Method for converting metrology data

Abstract

A metrology system and a method for converting metrology data via a trained machine learning (ML) model are described herein. The method includes accessing a first scanning electron metrology (SEM) data set (e.g., images, profile lamps) acquired by a first SEM system and a second SEM data set acquired by a second SEM system, wherein the first SEM data set and the second SEM data set are associated with a patterned substrate. Using the first SEM data set and the second SEM data set as training data, a machine learning (ML) model is trained such that the trained ML model is configured to convert a metrology data set acquired by the second SEM system to a converted data set having characteristics comparable to metrology data acquired by the first SEM system. Further, a measurement can be determined based on the converted SEM data.

Description

TECHNICAL FIELD

[0001] The description herein generally relates to processing metrology data acquired by a metrology system, and more particularly to processing metrology data using a machine learning model. BACKGROUND

[0002] Lithographic projection apparatuses can be used, for example, in the manufacture of integrated circuits (ICs). In such cases, a patterning device (e.g., a mask) can contain or provide the desired pattern (‘design layout’) corresponding to a single layer of the IC and this pattern can be transferred to a target portion (e.g., comprising one or more dies) on a substrate (e.g., silicon wafer), on which the pattern is imprinted, e.g., by illuminating the pattern on the patterning device onto a target portion. Typically, multiple substrates are imprinted together, so as to make full use of the size of the target portion. Transferring the patterned radiation onto the substrate can involve a number of parameters, such as the wavelength of the radiation used, the pattern provided by the patterning device, the design layout, the number of patterning devices used, the type of substrate used, the type of radiation used, and so on. The parameters are typically chosen in order to make the best possible imprint of the pattern on the substrate. Generally, the pattern on the patterning device is transferred to the target portion at a desired scale. In other words, the pattern on the patterning device is reduced in size by a factor of M (e.g., M = 4) and transferred to the target portion. More information on lithographic apparatuses can be gleaned, for example, from US 6, 046,792, incorporated herein by reference.

[0003] Before the pattern is transferred from the patterning device to the substrate, it can undergo various procedures, such as coating, resist coating and soft baking. After exposure, the substrate can undergo other procedures (‘post-exposure procedures’), such as post-exposure baking (PEB), development, hard baking and measurement / inspection of the transferred pattern. This procedure array is used as a basis to produce a device (e.g., an IC), which has one or more layers that are patterned on a substrate. Then, the substrate can go through various processes, such as etching, ion implantation (doping), metallization, oxidation, chemical mechanical polishing, etc., all intended to finish the device in the substrate. If multiple layers are needed in the device, the whole procedure or variations thereof are repeated for each layer. Finally, the device present in the substrate is separated from the substrate, typically by cutting or sawing, etc., and individual devices are obtained.

[0004] Therefore, manufacturing devices such as semiconductor devices typically involve using many manufacturing processes to process a substrate (e.g., a semiconductor wafer) to form various features and multiple layers of the device. Such layers and features are typically created using, for example, deposition, lithography, etching, chemical-mechanical polishing, and ion implantation. Multiple devices can be made on a plurality of dies on the substrate and then separated into individual devices. The device manufacturing process can be considered a patterning process. Patterning processes involve a patterning step such as optical and / or euvlithography lithography using a patterning device in a lithography apparatus to transfer a pattern on the patterning device to a substrate, and typically but optionally involve one or more associated pattern processing steps such as resist development by a development apparatus, baking of the substrate using a bake tool, etching using an etching apparatus using the pattern, etc. SUMMARY

[0005] In semiconductor manufacturing, multiple metrology tools operating with similar measurement principles or similar mechanisms of generating signals from a patterned substrate can be employed. For example, multiple scanning electron microscopes (SEMs) that determine measurement values based on interactions of an electron beam with a patterned substrate can be used during semiconductor manufacturing. Multiple metrology systems can be used to improve metrology throughput, measurement accuracy, cost efficiency, or for other reasons. For example, measurement values can be used to improve a patterning process. While operating similarly, one metrology system can differ from another metrology system in various aspects, such as mechanical subsystems, electrical subsystems, electronics, signal detection, image acquisition algorithms, image processing algorithms, profile extraction algorithms, or other structures and software. Therefore, using multiple metrology systems can result in inconsistent measurement results for the same pattern on the same substrate. The present disclosure provides a mechanism to convert measurement data (e.g., SEM images or profiles) obtained from a particular metrology tool using a trained machine learning model. The converted measurement data matches or is comparable to another metrology tool (e.g., a reference metrology system), allowing measurement consistency between different metrology tools. For example, the converted signal and / or CD associated with features of a patterned substrate matches the signal and / or CD obtained by another metrology tool. In other words, the converted measurement data or measurement values derived from the converted measurement data is comparable to obtaining measurement data or measurement values using another metrology tool.

[0006] The metrology system can be, for example, a plurality of scanning electron microscopes (SEMs). In one embodiment, the mechanisms herein include training a machine learning (ML) model to convert SEM images acquired by one SEM system to images as if acquired by another SEM system. Thus, physical property measurements of patterned features performed on the converted images will be similar to measurements performed on images acquired by the other SEM system. In one embodiment, differences between measurement data from different tools (e.g., CD mismatch and / or SEM image mismatch) can be incorporated into a cost function for training the ML model. In an example, measurement data including SEM signals can be used to guide the image conversion. Further, the mechanisms involve obtaining measurement settings (e.g., CD measurement settings) using the training data and applying it to the converted images to obtain measurement values (e.g., CD) associated with the patterned substrate. These measurement values are as if acquired by another SEM system (e.g., a reference SEM system).

[0007] In one embodiment, a method for training a machine learning model and converting metrology data using the trained machine learning model is provided. The method includes accessing a first scanning electron metrology (SEM) data set acquired by a first SEM system and a second SEM data set acquired by a second SEM system, wherein the first SEM data set and the second SEM data set are associated with a patterned substrate. Using the first SEM data set and the second SEM data set as training data, a machine learning (ML) model is trained such that the trained ML model is configured to convert a metrology data set acquired by the second SEM system to a converted data set having properties comparable to metrology data acquired by the first SEM system.

[0008] In one embodiment, the first SEM data set and the second SEM data set can be a set of images of the patterned substrate, profiles of features on the patterned substrate, physical properties associated with a pattern on the patterned substrate, or a combination thereof. In one embodiment, the physical properties include a critical dimension (CD) of the pattern on the patterned substrate.

[0009] In one embodiment, training the ML model involves: comparing the first SEM data set and the second SEM data set; and adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model.

[0010] In one embodiment, training the ML model involves signal-to-signal matching or CD-to-CD matching. For example, training the ML model involves: comparing first CD values of the first SEM data set and second CD values of the second SEM data set; and adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model as a function of the first CD values and the second CD values to improve CD matching between the first SEM data set and the second SEM data set.

[0011] In one embodiment, the method can further involve: receiving a metrology measurement recipe for the metrology system based on the first SEM data set and a physical property measurement of the substrate from the first SEM system; and applying the metrology measurement recipe to the converted metrology data to determine another physical property measurement. In one embodiment, the metrology measurement recipe includes a CD threshold that indicates where a CD measurement on the captured metrology data is taken.

[0012] In one embodiment, determining the metrology measurement recipe involves: extracting a contour from an image of the first SEM data set via a first contour extraction algorithm; drawing a cut line at a location on the contour to measure a CD; and determining a CD threshold corresponding to the measured CD based on a signal along the cut line.

[0013] In one embodiment, a metrology system is provided. The metrology system includes a process or computer system that includes one or more processors having a trained machine learning (ML) model stored thereon and programmed with computer program instructions that, when executed, cause the computer system to: capture metrology data of a patterned substrate; and convert the captured metrology data to converted metrology data via the trained ML model, the converted metrology data having characteristics as if captured by another metrology system.

[0014] According to embodiments, a computer system is provided that includes a non-transitory computer readable medium having instructions recorded thereon. When executed by a computer, the instructions implement the above method steps. BRIEF DESCRIPTION OF DRAWINGS

[0015] The above aspects and other aspects and features will become apparent to those of ordinary skill in the art from the following description, taken in conjunction with the accompanying drawings, in which:

[0016] Figure 1 A block diagram showing various subsystems of a lithography system according to embodiments is shown;

[0017] Figure 2This is an exemplary flowchart of a method for converting an image acquired by a particular measurement system into an image as if acquired by another measurement system, according to an embodiment.

[0018] Figure 3 The illustration shows exemplary measurement data acquired by a first measurement system and a second measurement system according to an embodiment;

[0019] Figure 4 The illustration shows an embodiment of the invention. Figure 3 Signals within a portion of the image obtained from the image;

[0020] Figure 5 The illustration shows an exemplary training of an ML model using a generator adversarial network (GAN) according to an embodiment;

[0021] Figure 6A The illustration shows the input of a second image acquired by a second measurement system into an input according to an embodiment. Figure 2 The trained model is used to generate the transformed image;

[0022] Figure 6B The illustration shows an exemplary first image acquired by a first measurement system according to an embodiment, the exemplary image being positioned to... Figure 6A The transformed images are adjacent for comparison;

[0023] Figure 7 The illustration shows an embodiment. Figure 6A and 6B The signals within a portion of the first image, the second image, and the transformed image, wherein the signals of the first image and the transformed image overlap each other, indicating that the transformed image has similar characteristics to the first image;

[0024] Figure 8 This is a block diagram of an exemplary training of a machine learning (ML) model according to an embodiment and the application of the trained ML model to determine measurements, the machine learning (ML) model being configured to convert an image acquired by a second measurement system into an image as acquired by a first measurement system;

[0025] Figure 9 An embodiment of a scanning electron microscope (SEM) according to an example is schematically depicted;

[0026] Figure 10 An embodiment of an electron beam inspection apparatus according to an embodiment is schematically depicted; and

[0027] Figure 11 This is a block diagram of an example computer system according to an embodiment. Detailed Implementation

[0028] Before describing the embodiments in detail, it is helpful to present an example environment in which the embodiments can be implemented.

[0029] Although specific reference can be made in this text to the manufacture of ICs, it should be explicitly understood that the description herein has many alternatives that can be practiced. For example, it can be

[0030] In the present document, the terms“radiation” and“beam” can be used to encompass all types of electromagnetic radiation, including ultraviolet radiation (e.g. with a wavelength of 365, 248, 193, 157 or 126 nm) and EUV (extreme ultra-violet radiation, e.g. having a wavelength in the range of about 5-100 nm).

[0031] The patterning device can comprise or can form one or more design layouts. The design layout can be generated using a CAD (computer-aided design) program, a process that is often referred to as EDA (electronic design automation). Most CAD programs follow the same general routine of creating a functional design layout / patterning device. These routines are set by processing and design limitations. For example, design rules define the space tolerance between devices such as gates, capacitors, etc. or interconnect lines, to ensure that the devices or lines do not interact in an undesirable way. One or more design rule limitations can be referred to as“critical dimensions” (CDs). A critical dimension of a device can be defined as the smallest width of a line or hole or the smallest space between two lines or two holes. Thus, the CDs determine the overall size and density of the designed devices. Of course, one of the goals in device fabrication is to faithfully reproduce the original design intent on the substrate (via the patterning device).

[0032] As an example, the pattern layout can include the application of resolution enhancement techniques, such as optical proximity correction (OPC). OPC addresses the fact that the final size and placement of the image of the design layout projected onto the substrate will not be the same as, or depend solely on, the size and placement of the design layout on the patterning device. It is noted that the terms "mask", "reticle", "patterning device" are used interchangeably herein. Also, the skilled person will recognize that the terms "mask", "patterning device" and "design layout" can be used interchangeably, as in the context of RET, a physical patterning device does not necessarily have to be used, but a design layout can be used to represent the physical patterning device. For small feature sizes and high feature densities present on a design layout, the position of a particular edge of a given feature will be influenced to some extent by the presence or absence of other neighboring features. These proximity effects are caused by microscopic radiation coupling from one feature to another, or non-geometrical optical effects such as diffraction and interference. Similarly, proximity effects can be caused by diffusion and other chemical effects during post-exposure bake (PEB), resist development, and etching, typically performed after lithography.

[0033] To increase the chances that the projected image of the design layout meets the requirements of a given target circuit design, proximity effects can be predicted and compensated for using, for example, complex numerical models, corrections or pre-distortions of the design layout. C. Spence, in "Full-Chip Lithography Simulation and Design Analysis - How OPC Is Changing IC Design", SPIE Proceedings Vol. 5751, pp. 1-14, 2005, provides an overview of current "model-based" optical proximity correction processes. In a typical high-end design, almost every feature of the design layout is modified in some way to achieve high fidelity of the projected image to the target design. These modifications can include shifts or biases of edge positions or line widths, and the application of "assist" features intended to assist the projection of other features.

[0034] An assist feature can be seen as a difference between a feature on the patterning device and a feature in the design layout. The terms "main feature" and "assist feature" do not imply that a particular feature on the patterning device has to be labeled as one or the other.

[0035] The term "mask" or "patterning device" as employed herein can be broadly interpreted as referring to a generic patterning device that can be used to impart a pattern to a beam of radiation passing through or onto it, corresponding to a desired patterned shape to be created in a target portion of the substrate; in this context, the term "light valve" can also be used. Examples of other such patterning devices include:

[0036] - programmable mirror arrays. An example of such a device is a matrix- addressable surface having a programmable reflection of each addressable area. The underlying principle is that the reflected amount of radiation coming from a surface cell of the patterning device will depend on the light incidence angle of the incoming radiation in that area. In other words, different from the case of a transmission type mask as used above, more or less of the highly directional radiation will come through the mask pattern relating to the different cells. An example of a programmable mirror array is a matrix- addressable surface having a programable reflection of each addressable area. For purposes of this description, the term "addressable area" shall also mean a single spatial beam that is capable of independent manipulation. The matrix- addressable surface has a matrix addressable surface of pixels that can each be individually addressed. Each addressable pixel is independently modifiable, e.g., by a programmable mirror that can be set to a "reflect" or "do not reflect" state. In this manner, the reflected radiation pattern is created by setting the pixels of the matrix- addressable surface in a desired pattern. The matrix- addressable surface is then used to determine the pattern imparted on the radiation.

[0037] - programmable LCD array. An example of such a construction is given in U.S. Patent No. 5,229,872, which is incorporated by reference herein.

[0038] As a brief introduction, Figure 1 An example lithographic projection apparatus 10A is illustrated in FIG. 1. Major components are a radiation source 12A, which can be a deep-ultraviolet (DUV) excimer laser source or other type of source including an extreme ultra violet (EUV) source (as discussed above, the lithographic projection apparatus itself need not have the radiation source), illumination optics which, e.g., define the partial coherence (denoted as Sigma, S) and which can include optics 14A, 16Aa and 16Ab that shape radiation from the source 12A; a patterning device 18A; and transmission optics 16Ac to project an image of the patterning device pattern onto a substrate plane 22A. An adjustable filter or aperture 20A at the pupil plane of the projection optics can limit the range of beam angles that impinge on the substrate plane 22A, with the maximum possible angle defining the numerical aperture NA = n sin (Θ max ) of the projection optics, where n is the refractive index of the medium between the substrate and the last element of the projection optics, and Θ max is the maximum angle of the beam leaving the projection optics that can still impinge on the substrate plane 22A.

[0039] In a lithographic projection apparatus, a source provides illumination (i.e., radiation), and a patterning device provides a patterned illumination to a substrate via the patterning device. The projection optics can include at least some of the components 14A, 16Aa, 16Ab, and 16Ac. The aerial image (AI) is the intensity distribution of the radiation at the substrate level. A resist layer on the substrate is exposed, and the aerial image is transferred as a latent "resist image" (RI) in the resist layer. The resist image (RI) can be defined as the spatial distribution of solubility of the resist in the resist layer. A resist model can be used to calculate the resist image from the aerial image, examples of which can be found in U.S. Patent Application Publication No. US 2009-0157360, the disclosure of which is incorporated herein in its entirety by this reference. The resist model is related to the properties of the resist layer (e.g., the influence of chemical processes that occur during exposure, PEB, and development). The optical properties of the lithographic projection apparatus (e.g., the properties of the source, the patterning device, and the projection optics) dictate the aerial image. Since the patterning device used in the lithographic projection apparatus can be changed, it can be desirable to separate the optical properties of the patterning device from the optical properties of the rest of the lithographic projection apparatus, which includes at least the source and the projection optics.

[0040] Although specific reference can be made in this text to the use of lithographic apparatus in the manufacture of ICs, it should be understood that the lithographic apparatus described herein can have other applications, such as the manufacture of integrated optical systems, guidance and detection patterns for magnetic domain memories, liquid crystal displays (LCDs), thin-film magnetic heads, etc. The skilled artisan will appreciate that, in the context of such alternative applications, any use of the terms "wafer" or "die" in this text can be considered as synonymous with the more general terms "substrate" or "target portion", respectively. The substrate referred to herein can be processed, before or after exposure, in one or more treatment tools, e.g. a track (typically a tool to which the substrate is transferred after exposure for example for development), and a metrology or inspection tool. Where applicable, the disclosure herein can be applied to such and other substrate processing tools. Further, the substrate can be processed more than once, e.g. update layers of a die are created such that the term substrate as used herein can also refer to a substrate that already contains multiple processed layers.

[0041] The terms "radiation" and "beam" used herein encompass all types of electromagnetic radiation, including ultraviolet (UV) radiation (e.g. having a wavelength of 365, 248, 193, 157 or 126 nm) and extreme ultra-violet (EUV) radiation (e.g. having a wavelength in the range of 5-20 nm), as well as particle beams, such as ion beams or electron beams.

[0042] In semiconductor manufacturing, multiple metrology systems operating on similar principles can be used to measure one or more physical characteristics (e.g., CD, EPE, overlay, etc.) of a patterned substrate. Although the operating principles can be similar, the multiple metrology systems differ from one another in various aspects, such as mechanical subsystems, electrical subsystems, electronics, image processing algorithms, profile extraction algorithms, or other structural and software components that can cause variations in measurement values. However, it is desirable that these measurement values of a pattern on the same patterned substrate or different patterned substrates performed using different tools should match. To obtain consistent measurement values between different metrology systems, metrology data matching that acquired by a reference (or reference) metrology system can be needed. For example, metrology data of a patterned substrate obtained using a first tool (e.g., a reference metrology system) should match metrology data of the same patterned substrate obtained using a second tool. In this way, variations caused by tool-to-tool differences are significantly reduced or eliminated from the measurement result variations, and the measurement result variations are indicative of substrate variations favorably and accurately. In one embodiment, the reference metrology system can be the same vendor, or a different vendor than the second metrology system. For example, the tools can include different process models or designs supplied by different manufacturers or the same manufacturer. In some cases, the tools can include the same process model, but still behave differently, i.e., have tool-to-tool variations.

[0043] In existing approaches, after acquiring metrology data (e.g., SEM images or profiles of patterned features), metrology system settings (e.g., tunable parameters of the second metrology tool) can be determined to extract measurement values (e.g., CD) from the metrology data. For example, the metrology system settings can be determined to obtain the best CD match results to the CD obtained using the reference metrology system. In many cases, even if multiple or all available parameters are tuned, the measurement values (e.g., CD) matching the specification can not be satisfied. In one embodiment, the metrology settings can include tunable parameters such as dose, field of view (FOV), or other parameters of the second metrology tool. In one embodiment, the tunable parameters such as dose, FOV, etc. can not be modified to obtain CD-to-CD match between measurement values from the first tool and the second tool. For example, the second metrology tool can use a higher dose, have a faster image acquisition speed, or a larger FOV compared to the first metrology tool. These tunable parameters favor faster metrology measurements. Also, changing such parameters can have different charging effects, resulting in additional differences in measurement values. Therefore, in one embodiment, favorable tunable parameters such as speed and FOV can not be modified, while other parameters such as CD threshold values applied during CD measurement by SEM images can be modified. Thus, matching of measurement results from the first metrology tool can be achieved while maintaining the favorable settings of the second metrology system.

[0044] The present disclosure provides mechanisms for converting metrology data (e.g., SEM images or profiles) obtained from a second metrology tool using machine learning models. The converted metrology data matches or is comparable to metrology data acquired from another metrology tool (e.g., a reference metrology system). For example, the mechanisms herein include training a machine learning (ML) model to convert images acquired by one SEM system to images that would be acquired by another SEM system. In one embodiment, the ML model can convert signals, CD values, or other characteristics determined by a SEM system to what would be acquired by another SEM system. Thus, any measurements of physical characteristics of patterned features performed on the converted images would be comparable (e.g., match) to measurements performed on images acquired by the other SEM system. In one embodiment, differences between metrology data from different tools (e.g., CD mismatch and / or SEM image mismatch) can be incorporated into a cost function for training the ML model. As an example, metrology data including SEM image signals (e.g., intensity values) can be used to guide the conversion.

[0045] Furthermore, the mechanisms involve obtaining measurement settings (e.g., metrology system settings, such as CD thresholds for obtaining CD measurements) using reference metrology data (e.g., from a reference metrology system). These settings can be applied to the converted images to obtain measurements (e.g., CD) associated with the patterned substrate such that the measurements are as if acquired by, for example, the reference metrology system. In one embodiment, the metrology systems can be different SEM systems, e.g., a reference SEM system and another different SEM system. Example metrology systems relative to Figure 9 and 10 are illustrated and discussed. In one embodiment, the mechanisms herein can be used with images (e.g., SEM images) captured by such metrology systems.

[0046] The mechanisms of the present disclosure have several advantages. For example, consistent measurements associated with similar patterns can be obtained using different metrology tools. Even if the metrology tools use different algorithms to acquire measurements, extract profiles from images, image enhancement to identify features or profiles, image or profile segmentation, image or profile alignment to a reference, derive measurements, etc., the final measurements obtained after applying the mechanisms herein will provide a tight match to measurements from a reference metrology system. In one embodiment, the match between the first metrology data and the second metrology data can be characterized by a difference between the first metrology data and the second metrology data, statistics associated with the metrology data, detected intensity values within the metrology data, or other matching parameters. As an example, a tight match refers to a match relative to a difference threshold, a reference statistic, a reference intensity value, or other ways of characterizing matching reference data.

[0047] The mechanisms herein can also enable different metrology systems to be employed to produce semiconductor chips faster without causing significant variations in the measurements of the patterned substrates. Thus, adjustments made to the patterning process based on measurements from the reference metrology system can remain substantially the same, thereby maintaining similar production settings of the patterning process to achieve the desired throughput.

[0048] Figure 2 is an exemplary flowchart of a method 300 for converting metrology data acquired by a particular metrology system to have characteristics as if acquired by another metrology system, according to an embodiment. In one embodiment, the method 300 involves training an ML model to convert metrology data. The ML model can be a convolutional neural network (CNN), a deep convolutional neural network (DNN), a generative adversarial network (GAN), or other types of neural networks. The training can be based on metrology data sets of the same patterned substrate but acquired by different metrology systems. The ML model learns the differences in metrology data from the different systems, which can be due to unknown algorithms, different measurement recipes, different metrology tool configurations or working principles, etc. Upon training, the trained ML model can predict metrology data measured using a different system, e.g., a reference metrology system. Thus, the trained ML model can be advantageously used to improve the consistency of the measurements of the patterned substrates when different metrology systems are used. Exemplary implementations of the method 300 involve processes P301 and P303 for training the ML model. In addition, processes P305, P307, P309, and P311 can be included to apply the trained ML model to convert metrology data and predict measurements of the patterned substrates. These processes are discussed in more detail below.

[0049] The process P301 involves acquiring training data for the same patterned substrate, e.g., a patterned training substrate, from different metrology tools. In one embodiment, the process P301 involves accessing a first metrology data set MD1 acquired by a first metrology system TS1 and a second metrology data set MD2 acquired by a second metrology system TS2. For example, accessing a first SEM data set acquired by a first scanning electron metrology (SEM) system and a second SEM data set acquired by a second SEM system. The first metrology data set MD1, e.g., the first SEM data set, and the second metrology data set MD2, e.g., the second SEM data set, are associated with the same patterned substrate. In one embodiment, the metrology data sets used to train the ML model can be referred to as training data, and the patterned substrate can be referred to as a training substrate.

[0050] In one embodiment, the first metrology dataset MD1 and the second metrology dataset MD2 can include sets of images (e.g., SEM images) of a patterned substrate (or training substrate) acquired from a first metrology system TS1 (e.g., a reference metrology system such as a SEM) and a second metrology system TS2 (such as another SEM), respectively. Thus, the first metrology dataset MD1 can be referred to as a first set of SEM images, and the second metrology dataset MD2 can be referred to as a second set of SEM images. In one embodiment, the first metrology dataset MD1 and the second metrology dataset MD2 can include profiles of features on the patterned substrate. In one embodiment, the profiles can be extracted from images (e.g., SEM images) of the patterned substrate. In one embodiment, the first profiles extracted by the first metrology system TS1 can employ a first extraction algorithm (e.g., an unknown algorithm with fixed or non-tunable parameters), and the second profiles extracted by the second metrology system TS2 can employ a second extraction algorithm different from the first algorithm (e.g., a known algorithm with tunable parameters). In one embodiment, the first metrology dataset MD1 and the second metrology dataset MD2 can include physical characteristics (e.g., CD, overlay, etc.) associated with a pattern on the patterned substrate.

[0051] In one embodiment, the first metrology system can output CD measurements, while the implementation details (e.g., algorithms, image processing, etc.) used to determine the CD measurements can be unknown. The present disclosure provides mechanisms to adjust one or more parameters of the second metrology system to match the CD measurements from the first metrology system (e.g., by generating transformed images through a trained ML model). For example, if the SEM images captured using the second metrology system are different from the first metrology system, simply tuning the parameters of the second metrology system can not achieve satisfactory CD-to-CD matching results. However, transforming the captured images of the second metrology system using the trained ML model herein, and tuning the parameters (e.g., CD threshold) based on such transformed images can advantageously achieve the desired CD-to-CD matching or other measurement matching results.

[0052] Figure 3FIGURE illustrates exemplary metrology data of a patterned substrate W acquired by a first metrology system TS1 and a second metrology system TS2, according to an embodiment. In this example, the first metrology system TS1 can be a reference metrology system, such as a first SEM. The first metrology system TS1 can capture a first image IMG1 (e.g., a SEM image) of the patterned substrate W or a portion of the patterned substrate W. As an example, the SEM image IMG1 includes a feature F1 (represented by the light-colored elliptical portion in the image IMG1) corresponding to a feature (e.g., F, not labeled on the substrate W) patterned on the substrate W. In one embodiment, a cutline CL1 can be drawn across the feature F1 to measure intensity values of pixels along the cutline CL1. These intensity values are represented as a first signal S1 (see Figure 5 ).

[0053] Similarly, the second metrology system TS2 can be another metrology system different from the first SEM, such as a second SEM. The second metrology system TS2 can capture a second image IMG2 (e.g., another SEM image) of the patterned substrate W or a portion of the patterned substrate W. For example, the SEM image IMG2 includes a feature F1’ (represented by the light-colored elliptical portion in the image IMG2) corresponding to the same feature (e.g., F, not labeled on the substrate W) patterned on the substrate W. In one embodiment, a cutline CL1 can be drawn across the feature F1’ to measure intensity values of pixels along the cutline CL1. These intensity values are represented as a second signal S2 (see Figure 5 ).

[0054] In Figure 5In particular embodiments, comparing the first signal S1 and the second signal S2 for the same feature patterned on the substrate W indicates that the first metrology system TS1 and the second metrology system TS2 can produce different measurements for the same feature. For example, when CD is measured along the cut line CL1 based on the first signal S1, the measured CD will be different from the CD measured based on the second signal S2. As an example, CD can be measured based on a CD threshold defined as a function of the distance between adjacent peaks in the signal. In some embodiments, the systems can have different design configurations (e.g., different product models) supplied by the same or different manufacturers. The difference in signals for the same feature can be associated with structural differences in the tools, such as differences in the systems designed in mechanical, electrical, and electronic configurations, differences in algorithms employed by the systems TS1 and TS2, such as signal acquisition mechanisms, signal processing algorithms, image enhancement mechanisms, and the like, differences in recipe employed by each system TS1 and TS2 during measurement, such as profile extraction algorithms, cut line placement algorithms, and the like, or other known or unknown differences between the systems TS1 and TS2. In some embodiments, the systems can have the same design configuration (e.g., the same product model), and the measurement difference can be attributed to unintended system variations caused by variations in manufacturing, assembly, and / or system drift, and the like.

[0055] Referring back to Figure 3 A profile C1 corresponding to the feature F1 in the image IMG1 can be extracted. In one embodiment, the profile C1 can be extracted using a first algorithm implemented in the first metrology system TS1. In one embodiment, the implementation details of the first profile extraction algorithm can be unknown or fixed, as such parameters within the algorithm can not be adjustable when extracting the profile C1. In one embodiment, the metrology system configuration can be fixed as it is considered a reference, and knowledge of specific differences or difference causes can not be required. In one embodiment, such profile C1 can be used as the first metrology data. Similarly, another profile C2 corresponding to the feature F1’ in the image IMG2 can be extracted. In one embodiment, the profile C2 can be extracted using a second algorithm implemented in the second metrology system TS2. In one embodiment, the implementation details of the extraction algorithm can be known, and one or more parameters of the second algorithm (e.g., an intensity threshold characterizing the feature edge, extraction model parameters such as Gaussian sigma, and the like) can be adjustable during the adjustment of the profile.

[0056] In one embodiment, a value of a physical characteristic such as CD of the feature Fl can be determined based on the extracted profile Cl. In one embodiment, the CD can be determined based on an average of the distance between the two ends along the length of the feature Fl. Similarly, a value of a physical characteristic such as CD of the feature Fl' can be determined based on the extracted profile C2. In one embodiment, the same algorithm can be used to determine the CD values of the features Fl and Fl'. The present disclosure is not limited to a particular measurement method. For example, the CD can be measured in a linear mode or a threshold mode of the metrology system. In another example, a maximum, a minimum, or an average of several neighboring locations of the feature can be used to determine the CD value. In yet another example, a shape fitting of the profiles Cl and C2 can be performed. In yet another example, parameters such as an intensity threshold, a signal smoothing window size, a starting point to find the edge position of a feature, or other tunable parameters can be performed for a desired CD to CD matching result.

[0057] The present disclosure is not limited to extracting measurement values using a particular technique. The CD measurement values determined based on the profile extraction and adjustment algorithm are merely exemplary and do not limit the scope of the present disclosure. In one embodiment, the metrology system can employ an image processing algorithm and / or a CD measurement algorithm to determine the CD of a feature without extracting a profile of the feature patterned on the substrate from the image.

[0058] However, due to the inherent differences of the images IMG1 and IMG2 from which the profiles Cl and C2 are derived respectively, the measured CD of the same feature patterned on the substrate W will be different.

[0059] In the present disclosure, the trained ML model is configured to transform the second metrology data (e.g., the second signal S2) such that it closely matches the first metrology data (e.g., the first signal S1). Therefore, any measurement made using the transformed metrology data (e.g., a transformed version of the second signal S2) can have a similar measurement value.

[0060] Referring back to Figure 2 , the process P303 involves training an ML model based on the first metrology data set MD1 and the second metrology data set MD2. After the training process is completed, the trained ML model TML is configured to transform a metrology data set acquired by the second metrology system into a transformed data set that has characteristics as if acquired by the first metrology system TS1 (e.g., a reference metrology system).

[0061] In one embodiment, training the ML model involves comparing the first metrology dataset MD1 and the second metrology dataset MD2 of the patterned training substrate. Based on the comparison, parameters of the ML model can be adjusted to affect a cost function used to train the ML model. For example, the cost function can be a function of a difference between the first metrology dataset MD1 and the second metrology dataset MD2. As the ML model parameters (e.g., weights) are adjusted, the cost function value (e.g., difference) is progressively reduced. In one embodiment, the cost function is minimized. In one embodiment, the adjusting of the ML model parameters is stopped when a given number of iterations is reached, when the cost function value is within a desired threshold, when the cost function value is not significantly reduced in subsequent iterations, or other stopping criteria.

[0062] In one embodiment, the training of the ML model is performed to obtain a CD-to-CD or signal-to-signal match between the first SEM dataset and the second SEM dataset. For example, training the ML model involves comparing first CD values of the first SEM dataset and second CD values of the second SEM dataset, and adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model to improve the CD match between the first SEM dataset and the second SEM dataset, the cost function being a function of the first CD values and the second CD values.

[0063] In one embodiment, training the ML model involves using training datasets that include metrology data aligned to the same design layout. For example, the first image set (e.g., first SEM image) of the first metrology dataset MD1 can be aligned to the design layout image. As another example, the first profile can be aligned to a design profile of the design layout. Similarly, the second image set (e.g., another SEM image) of the second metrology dataset MD2 can be aligned to the design layout image, or the second profile can be aligned to a design profile of the design layout. In one embodiment, the aligned first image set (or aligned first profile) and the aligned second image set (or aligned second profile) can be used as training data to train the ML model.

[0064] As an example, training the ML model involves comparing intensity values from the first image set and the second image set, and adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model. In one embodiment, the intensity value of each pixel of the first image set can be compared to the intensity value of a corresponding pixel of the second image set.

[0065] In one embodiment, training the ML model is an iterative process. Each iteration involves (i) applying adjusted parameters to use the ML model to convert the second metrology dataset MD2 to a converted metrology dataset MD2'; (ii) comparing intensity values from the first metrology dataset MD1 and the converted metrology dataset MD2'; (iii) further adjusting the parameters of the ML model based on the comparison to affect the cost function (e.g. within a desired threshold, or minimized); (iv) determining whether the cost function is within the desired threshold (or minimized); and (v) in response to the cost function not being within the desired threshold (or minimized), repeating steps (i) to (iv).

[0066] In one embodiment, the cost function can comprise a difference between the metrology datasets MD1 and MD2. For example, the cost function can comprise a difference between a first intensity value from the first metrology dataset MD1 and a corresponding second intensity value from the second metrology dataset MD2 or converted dataset MD2'. In one embodiment, the intensity values can be determined by a cutline drawn along the feature on the first metrology dataset MD1 and the second metrology dataset MD2. In one embodiment, determining the intensity values comprises determining pixel intensity values by the cutline on the first metrology dataset MD1 and the second metrology dataset MD2.

[0067] In one embodiment, determining the intensity values involves: applying a first profile extraction algorithm associated with the first metrology system TS1 to the first metrology dataset MD1; and applying a second profile extraction algorithm associated with the second metrology system TS2 to the second metrology dataset MD2.

[0068] In one embodiment, the process P303 of training the ML model employs a generative adversarial network. In this example, the training process involves training a generator model in conjunction with a discriminator model. The generator model can be trained using the second metrology dataset MD2 as input to generate data similar to the first metrology dataset MD1; and the discriminator model is trained to distinguish the generated data from the first metrology dataset MD1. The generator model and the discriminator model can be trained in cooperation with each other, such that the generator model generates realistic data (e.g. matching the reference data), while the discriminator model classifies such generator model data as likely not realistic.

[0069] An example training process for a GAN is an iterative process. Each training iteration can comprise the steps of (i) randomly selecting a training sample (e.g. MD2 and MD1) from the training data; (ii) training the generator model using the selected training sample; (iii) training the discriminator model using the selected training sample; (iv) determining whether the training process has converged; and (v) in response to the training process not having converged, repeating steps (i) to (iv). Figure 4some of the aligned images IMG1 and IMG2), and step (ii) is used to train the generator model to convert the selected input data from MD2 to converted data MD2', and train the discriminator model using the converted data MD2' and the reference data MD1. The cost function for training the discriminator model can be configured to adjust the weights of the discriminator model such that the discriminator model classifies the converted data MD2' as fake (or false), and classifies the reference data MD1 as real (or true). For example, the discriminator-related cost function can maximize the probabilities assigned to real and fake images. The cost function for the generator model can include two parts, for example configured to minimize the probability that the generator output is classified as fake by the discriminator. The first part can include a configured term that causes the discriminator model to label the converted data MD2' as real (or true). The second part can include a term such as the difference in intensity between each pixel of the images MD2' and MD1. The second part can be reduced (e.g. minimized) to cause the generator model to generate realistic converted data MD2'. The above training process (e.g. steps (i) and (ii)) is repeated until a desired result is obtained.

[0070] Figure 5 An exemplary training of a ML model using a generative adversarial network (GAN) is illustrated in accordance with an embodiment. As an example, the ML model includes a generator model GM and a discriminator model DM. The generator model GM can be configured to receive metrology data as input, and generate data similar to reference data. For example, the generator model GM is configured to receive metrology data (e.g. SEM images IMG2 or contours) acquired by a second metrology system (e.g. TS2 in Figure 3 TS2) and generate metrology data (e.g. PIMG2 or contours). In one embodiment, the generated metrology data has characteristics (e.g. intensity, shape of features, size of features, etc.) as if the generated metrology data was acquired by a first metrology system (e.g. TS1 in Figure 3 TS1). The discriminator model DM is configured to receive the generated data output by the generator model GM, and determine whether the generator data is similar to the reference data. For example, the discriminator model DM is configured to receive the generated metrology data (e.g. PIMG2), and determine whether the generated metrology data (e.g. PIMG2) is similar to the reference data (e.g. SEM images IMG1). During the training process, the generator model GM and the discriminator model DM compete with each other, such that the generator model GM gradually generates realistic images, and the discriminator model attempts to classify these generated images as fake (false).

[0071] In Figure 5In the example shown in (A), the discriminator model DM can be trained using the generated second image PIMG2 as input and the first image IMG1 (e.g., W in Figure 3 In one embodiment, the discriminator model DM classifies input images into a first class (e.g., labeled as real) or a second class (e.g., labeled as fake). For example, the first class refers to reference data, and the second class refers to data generated by a model (e.g., GM). In one embodiment, the weights WTS of the discriminator model DM can be adjusted until the DM model determines that the generated image PIMG2 is real. In one embodiment, real refers to a label assigned to the generated image PIMG2 by the discriminator model DM, where real indicates that the generated image PIMG2 is similar to the first image IMG1. In one embodiment, the adjustment of the weights WTS of the discriminator model DM is guided by a cost function CF. In one embodiment, the cost function CF can be a function of the difference between the first image IMG1 and the generated image PIMG2. In one embodiment, the difference can be characterized by a label (e.g., real vs. fake, difference in intensity in the images, etc.).

[0072] In Figure 5 In the example shown in (B), the generator model GM can be trained using the second image IMG2 (e.g., W in Figure 3 as input and the first image IMG1 (e.g., W in Figure 3 as reference data. In one embodiment, the generator model GM converts input images (e.g., IMG2) to images PIMG2 that have similar characteristics (e.g., image intensity, intensity slope, or other image characteristics) as the reference data (e.g., IMG1).

[0073] In one embodiment, the weights WTS’ of the generator model GM can be adjusted based on a cost function CF. In one embodiment, the cost function CF can be a function of the difference between the first image IMG1 and the generated image PIMG2 and whether the discriminator model DM classifies the generated image PIMG2 as the second class or the first class. In one embodiment, the cost function guides the adjustment of the weights WTS’ of the generator model GM such that the difference between the image PIMG2 and IMG1 is reduced (in one embodiment, minimized) and the generated image PIMG2 is classified as the first class (e.g., real) by the discriminator model DM.

[0074] In one embodiment, the trained ML model TML can be applied to convert captured metrology data by a metrology tool that trains the ML model. As an example, the method 300 can also involve processes P305 and P307. Process P305 involves capturing metrology data 310 of a patterned substrate (e.g. different from the substrate used to train the ML model) via the second metrology system TS2. Process P307 involves converting the captured metrology data 310 via the trained ML model TML to converted metrology data 311. The converted metrology data 311 of the patterned substrate has characteristics as if captured by the first metrology system TS1.

[0075] Figure 6A Figures illustrate inputting of a second image 601 acquired by a second metrology system (e.g. TS2) to a trained ML model TML (e.g. trained according to method 300) to generate a converted image 605, according to an embodiment. Figure 6B An exemplary first image 610 acquired by a first metrology system is shown. The first image 610 and the second image 601 are associated with the same substrate. Comparing the converted image 605 and the first image 610 indicates that these images closely match each other. For example, Figure 7 Figures illustrate signals (e.g. intensity values) plotted along a cutline on a feature in each of the images 601, 605 and 610. Comparing the signal S10 obtained from the second image 601 (in Figure 6A the signal S30 obtained from the first image 610 shows a significant difference in intensity values. On the other hand, the signal S20 obtained from the converted image 605 and the signal S30 closely match each other. For example, whether the signals closely match can be determined by taking a difference between the signals. Thus, advantageously, the converted image 605 can be used to make measurements associated with physical characteristics of the features in the image 605.

[0076] In one embodiment, the metrology recipe can be determined based on the first metrology data acquired from the first metrology system TS1. The metrology recipe can also be used to obtain the converted metrology data from the trained ML model TML to determine the measurement of the physical characteristics (e.g., CD, overlay, etc.) of the pattern on the patterned substrate. As an example, the method 300 can also involve processes P307 and P311. Process P309 involves determining a metrology measurement recipe 315 for the second metrology system TS2 based on the first metrology data set MD1 and the physical characteristics measurements PC1 of the patterned substrate from the first metrology system TS1. For example, the metrology measurement recipe 315 includes a CD threshold that indicates a location on the captured metrology data 311 where a CD measurement is acquired. In one embodiment, determining the metrology measurement recipe 315 involves: extracting a contour from an image of the first metrology data set MD1 via a first contour extraction algorithm; measuring a CD at a location on the contour (e.g., line CL1 in Figure 3 Figure 4

[0077] In one embodiment, the metrology data (e.g., 310) of the patterned substrate can be captured using the second metrology system TS2 and the captured metrology data (e.g., 311) can be converted using the trained ML model TML. Process P311 can be performed on the converted metrology data 311. Process P311 involves applying the metrology measurement recipe 315 to the converted metrology data 311 to determine the physical characteristics measurements PC2 of the patterned substrate. For example, the physical characteristics measurements PC2 can be a critical dimension (CD) measurement, an overlay measurement, an edge placement error, or other characteristics associated with the patterned substrate.

[0078] Figure 8 is a block diagram of exemplary operations performed to convert metrology data and determine measurements using the converted metrology data according to embodiments. For example, metrology data acquired by a metrology system (e.g., SEM) is converted to match the measurement behavior of a reference metrology system. In one embodiment, operation 801 involves obtaining training data used to train a ML model, where the ML model converts metrology data to be comparable to metrology data acquired by a reference metrology tool. In one embodiment, the training data includes metrology data acquired by a metrology system (e.g., SEM) and metrology data acquired by a reference metrology tool (e.g., a reference SEM). In one embodiment, the metrology data acquired by the metrology system and the reference metrology tool are of the same patterned substrate. In one embodiment, the metrology data acquired by the metrology system and the reference metrology tool are of the same patterned substrate and the same physical characteristics (e.g., CD, overlay, etc.). In one embodiment, the metrology data acquired by the metrology system and the reference metrology tool are of the same patterned substrate and the same physical characteristics (e.g., CD, overlay, etc.) at the same location on the patterned substrate. Figure 3 ​​The metrology data set of the patterned substrate (e.g. training substrate) obtained as discussed in the background section. For example, the metrology data set includes images IMG1 and IMG2 acquired from the first metrology system TS1 and the second metrology system TS2, respectively. In one embodiment, the images IMG1 and IMG2 can be aligned with the design layout to generate aligned images AIMG1 and AIMG2 to be used as training data.

[0079] In operation 803, the training data is used to train the ML model. The ML model can be trained according to the process P303 (in the background section) or the process P304 (in the background section). Figure 3

[0080] In operation 805, metrology data 810 (e.g. SEM images or profiles) of a patterned substrate to be measured can be obtained. For example, the metrology data 810 is obtained from a metrology system other than the reference metrology system. Further, the metrology data 810 can be input to the trained ML model TML to convert the metrology data 810 to converted metrology data 820.

[0081] In operation 811, reference metrology data (e.g. IMG1) and reference measurement values (e.g. CD values associated with IMG1) can be obtained from the reference metrology system. Based on this data, a metrology recipe R1 can be determined. For example, the metrology recipe includes CD threshold values to be applied to signals along a cutline across a feature in the image IMG1 to determine the CD value of the feature.

[0082] In operation 813, the metrology recipe R1 can be applied to the converted metrology data 820 to determine measurement values 830 of physical properties (e.g. CD) of the patterned substrate. Since the converted metrology data is comparable to the data acquired by the reference metrology system, and the metrology recipe R1 also corresponds to the reference metrology system, the measurement values 830 will be comparable to the data acquired by the reference metrology tool. Thus, advantageously, the combination of the converted metrology data 820 and the metrology recipe R1 provides measurement values that are consistent with the reference metrology system.

[0083] In some embodiments, the inspection apparatus or metrology apparatus can be a scanning electron microscope (SEM) which produces an image of a structure (e.g. some or all of the structures of a device) exposed onto or transferred to a substrate. Figure 9 An embodiment of a SEM tool is depicted. A primary electron beam EBP emitted from an electron source ESO is converged by a condenser lens CL and then passed through a beam deflector EBD1, an ExB deflector EBD2 and an objective lens OL to illuminate a substrate PSub on a substrate table ST at a focal point.

[0084] ​When the substrate PSub is irradiated with the electron beam EBP, secondary electrons are generated from the substrate PSub. The secondary electrons are deflected by the ExB deflector EBD2 and detected by the secondary electron detector SED. By synchronizing the detection of the electrons generated from the sample with a two-dimensional scanning of the electron beam by the electron beam deflector EBD1 or a repeated scanning of the electron beam EBP in the X or Y direction by the electron beam deflector EBD1 and a continuous movement of the substrate PSub in the other direction in the X or Y direction by the substrate table ST, a two-dimensional electron beam image can be obtained.

[0085] The signal detected by the secondary electron detector SED is converted to a digital signal by an analog / digital (A / D) converter ADC and the digital signal is sent to an image processing system IPU. In one embodiment, the image processing system IPU can have a memory MEM to store the entire or a part of the digital image for processing by a processing unit PU. The processing unit PU, e.g. a specially designed hardware or a combination of hardware and software, is configured to convert or process the digital image into a data set representing the digital image. Further, the image processing system IPU can have a storage medium STOR configured to store the digital image and the corresponding data set in a reference database. A display device DIS can be connected to the image processing system IPU so that an operator can perform the necessary operations on the device by means of a graphical user interface.

[0086] As mentioned above, SEM images can be processed to extract contours describing the edges of objects representing device structures in the images. These contours are then quantified via metrics such as CD. Thus, typically, images of device structures are compared and quantified via simple metrics such as edge-to-edge distance (CD) or simple pixel difference between images. Typical contour models that detect object edges in images to measure CD use image gradients. In fact, these models rely on strong image gradients. However, in practice, images are often noisy and have discontinuous boundaries. Techniques such as smoothing, adaptive thresholding, edge detection, erosion and dilation can be used to process the results of image gradient contour models to address noisy and discontinuous images, but ultimately result in low resolution quantification of high resolution images. Thus, in most instances, mathematical operations to reduce noise and automatic edge detection of device structure images result in a loss of image resolution, resulting in a loss of information. Thus, the result is a low resolution quantification that is equivalent to a simple representation of complex high resolution structures.

[0087] Thus, it is desirable to have a mathematical representation of a structure (e.g., a circuit feature, an alignment mark, or a metrology target portion (e.g., a grating feature), etc.) that is produced or is expected to be produced using a patterning process, whether it is, for example, in a latent resist image, in a developed resist image, or in a layer transferred onto a substrate, e.g., by etching, that can preserve resolution and also describe the general shape of the structure. In the context of a lithographic or other patterning process, the structure can be a device or a portion thereof that is being manufactured, and the image can be an SEM image of the structure. In some examples, the structure can be a feature of a semiconductor device (e.g., an integrated circuit). In this case, the structure can be referred to as a pattern or a desired pattern that includes a plurality of features of the semiconductor device. In some examples, the structure can be an alignment mark or a portion thereof (e.g., a grating of an alignment mark) that is used in an alignment measurement process to determine an alignment of an object (e.g., a substrate) to another object (e.g., a patterning device) or a metrology target or a portion thereof (e.g., a grating of a metrology target) that is used to measure a parameter (e.g., overlay, focus, dose, etc.) of the patterning process. In one embodiment, the metrology target is a diffractive grating used to measure, for example, overlay.

[0088] Figure 10 A further embodiment of an inspection apparatus is schematically illustrated. The system is used to inspect a sample 90 (such as a substrate) on a sample stage 88 and comprises a charged particle beam generator 81, a condenser lens module 82, a probe forming objective lens module 83, a charged particle beam deflection module 84, a secondary charged particle detector module 85 and an image forming module 86.

[0089] The charged particle beam generator 81 generates a primary charged particle beam 91. The condenser lens module 82 converges the generated primary charged particle beam 91. The probe forming objective lens module 83 focuses the converged primary charged particle beam into a charged particle beam probe 92. The charged particle beam deflection module 84 scans the formed charged particle beam probe 92 over a surface of a region of interest on the sample 90 fixed on the sample stage 88. In one embodiment, the charged particle beam generator 81, the condenser lens module 82 and the probe forming objective lens module 83 or an equivalent design thereof, a substitute thereof or any combination thereof together form a charged particle beam probe generator which generates the scanning charged particle beam probe 92.

[0090] The secondary charged particle detector module 85 detects secondary charged particles 93 emitted from the sample surface (and possibly other reflected or scattered charged particles from the sample surface) when bombarded by the charged particle beam probe 92, generating a secondary charged particle detection signal 94. An image forming module 86 (e.g., a computing device) is coupled to the secondary charged particle detector module 85 to receive the secondary charged particle detection signal 94 from the secondary charged particle detector module 85 and thus form at least one scan image. In one embodiment, the secondary charged particle detector module 85 and the image forming module 86, or their equivalents, alternatives, or any combination thereof, together form an image forming apparatus that forms a scan image from the detected secondary charged particles emitted from the sample 90 bombarded by the charged particle beam detector 92.

[0091] In one embodiment, the monitoring module 87 is coupled to the image forming module 86 of an image forming apparatus to monitor, control, and / or derive parameters for patterning process design, control, monitoring, etc., using scanned images of the sample 90 received from the image forming module 86. Therefore, in one embodiment, the monitoring module 87 is configured or programmed to cause the execution of the methods described herein. In one embodiment, the monitoring module 87 includes a computing device. In one embodiment, the monitoring module 87 includes a computer program to provide the functionality described herein and is encoded on a computer-readable medium that forms or is disposed within the monitoring module 87.

[0092] In one embodiment, compared with using a probe to inspect the substrate... Figure 9 Similar to electron beam inspection tools, Figure 10 Electronic current in a system and, for example, such as Figure 9 The depicted CD SEM is significantly larger, resulting in a sufficiently large probe spot for rapid detection. However, due to the larger probe spot, the resolution may not be as high as that of a CD SEM. In one embodiment, the inspection device discussed above can be a single-beam or multi-beam device, without limiting the scope of this disclosure.

[0093] From, for example Figure 9 and / or Figure 10 The system's SEM images can be processed to extract contours describing the edges of objects representing the device structure in the image. These contours are then typically quantized using metrics such as CD at user-defined cut lines. Therefore, images of the device structure are typically compared and quantized using metrics such as edge-to-edge distance (CD) measured on the extracted contours or simple pixel differences between images.

[0094] In one embodiment, one or more procedures of the method 300 can be implemented as instructions (e.g., program code) in a processor of a computer system (e.g., process 104 of computer system 100). In one embodiment, the procedures can be distributed over multiple processors (e.g., parallel computing) to improve computing efficiency. In one embodiment, a computer program product includes a non-transitory computer-readable medium having instructions recorded thereon, which, when executed by a computer, implement the methods described herein.

[0095] Figure 11 is a block diagram that illustrates a computer system 100 that can aid in the implementation of the methods, procedures or devices disclosed herein. Computer system 100 includes a bus 102 or other communication mechanism for communicating information, and a processor 104 (or multiple processors, processors 104 and 105) coupled with bus 102 for processing information. Computer system 100 also includes a main memory 106, such as a random access memory (RAM) or other dynamic storage device, coupled to bus 102 for storing information and instructions to be executed by processor 104. Main memory 106 also can be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 104. Computer system 100 further includes a read only memory (ROM) 108 or other static storage device coupled to bus 102 for storing static information and instructions for processor 104. A storage device 110, such as a magnetic disk or optical disk, is provided and coupled to bus 102 for storing information and instructions.

[0096] Computer system 100 can be coupled via bus 102 to a display 112, such as a cathode ray tube (CRT) or flat panel or touch panel display for displaying information to a computer user. An input device 114, including alphanumeric and other keys, is coupled to bus 102 for communicating information and command selections to processor 104. Another type of user input device is cursor control 116, such as a mouse, a trackball or cursor direction keys for communicating direction information and command selections to processor 104 and for

[0097] According to one embodiment, portions of one or more methods described herein can be performed by computer system 100 in response to processor 104 executing one or more sequences of one or more instructions contained in main memory 106. Such instructions can be read into main memory 106 from another computer-readable medium, such as storage device 110. Execution of the sequences of instructions contained in main memory 106 causes processor 104 to perform the process steps described herein. One or more processors in a multi-processing arrangement can also be employed to execute the sequences of instructions contained in main memory 106. In alternative embodiments, hard-wired circuitry can be used in place of or in combination with software instructions. Thus, the description herein is not limited to any specific combination of hardware circuitry and software.

[0098] The term "computer-readable medium" as used herein refers to any medium that participates in providing instructions to processor 104 for execution. Such a medium can take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. Non-volatile media include, for example, optical or magnetic disks, such as storage device 110. Volatile media include dynamic memory, such as main memory 106. Transmission media include coaxial cables, copper wire and fiber optics, including the wires that comprise bus 102. Transmission media can also take the form of acoustic or light waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media include, for example, floppy disk, flexible disk, hard disk, magnetic tape, any other magnetic medium, CD-ROM, DVD, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, RAM, PROM, and EPROM, a FLASH- EPROM, any other memory chip or cartridge, a carrier wave as described hereinafter, or any other medium from which a computer can read.

[0099] Various forms of computer-readable media can be involved in carrying one or more sequences of one or more instructions to processor 104 for execution. For example, the instructions can initially be carried on a magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A modem local to computer system 100 can receive the data on the telephone line and use an infrared transmitter to convert the data to an infrared signal. An infrared detector coupled to bus 102 can receive the data carried in the infrared signal and place the data on bus 102. Bus 102 carries the data to main memory 106, from which processor 104 retrieves and executes the instructions. The instructions can optionally be stored on storage device 110, either before or after execution by processor 104.

[0100] Computer system 100 can also include a communication interface 118 coupled to bus 102. Communication interface 118 provides a two-way data communication coupling to a network link 120 that is connected to a local network 122. For example, communication interface 118 can be an integrated services digital network (ISDN) card or a modem to provide a data communication connection to a corresponding type of telephone line. As another example, communication interface 118 can be a local area network (LAN) card to provide a data communication connection to a compatible LAN. Wireless links can also be implemented. In any such implementation, communication interface 118 sends and receives electrical, electromagnetic or optical signals that carry digital data streams representing various types of information.

[0101] Network link 120 typically provides data communication through one or more networks to other data devices. For example, network link 120 can provide a connection through local network 122 to a host computer 124 or to data equipment operated by an Internet Service Provider (ISP) 126. ISP 126 in turn provides data communication services through the worldwide packet data communication network now commonly referred to as the "Internet" 128. Local network 122 and Internet 128 both use electrical, electromagnetic or optical signals that carry digital data streams. The signals through the various networks and the signals on network link 120 and through communication interface 118, which carry the digital data to and from computer system 100, are exemplary forms of carrier waves transporting the information.

[0102] Computer system 100 can send messages and receive data, including program code, through the network(s), network link 120, and communication interface 118. In the Internet example, a server 130 might transmit a requested code for an application program through Internet 128, ISP 126, local network 122 and communication interface 118. One such downloaded application can provide all or part of the functionality described herein for the methods. The received code can be executed by processor 104 as it is received, and / or stored in storage device 110, or other non-volatile storage for later execution. In this manner, computer system 100 can obtain application code in the form of a carrier wave.

[0103] Embodiments of the present disclosure can be further described by the following clauses.

[0104] 1. A non-transitory computer-readable medium having instructions recorded thereon, the instructions, when executed by a computer, implement a method for transforming data associated with a metrology system, the method comprising:

[0105] accessing a first scanning electron metrology (SEM) dataset acquired by a first SEM system and a second SEM dataset acquired by a second SEM system, the first and second SEM datasets being associated with a patterned substrate; and

[0106] training a machine learning (ML) model using the first and second SEM datasets as training data, such that the trained ML model is configured to convert a metrology dataset acquired by the second SEM system into a converted dataset having characteristics comparable to metrology data acquired by the first SEM system.

[0107] 2. The medium of clause 1, wherein training the ML model comprises:

[0108] comparing the first and second SEM datasets; and

[0109] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model.

[0110] 3. The medium of clause 2, wherein the first and second SEM datasets comprise:

[0111] 4. The medium of clause 3, wherein training the ML model comprises:

[0112] comparing a first set of images acquired by the first SEM system and a second set of images acquired by the second SEM system; and

[0113] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model to improve a match between the first set of images and ML generated images using the second set of images as input to the ML model.

[0114] 5. The medium of clause 2, wherein the first and second SEM datasets comprise:

[0115] a profile of a feature on the patterned substrate; and / or

[0116] a physical characteristic associated with a pattern on the patterned substrate.

[0117] 6. The medium of clause 4, wherein the physical characteristic comprises a critical dimension (CD) of a pattern on the patterned substrate.

[0118] 7. The medium of clause 5, wherein training the ML model comprises:

[0119] comparing a first CD value of the first SEM dataset and a second CD value of the second SEM dataset; and

[0120] adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model, the cost function being a function of the first CD values and the second CD values, to improve the CD matching between the first SEM data set and the second SEM data set.

[0121] 8. The medium of clause 3, wherein training the ML model comprises:

[0122] aligning the first set of images or the first contours of the first SEM data set with design layout images or design contours of the design layout;

[0123] aligning the second set of images or the second contours of the second SEM data set with design layout images or design contours of the design layout; and

[0124] using the aligned first set of images and the aligned second set of images as training data for training the machine learning model.

[0125] 9. The medium of clause 6, wherein training the ML model comprises:

[0126] comparing intensity values from the first set of images and the second set of images; and

[0127] adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model.

[0128] 10. The medium of clause 7, wherein training the ML model is an iterative process, each iteration comprising:

[0129] (i) transforming, via the ML model with the adjusted parameters and using the first SEM data set, the second SEM data set;

[0130] (ii) comparing intensity values from the first SEM data set with the transformed data set;

[0131] (iii) further adjusting parameters of the ML model based on the comparison to influence the cost function to be within a desired threshold;

[0132] (iv) determining whether the cost function is within the desired threshold; and

[0133] (v) responsive to the cost function not being within the desired threshold, repeating steps (i) through (iv).

[0134] 11. The medium of clause 8, wherein the cost function is a difference between first intensity values from the first SEM data set and corresponding second intensity values from the second SEM data set or the transformed data set.

[0135] 12. The medium of clause 7, wherein training the ML model comprises:

[0136] Determine intensity values along the cutline plotted on the feature from the first SEM dataset and the second SEM dataset.

[0137] 13. The medium of clause 10, wherein determining intensity values comprises: determining pixel intensity values along the cutline from the first SEM dataset and the second SEM dataset.

[0138] 14. The medium of clause 11, wherein determining intensity values comprises:

[0139] applying a first profile extraction algorithm associated with the first SEM system to the first SEM dataset; and

[0140] applying a second profile extraction algorithm associated with the second SEM system to the second SEM dataset.

[0141] 15. The medium of clause 1, further comprising:

[0142] capturing, via the second SEM system, metrology data of another patterned substrate; and

[0143] converting, via the trained ML model, the captured metrology data to converted metrology data, the converted metrology data of the other patterned substrate having characteristics as if captured by the first SEM system.

[0144] 16. The medium of clause 1, further comprising:

[0145] determining, based on the first SEM dataset and physical property measurements of the patterned substrate from the first SEM system, a metrology measurement recipe for the second SEM system;

[0146] capturing, using the second SEM system, metrology data of the patterned substrate;

[0147] converting, using the trained machine learning model, the captured metrology data; and

[0148] applying the metrology measurement recipe to the converted metrology data to determine another physical property measurement.

[0149] 17. The medium of clause 14, wherein the physical property measurements comprise at least one of: critical dimension (CD) measurements, overlay measurements, and edge placement errors.

[0150] 18. The medium of clause 15, wherein the metrology measurement recipe comprises a CD threshold that indicates a location on the captured metrology data to take CD measurements.

[0151] 19. The medium of clause 16, wherein determining the metrology measurement recipe comprises:

[0152] extracting contours from images of the first SEM dataset via a first contour extraction algorithm;

[0153] drawing a cutline at a location on the contours to measure a CD; and

[0154] determining a CD threshold corresponding to the measured CD based on signals along the cutline.

[0155] 20. The medium of clause 1, wherein the first SEM system is manufactured by a first manufacturer and the second metrology system is manufactured by a second manufacturer.

[0156] 21. The medium of clause 1, wherein the ML model is a convolutional neural network.

[0157] 22. The medium of clause 1, wherein the ML model is trained using a generative adversarial network architecture, the ML model comprising a generator model and a discriminator model.

[0158] 23. The medium of clause 20, wherein training the ML model comprises:

[0159] training the generator model in conjunction with the discriminator model using the second SEM dataset as input to generate data similar to the first SEM dataset; and

[0160] training the discriminator model to distinguish the generated data from the first SEM dataset.

[0161] 24. A metrology system comprising:

[0162] a computer system comprising one or more processors having stored thereon a trained machine learning (ML) model and programmed with computer program instructions that, when executed, cause the computer system to:

[0163] capture metrology data of a patterned substrate; and

[0164] convert the captured metrology data to converted metrology data via the trained ML model, the converted metrology data having characteristics as if captured by another metrology system.

[0165] 25. The metrology system of clause 22, wherein the computer system is used to train the ML model, the training comprising:

[0166] accessing a first SEM dataset acquired by a first SEM system and a second SEM dataset acquired by the metrology system, the first SEM dataset and the second SEM dataset associated with a training substrate; and

[0167] using the first SEM data set and the second SEM data set as training data, training a machine learning (ML) model such that the trained ML model is configured to convert a metrology data set acquired by a metrology system into a converted data set having characteristics comparable to metrology data acquired by the first SEM system.

[0168] 26. The metrology system of clause 23, wherein training the ML model comprises:

[0169] comparing the first SEM data set and the second SEM data set; and

[0170] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model.

[0171] 27. The metrology system of clause 24, wherein the first SEM data set and the second SEM data set comprise:

[0172] a set of images of a training substrate.

[0173] comparing the first set of images acquired by the first SEM system and the second set of images acquired by the second SEM system; and

[0174] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model to improve a match between the first set of images and ML generated images using the second set of images as input to the ML model.

[0175] 29. The metrology system of clause 25, wherein the first SEM data set and the second SEM data set comprise:

[0176] a profile of a feature on a training substrate; and / or

[0177] a physical characteristic associated with a pattern on the training substrate.

[0178] 30. The metrology system of clause 27, wherein training the ML model comprises:

[0179] aligning the first set of images or the first profile of the first SEM data set to a design layout image or a design profile of a design layout;

[0180] aligning the second set of images or the second profile of the second SEM data set to the design layout image or the design profile of the design layout; and

[0181] using the aligned first set of images and the aligned second set of images as training data for training the machine learning model.

[0182] 31. The metrology system of clause 30, wherein training the ML model comprises:

[0183] comparing intensity values from the first image set and the second image set; and

[0184] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model.

[0185] 32. The metrology system of clause 31, wherein training the ML model is an iterative process, each iteration comprising:

[0186] (i) transforming, via the ML model with the adjusted parameters and using the first SEM data set, the second SEM data set;

[0187] (ii) comparing intensity values from the first SEM data set to the transformed data set;

[0188] (iii) further adjusting parameters of the ML model based on the comparison to affect the cost function to be within a desired threshold;

[0189] (iv) determining whether the cost function is within the desired threshold; and

[0190] (v) in response to the cost function not being within the desired threshold, repeating steps (i) through (iv).

[0191] 33. The metrology system of clause 32, wherein the cost function is a difference between first intensity values from the first SEM data set and corresponding second intensity values from the second SEM data set or the transformed data set.

[0192] 34. The metrology system of clause 31, wherein training the ML model comprises:

[0193] determining intensity values along a cutline drawn on the feature by the first SEM data set and the second SEM data set.

[0194] 35. The metrology system of clause 34, wherein determining intensity values comprises determining pixel intensity values along the cutline by the first SEM data set and the second SEM data set.

[0195] 36. The metrology system of clause 35, wherein determining intensity values comprises:

[0196] applying a first profile extraction algorithm associated with the first SEM system to the first SEM data set; and

[0197] applying a second profile extraction algorithm associated with the metrology system to the second SEM data set.

[0198] 37. The metrology system of clause 36, further comprising:

[0199] determining a metrology measurement recipe for the metrology system based on the first SEM data set and the physical property measurement from the substrate from the first SEM system; and

[0200] applying the metrology measurement recipe to the converted metrology data to determine another physical property measurement.

[0201] 38. The metrology system of clause 37, wherein the physical property measurement comprises at least one of: a critical dimension (CD) measurement, an overlay measurement, and an edge placement error.

[0202] 39. The metrology system of clause 38, wherein the metrology measurement recipe comprises a CD threshold that indicates a location on the captured metrology data at which a CD measurement is taken.

[0203] 40. The metrology system of clause 39, wherein determining the metrology measurement recipe comprises:

[0204] extracting a contour from an image of the first SEM data set via a first contour extraction algorithm; drawing a cutline at a location on the contour to measure a CD; and

[0205] determining a CD threshold corresponding to the measured CD based on a signal along the cutline.

[0206] 41. The metrology system of clause 24, wherein the metrology system is a scanning electron microscope.

[0207] 42. The metrology system of clause 24, wherein the trained ML model is a convolutional neural network.

[0208] 43. The metrology system of clause 24, wherein the ML model is trained using a generative adversarial network architecture, the ML model comprising a generator model and a discriminator model.

[0209] 44. The metrology system of clause 43, wherein training the ML model comprises:

[0210] training the generator model in conjunction with the discriminator model using a second SEM data set as input to generate data similar to the first SEM data set; and

[0211] training the discriminator model to distinguish the generated data from the first SEM data set.

[0212] 45. A method for converting data associated with a metrology system, the method comprising:

[0213] accessing a first SEM data set acquired by a first scanning electron metrology (SEM) system and a second SEM data set acquired by a second SEM system, the first and second SEM data sets being associated with a patterned substrate; and

[0214] training a machine learning (ML) model using the first and second SEM data sets as training data, such that the trained ML model is configured to convert a metrology data set acquired by the second SEM system into a converted data set having characteristics comparable to metrology data acquired by the first SEM system.

[0215] 46. The method of clause 45, wherein training the ML model comprises:

[0216] comparing the first and second SEM data sets; and

[0217] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model.

[0218] 47. The method of clause 46, wherein the first and second SEM data sets comprise:

[0219] 48. The method of clause 47, wherein training the ML model comprises:

[0220] comparing a first set of images acquired by the first SEM system and a second set of images acquired by the second SEM system; and

[0221] adjusting parameters of the ML model based on the comparison to affect a cost function used to train the ML model to improve a match between the first set of images and ML generated images using the second set of images as input to the ML model.

[0222] 49. The method of clause 42, wherein the first and second SEM data sets comprise:

[0223] a profile of a feature on the patterned substrate; and / or

[0224] a physical characteristic associated with a pattern on the patterned substrate.

[0225] 50. The method of clause 47, wherein the physical characteristic comprises a critical dimension (CD) of a pattern on the patterned substrate.

[0226] 51. The method of clause 50, wherein training the ML model comprises:

[0227] comparing first CD values of the first SEM data set and second CD values of the second SEM data set; and

[0228] adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model, the cost function being a function of the first CD value and the second CD value, to improve the CD matching between the first SEM data set and the second SEM data set.

[0229] 52. The method of clause 49, wherein training the ML model comprises:

[0230] aligning the first set of images or the first contours of the first SEM data set with design layout images or design contours of the design layout;

[0231] aligning the second set of images or the second contours of the second SEM data set with design layout images or design contours of the design layout; and

[0232] using the aligned first set of images and the aligned second set of images as training data for training the machine learning model.

[0233] 53. The method of clause 52, wherein training the ML model comprises:

[0234] comparing intensity values from the first set of images and the second set of images; and

[0235] adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model.

[0236] 54. The method of clause 53, wherein training the ML model is an iterative process, each iteration comprising:

[0237] (i) transforming, via the ML model with the adjusted parameters and using the first SEM data set, the second SEM data set;

[0238] (ii) comparing intensity values from the first SEM data set with the transformed data set;

[0239] (iii) further adjusting parameters of the ML model based on the comparison to influence the cost function to be within a desired threshold;

[0240] (iv) determining whether the cost function is within the desired threshold; and

[0241] (v) responsive to the cost function not being within the desired threshold, repeating steps (i) through (iv).

[0242] 55. The method of clause 54, wherein the cost function is a difference between first intensity values from the first SEM data set and corresponding second intensity values from the second SEM data set or the transformed data set.

[0243] 56. The method of clause 53, wherein training the ML model comprises:

[0244] The intensity values are determined along the cutline drawn on the feature by the first SEM dataset and the second SEM dataset.

[0245] 57. The method of clause 56, wherein determining the intensity values comprises determining pixel intensity values along the cutline by the first SEM dataset and the second SEM dataset.

[0246] 58. The method of clause 57, wherein determining the intensity values comprises:

[0247] applying a first profile extraction algorithm associated with the first SEM system to the first SEM dataset; and

[0248] applying a second profile extraction algorithm associated with the second SEM system to the second SEM dataset.

[0249] 59. The method of clause 45, further comprising:

[0250] capturing, via the second SEM system, metrology data of another patterned substrate; and

[0251] converting, via the trained ML model, the captured metrology data to converted metrology data, the converted metrology data of the other patterned substrate having characteristics as if captured by the first SEM system.

[0252] 60. The method of clause 45, further comprising:

[0253] determining, based on the first SEM dataset and physical property measurements of the patterned substrate from the first SEM system, a metrology measurement recipe for the second SEM system;

[0254] capturing, using the second SEM system, metrology data of the patterned substrate;

[0255] converting, using the trained machine learning model, the captured metrology data; and

[0256] applying the metrology measurement recipe to the converted metrology data to determine another physical property measurement.

[0257] 61. The method of clause 60, wherein the physical property measurements comprise at least one of: critical dimension (CD) measurements, overlay measurements, and edge placement errors.

[0258] 62. The method of clause 61, wherein the metrology measurement recipe comprises a CD threshold that indicates where on the captured metrology data a CD measurement is to be taken.

[0259] 63. The method of clause 62, wherein determining a metrology measurement recipe comprises:

[0260] extracting contours from images of the first SEM dataset via a first contour extraction algorithm;

[0261] drawing a cutline at a location on the contours to measure a CD; and

[0262] determining a CD threshold corresponding to the measured CD based on signals along the cutline.

[0263] 64. The method of clause 45, wherein the first SEM system is manufactured by a first manufacturer and the second metrology system is manufactured by a second manufacturer.

[0264] 65. The method of clause 45, wherein the ML model is a convolutional neural network.

[0265] 66. The method of clause 45, wherein the ML model is trained using a generative adversarial network architecture, the ML model comprising a generator model and a discriminator model.

[0266] 67. The method of clause 66, wherein training the ML model comprises:

[0267] training the generator model in conjunction with the discriminator model using the second SEM dataset as input to generate data similar to the first SEM dataset; and

[0268] training the discriminator model to distinguish the generated data from the first SEM dataset.

[0269] 68. A method for converting SEM images acquired by a scanning electron microscope (SEM) system, the method comprising:

[0270] accessing a first set of SEM images acquired by a first SEM system and a second set of SEM images acquired by a second SEM system, the first and second sets of SEM images being associated with a patterned substrate; and

[0271] training a machine learning (ML) model using the first and second sets of SEM images as training data, such that the trained ML model is configured to convert a set of SEM images acquired by the second SEM system into a set of converted images having characteristics comparable to SEM images acquired by the first SEM system.

[0272] 69. The method of clause 68, wherein training the ML model comprises:

[0273] comparing the first and second sets of SEM images; and

[0274] adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model.

[0275] 70. The method of clause 68, wherein training the ML model comprises:

[0276] aligning the first set of SEM images to the design layout image;

[0277] aligning the second set of SEM images to the design layout image; and

[0278] using the aligned first set of SEM images and the aligned second set of SEM images as training data for training the machine learning model.

[0279] 71. The method of clause 69, wherein training the ML model comprises:

[0280] comparing intensity values from the first set of SEM images and the second set of SEM images; and

[0281] adjusting parameters of the ML model based on the comparison to influence a cost function to reduce intensity value differences between the first set of images and ML generated images using the second set of SEM images as input to the ML model.

[0282] 72. The method of clause 71, wherein training the ML model is an iterative process, each iteration comprising:

[0283] (i) converting the second set of SEM images via the ML model with the adjusted parameters and using the first set of SEM images;

[0284] (ii) comparing intensity values from the first set of SEM images to the converted set of images;

[0285] (iii) further adjusting parameters of the ML model based on the comparison to influence the cost function to be within a desired threshold;

[0286] (iv) determining whether the cost function is within the desired threshold; and

[0287] (v) in response to the cost function not being within the desired threshold, repeating steps (i) through (iv).

[0288] 73. The method of clause 72, wherein the cost function is a difference between first intensity values from the first set of SEM images and corresponding second intensity values from the second set of SEM images or the converted set of images.

[0289] 74. The method of clause 68, further comprising:

[0290] capturing SEM images of another patterned substrate via a second SEM system; and

[0291] convert the captured SEM images to converted SEM images via the trained ML model, the converted SEM images of the other patterned substrate having characteristics comparable to images captured by the first SEM system.

[0292] 75. The method of clause 68, further comprising:

[0293] determining a metrology measurement recipe for the second SEM system based on the first set of SEM images and CD measurements from the patterned substrate by the first SEM system;

[0294] capturing SEM images of the patterned substrate using the second SEM system;

[0295] converting the captured SEM images using the trained ML model; and

[0296] applying the metrology measurement recipe to the converted SEM images to determine another CD measurement.

[0297] 76. The method of clause 75, wherein the metrology measurement recipe includes a CD threshold that indicates where on the captured SEM images the CD measurements are taken.

[0298] 77. The method of clause 68, wherein the ML model is a convolutional neural network.

[0299] 78. The method of clause 68, wherein the ML model is trained using a generative adversarial network architecture, the ML model including a generator model and a discriminator model.

[0300] While the concepts disclosed herein can be used for imaging on substrates such as silicon wafers, it should be appreciated that the disclosed concepts can be used with any type of lithographic imaging system, such as those used for imaging on substrates other than silicon wafers.

[0301] The above description is intended to be illustrative, and not restrictive. Thus, it will be apparent to those skilled in the art that modifications can be made to the described embodiments without departing from the scope of the claims set forth below.

Claims

1. A non-transitory computer readable medium having instructions recorded thereon, the instructions when executed by a computer implement a method for converting data associated with a metrology system, the method comprising: accessing a first SEM data set acquired by a first scanning electron metrology (SEM) system and a second SEM data set acquired by a second SEM system, the first SEM data set and the second SEM data set associated with a patterned substrate; and training a machine learning (ML) model using the first SEM data set and the second SEM data set as training data, such that the trained ML model is configured to convert a metrology data set acquired by the second SEM system into a converted data set having characteristics comparable to metrology data acquired by the first SEM system.

2. The medium of claim 1, wherein the first SEM data set and the second SEM data set comprise SEM images of the patterned substrate, wherein training the ML model comprises: comparing a first image set acquired by the first SEM system and a second image set acquired by the second SEM system; and adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model to improve a match between the first image set and images generated by the ML using the second image set as input to the ML model.

3. The medium of claim 1, wherein the first SEM data set and the second SEM data set comprise: a profile of a feature on the patterned substrate; and / or a physical characteristic associated with a pattern on the patterned substrate.

4. The medium of claim 3, wherein the physical characteristic comprises a critical dimension (CD) of the pattern on the patterned substrate.

5. The medium of claim 1, wherein training the ML model comprises: comparing a first CD value of the first SEM data set and a second CD value of the second SEM data set; and adjusting parameters of the ML model based on the comparison to influence a cost function used to train the ML model, the cost function a function of the first CD value and the second CD value, to improve a CD match between the first SEM data set and the second SEM data set.

6. The medium of claim 1, wherein training the ML model comprises: aligning a first image set or a first profile of the first SEM data set to a design layout image or a design profile of a design layout; aligning a second image set or a second profile of the second SEM data set to the design layout image or the design profile of the design layout; and using the aligned first image set and the aligned second image set as training data for training the machine learning model.

7. The medium of claim 2, wherein training the ML model comprises: comparing pixel intensity values from the first image set and the second image set; and adjusting parameters of the ML model based on the comparison to influence the cost function used to train the ML model.

8. The medium of claim 7, wherein the training the ML model further comprises: determining intensity values from the first and second image sets by performing the following steps: applying a first profile extraction algorithm associated with the first SEM system to the first SEM data set; and applying a second profile extraction algorithm associated with the second SEM system to the second SEM data set.

9. The medium of claim 1, further comprising: capturing, via the second SEM system, metrology data of another patterned substrate; and converting, via the trained ML model, the captured metrology data to converted metrology data, the converted metrology data of the other patterned substrate having characteristics as if captured by the first SEM system.

10. The medium of claim 1, further comprising: determining a metrology measurement recipe for the second SEM system based on the first SEM data set and physical property measurements of the patterned substrate from the first SEM system; capturing, using the second SEM system, metrology data of the patterned substrate; converting, using the trained machine learning model, the captured metrology data; and applying the metrology measurement recipe to the converted metrology data to determine another physical property measurement.

11. The medium of claim 10, wherein the physical property measurements include at least one of: critical dimension (CD) measurements, overlay measurements, and edge placement errors.

12. The medium of claim 10, wherein the metrology measurement recipe includes a CD threshold that indicates a location on the captured metrology data where a CD measurement is taken.

13. The medium of claim 10, wherein determining the metrology measurement recipe includes: extracting, via a first profile extraction algorithm, profiles from images of the first SEM data set; plotting cut lines at locations across the profiles to measure CD; and determining, based on signals along the cut lines, a CD threshold corresponding to the measured CD.

14. The medium of claim 1, wherein the first SEM system is manufactured by a first manufacturer and the second SEM system is manufactured by a second manufacturer.

15. The medium of claim 1, wherein the ML model is trained using a generative adversarial network architecture, the ML model including a generator model and a discriminator model.

Images