more temperature sensors configured to measure one or more temperatures within the system. Generally, the temperature sensors included in the system, whose measured temperatures are used for position compensation as further described herein, may include any or all of the temperature sensors already installed in the system. In one such example, the temperatures may include any or all of the following: Y front temperature recorded at the tool (also referred to herein as “Y_TEMP_FRONT”), Y back temperature recorded at the tool (also referred to herein as “Y_TEMP_BACK”), X temperature recorded at the tool (also referred to herein as “X_TEMP”), carriage temperature recorded at the tool (also referred to herein as “CARRIAGE_TEMP”), theta temperature recorded at the tool (also referred to herein as “T_TEMP”), and Y plate temperature recorded at the tool (also referred to herein as “Y_PLATE_TEMP”).

[0097] Y_TEMP_FRONT is the temperature of the first of two linear motors moving along the Y direction of the sample handling stage. Y_TEMP_BACK is the temperature of the second of two linear motors. X_TEMP is the temperature of the X-axis linear motor moving along the X direction of the sample handling stage; if the X-axis motor is a single linear motor, there can be only one temperature in the X direction (unlike the Y-axis motor, which can have two linear motors moving along it). CARRIAGE_TEMP can be measured by a temperature sensor placed near and somewhat below the sample chuck, which may be on the sample handling stage. T_TEMP can be measured with a resistance temperature detector (RTD) near the theta-axis winding. Y_PLATE_TEMP is the temperature of certain components of some sample handling stages.

[0098] While these temperature measurements and temperature sensors may be useful in many tools, they are included herein merely as illustrative examples, and due to all the variations in motors and other components that may be included in the sample handling subsystem, the inventors have purposely chosen to depict the scanning subsystem and temperature sensors in a relatively abstract manner in the figures described further herein. In particular, the exact locations of the various temperature sensors relative to the various temperatures described above are unlikely to significantly impact the implementation of the embodiments described herein. As long as temperatures are measured at or near at least some (i.e., one or more) of the hardware elements of the scanning subsystem primarily responsible for positioning the sample, a DL model should be trained to predict target position errors from temperature measurements. If the DL model does not perform satisfactorily through training and testing, the temperature sensor(s) may be changed or additional temperature sensor(s) may be provided until a sensor configuration is found that provides sufficient data for the DL model and enables sufficiently accurate predictions. Such experimentation and testing is within the skill of one of ordinary skill in the art and will vary depending on the tool configuration.

[0099] In one embodiment, one or more temperature sensors are configured to measure one or more temperatures during a process performed on the specimen using the output acquisition subsystem. For example, after a DL model has been trained as described herein and is ready for use at runtime, one or more temperature sensors can collect temperature measurements while the specimen is being scanned or measured with the output acquisition subsystem. The sampling rate of the temperature sensors can be set based on any information regarding how temperature changes may cause changes in positioning accuracy and, in some cases, how quickly temperatures may change near the scanning subsystem. In one such non-limiting example, the sampling rate of the temperature sensor(s) may be set to detect a 0.1° C. change in temperature within the tool, which in this case means a sampling rate of about one sample every 30 seconds.

[0100] Measuring temperature during a process can mean several different things, depending on the tool's configuration and capabilities. For example, the temperature(s) may be measured continuously once the sample is loaded into the tool and ready to be measured or scanned. However, because temperatures can often be expected to change relatively slowly, continuous temperature measurement is likely not required. As mentioned above, a sampling rate of about one sample every 30 seconds may be sufficient. Whether these temperature measurements and error predictions are performed while the system is performing other tasks, such as measurement, output acquisition (e.g., imaging), or sample positioning, may depend on the system's data collection and processing capabilities. For example, temperature measurements and error predictions may alternate between sample positioning and measurement in metrology, because temperature measurements and error predictions are performed much less frequently than other functions. In other applications, such as inspection, where a relatively large area of ​​the sample can be scanned and output is generated continuously (or nearly continuously), it may be most practical to obtain temperature measurements during scanning, i.e., at the same time that output acquisition is occurring. Thus, measuring the temperature during a process can mean measuring the temperature at the same time as the sample is measured, inspected, imaged, etc., or between the start and end of the process.

[0101] In another embodiment, one or more temperature sensors are configured to measure one or more temperatures of or within one or more hardware elements of the system configured to change the position of the sample within the system. Thus, the temperature sensor(s) may be configured to measure a temperature(s) at one or more locations near or within the scanning subsystem described herein. In one such example, the system may include temperature sensors 40, 42, 44, and 46 shown in FIG. 1 . Temperature sensors 40 and 42 are positioned below stage 22 (i.e., on the opposite side of stage 22 from sample 14). These temperature sensors may be directly coupled to the stage or may be located within one or more additional elements (not shown) coupled to the stage, which may include mechanical and / or robotic elements configured to move the stage. These temperature sensors may be positioned opposite each other along a diameter of the stage, which diameter may define, for example, the X- or Y-axis of the stage. These temperature sensors may also be positioned at other locations, for example, along different diameters and therefore different axes of the stage. The temperature sensors may also be positioned at different distances from the center of the stage. The exact location of the sensor may be determined in some respects based on the exact hardware configuration of the scanning subsystem, where there is space within the hardware that can accommodate a temperature sensor.

[0102] Temperature sensor 44 is shown located within stage 22. This temperature sensor location is intended to illustrate how a temperature sensor may be located within a hardware element of the scanning subsystem. In some cases, having the temperature sensor within the stage may not actually be useful. If the temperature sensor is located within the stage, it is clearly desirable to configure it to minimize contact between the temperature sensor and the sample to prevent the temperature sensor from altering the sample in any way. A temperature sensor that is "within a hardware element" of the scanning subsystem may not be entirely contained within a single hardware element of the scanning subsystem. For example, such a temperature sensor may be only partially surrounded by a hardware element of the scanning subsystem. Additionally, such a temperature sensor may be entirely enclosed within a space formed between multiple hardware elements of the system. Thus, while a temperature sensor may be within the scanning subsystem, it need not (but can) be within any single hardware element of the scanning subsystem.

[0103] Temperature sensor 46 is shown in FIG. 1 as coupled to the outer edge of stage 22. This temperature sensor may be fixedly attached to the outer edge of the stage, if the tool configuration allows. Locating the temperature sensor on the outer edge of the stage (or on the outer edge of the scanning subsystem by attaching it to another hardware element of the scanning subsystem) may be advantageous because it may be suitable and even advantageous to measure one or more temperatures that are somewhat external to the scanning subsystem and / or that are exposed to the atmospheric environment within the tool housing (and not just the scanning subsystem). While such a temperature sensor is shown attached to the stage, the temperature sensor may alternatively be located near the stage or another scanning subsystem hardware element and need not necessarily be attached to the stage.

[0104] The temperature sensors shown in FIG. 1 are intended to illustrate some possible advantageous placements of temperature sensors, but are not intended to limit the embodiments described herein in terms of temperature sensor configurations that may be used therewith. For example, in addition to the temperature sensors shown in FIG. 1 , the system may include one or more temperature sensors (not shown) that are remote from the scanning subsystem, attached to the system housing, attached to one or more imagers of the system, configured to measure the temperature of the sample in a non-contact or contact manner, etc. Additionally, while four temperature sensors are shown in FIG. 1 and example numbers of measured temperatures are described further herein, it should be apparent that the system may include any suitable number of temperature sensors arranged in any suitable configuration. Furthermore, the temperature sensors described herein may be any suitable commercially available sensors and may be selected based on various information about the tool, such as expected temperature ranges, size considerations, expected measurement frequency, environmental conditions within the tool, etc. In some cases, temperature sensors attached to the tool for other purposes may be used to measure and record the various temperatures described herein and / or other temperatures that may be suitable for use with the embodiments described herein. For tools with existing temperature sensors, training of a DL model can be attempted using the temperatures measured by those existing sensors, and if the trained DL model does not pass testing and validation, adding additional temperature sensors can be explored as a way to improve the performance of the DL model.

[0105] The system further includes a computer subsystem configured to acquire one or more temperature measurements from the one or more temperature sensors. The computer subsystem may include any of the computer subsystem(s) or system(s) described herein and may have any configuration thereof. Acquiring the one or more temperature measurements may be performed in the same manner as described above with respect to acquiring the output generated by the output acquisition subsystem. For example, the computer subsystem 36 shown in FIG. 1 may be coupled to the temperature sensors in any suitable manner (e.g., via one or more transmission media, which may include “wired” and / or “wireless” transmission media) so that the computer subsystem receives the temperature measurements generated by the temperature sensors. The computer subsystem 36 may be configured to perform several functions, including steps and functions described further herein, with or without the temperature sensor output. As described further herein, the one or more temperature measurements acquired by the computer subsystem may include various information, such as time and spatial information associated with the temperature measurements, in addition to the actual temperature measurements. This additional information may be generated and acquired as described further herein.

[0106] The system includes one or more components executed by a computer subsystem. For example, as shown in FIG. 1 , the system includes one or more components 104 executed by computer subsystem 36 and / or computer system(s) 102. Systems shown in other figures described herein may be configured to include similar elements. The one or more components may be executed by the computer subsystem as described further herein or in any other suitable manner known in the art. At least part of executing one or more components may include inputting one or more inputs, such as acquired temperatures, data, etc., to one or more components. The computer subsystem may be configured to input any measured temperature data, etc., to one or more components in any suitable manner.

[0107] One or more components include a deep learning (DL) model configured to predict errors in at least one of the one or more target locations based on at least one of one or more measured temperatures input to the DL model by a computer subsystem. For example, as shown in FIG. 1 , one or more components 104 include a DL model 106. The embodiments described herein leverage domain knowledge understanding, including correlations between temperatures within the tool and X and / or Y position errors. The one or more measured temperatures input to the DL model may include any information generated by or associated with the temperature measurements, such as time, space, and sensor ID information associated with the measured temperatures, or the measured temperatures themselves. Additionally, embodiments provide methods and systems for estimating position corrections by using domain-specific knowledge of the process without using external tools. The embodiments described herein enable position corrections using machine learning (ML) recipes trained based on temperature domain knowledge. Using location (i.e., target location) information as input to the DL model, the trained DL recipe can provide location-specific position corrections. Additionally, the DL model can estimate a position correction map at runtime to improve specimen placement accuracy. Some or all of the temperature domain knowledge described herein may be used to train DL recipes.

[0108] The "prediction error at at least one of the one or more target locations" can be defined in one of two ways, depending on which target location information is input to the DL model. If the target location information is an intended target location, the prediction error is defined as the difference between the intended target location and the predicted location at which output generation would occur without any temperature-based correction. If the target location is a reported target location, the prediction error is defined as the difference between the reported target location and the actual sample location at which output would be generated. The "prediction error" is also referred to herein as a "position error," and these terms are used interchangeably herein. The correction made as further described herein based on the prediction error is referred to herein as a "position correction," and results in a corrected location or "corrected target location."

[0109] Obviously, information about at least one target location is input into the DL model along with temperature information, since the DL model predicts an error for at least one target location. The number of target locations for which errors are predicted may vary significantly from sample to sample and / or process to process. Generally, errors may be predicted for a relatively sparse number of target locations across the sample, e.g., 100 target locations versus several hundred or 1,000 target locations. The predicted error for any one target location may be used to determine a corrective target location for that one target location, and possibly other target locations near that one target location, as described further herein.

[0110] Thus, the target location information input to a DL model may vary depending on the target locations themselves and sometimes depending on whether the DL model is being trained or used at run time. For example, when predicting errors for target locations specified in a process recipe, such as where the target locations are scanned, imaged, measured, etc., the target location information input to the DL model during training and run time may be a subset of all target locations in the process recipe. The target location information input to the DL model during training and run time may be the same for any one specimen and process, but this need not be the case. For example, more target location information may be input for training purposes than run time. The reverse may also be true, meaning that the target location information used for training purposes may be for fewer target locations than the number of target locations for which errors are predicted at run time. The target location information input for training may be arbitrary, in the sense that it may be for an arbitrarily selected set of specimen locations. Such target location information may be used for training when the exact target locations are unknown before the process is run, such as in the case of a process such as inspection, where a relatively large area on the specimen is scanned and errors are predicted only for the reported scanned locations or locations where defects are detected.

[0111] The target location information input to the DL model during training and execution may be in any suitable file or format. For example, the target location information may be coordinates defined relative to the specimen, the output acquisition subsystem, an alignment target on the specimen, an alignment target on an element of the output acquisition subsystem, a design or patterned feature formed on the specimen, etc. In general, it may be useful to train a DL model to predict errors in target location coordinates used in a process performed on the specimen. However, as with any other target location system or method, coordinates of a target location in one space may be transformed to coordinates of a target location in another space in any suitable manner using any suitable coordinate transformation relationship.

[0112] The inventors have numerically proven that the positional errors recorded during movement along the X and Y axes of the stage correlate with the temperature changes within the tool. In particular, the inventors have established a correlation between the various temperatures listed herein and the errors in the X and Y directions. The correlation between the measured temperature and the measured positional error in X for various locations on the sample ("Location A" and "Location B") is shown in Table 1. [Table 1]

[0113] The correlation between the measured temperature and the measured position error in Y for various locations on the sample ("Location A" and "Location B") is shown in Table 2. [Table 2]

[0114] As can be seen from Tables 1 and 2, the inventors found that the correlation assumption holds across locations, with sample variability in the correlation. The correlation between temperature and error in the Y direction is generally stronger than in the X direction. There are two temperatures that are relatively strongly correlated with error in the X direction: carriage temperature and Y plate temperature. In the Y direction, all temperatures show a relatively strong relationship with error.

[0115] In one embodiment, the DL model is configured to predict errors at first and second of the one or more target locations based on at least one of the one or more measured temperatures input into the DL model, and the computer subsystem is configured to determine first and second corrected target positions for the first and second of the one or more target locations by applying the predicted errors at the first and second of the one or more target locations to the first and second of the one or more target locations, respectively. For example, the errors may be predicted in several different ways and then used to determine the corrected target positions. In one example, an overall or global error may be predicted for multiple (or all) target locations on the specimen, and then the same predicted error may be applied to each target location to determine a corrected target position for each target location. However, different errors may be predicted for different target locations or different subsets of target locations, and then a corrected target position for any one target location may be determined using the predicted error for only that target location or only the subset of target locations in which that target location is included. How the error is predicted and used to determine the corrected target position may depend on the configuration of the temperature sensor(s) and the method for measuring the temperature during the scan or process.

[0116] In some cases, different measured temperatures may be used to predict different errors for different targets or subsets of targets. For example, one temperature type may be useful for predicting errors at a target location on one portion of the sample, while another temperature type may be useful for predicting errors at a target location on another portion of the sample. However, in other examples, the same temperature type may be used to predict errors for multiple or all target locations, and some of those predicted errors may be determined using the same temperature type measured at different locations. The different locations may be defined relative to the sample, one or more hardware elements of the output acquisition subsystem, or other reference coordinate system. For example, temperature may be measured by a temperature sensor as the sample is scanned by the output acquisition subsystem, and position information for the temperature measurements may be defined based on the position of the temperature sensor relative to the sample at the time of the temperature measurement, the position of the sample within the output acquisition subsystem at the time of the temperature measurement, etc. In this manner, temperature may be measured as a function of time and position and used to predict errors in the target locations. In such examples, the DL model may be trained using the position information and measurement information generated during the training scan, as described further herein.

[0117] To accurately estimate the position error at the location level, position data must be input to the DL estimator, similar to the correlated temperature data. In another embodiment, the computer subsystem is configured to acquire position information for one or more measured temperatures according to one or more locations at which the one or more measured temperatures are measured, and the position information for at least one of the one or more measured temperatures is input to the DL model by the computer subsystem along with at least one of the one or more measured temperatures. The computer subsystem may acquire the position information in several different ways. One is that a temperature sensor may generate the position information in some way, for example, if it measures the temperature in the system as a function of the temperature sensor's position. In this manner, the computer subsystem may acquire the position information simply by receiving position information from the temperature sensor(s). In another example, the temperature sensor may be in a fixed position relative to the output acquisition hardware. Then, during scanning of the sample, temperature measurements generated by the temperature sensor may be recorded as a function of time, which can be used to correlate the temperature measurements with the position of the sample relative to the output acquisition subsystem hardware (or vice versa) and thus with the temperature sensor. The computer subsystem may input the position information along with the measured temperature(s) into the DL model, as further described herein. A DL model can be trained to use the position information along with the measured temperature(s) to predict the error in the target position, as further described herein.

[0118] In one such embodiment, the DL model is configured to predict errors at first and second of the one or more target locations based on at least one of one or more measured temperatures and location information input to the DL model, and the computer subsystem is configured to determine first and second corrected target locations for the one or more first and second target locations by applying the predicted errors at the first and second of the one or more target locations to the one or more first and second target locations, respectively. The DL model described herein may have great flexibility in predicting errors from measured temperature(s) alone or together with location information. For example, if the input to the DL model includes only temperature information, the corrections during estimation may not be location-based. However, if the input to the DL model during training includes location data (e.g., in X and Y) and temperature, the DL model may provide location-based corrections during estimation. The location information input to the DL model can vary widely, but generally, the location data need not be extremely dense (e.g., thousands or hundreds), as long as reasonably sparse data, such as 100 temperatures measured at different locations, is input.

[0119] Machine learning (ML) can be broadly defined as a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. ML focuses on developing computer programs that can learn, grow, and change on their own when exposed to new data. In other words, ML can be defined as a subfield of computer science that "gives computers the ability to learn without being explicitly programmed." ML explores the study and construction of algorithms that can learn from and make predictions on data; such algorithms overcome strictly static program instructions by making data-driven predictions or decisions through the construction of models from sample inputs.

[0120] The ML described herein may be further implemented as described in "Introduction to Statistical Machine Learning," by Sugiyama, Morgan Kaufmann, 2016, 534 pages; "Discriminative, Generative, and Imitative Learning," Jebara, MIT Thesis, 2002, 212 pages; and "Principles of Data Mining (Adaptive Computation and Machine Learning)," Hand et al., MIT Press, 2001, 578 pages, which are incorporated by reference as if fully set forth herein. The embodiments described herein may be further configured as described in these references.

[0121] Generally speaking, "deep learning" (DL) (also known as deep structural learning, hierarchical learning, or deep ML) is a branch of ML based on a set of algorithms that attempt to model high-level abstractions of data. In simple cases, there may be two sets of neurons: neurons that receive input signals and neurons that send output signals. An input layer receives an input and passes a modified version of the input to the next layer. In DL-based models, there are typically many layers between the input and output (layers are not composed of neurons, although it can be helpful to think of them that way), allowing the algorithm to use multiple processing layers composed of multiple linear and nonlinear transformations.

[0122] In one embodiment, the DL model includes a linear model. For example, the DL model may be a multi-output linear regression. The linear relationship between temperature and positioning error may be described by the following function:

number

[0123] In another embodiment, the DL model includes a neural network. The neural network may be configured as a generative model. A "generative" model may be generally defined as a model that is probabilistic in nature. In other words, a "generative" model is not a model that performs forward simulation or a rule-based approach, and therefore does not require a physical model of the process involved. Instead, as described further herein, a generative model may be trained (its parameters may be learned) based on an appropriate training dataset.

[0124] In one embodiment, the neural network is configured as a deep generative model. For example, the model may be configured to have a DL architecture in that it can include multiple layers that perform numerous algorithms or transformations. In a further embodiment, the neural network may be a deep neural network with a set of weights that model the world according to data input for training. A neural network may generally be defined as a computational method based on a relatively large collection of neural units that loosely model the way biological brains solve problems using relatively large clusters of biological neurons connected by axons. Each neural unit is connected to many other neural units, and the links can influence or inhibit the activation states of the connected neural units. These systems are self-learning and trained rather than explicitly programmed, and excel in areas where solutions or feature detection are difficult to express using traditional computer programs.

[0125] A neural network typically consists of multiple layers, with signal paths traveling from front to back. The goal of a neural network is to solve problems in the same way as the human brain, although some neural networks are much more abstract. Modern neural network projects typically operate with thousands to millions of neural units and millions of connections. A neural network may have any suitable architecture and / or configuration known in the art.

[0126] In further embodiments, the DL model includes a convolutional neural network (CNN). The CNN may include layers of any suitable type, e.g., convolutional, pooling, fully connected, softmax, etc., and the layers may have any suitable configuration known in the art. The CNN may be trained as described herein or in any other suitable manner known in the art. The DL model may include any suitable CNN configuration or architecture known in the art.

[0127] In another embodiment, the computer subsystem is configured to train the DL model using a training set including training inputs and training outputs. The training inputs include one or more temperatures measured by one or more temperature sensors during a training scan, and the training outputs include measurement errors at at least a portion of one or more target locations during the training scan. For example, the computer subsystem may build a DL model recipe for estimating position errors as described herein.

[0128] FIG. 4 shows one flow diagram for training a DL model. These steps may be performed during the setup phase and / or the calibration phase. As shown in FIG. 4, the computer subsystem may generate training inputs in step 400. Generating the training inputs may include two steps. The first step is changing the temperature(s) in the system, as shown in step 402. This step may include, for example, heating and / or cooling a robot that is at least partially responsible for changing and controlling the position of the specimen within the tool. Changing the temperature(s) in the system may be performed in any suitable manner known in the art. Generating the training inputs may also include measuring the temperature(s) in the system, as shown in step 404. For example, intentionally changing the temperature in the system by heating and / or cooling the robot may result in the temperature being measured continuously and may include one or more of the temperatures described herein. In this manner, the computer subsystem may obtain temperature information from the temperature sensor(s) and measure various temperature data within the tool at various time frames. Ideally, the temperatures measured and recorded at this stage would include the temperatures measured by all temperature sensors included in the system.

[0129] In some embodiments, the training input includes location information for one or more temperatures measured during the training scan, corresponding to one or more locations at which the one or more temperatures were measured during the training scan. For example, as shown in step 404, measuring a temperature(s) in the system may include measuring the temperature(s) as a function of location within the system or as a function of location relative to the sample. Thus, the results of measuring a temperature(s) in the system during the training scan may include the measured temperature(s), location information associated with the measured temperature(s), and any other data generated during the temperature measurement. In this manner, generating the training input may include any information associated with the temperature measurement, including the measured temperature itself and other data such as location, time, and other information. Thus, the training input may include X and Y location data and the measured temperature. All of this information may be included in the training input, which is added to a training set used to train the DL model.

[0130] As shown in step 406, the computer subsystem may generate training outputs. This step may include multiple steps, including moving over known locations on the sample, as shown in step 408, and determining the offset (error) from the measurement location to the target location, as shown in step 410. These steps are intended to measure the robot movement error by moving over known locations on the sample and finding the offset (or position error) from the measurement location to the target location. This step may also include collecting stage-sample map data at different time frames. For example, the error may be measured using repeated stage maps, which may be performed using a stage mapping application described further herein. Differences between successive stage maps while the tool is changing temperature may be due to thermal differences and are indicative of thermal errors in positioning the target location. The position information used for training is input into the DL model as described above and may include any of the position information described above.

[0131] The generation of training inputs and training outputs can be repeated n times depending on how much data is expected to be needed to train the DL model. The amount of data needed for training may vary from tool to tool and from DL model architecture to DL model architecture. For example, if the DL model uses the linear relationship above, in the simplest case, training involves finding the coefficients c that minimize the sum of squared errors when predicting δx from the calibration data. i To perform this least-squares fit, at least n calibration data points may need to be fitted. Generally speaking, training of the DL model may be performed, at least initially, using training data that can be reasonably or practically collected using existing temperature sensors or an initial temperature sensor configuration. The testing and validation phase of training will indicate whether the amount of training data is sufficient. If not, additional training data can then be collected as described above, and the additional training data can be used for retraining or added to the initial training data for retraining.

[0132] The result of generating the training inputs and training outputs is a training set 412, which may then be used to train a DL model in step 414. Training may be performed in this step as further described herein, but generally includes training the DL model to estimate position errors from inputs including the intended target location and temperature. In addition, training may include training the DL model to output position corrections at the location level. Depending on how well the trained DL model performs in testing and validation or use, the computer subsystem may measure a position correction map in an unknown time frame and improve the sample placement accuracy using the errors predicted by the DL model. This step may be performed before the trained DL model is released for use and / or after the DL model has been used for a period of time for monitoring and / or recalibration purposes. The training step may generate a trained DL model 416, which may be used as further described herein.

[0133] In one embodiment, at least one of the one or more measured temperatures input by the computer subsystem into the DL model includes all of the one or more temperatures measured by all of the one or more temperature sensors during a process performed by the output acquisition subsystem on the specimen. For example, the temperature data is used to calculate the position correction D x ,D y To demonstrate the effect of temperature on accuracy, two different ML models were tested: Model A is a function of all temperatures and locations X and Y, and Model B is a function of only the two highly correlated temperatures and locations X and Y.

number

[0134] Error maps (contour plots of error in X and Y) of the position correction for the two different models were generated but are not included here due to issues of reproducibility and clarity. The error maps show that the estimated accuracy of the position correction on unknown timeframe data (i.e., blind test data, not used to train the DL model but used to show that the trained model actually meets the performance criteria on new data) is degraded compared to when all temperature information is used.

[0135] Training a DL model using all temperatures measured by all temperature sensors can be advantageous for several reasons. One is that the mechanisms by which temperature affects position accuracy are not necessarily well understood with the tools described herein. This is also one of the reasons why DL models are such an advantageous method for predicting temperature-induced position errors. More particularly, DL models as described herein can learn relationships between temperature and position errors that are not currently understood. In other words, DL models learn how to predict target position errors from temperature measurements without requiring physics- and / or thermodynamics-based models.

[0136] Additionally, by using all available temperature measurements from all available temperature sensors for training, the DL model can learn which temperature measurements from which sensors are useful for predicting target position errors. In this way, the user does not need to establish which temperatures are useful and which are not, and it is fine for non-useful temperature measurements to be input into the DL model as long as data collection from the temperature sensors does not interfere with the tool's performance. Thus, by utilizing existing temperature sensors and / or by adding several additional temperature sensors in judicious locations (e.g., near or inside hardware elements responsible for controlling the sample position, rather than at every possible location within the tool), the DL model itself can learn which temperature measurements are useful and which are not.

[0137] The parameters of the trained DL model may indicate which temperature measurements are useful and which are not when temperature measurement data collection is a concern. For example, by examining the DL model weights assigned to various inputs corresponding to various temperature sensors, a user may identify temperature measurements that are not being used or are of little use in predicting target position errors and optionally choose to remove them from the DL model inputs. Thus, the embodiments described herein advantageously eliminate the need for a user to individually verify many possible paths from temperature changes to sample positioning errors. Instead, all temperatures that may have an impact on positioning errors may be measured collectively and corrected all at once.

[0138] Despite the fact that the DL model eliminates the need for science-based models to predict positioning error from temperature measurements, the following examples of how temperature changes can affect positioning error were established by the inventors for some of the commercially available tools used in the experiments and are included herein to help understand how temperature can affect positioning accuracy and why the embodiments described herein are important for tools such as those described herein. Two of the most important connections identified between temperature changes in microscope components and microscope positioning accuracy are 1) the deflection of the Y-plate (part of the sample-positioning robot) relative to the Y-encoder and 2) the displacement of the optical system's support structure relative to the sample handling stage. In the first example, the Y-plate (a component that supports the movement of the sample aligned with the tool's Y-axis) expands when heated, but is constrained in some locations where it is fixed to other components. Because the Y-plate is restricted in expansion, it may bend off-center, causing it to become misaligned with the encoder (a component used to measure the motor's position along this axis). In the second example, the support structure for the camera that measures its position on the sample may move / deform slightly as temperature changes. The positioning of the sample is measured using a camera (e.g., by imaging a defined reference mark on the sample and using image processing to calculate where that mark is relative to where it should be), and any relative movement between the camera and the sample handling stage is converted into a positioning error.

[0139] The inventors have also discovered that the quality of the training data can affect the performance of a DL model. For example, a DL model can be affected by different approaches or tools used to collect stage map data (i.e., data quality issues). By improving data quality, very large errors in DL error predictions, which may occur, for example, near the edge of a sample, can be brought within a specified target range for the prediction error. Thus, in some cases, training a DL model may include initial training using appropriate training data, testing the initially trained DL model, and determining whether all error ranges on the sample map are within specification or whether all locations on the sample are within specification. If the error ranges are out of specification, additional training data may be generated, which may include training data for the entire sample or only training data for the portion of the sample where the out-of-specification error ranges were found. The DL model may then be retrained using the additional training data and tested again to confirm whether all predicted error ranges are within specification. Thus, training a DL model may include one or more cycles of training, testing, and validation, each cycle being performed using different or additional training data.

[0140] The DL model may or may not be trained by the computer subsystem and / or one of the component(s) executed by the computer subsystem. For example, another method or system may train the DL model, which may then be stored for use as one of the component(s) executed by the computer subsystem. In this manner, the DL model may be trained on-tool or off-tool. In either case, training may include inputting training inputs to the DL model and changing one or more parameters of the DL model until outputs produced by the DL model match (or substantially match) the training outputs. Training may include changing one or more trainable parameters of the DL model. The one or more parameters of the DL model that are trained may include one or more weights of any layer of the DL model that has trainable weights. In one such example, the weights may include weights of convolutional layers but not weights of pooling layers.

[0141] FIG. 5 illustrates steps that may be performed during runtime by embodiments described herein. Temperature measurements, error predictions, and determination of corrective target positions may all be performed on the tool to monitor offset changes and specimen placement improvements. As shown in FIG. 5, the computer subsystem may calculate the target position(s), as shown in step 500. The target positions may or may not need to be calculated in this step, depending on how they are input to the computer subsystem. For example, in some cases, the computer subsystem actually calculates the target positions during runtime. In such cases, the process recipe includes the target positions expressed in specimen coordinates, which the computer subsystem converts to specimen handling coordinates. This step may include, for example, a 30° rotation of the axes and one or more other alignment techniques described further herein.

[0142] Temperature sensor(s) measure temperature(s) within the system, as shown in step 502. The temperature sensor(s) may measure temperatures on-the-fly, as described further herein. The calculated target location(s) and measured temperature(s) are input to a trained DL model 504, which outputs a predicted error for the target location, as shown in step 506. In this manner, the trained DL model can be applied to estimate the expected error when moving to nearby locations.

[0143] The computer subsystem is configured to determine a corrected target position for at least one of the one or more target positions by applying the prediction error to at least one of the one or more target positions. For example, as shown in step 508 of FIG. 5 , the computer subsystem may determine the corrected target position by applying the prediction error from step 506 to the target position calculated in step 500. Applying the prediction error to the at least one target position may include, for example, subtracting the prediction error from the target position, but may include more advanced functions if simple subtraction does not produce a sufficiently accurate result. Determining the corrected target position step may be performed by the computer subsystem using an algorithm, which may be part of one or more components executed by the computer subsystem or may be separate from those components.

[0144] In some embodiments, the computer subsystem is configured to determine a corrected target position for a second of the one or more target positions by interpolating the prediction error determined for at least one of the one or more target positions to the second of the one or more target positions and applying the interpolated prediction error to the second of the one or more target positions. For example, as shown in step 510, the computer subsystem may interpolate the prediction error for one target position to another target position. The computer subsystem may also determine a corrected target position by applying the interpolated prediction error to the other target position, as shown in step 512. In this manner, the computer subsystem may be configured to interpolate error estimates from nearby location(s) to the target position. Interpolating the prediction error from one location to another may be performed in any suitable manner known in the art. Determining the corrected target position from the interpolated prediction error may be performed in the same manner as described herein for the prediction error.

[0145] The correction target positions determined with or without interpolation may be combined into a set of correction target positions, as shown in step 514, which may then be used for position correction in one or more of the various ways described herein.

[0146] In one embodiment, the computer subsystem is configured such that after determining the correction target positions, the output acquisition subsystem generates an output at at least one of the one or more target positions. For example, as shown in step 516 of FIG. 5 , the system may move to the correction target positions and generate an output. In this manner, the system may move to the correction target positions, e.g., target position-prediction error (or interpolated prediction error), after which the output acquisition subsystem may perform imaging, scanning, measurement, etc., at the correction target positions. This sequence of steps may be most useful in move-acquire-measure type systems, where measurements are intended to be performed at specific, discrete locations on the sample and do not involve scanning a relatively large area on the sample.

[0147] In one such example, assume that a process involves measuring the space between two specific features of a die formed on a wafer and then measuring the dimensions of another feature spaced from the two specific features. In this example, ideally, the scanning subsystem moves to a location of the space between the two specific features, the output acquisition subsystem generates output at that location, and then the scanning subsystem moves the specimen to position the next feature or features to be measured within the field of view of the output acquisition subsystem. If the target position is not temperature-compensated and a measurement is performed on the space between the two features, the measurement may be performed at a different location. The process results may then be missing that measurement or may include a measurement for a different location that could be misinterpreted as a measurement of the space between the two features. The process results may then be incomplete or erroneous, which is disadvantageous for several obvious reasons. Therefore, compensating the target position for positioning errors caused by temperature changes, as described herein, can be advantageously used to prevent problems such as incomplete or erroneous specimen measurements and other quality control-related process results.

[0148] In another embodiment, the computer subsystem is configured such that the output acquisition subsystem generates an output at at least one of the one or more target locations before determining the corrected target location. For example, as shown in step 518 of FIG. 5 , the computer subsystem may modify the target location reported for a previously generated output. In this manner, the output acquisition subsystem may generate an output for target location 1 on the specimen, the DL model may predict an error at that target location from one or more input temperature measurements (or the error may be determined, e.g., interpolated, from predicted errors for nearby target locations), and the coordinates of target location 1 may then be corrected using the predicted error.

[0149] Using the output acquisition subsystem to generate output on the specimen and then correcting the coordinates of the target location where the output was generated may be more appropriate for applications such as inspection, in which a relatively large area on the specimen is scanned. For example, during inspection, the coordinates of any defects detected during the process may be recorded, the DL model may predict an error in the recorded defect location based on the temperature measured at or near where the defect was detected, and the reported defect coordinates may be corrected using the predicted error. Such position correction may be less useful in applications where it is important to generate output at specific, individual locations on the specimen, as it is more advantageous to correct the target location prior to output generation and ensure that specific locations are imaged, measured, inspected, defect reviewed, etc.

[0150] In one embodiment, the DL model is configured to predict errors at one or more additional target locations on the specimen or additional specimens, where additional outputs are generated by additional output acquisition subsystems of the additional systems, based on one or more additional temperatures measured by one or more additional temperature sensors of the additional systems that are input into the DL model by the computer subsystem. For example, if the DL model is trained for multiple tools, the DL model may be used to estimate specimen-level position errors and corrections for each tool. In this case, each output acquisition subsystem and other components of the system, such as temperature sensors, may be configured as described herein. Each system may have the same or at least a similar configuration. For example, each system may be configured as shown in FIG. 3. Training of the DL model for multiple tools may be performed as described herein.

[0151] In further embodiments, the training input includes one or more additional temperatures measured by one or more additional temperature sensors included in the additional system during an additional training scan of the sample or additional sample performed in the additional system, and the training output includes additional measurement errors at one or more additional target locations on the sample or additional sample, the additional target locations being locations at which outputs are generated by an additional output acquisition subsystem of the additional system during the additional training scan. For example, in the more general case, a DL model may be trained with stage map and temperature data from multiple tools simultaneously. Training a DL model for multiple tools may include preparing sample stage map data from the multiple tools and training the DL model on all of the tools. In this way, if training data is generated on multiple tools to train a model (a single model), the same model may be used to estimate position errors for all of the tools.

[0152] In some embodiments, the computer subsystem is configured to monitor the drift of the system and the additional system relative to each other based on the predicted error and the additional predicted error. In this manner, embodiments described herein can be configured to estimate tool drift in position error based on the error predicted by the DL model from the input measured temperature(s). For example, if the DL model is trained with stage map and temperature data from multiple tools simultaneously, a user can estimate how much a particular tool is drifting by estimating the amount of position correction for each tool.

[0153] In another embodiment, the computer subsystem is configured to calibrate the system and the additional system relative to one another based on the predicted error and the additional predicted error. In this manner, the embodiments described herein may be configured to perform inter-tool calibration. For example, if a DL model is trained simultaneously with stage map and temperature data from multiple tools, a user may estimate inter-tool calibration by estimating the amount of position correction for each tool.

[0154] The computer subsystem is also configured to generate results including the determined information, which may include any of the results or information described herein. The results of the information determination may be generated by the computer subsystem in any suitable manner. All embodiments described herein may be configured to store the results of one or more steps of the embodiment on a computer-readable storage medium. The results may include any of the results described herein and may be stored in any manner known in the art. The results including the determined information may have any suitable form or format, such as a standard file type. The storage medium may include any storage medium described herein or any other suitable storage medium known in the art.

[0155] After the results are stored, they may be accessed in the storage medium, used by any of the method or system embodiments described herein, formatted for display to a user, used by another software module, method, system, etc., to perform one or more functions on the specimen or another specimen of the same type. For example, the measured temperature, predicted error, and correction target position may be stored and used as described herein or in any other suitable manner. Additionally, depending on the process performed on the specimen by the system, the results may include any information about the specimen determined from output generated for the specimen by the output acquisition subsystem. Such results generated by the computer subsystem may include information about any defects detected on the specimen, such as the location of a bounding box for the detected defect, information about the defect classification, such as a detection score, a class label or ID, any defect attributes determined from any image, etc., measurements, dimensions, shapes, etc. of specimen structures, or any other suitable information known in the art. The information may be used by the computer subsystem or another system or method to perform additional functions on the specimen and / or the detected defects, such as sampling the defects for defect review or other analysis, determining the root cause of the defects, etc.

[0156] Such functionality also includes, but is not limited to, modifying a process, such as a manufacturing process or step, performed or being performed on the specimen in a feedback or feedforward manner. For example, the computer subsystem may be configured to determine one or more modifications to a process performed on and / or to be performed on the specimen based on the determined information. The modifications to the process may include any suitable modifications to one or more parameters of the process. In one such example, the computer subsystem may desirably determine the modifications so that the defect may be mitigated or prevented on other specimens on which the modified process is performed, so that the defect may be corrected or eliminated in a specimen of another process performed on the specimen, so that the defect may be compensated for in another process performed on the specimen, etc. The computer subsystem may determine such modifications in any suitable manner known in the art.

[0157] These changes may then be transmitted to a semiconductor manufacturing system (not shown) or to a storage medium (not shown) accessible to both the computer subsystem and the semiconductor manufacturing system. The semiconductor manufacturing system may or may not be part of the system embodiments described herein. For example, the output acquisition subsystem and / or the computer subsystem described herein may be coupled to the semiconductor manufacturing system via one or more common elements, such as a housing, a power supply, a sample handling device or mechanism, etc. The semiconductor manufacturing system may include any semiconductor manufacturing system known in the art, such as a lithography tool, an etching tool, a chemical-mechanical polishing (CMP) tool, a deposition tool, etc.

[0158] The embodiments described herein provide several advantages in addition to those already mentioned. For example, the embodiments described herein correct the sample position without the need for external metrology tools. In particular, once the DL model is trained, the only data required for position correction is information about the sample and measured temperature information. No other external tools or methods are required to predict position errors, which is a major advantage of the DL model described herein. Additionally, once the DL model is trained, inferences over different time frames are essentially instantaneous. Furthermore, the DL models and embodiments described herein can be generalized across multiple tools.

[0159] While the embodiments described herein do not require an external metrology tool for specimen position correction, the embodiments described herein may be used in combination with any other position control / correction method or subsystem available in the tool. One such method may be a stage mapping application. In stage mapping, during tool calibration, position errors are recorded when moving to multiple positions on the specimen in an open loop (this map of positions and errors is called a "stage map"), and these (invariant) error measurements can then be corrected during runtime. Another such method is the alignment technique, in which each time a specimen is loaded into the tool, the scanning subsystem moves to several reference marks near the center of the specimen and across the Y-axis, measures the specimen's exact angle relative to the chuck, and takes that angle into account when moving across the specimen. A further such method is the navigation correction technique, in which the positions of several reference marks around the edge of the specimen are recorded and used to calculate Euclidean transformations (e.g., translation, scale, rotation, tilt) for subsequent movements.

[0160] The embodiments of each of the systems described above may be combined into one single embodiment.

[0161] Another embodiment relates to a method for determining information about a specimen. The method includes measuring one or more temperatures in a system using one or more temperature sensors. The system includes an output acquisition subsystem configured to generate a specimen output at one or more target locations on the specimen. The method includes predicting an error at at least one of the one or more target locations by inputting at least one of the one or more measured temperatures into a DL model included in one or more components executed by a computer system. Additionally, the method includes determining a corrected target location for at least one of the one or more target locations by applying the predicted error to at least one of the one or more target locations. This inputting and determination is performed by the computer system.

[0162] Each of the steps of the method may be performed as further described herein. The method may also include any other step(s) that may be performed by the system, temperature sensor(s), output acquisition subsystem, DL model, and computer system described herein. The system, temperature sensor(s), output acquisition subsystem, DL model, and computer system may be configured according to any of the embodiments described herein. The method may be performed by any of the system embodiments described herein.

[0163] An additional embodiment relates to a non-transitory computer-readable medium having stored thereon program instructions executable on a computer system to perform a computer-implemented method for determining information about a sample. One such embodiment is shown in FIG. 6. In particular, as shown in FIG. 6, a non-transitory computer-readable medium 600 includes program instructions 602 executable on a computer system(s) 604. The computer-implemented method includes obtaining one or more temperatures measured within the system using one or more temperature sensors. The system is configured as further described herein. The computer-implemented method also includes the predicting and determining steps described above. The computer-implemented method may further include any step(s) of any method(s) described herein.

[0164] Program instructions 602 implementing methods such as those described herein may be stored on a computer-readable medium 600. The computer-readable medium may 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.

[0165] The program instructions may be implemented in any of a variety of ways, including procedure-based, component-based, 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"), SSE (Streaming SIMD Extensions), or other technologies or methodologies, as appropriate.

[0166] The computer system(s) 604 may be configured according to any of the embodiments described herein.

[0167] Further modifications and alternative embodiments of various aspects of the invention will be apparent to those skilled in the art in view of this description. For example, methods and systems for determining information about a sample are provided. Accordingly, this description is to be construed as merely illustrative and is intended to teach those skilled in the art the general manner of carrying out the invention. It is to be understood that the forms of the invention shown and described herein are to be taken as the presently preferred embodiments. Elements and materials may be substituted for those shown and described herein, parts and processes may be reversed, and certain features of the invention may be utilized independently, all as would be apparent to one skilled in the art having the benefit of this description of the invention. Changes may be made in 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 determine information about a sample, comprising: an output acquisition subsystem configured to generate an output of the specimen at one or more target locations on the specimen; one or more temperature sensors configured to measure one or more temperatures within the system; a computer subsystem configured to obtain the one or more temperature measurements from the one or more temperature sensors; and One or more components executed by the computer subsystem Equipped with wherein the one or more components include a deep learning model configured to predict an error in at least one of the one or more target locations based on at least one of the one or more measured temperatures input into the deep learning model by the computer subsystem; the computer subsystem is further configured to determine a corrected target position for the at least one of the one or more target positions by applying the prediction error to the at least one of the one or more target positions. system.

2. 10. The system of claim 1, wherein the one or more temperature sensors are further configured to measure the one or more temperatures during a process performed on the specimen using the output acquisition subsystem.

3. 10. The system of claim 1, wherein the one or more temperature sensors are further configured to measure the one or more temperatures of or within one or more hardware elements of the system configured to alter the position of the specimen within the system.

4. 10. The system of claim 1, wherein the at least one of the one or more measured temperatures input by the computer subsystem to the deep learning model includes all of the one or more temperatures measured by all of the one or more temperature sensors during a process performed in the output acquisition subsystem of the specimen.

5. 2. The system of claim 1, wherein the computer subsystem is further configured such that after determining the corrected target location, the output acquisition subsystem generates the output at the at least one of the one or more target locations.

6. The system of claim 1 , wherein the computer subsystem is further configured such that, prior to determining the corrected target position, the output acquisition subsystem generates the output at the at least one of the one or more target positions.

7. 2. The system of claim 1, wherein the deep learning model is further configured to predict the errors at first and second of the one or more target locations based on the at least one of the one or more measured temperatures input to the deep learning model, and the computer subsystem is further configured to determine first and second corrected target locations for the first and second of the one or more target locations by applying the predicted errors at the first and second of the one or more target locations to the first and second of the one or more target locations, respectively.

8. 2. The system of claim 1, wherein the computer subsystem is further configured to obtain location information for the one or more measured temperatures responsive to one or more locations at which the one or more measured temperatures were measured, and wherein the location information for the at least one of the one or more measured temperatures is input into the deep learning model by the computer subsystem along with the at least one of the one or more measured temperatures.

9. 9. The system of claim 8, wherein the deep learning model is further configured to predict the errors at first and second of the one or more target locations based on at least one of the one or more measured temperatures and the location information input to the deep learning model, and the computer subsystem is further configured to determine first and second corrected target locations for the first and second of the one or more target locations by applying the predicted errors at the first and second of the one or more target locations, respectively, to the first and second of the one or more target locations.

10. 2. The system of claim 1, wherein the computer subsystem is configured to determine a corrected target position for the second of the one or more target positions by interpolating the prediction error determined for the at least one of the one or more target positions to the second of the one or more target positions and applying the interpolated prediction error to the second of the one or more target positions.

11. 2. The system of claim 1, wherein the computer subsystem is further configured to train the deep learning model using a training set including training inputs and training outputs, the training inputs including the one or more temperatures measured by the one or more temperature sensors during a training scan, and the training outputs including measurement errors at at least a portion of the one or more target locations during the training scan.

12. 12. The system of claim 11, wherein the training input further includes location information of the one or more temperatures measured during the training scan, where the one or more temperatures correspond to one or more locations at which the one or more temperatures were measured during the training scan.

13. 12. The system of claim 11, wherein the training input further includes one or more additional temperatures measured by one or more additional temperature sensors included in an additional system during an additional training scan of the sample or additional sample performed in the additional system, and the training output further includes additional measurement errors at one or more additional target locations on the sample or additional sample, the additional target locations being locations at which outputs are generated by an additional output acquisition subsystem of the additional system during the additional training scan.

14. 2. The system of claim 1, wherein the deep learning model is further configured to predict additional errors at one or more additional target locations on the sample or additional samples, where additional outputs are generated by an additional output acquisition subsystem of an additional system, based on one or more additional temperatures measured by one or more additional temperature sensors of the additional system that are input to the deep learning model by the computer subsystem, and wherein the computer subsystem is further configured to monitor drift of the system and the additional system relative to each other based on the predicted errors and the additional predicted errors.

15. 2. The system of claim 1, wherein the deep learning model is further configured to predict additional errors at one or more additional target locations on the sample or additional samples, where additional outputs are generated by an additional output acquisition subsystem of an additional system, based on one or more additional temperatures measured by one or more additional temperature sensors of the additional system that are input to the deep learning model by the computer subsystem, and wherein the computer subsystem is further configured to calibrate the system and the additional system relative to each other based on the predicted errors and the additional predicted errors.

16. The system of claim 1 , wherein the deep learning model comprises a linear model.

17. The system of claim 1 , wherein the deep learning model comprises a neural network.

18. The system of claim 1 , wherein the deep learning model comprises a convolutional neural network.

19. The system of claim 1 , wherein the output acquisition subsystem is further configured as a metrology subsystem.

20. The system of claim 1 , wherein the output acquisition subsystem is further configured as an inspection subsystem.

21. The system of claim 1 , wherein the output acquisition subsystem is an optical-based subsystem.

22. The system of claim 1 , wherein the output acquisition subsystem is an electronic-based subsystem.

23. A non-transitory computer-readable medium storing program instructions executable on a computer system to perform a computer-implemented method for determining information about a sample, the computer-implemented method comprising: acquiring one or more temperatures measured in a system using one or more temperature sensors, the system including an output acquisition subsystem configured to generate an output of the specimen at one or more target locations on the specimen; predicting an error in at least one of the one or more target locations by inputting at least one of the one or more measured temperatures into a deep learning model included in one or more components executed by the computer system; and determining a corrected target position for the at least one of the one or more target positions by applying the prediction error to the at least one of the one or more target positions; 1. A non-transitory computer-readable medium, comprising:

24. 1. A method for determining information about a sample, comprising: measuring one or more temperatures in a system using one or more temperature sensors, the system including an output acquisition subsystem configured to generate an output of the sample at one or more target locations on the sample; predicting an error in at least one of the one or more target locations by inputting at least one of the one or more measured temperatures into a deep learning model included in one or more components executed by a computer system; and determining a corrected target position for the at least one of the one or more target positions by applying the prediction error to the at least one of the one or more target positions, wherein the inputting and determining are performed by the computer system; A method comprising:

Images