SEM image enhancement method and system

Abstract

Systems and methods for image enhancement using a multi-beam device are disclosed. A method for enhancing an image includes acquiring a first scanning electron microscope (SEM) image by using coaxial beams of a multi-beam device. The method further includes acquiring a second SEM image by using an off-axis beam of the multi-beam device. The method further includes providing an enhanced image by enhancing the second SEM image using the first SEM image as a reference. The enhanced image may be provided by enhancing the second SEM image using one or more features extracted from the first image or by numerically enhancing the second SEM image using the first SEM image as a reference.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is a divisional application of the Chinese national phase patent application of the PCT application with international application number PCT / EP2019 / 067555, Chinese national application number 201980046684.X, and invention name “SEM image enhancement method and system”. Technical Field

[0003] The description herein relates to the field of image enhancement, and more particularly to scanning electron microscope (SEM) image enhancement. Background Art

[0004] In the manufacturing process of integrated circuits (ICs), unfinished or completed circuit components are inspected to ensure that they are manufactured according to the design and are free of defects. Inspection systems using optical microscopes or charged particle (e.g., electron) beam microscopes such as scanning electron microscopes (SEMs) can be used. SEMs deliver low energy electrons (e.g., <1 keV) to a surface and use detectors to record secondary and / or backscattered electrons that leave the surface. By recording such electrons for different excitation locations on the surface, images with nanometer-scale spatial resolution can be produced.

[0005] As the physical size of IC components continues to shrink, the accuracy and yield of defect detection become increasingly important. However, the imaging resolution and yield of inspection tools are difficult to keep up with the ever-decreasing feature size of IC components. There are several techniques that can be used to increase yield, including, for example, 1) reducing the amount of data averaging, 2) increasing the inspection pixel size, 3) increasing the beam current, and 4) using multiple beams to perform inspections. However, it should be noted that when these techniques are used, the image quality inevitably deteriorates. Specifically, it should be noted that the use of techniques (such as reducing the amount of data averaging or increasing the inspection pixel size) will reduce the amount of sampling, which in turn reduces the image quality. The use of techniques (such as increasing the beam current or using multiple beams) will increase the spot size, which will also reduce the image quality. In addition, when multiple beams are used to perform inspections, off-axis beams may suffer resolution loss due to aberrations, thereby further reducing the image quality. Further improvements in this area are expected. Summary of the invention

[0006] Embodiments of the present disclosure provide systems and methods for image enhancement. In some embodiments, the method for enhancing an image may include acquiring a first scanning electron microscope (SEM) image at a first resolution. The method may also include acquiring a second SEM image at a second resolution. The method may also include providing an enhanced image by enhancing the second SEM image using the first SEM image as a reference. In some embodiments, the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0007] In some embodiments, an inspection system is disclosed. The inspection system may include a memory storing an instruction set and a processor configured to execute the instruction set. The processor may execute the instruction set to cause the inspection system to acquire a first scanning electron microscope (SEM) image at a first resolution and acquire a second SEM image at a second resolution. The processor may also execute the instruction set to cause the inspection system to provide an enhanced image by enhancing the second SEM image using the first SEM image as a reference. In some embodiments, the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0008] In some embodiments, a non-transitory computer-readable medium is disclosed. The non-transitory computer-readable medium may store a set of instructions executable by at least one processor of a device to cause the device to perform a method. The method may include acquiring a first scanning electron microscope (SEM) image at a first resolution, and acquiring a second SEM image at a second resolution. The method may also include providing an enhanced image by enhancing the second SEM image using the first SEM image as a reference. In some embodiments, the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a schematic diagram illustrating an exemplary electron beam inspection (EBI) system consistent with embodiments of the present disclosure.

[0010] Figure 2 The embodiment consistent with the present disclosure may be Figure 1 A schematic diagram of an exemplary electron beam tool that is a portion of an exemplary electron beam inspection system.

[0011] Figure 3is a block diagram of an exemplary image enhancement system consistent with embodiments of the present disclosure.

[0012] Figure 4 is a block diagram illustrating an exemplary image enhancement system consistent with embodiments of the present disclosure.

[0013] Figure 5 is a flow chart illustrating an exemplary image enhancement method consistent with embodiments of the present disclosure.

[0014] Figure 6 is a flow chart illustrating an exemplary image enhancement method using a machine learning network consistent with embodiments of the present disclosure.

[0015] Figure 7 is a flow chart illustrating an exemplary image enhancement method consistent with embodiments of the present disclosure.

[0016] Figure 8 is a diagram depicting an exemplary deconvolution process consistent with embodiments of the present disclosure.

[0017] Fig. 9 is a flow chart illustrating an exemplary image enhancement method consistent with embodiments of the present disclosure.

[0018] Fig.10 is a diagram depicting an exemplary feature identification process consistent with an embodiment of the present disclosure. DETAILED DESCRIPTION

[0019] Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings, wherein, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments set forth in the following description of the exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, these embodiments are merely examples of apparatus and methods consistent with aspects related to the subject matter described in the accompanying claims. For example, although some embodiments are described in the context of utilizing electron beams, the present disclosure is not limited thereto. Other types of charged particle beams may be similarly applied. In addition, other imaging systems such as optical imaging, light detection, x-ray detection, etc. may be used.

[0020] The increased computing power of electronic devices (while reducing the physical size of the devices) can be achieved by significantly increasing the packaging density of circuit components (such as transistors, capacitors, diodes, etc.) on IC chips. For example, in a smart phone, an IC chip (which is the size of a thumb) can include more than 2 billion transistors, each of which is less than 1 / 1000 the size of a human hair. Not surprisingly, semiconductor IC manufacturing is a complex process with hundreds of individual steps. Even an error in one step can greatly affect the functionality of the final product. Even one "fatal defect" can cause device failure. The goal of the manufacturing process is to improve the overall yield of the process. For example, for a 50-step process to obtain a 75% yield, each individual step must have a yield greater than 99.4%, and if the individual step yield is 95%, the yield of the overall process will drop to 7%.

[0021] While high process yields are required in IC chip manufacturing facilities, it is also important to maintain high wafer throughput, which is defined as the number of wafers processed per hour. High process yields and high wafer throughput can be affected by the presence of defects, especially when operator intervention is involved. Therefore, detection and identification of micron and nanometer-sized defects by inspection tools (such as SEM) is critical to maintaining high yields and low costs.

[0022] In order to maintain high throughput, SEM inspection tools can often be operated in high throughput mode. However, compared to images obtained in normal mode, SEM inspection tools operating in high throughput mode may reduce the amount of data averaging or increase the inspection pixel size, or a combination of both, which can lead to image quality degradation. In high throughput mode, image enhancement and restoration become challenging, especially because small-scale features may be severely distorted, lost, or misrepresented. On the other hand, traditional single-image-based image processing methods may be inappropriate because it is time-consuming. Therefore, further improvements in the field of defect detection and identification need to be accompanied by maintaining high throughput and high yield.

[0023] In one aspect of some embodiments of the present disclosure, when in high throughput mode, the SEM inspection tool may be used to acquire low resolution inspection images (such as Figure 4 Using the features of the low-resolution inspection image, the inspection tool can identify one or more stored high-resolution inspection images (such as Figure 4 Using pattern information from the high-resolution inspection image, the SEM inspection tool can improve the low-resolution inspection image (such as Figure 4 Thus, low resolution inspection images can be enhanced while maintaining high throughput of the SEM inspection tool.

[0024] Further, some embodiments of the SEM inspection tool may be configured to: analyze data representing the first SEM image (such as Figure 8 810) to obtain one or more spatial-spectral characteristics (e.g., phase and amplitude characteristics), which can be applied to data representing a second SEM image (such as Figure 8 820) to numerically enhance the second SEM image (to obtain, for example, Figure 8 Data 830 representing a portion of the enhanced image).

[0025] The SEM inspection tool may also be configured to analyze data representing the first SEM image (such as Fig.10 The inspection tool may numerically blur the data representing the first SEM image to produce simulated data (such as data representing a portion of the high resolution image acquired at a second resolution) to identify features (such as edges) in the first SEM image. The inspection tool may numerically blur the data representing the first SEM image to produce simulated data (such as data representing a blurred SEM image acquired at a second resolution) Fig.10 The inspection tool may also determine a portion of the simulated data that represents a blurred SEM image (e.g., Fig.10 Does the simulated data 1020) fit the data representing the second SEM image (such as Fig.10 If the inspection tool determines that there is a fit, the inspection tool can identify the portion of the data representing the second SEM image as containing the location of a feature (e.g., an edge) identified in the first SEM image.

[0026] In some embodiments, the inspection tool can be configured to focus the inspection on certain features, such as edge locations, etc. In such embodiments, rather than attempting to enhance the image quality of the entire second SEM image, the inspection tool can be configured to enhance the image quality of only certain areas of interest, which can also improve its accuracy and yield.

[0027] It is contemplated that embodiments of the present disclosure may provide a charged particle system that may be used for charged particle imaging. The charged particle system may be used as a SEM inspection tool for imaging and inspecting a sample for detecting defects. A defect may refer to an abnormal condition of a component on a sample or wafer that may cause a failure. In some embodiments, a defect may refer to an aberration, such as a photoresist profile, particle contamination, a surface defect, etc., compared to a standard.

[0028] For the sake of clarity, the relative sizes of the components in the drawings may be exaggerated. Within the following description of the figures, identical or similar reference signs refer to identical or similar components or entities and only the differences with respect to individual embodiments are described.

[0029] As used herein, unless otherwise expressly stated, the term "or" encompasses all possible combinations unless not feasible. For example, if it is stated that a database may include A or B, then unless otherwise expressly stated or not feasible, the database may include A, or B, or A and B. As a second example, if it is stated that a database may include A, B, or C, then unless otherwise expressly stated or not feasible, the database may include A, or B, or C, or A and B, or A and C, or B and C, or A and B and C.

[0030] Embodiments of the present disclosure include methods and systems for enhancing SEM images. For example, a SEM inspection tool may acquire a first SEM image at a first resolution and acquire a second SEM image at a second resolution. Assuming that the second resolution is lower than the first resolution, the SEM inspection tool may enhance the second SEM image using the first SEM image as a reference. In some embodiments, an enhanced image may be provided by enhancing the second SEM image using one or more features extracted from the first image. Features may be extracted using image processing and / or machine learning. Alternatively or additionally, an enhanced image may be provided by numerically enhancing and restoring the second SEM image using the first SEM image as a reference.

[0031] Reference now Figure 1 , which illustrates an exemplary electron beam inspection (EBI) system 100 consistent with embodiments of the present disclosure. The EBI system 100 can be used for imaging. Figure 1 As shown in , the EBI system 100 includes a main chamber 101, a load / lock chamber 102, an electron beam tool 104, and an equipment front end module (EFEM) 106. The electron beam tool 104 is positioned within the main chamber 101. The EFEM 106 includes a first load port 106a and a second load port 106b. The EFEM 106 may include (multiple) additional load ports. The first load port 106a and the second load port 106b receive wafer front opening wafer transport boxes (FOUPs) that contain wafers to be inspected (e.g., semiconductor wafers or wafers made of (multiple) other materials) or samples (wafers and samples can be used interchangeably). A "batch" is a plurality of wafers that can be loaded for batch processing.

[0032] One or more robotic arms (not shown) in the EFEM 106 can transport the wafer to the load / lock chamber 102. The load / lock chamber 102 is connected to a load / lock vacuum pump system (not shown), which removes gas molecules in the load / lock chamber 102 to reach a first pressure below atmospheric pressure. After reaching the first pressure, one or more robotic arms (not shown) can transport the wafer from the load / lock chamber 102 to the main chamber 101. The main chamber 101 is connected to a main chamber vacuum pump system (not shown), which removes gas molecules in the main chamber 101 to reach a second pressure below the first pressure. After reaching the second pressure, the wafer is inspected by the electron beam tool 104. The electron beam tool 104 can be a single beam system or a multi-beam system.

[0033] The controller 109 is electrically connected to the electron beam tool 104. The controller 109 may be a computer configured to perform various controls of the EBI system 100. Figure 1 106 , but it should be understood that the controller 109 may be part of the structure.

[0034] Figure 2 An exemplary imaging system 200 is illustrated in accordance with an embodiment of the present disclosure. Figure 2 The electron beam tool 104 can be configured for use in the EBI system 100. The electron beam tool 104 can be a single beam device or a multi-beam device. Figure 2 As shown in FIG. 1 , the electron beam tool 104 includes a motorized sample stage 201, and a wafer holder 202 supported by the motorized sample stage 201 to hold a wafer 203 to be inspected. The electron beam tool 104 also includes an objective lens assembly 204, an electron detector 206 (which includes electron sensor surfaces 206a and 206b), an objective lens aperture 208, a condenser lens 210, a beam limiting aperture 212, an electron gun aperture 214, an anode 216, and a cathode 218. In some embodiments, the objective lens assembly 204 may include a modified swing objective delayed immersion lens (SORIL) including a pole piece 204a, a control electrode 204b, a deflector 204c, and an excitation coil 204d. The electron beam tool 104 may additionally include an energy dispersive X-ray spectrometer (EDS) detector (not shown) to characterize the material on the wafer 203.

[0035] The primary electron beam 220 is emitted from the cathode 218 by applying a voltage between the anode 216 and the cathode 218. The primary electron beam 220 passes through the electron gun aperture 214 and the beam limiting aperture 212, both of which can determine the size of the electron beam entering the condenser lens 210, which resides below the beam limiting aperture 212. The condenser lens 210 focuses the primary electron beam 220 before the beam enters the objective lens aperture 208 to set the size of the electron beam before entering the objective lens assembly 204. The deflector 204c deflects the primary electron beam 220 to facilitate beam scanning on the wafer. For example, during the scanning process, the deflector 204c can be controlled to sequentially deflect the primary electron beam 220 to different locations on the top surface of the wafer 203 at different points in time to provide data for image reconstruction of different portions of the wafer 203. In addition, the deflector 204c can also be controlled to deflect the primary electron beam 220 to different sides of the wafer 203 at different time points at a specific location to provide data for stereoscopic image reconstruction of the wafer structure at that location. In addition, in some embodiments, the anode 216 and the cathode 218 can be configured to generate multiple primary electron beams 220, and the electron beam tool 104 can include multiple deflectors 204c to simultaneously project multiple primary electron beams 220 to different parts / sides of the wafer to provide data for image reconstruction for different parts of the wafer 203.

[0036] The excitation coil 204d and the pole piece 204a generate a magnetic field that starts at one end of the pole piece 204a and ends at the other end of the pole piece 204a. A portion of the wafer 203 scanned by the primary electron beam 220 can be immersed in the magnetic field and can be charged, which in turn creates an electric field. The electric field reduces the energy of the primary electron beam 220 that strikes the wafer 203 near the surface before colliding with the wafer 203. The control electrode 204b, which is electrically isolated from the pole piece 204a, controls the electric field on the wafer 203 to prevent the wafer 203 from micro-bowing and ensure proper beam focusing.

[0037] After receiving the primary electron beam 220, the secondary electron beam 222 may be emitted from a portion of the wafer 203. The secondary electron beam 222 may form a beam spot on the sensor surfaces 206a and 206b of the electron detector 206. The electron detector 206 may generate a signal (e.g., voltage, current, etc.) representing the intensity of the beam spot and provide the signal to the image processing system 250. The intensity of the secondary electron beam 222 and the resulting beam spot may vary according to the external or internal structure of the wafer 203. In addition, as discussed above, the primary electron beam 220 may be projected onto different locations on the top surface of the wafer or onto different sides of the wafer at a specific location to generate secondary electron beams 222 (and resulting beam spots) of different intensities. Therefore, by mapping the intensity of the beam spot using the location of the wafer 203, the processing system may reconstruct an image reflecting the internal or surface structure of the wafer 203.

[0038] The imaging system 200 can be used to inspect a wafer 203 on a sample stage 201 and includes an electron beam tool 104, as discussed above. The imaging system 200 can also include an image processing system 250, which includes an image acquisition device 260, a storage device 270, and a controller 109. The image acquisition device 260 can include one or more processors. For example, the image acquisition device 260 can include a computer, a server, a mainframe, a terminal, a personal computer, any kind of mobile computing device, etc., or a combination thereof. The image acquisition device 260 can be connected to the detector 206 of the electron beam tool 104 through a medium (such as an electrical conductor, a fiber optic cable, a portable storage medium, IR, Bluetooth, the Internet, a wireless network, a wireless radio, or a combination thereof). The image acquisition device 260 can receive a signal from the detector 206 and can construct an image. Therefore, the image acquisition device 260 can acquire an image of the wafer 203. The image acquisition device 260 can also perform various post-processing functions, such as generating a contour, superimposing an indicator on the acquired image, etc. The image acquirer 260 may be configured to perform adjustments to the brightness and contrast of the acquired image, etc. The storage device 270 may be a storage medium such as a hard disk, a cloud storage device, a random access memory (RAM), other types of computer-readable memory, etc. The storage device 270 may be coupled to the image acquirer 260 and may be used to save the scanned raw image data as an initial image and a post-processed image. The image acquirer 260 and the storage device 270 may be connected to the controller 109. In some embodiments, the image acquirer 260, the storage device 270, and the controller 109 may be integrated together into a control unit.

[0039] In some embodiments, the image acquirer 260 can acquire one or more images of the sample based on the imaging signal received from the detector 206. The imaging signal can correspond to a scanning operation for performing charged particle imaging. The acquired image can be a single image including multiple imaging regions. The single image can be stored in the storage device 270. The single image can be an initial image, and the initial image can be divided into multiple regions. Each region can include an imaging region containing a feature of the wafer 203.

[0040] Reference now Figure 3 , which is a schematic diagram of an exemplary image enhancement system 300 consistent with an embodiment of the present disclosure. In some embodiments, the image enhancement system 300 may be Figure 2 The image enhancement system 300 may include the image processing system 250, which includes the controller 109, the image acquirer 260, the storage device 270, and the like.

[0041] The image enhancement system 300 may include a high-resolution image 310 , an information file 315 , a machine learning network 320 , an inspection image 330 , a pattern extractor 340 , a storage module 350 , an image enhancement module 360 ​​, a comparator 370 , an image enhancer 380 , and a display device 390 .

[0042] The high resolution image 310 may be a high resolution image of a portion of a sample or wafer. As used herein, a high resolution image refers to, but is not limited to, an image having a resolution high enough to resolve two different features in the image having a spacing of less than 20 nm. It should be understood that the image resolution may depend on various factors, including, but not limited to, the amount of signal averaging used for image acquisition, the noise ratio of the SEM image frame, the pixel size, the SEM beam width of the coaxial beam of a multi-beam system, the SEM beam width of a single beam system, or the current supplied to the (multiple) SEM beams, etc. It is contemplated that one or more of the factors listed above may be adjusted to provide a desired resolution for acquiring the high resolution image 310. For example, a small spot width, a small pixel size, a low current, minimal beam profile aberrations, a high signal averaging, etc. are all factors that may contribute to improving the resolution of the high resolution image 310.

[0043] The high resolution image 310 may be acquired using the image acquirer 260 of the EBI system 100 or any such system capable of acquiring a high resolution image. The high resolution image 310 may be acquired by any inspection system that can generate an inspection image of a wafer. For example, the wafer may be a semiconductor wafer substrate, or a semiconductor wafer substrate having one or more epitaxial layers or process films. The embodiments of the present disclosure are not limited to a particular type of inspection system, so long as the system can generate a wafer image with sufficient resolution.

[0044] In some embodiments, the high resolution image 310 may be acquired in an offline mode and used as a training image. As used herein, an offline mode refers to an operating mode of the image acquirer 260 or the EBI system 100 when the system is not being used for wafer processing in a production run. For example, the offline mode may include operation of the image acquirer 260 or the EBI system 100 before an actual inspection begins or before an actual processing run. In some embodiments, the high resolution image 310 is acquired by an image acquirer positioned separately from the processing equipment, which includes, for example, a stand-alone EBI system or image acquirer. Since image enhancement can be performed by comparing the acquired high throughput inspection image 330 with an already existing high resolution image 310, acquiring the high resolution image 310 in an offline mode can help significantly increase the throughput of the SEM inspection tool.

[0045] Alternatively, in some embodiments, the high-resolution image 310 may be acquired during the inspection process along with acquiring the high-throughput inspection image(s) 330. For example, if the image acquirer 260 or the EBI system 100 implements a multi-beam system, the image acquirer 260 or the EBI system 100 may be configured to acquire the high-resolution image 310 using the coaxial beam of the multi-beam system and to acquire the high-throughput inspection image(s) 330 using the off-axis beam(s). The image acquirer 260 or the EBI system 100 may also be configured to supply different levels of current to the single beam system in an attempt to acquire the high-resolution image 310 (using a low-current beam, or when a lower current is supplied to the beam system) and the high-throughput inspection image(s) 330 (using a high-current beam, or when a higher current is supplied to the beam system). Thus, it should be understood that the high-resolution image 310 may be acquired before, after, or simultaneously with acquiring the high-throughput inspection image(s) 330.

[0046] It should also be understood that the image acquirer 260 or the EBI system 100 can be configured to acquire more than one high resolution image 310. Such high resolution images 310 may include, for example, reference images of locations on the wafer 203, or reference images or post-processing reference images of features on the wafer 203, etc. In some embodiments, the (multiple) high resolution images 310 may be one or more reference images of locations or features of wafers from multiple product types. For example, when multiple product types are manufactured using the same process in the same fab line, a reference image of a specific feature from a first product type may be used as a high resolution image for a specific feature from a second product type that has a different build than the first product type.

[0047] In some embodiments, high-resolution image 310 may be a review-mode image, which is an image acquired under optimal acquisition conditions. Review-mode images may have higher resolution, such as optimal magnification, optimal contrast and brightness, optimal electron beam intensity, etc. The settings of detector 206 may also be optimized to acquire high-resolution images.

[0048] In some embodiments, the image enhancement system 300 may include an information file 315 containing reference feature information. The information file 315 may include a chip design layout in a Graphic Database System (GDS) format, a Graphic Database System II (GDS II) format (including a graphical representation of features on the surface of the chip), or an Open Source System Interchange Standard (OASIS) format or a Caltech Intermediate Format (CIF), etc. The chip design layout may be based on a pattern layout used to construct a chip. For example, the chip design layout may correspond to one or more photolithography masks or masks used to transfer features from a photolithography mask or mask to the chip 203. Among other things, a GDS information file or an OASIS information file, etc. may include feature information representing planar geometry, text, and other information related to the chip design layout stored in a binary file format. The OASIS format can help significantly reduce the amount of data, resulting in a more efficient data transfer process. A large number of GDS or OASIS format images may have been collected and may constitute a large number of data sets for comparing features.

[0049] In some embodiments, the image enhancement system 300 may further include a machine learning network 320. The machine learning network 320 may be configured to extract feature information from the high-resolution image 310. The machine learning network 320 may also extract relevant features from an information file 315 including a GDS format file or an OASIS format file. The machine learning network 320 may include, for example, an artificial intelligence system, a neural network or deep learning technology, a software-implemented algorithm, etc. The feature extraction architecture of the machine learning network 320 may include, for example, a convolutional neural network. In some embodiments, a linear classifier network of a deep learning architecture may be used as a starting point to train and construct a feature extraction architecture for the machine learning network 320.

[0050] In some embodiments, the machine learning model may include multiple layers. For example, the architecture of a convolutional neural network may include an input, a first convolution, a first pooling, a second convolution, a second pooling, one or more hidden layers, an activation layer, and an output layer. Based on the nature and complexity of the features, each layer of the architecture may produce a different number of subsamples. After the first convolution operation, less than 10 subsamples may be generated in the first pool. However, after the second convolution operation, the second layer may have more than 10 subsamples generated in the second pool. In some embodiments, variations between layers may be introduced by the complexity of the geometric features of the layout. Features with more geometric information may have a higher probability of generating more subsamples. For example, complex features may exhibit various subshapes that can be decomposed and analyzed into separate attributes.

[0051] In some embodiments, the machine learning network 320 may receive the high-resolution image 310 or the information file 315 to extract relevant features and knowledge. The machine learning network 320 may include a temporary storage medium (not shown) to store the received information file or high-resolution image. The temporary storage medium may also be configured to store post-processing data (e.g., such as the extracted relevant features). In some embodiments, the feature extraction algorithm or the deep learning algorithm or the neural network may be configured to include the following steps: Retrieve the high-resolution image or information file from the temporary storage medium or storage module 350 (discussed later).

[0052] The machine learning network 320 may also receive additional training images as input. Such training images may include, for example, a wafer design plan based on a GDS or OASIS design or a review mode image with high resolution, additional SEM images acquired at high resolution, and the like. Such training images, together with the high resolution images 310, may be collectively referred to as training data and may be stored in a user-defined storage device, database, or storage module 350 accessible by a user. The training data may be fed into a machine learning network 320 designed to extract trained features and knowledge from the training data. The extracted feature information and knowledge may be stored in a storage module 350, which may be configured to access other components of the image enhancement system 300. In some embodiments, the extraction of trained features using the machine learning network 320 may be performed offline so that these steps do not adversely affect the overall inspection yield.

[0053] In some embodiments, the machine learning network 320 is a self-supervised network configured to extract one or more trained features from training data. The machine learning network 320 can be configured to train itself to extract one or more trained features from the high-resolution image 310, the additional training images, and from the information file 315 based on previously identified trained features. In some embodiments, the machine learning network 320 can extract trained features from the high-resolution image 310, the additional training images, and from the information file 315 in an offline mode.

[0054] In some embodiments, the image enhancement system 300 can acquire an inspection image 330 as a low-resolution image of a sample, a feature of a wafer 203, an area of ​​interest on a wafer 203, or the entire wafer 203. The inspection image 330 can be acquired using the image acquirer 260 of the EBI system 100 or any such inspection system capable of acquiring a low-resolution image. The inspection image 330 can be acquired by any inspection system that can generate an inspection image of a wafer or an area of ​​interest on a wafer. For example, the wafer 203 can be a semiconductor wafer substrate or a semiconductor wafer substrate having one or more epitaxial layers or process films. The embodiments of the present disclosure do not limit the specific type of inspection system, as long as the system can generate a wafer image. In some embodiments, the inspection image 330 can be an optical image acquired, for example, using an optical microscope.

[0055] In some embodiments, the inspection image 330 is a high throughput mode image acquired in-line during wafer processing. The inspection image 330 acquired in this way may be a degraded, distorted, inferior or artifact image of the features of the wafer.

[0056] In some embodiments, the image enhancement system 300 may include a pattern extractor 340. The pattern extractor 340 may be configured to extract global structural information or patterns from the inspection image 330 in real time, such as online, during wafer processing. In some embodiments, the pattern extractor 340 may be a mathematical algorithm, a software-implemented algorithm, an image processing algorithm, etc. The pattern extractor 340 may be integrated into the image acquirer 260, or may be configured to operate as a separate independent unit that is configured to process the inspection image 330. In some embodiments, the pattern extractor 340 may include an image processing unit (not shown) that is configured to adjust the brightness, contrast, saturation, flatness, noise filtering, etc. of the inspection image 330 before storage. In some embodiments, the pattern extractor 340 may extract pattern information from an already stored inspection image 330.

[0057] In some embodiments, the pattern extractor 340 may include a feature extraction algorithm to extract relevant pattern information from the inspection image 330. The extracted relevant pattern information may include global information, such as global structural features, global patterns, reference fiducials, etc. The extracted relevant pattern information may be stored in a storage module configured to be accessed by other components of the image enhancement system 300.

[0058] The pattern extractor 340 may also be configured to extract global structure information, feature information, etc. from the inspection image 330 .

[0059] In some embodiments, the image enhancement system 300 may include a storage module 350. The storage module 350 may be configured to store one or more high-resolution images 310, training images, information files 315, inspection images 330, relevant features extracted from the machine learning network 320, pattern information extracted from the pattern extractor 340, etc. The storage module 350 may also be configured to share the stored information with components of the image enhancement system 300 (including, for example, the machine learning network 320 and the pattern extractor 340). It should be understood that the storage module 350 may be Figure 2 A portion of the storage device 270.

[0060] In some embodiments, storage module 350 may be an integrated storage medium of image enhancement system 300, configured to connect with each component of image enhancement system 300. For example, storage module 350 may be a remote storage module accessible via wireless communication via the Internet, a cloud platform, or a suitable Wi-Fi communication path.

[0061] The storage module 350 may include a pattern and feature / knowledge base. The disclosed method may search for matching patterns in a pattern library based on correspondence. For example, if a feature design profile has a similarity of 90% or more with another pattern in the pattern library, a pattern match may be determined. The patterns in the pattern library may include previously extracted patterns, standard patterns (such as patterns of standard IC features), and the like.

[0062] The image enhancement system 300 may include an image enhancement module 360 ​​including one or more processors and a storage device (e.g., such as the storage module 350). The image enhancement module 360 ​​may include a comparator 370, an image enhancer 380, and a display device 390. It should be appreciated that the image enhancement module 360 ​​may also include one or more of the pattern extractor 340, the comparator 370, the image enhancer 380, and the display device 390, or a combination thereof.

[0063] The display device 390 can be configured to display the enhanced image from the image intensifier 380, or the inspection image 330, or the high-resolution image 310, etc. In some embodiments, the display device 390 can display pre-processed images and post-processed images of the wafer, the area of ​​interest, the features on the wafer, etc., including, for example, the enhanced inspection image. The image enhancement module 360 ​​can be connected to the storage module 350 as a single integrated unit, and the connection is configured to allow data to be shared between the image enhancement module 360 ​​and the storage module 350. In some embodiments, the comparator 370, the image intensifier 380, and the display device 390 can be connected to the storage module 350 separately. Other connection combinations are also possible.

[0064] In some embodiments, the comparator 370 is configured to compare the relevant information extracted from the machine learning network 320 with the pattern information extracted from the pattern extractor 340. In some embodiments, the comparator 370 is configured to identify trained features from the extracted trained features of the high-resolution image 310 based on the pattern information of the inspection image 330 from the pattern extractor 340, and compare the identified extracted trained features with the extracted pattern information.

[0065] The comparator 370 may include an image processing algorithm, a software-implemented algorithm, etc. In some embodiments, the comparator 370 may be configured to communicate with the storage module 350. The comparator 370 may generate an output file including the matching results between the identified extracted trained features and the extracted pattern information. The generated output file may be stored in the storage module 350. The comparator 370 may also be configured to access the high-resolution image 310, the training image, the information file 315, the extracted trained features, etc. from the storage module 350 for matching purposes.

[0066] In some embodiments, comparator 370 may include a processing unit, memory, a display, and communication paths to interact with other components of image enhancement system 300 (eg, machine learning network 320 and pattern extractor 340 ).

[0067] In some embodiments, image enhancer 380 may be configured to generate an enhanced image of inspection image 330 based on one or more matching results from comparator 370. In some embodiments, image enhancer 380 receives an output file generated by comparator 370 and inspection image 330. Alternatively, comparator 370 may be integrated with image enhancer 380 as a single unit configured to compare, identify, and generate an enhanced image based on the matching results generated by comparator 370.

[0068] In some embodiments, the image enhancer 380 may be configured to analyze the data representing the high-resolution image 310 to obtain one or more spatial-spectral characteristics (e.g., including phase and amplitude characteristics). The image enhancer 380 may apply the spatial-spectral characteristics to the data representing the high-throughput (or low-resolution) inspection images 330 to numerically enhance the inspection images 330.

[0069] In some embodiments, the image enhancer 380 may be configured to analyze the data representing the high resolution image 310 to identify features (e.g., edges) in the high resolution image 310. The image enhancer 380 may numerically blur the data representing the high resolution image 310 to simulate data representing a blurred image acquired at a lower resolution. The image enhancer 380 may determine whether a portion of the data representing the blurred image fits a portion of the data representing the low resolution inspection image 330. If the image enhancer 380 determines that there is a fit, the image enhancer 380 may identify the portion of the data representing the low resolution inspection image 330 as containing the location of the identified feature (e.g., edge).

[0070] In some embodiments, the image enhancer 380 can be configured to focus the enhancement on certain features, such as edges, etc. In such embodiments, rather than attempting to enhance the image quality of the entire low-resolution inspection image 330, the image enhancer 380 can be configured to enhance the image quality of only certain areas of interest, which can also improve its accuracy and yield.

[0071] In some embodiments, the image enhancer 380 is an executable application or software. The image enhancer 380 may include software-implemented algorithms, image processing algorithms, or mathematical algorithms, among others.

[0072] The image enhancement module 360 ​​may include an output device or display device 390 configured to display the generated enhanced inspection image. The display device 390 may be integrally connected to the EBI system 100. In some embodiments, the display device 390 may be a handheld display device, a wearable display device, a multi-screen display, an interactive display device, etc. Other suitable display devices may also be used.

[0073] The enhanced inspection image may be displayed on a display device 390 of the image enhancement system 300. The display device 390 may be integrated within the electron beam tool 104, or may be a separate output device located in a remote location.

[0074] In some embodiments, the display device 390 can be remotely located and operated via a wireless communication network (e.g., Wi-Fi, the Internet, or a cloud network). Other suitable wireless communication networks and platforms may also be used. The display device 390 may also be connected to the storage module 350 to store displayed images of samples or areas of interest on the wafer. The display device 390 may also be used to display the real-time pre-processed inspection image 330.

[0075] Figure 4 An exemplary image enhancement system 300 consistent with embodiments of the present disclosure is illustrated. In some embodiments, the image enhancement system 300 may be Figure 2 The image enhancement system 300 may include the image processing system 250, which includes the controller 109, the image acquirer 260, the storage device 270, and the like.

[0076] The image enhancement system 300 may store high-resolution images 310. As illustrated, the high-resolution images 310 are high-resolution images from which relevant trained features may be extracted. The image enhancement system 300 may use the high-resolution images 310 to train a machine learning network 320 before performing a high-throughput inspection. The high-resolution images 310 may include high-resolution optical microscope images, high-resolution secondary electron microscope images, backscattered electron beam images, atomic force microscope images, and the like.

[0077] As shown in the figure, Figure 4 The high resolution image 310 of the wafer may include trained features 430. The trained features 430 may include, for example, one or more circular structures arranged in a stripe arrangement, one or more strips of substrate material, metal interconnects, spaces between metal interconnects, contact pads, edges, etc., which are defined by features on a mask and transferred to a wafer or substrate through a photolithography process. The trained features 430 may include one or more shapes, sizes, arrangements, materials, orientations, etc. of the structures.

[0078] Information file 315 may include a layout of a desired wafer design or chip design to be transferred onto wafer 203. Information file 315 may include information stored in a suitable format, such as GDS, GDSII, OASIS, or CIF.

[0079] The information file 315 may include knowledge, which may include information associated with the trained feature 430, such as relative orientation of structures in the feature, physical location information of the feature on the wafer, xy location coordinates of the feature on the wafer, critical dimensions and critical dimension tolerances, etc. In some embodiments, the knowledge may include target GDS information, target GDSII information, target OASIS information, etc.

[0080] In some embodiments, information file 315 may contain critical dimension information. Critical dimension tolerances for optical lithography may be very difficult to achieve and may require an iterative correction process including adjustments to features on the mask. Some possible solutions include, but are not limited to, resolution enhancement techniques such as optical proximity correction, phase-shift masks, and off-axis illumination.

[0081] Optical proximity correction involves changing the actual chrome width on the mask to be different from the desired photoresist width on the wafer. For example, the difference between the size of isolated lines and the lines in a dense array of equal lines and spaces is the most commonly observed proximity effect in optical lithography. The magnitude of the print bias is affected by the optical parameters of the stepper and the contrast of the photoresist. If the optical parameters of the stepper and the contrast of the photoresist remain constant, the print bias can be characterized and corrected by biasing the mask. This type of geometry-dependent mask bias is typically used to maintain critical dimensions and critical dimension tolerances on actual wafers. Mask bias is defined as the actual chrome width minus the nominal (no bias) chrome width. Therefore, a positive bias means that the chrome is made larger. Phase-shift masks and off-axis illumination techniques can also be used to adjust the mask bias so that the features transferred to the wafer 203 by the photolithography process match the desired feature shape and size.

[0082] The information file 315 may include target bias-corrected mask information, uncorrected mask information, bias-corrected signature information of a reticle for a stepper, uncorrected signature information of a reticle, or a combination thereof.

[0083] In some embodiments, the machine learning network 320 is configured to extract trained features 430 from the high-resolution image 310 or knowledge from the information file 315. In some embodiments, training the machine learning network 320 is automated. For example, the automated machine learning network can silently receive the high-resolution image 310 or information from the information file 315. After receiving the high-resolution image 310 or information from the information file 315, the automated machine learning network can silently extract relevant trained features.

[0084] The extracted trained features 430 may be stored in the storage module 350 or temporarily stored in a repository (not shown). The comparator 370, the machine learning network 320, the image enhancer 380, the display device 390, etc. may access the repository.

[0085] In some embodiments, inspection image 330 is a low-resolution, high-throughput mode image acquired in real time by EBI system 100 during an inspection step in a manufacturing process. Figure 4 As illustrated in , inspection image 330 includes a degraded through-focus image that shows features and patterns, but does not clearly resolve features and patterns similar to those depicted in high-resolution image 310, for example. Inspection image 330 also shows defects toward the center of the image. Defects may include voids, particles, unstripped photoresist, over-etched surfaces of features, etc.

[0086] Inspection image 330 may be an optical image from an optical microscope. In some embodiments, one or more views of inspection image 330 may be acquired at a given inspection step in the process. A pattern extractor (e.g. Figure 3 The pattern extractor 340 ) may be configured to process one or more views of the inspection image 330 to determine a “best” image to extract relevant pattern information 410 therefrom.

[0087] Image enhancement module 360 ​​may be configured to process inspection image 330 to generate enhanced image 420. Image enhancement module 360 ​​may include Figure 3 The image enhancement module 360 ​​may further include a comparator 370, an image enhancer 380, and a display device 390. In some embodiments, the image enhancement module 360 ​​may further include a pattern extractor 340. The comparator 370 may be configured to identify the extracted trained features 430 or knowledge from the information file 315 based on the pattern information 410 extracted by the pattern extractor 340. After identifying the trained features, the comparator 370 may compare the extracted trained features 430 with the pattern information 410 and generate an output. The output may include information associated with the comparison results and the extracted trained features 430, which are then used to enhance the inspection image. For example, Figure 4 As shown in the enhanced image 420 of FIG. 4 , features of circular structures, bands, and defects have been enhanced using information derived from the high resolution image 310. The output may be temporarily stored in the storage module 350. The image enhancer 380 may be configured to receive the output from the comparator 370 or from the storage module 350 and the inspection image 330 to generate the enhanced image 420.

[0088] In some embodiments, the image enhancer 380 may perform a comparison of the identified extracted trained features 430 and the extracted pattern information 410. In some embodiments, the comparator 370 may be integrated with the image enhancer 380, operating as a single unit.

[0089] The enhanced image 420 may be displayed on the display device 390. In some embodiments, the enhanced image 420 is displayed on the display device 390 and stored in the storage module 350. The enhanced image 420 may include features and patterns that are enhanced compared to the features and patterns of the inspection image 330, for example, the features and patterns are more clearly resolved. In some embodiments, the brightness and contrast may be optimized compared to the inspection image 330. Defects in the inspection image 330 may be better resolved and focused, which may help the reviewer accurately identify defects and address related issues.

[0090] The enhanced image 420 can be used in real time for inspection, defect identification and analysis, process verification, quality control, yield improvement analysis, etc., while maintaining the high throughput required during the inspection step in the wafer manufacturing process. The enhanced image 420 can be displayed on multiple displays simultaneously. For example, the enhanced image 420 representing the inspection image of the wafer after the photoresist stripping step but before the following metal deposition step can be reviewed by multiple reviewers or users requesting information. In some embodiments, the enhanced image 420 can be obtained by the user prompt at a later time for review and in-depth analysis. The enhanced image 420 can be stored in a suitable format, such as a Joint Photographic Experts Group (JPEG) file, a Portable Network Graphics (PNG) file, a Portable Document Format (PDF) file, a Tagged Image File Format (TIFF) file, etc.

[0091] Figure 5 1 is a flow chart illustrating an exemplary image enhancement method consistent with an embodiment of the present disclosure. The image enhancement method may be performed by an image enhancement module that may be coupled to a charged particle beam device including the EBI system 100. For example, a controller (e.g. Figure 2The controller 109 of the embodiment of the present invention may include an image enhancement module and may be programmed to implement the image enhancement method. It should be understood that the charged particle beam device may be controlled to image the wafer or a region of interest on the wafer. Imaging may include scanning the wafer to image at least a portion of the wafer.

[0092] In step 510, one or more scanned raw images of the inspection area may be obtained. The one or more scanned raw images may include the entire surface of the wafer, or only a portion of the surface of the wafer. The image acquisition in step 510 may include: obtaining a detector of a charged particle beam device (such as an electron detector (e.g., Figure 2 electronic detector 206)) or receives a signal from a storage device (e.g. Figure 2 The image is loaded from the storage device 270).

[0093] In step 520, pattern information (e.g., pattern information 410) on the wafer is identified and extracted by a pattern extractor. The pattern information may include global structural information, such as a reference datum for a photolithography process on the wafer, reference features on the wafer, etc. Step 520 may include relevant feature extraction. The identification of relevant features and patterns may be performed by a feature extraction algorithm. For example, the feature extraction of step 520 may include performing image analysis on an acquired image of the wafer surface using a first parameter set or using a first image processing algorithm. The identification of relevant features may include determining a location, a group of locations, or a range of locations in an xy coordinate on the wafer surface. The location information may be stored in a wafer mapping database (e.g., set in a Figure 3 in the storage module 350).

[0094] In step 530, images from the training images (eg, Figure 3 The trained features of the high-resolution image 310 of FIG. 1 ) correspond to the extracted pattern information. The trained features may be based on an information file (such as information file 315) or a training image (such as Figure 3 The training image may be a high-resolution image acquired by the image acquisition system in an offline mode or a review mode, and the trained features are extracted from the wafer design plan. The training image may be a high-resolution image acquired by the image acquisition system in an offline mode or a review mode. The wafer design plan may be registered in advance. For example, the wafer design plan may be a graphical representation of features on the surface of the wafer. The wafer design plan may be based on a pattern layout for designing a wafer. The wafer design plan may correspond to a mask used to manufacture the wafer. The wafer design plan may be stored in a wafer design database (e.g., set in a Figure 3In the storage module 350 of the invention, the individual features corresponding to the identified global structure information or pattern information can be extracted according to the information data from the information file or the training image. For example, based on the xy coordinates of the pattern location identified in step 520, the relevant trained features including the geometric information of the initial feature design can be collected.

[0095] In some embodiments, a machine learning network (e.g. Figure 3 The machine learning network 320 of the present invention may perform identification of relevant trained features based on the pattern information extracted from the inspection image in step 520. The machine learning network may be automated to obtain trained feature information from an information file, or to obtain trained feature information from a training image. The machine learning network may also be automated to extract relevant trained features from an information file or a training image. The extracted trained features / knowledge may be stored in a storage device or repository configured to store information. The extracted trained features / knowledge may be temporarily stored or used for long-term purposes. The extracted trained features / knowledge may be used by an image enhancement module (e.g., Figure 3 image enhancement module 360) or an image enhancer (e.g. Figure 3 Image intensifier 380) access.

[0096] The disclosed method can search for matching patterns in a pattern library (stored, for example, in the storage module 350) based on the correspondence. For example, if the feature design profile has a similarity of 90% or more with another pattern in the pattern library, a pattern match can be determined. The patterns in the pattern library can include previously extracted patterns, standard patterns (such as patterns of standard IC features), etc.

[0097] In step 540, the inspection image is enhanced based on the comparison results between the trained relevant features / knowledge identified from the training image and the pattern information extracted from the inspection image. In step 540, image enhancement includes: generating an enhanced image (e.g., Figure 4 Then, the matched results and the trained features 430 are used to enhance the inspection image.

[0098] Figure 6 is a flow chart illustrating an exemplary image enhancement method consistent with embodiments of the present disclosure. The image enhancement method may be performed by an image enhancement module that may be coupled to a charged particle beam device (e.g., EBI system 100). For example, a controller (e.g., Figure 2The controller 109 of the embodiment of the present invention may include an image enhancement module and may be programmed to implement the image enhancement method. It should be understood that the charged particle beam device may be controlled to image the wafer or a region of interest on the wafer. Imaging may include scanning the wafer to image a portion of the wafer, which may include scanning the wafer to image the entire wafer. In some embodiments, the image enhancement method may be performed by an image intensifier module (e.g., Figure 3 The image enhancement system 300 may be used to enhance the image quality of the device.

[0099] In step 610, a training image (eg Figure 4 The training images may be high-resolution images 310, review mode images, high-quality images acquired under optimal acquisition conditions, etc. The training images may be acquired offline, for example, using different charged particle beam devices, optical microscopes, atomic force microscopes, etc. The training image acquisition may be performed at a remote location different from the wafer fabrication location and stored in a storage module (e.g., Figure 3 The image enhancement process may be performed offline to minimize the impact of the image enhancement process on the inspection throughput.

[0100] In step 620, a machine learning network (e.g. Figure 3 The machine learning network 320 of FIG. 3 extracts relevant trained features from the acquired training images. It is also possible to extract relevant trained features from information files (e.g., Figure 3 The information file 315) extracts relevant features, such as a wafer design plan or a mask layout, a graphic layout of one or more layers of an integrated circuit. The information file may include information stored in a GDS format, a GDSII format, an OASIS format, etc.

[0101] The machine learning network can be configured to automatically extract relevant trained features. The machine learning network can be automated to obtain trained feature information from an information file or to obtain trained feature information from a training image. Extracting relevant trained features using a machine learning network can be performed offline using different processors at different locations. The machine learning network can be integrated into the EBI system 100, or can be operated as a standalone unit operated remotely.

[0102] In step 630, after the wafer is loaded, an inspection image of the wafer is acquired (eg, Figure 3Inspection image 330). For example, the wafer may be placed on a sample stage and prepared for imaging. The inspection image may be a scanned raw image with low resolution. Inspection image 330 may be a high-throughput inspection mode image acquired using a non-optimal set of acquisition conditions. The inspection image may be a degraded or inferior image, or a low-resolution image, etc. The inspection image may be acquired online during the wafer manufacturing process. Steps 630 and 610 may use one or more image acquirers (e.g. Figure 2 image acquisition device 260) or EBI (e.g. Figure 1 The inspection image and the training image can be acquired by the same EBI at different times.

[0103] In step 640, global structure information or pattern information (eg, Figure 4 Step 640 may include performing image analysis on the acquired image of the wafer surface using a first parameter set or using a first image processing algorithm. In step 640, a pattern extractor (e.g. Figure 3 The pattern extractor 340 of the embodiment of the present invention identifies and extracts pattern information of the wafer. The pattern information may include global structural information, such as reference datums of photolithography processes on the wafer, reference features on the wafer, etc. Step 640 may also include extracting (multiple) relevant features using a feature extraction algorithm. The identification and extraction of relevant pattern information may include determining a location, a group of locations, or a range of locations in xy coordinates on the surface of the wafer. The location information may be stored in a wafer mapping database stored in the storage module.

[0104] In step 650, an image enhancement module or a comparator (eg, Figure 3 The comparator 370 of the chip 100 identifies the trained features from the training image based on the identified pattern information. The trained features can also be extracted from the chip design plan based on the information file or the training image. The training image can be a high-resolution image acquired by an image acquisition system (e.g., EBI system 100) in an offline mode or a review mode. The chip design plan can be registered in advance. For example, the chip design plan can be a graphical representation of features on the surface of the chip. The chip design plan can be based on a pattern layout for designing a chip. The chip design plan can correspond to a mask used to manufacture a chip. The chip design plan can be stored in a chip design database.

[0105] The mask can include basic two-dimensional shapes, including but not limited to rectangles, which can have sharp corners. In order to describe the error in the actual resist image from the target or desired resist image, a target image is defined. When printing in the photoresist, the corners from the initial design may be rounded to some extent. A certain amount of corner rounding is acceptable. Although there is no reason to round the corners of the designed mask layout, there is no reason to insist that the final printed pattern match the designed square corners. Therefore, the actual desired pattern may deviate from the designed pattern due to an acceptable amount of corner rounding. Alternatively, the mask bias can be used to modify the chrome features so that the final printed features on the wafer are the desired shape and size. Target GDS information (e.g. Figure 3 The information contained in the information file 315 may include mask bias data.

[0106] The wafer design plan stored in the information file may include target GDS or target GDSII information. As mentioned herein, the target GDS information may include updated or adjusted mask feature information based on wafer processing conditions. For example, as discussed above, transferring nanometer-sized features from a mask to a wafer may include a mask bias to accommodate optical proximity correction. The mask bias of the optical proximity correction may include changing the actual chrome width on the mask to be different from the desired photoresist width on the wafer. For example, the difference between the size of the isolated line and the line in a dense array of equal lines and intervals is the most commonly observed proximity effect in optical lithography. The magnitude of the printing bias may be greatly affected by the optical parameters of the stepper or the contrast of the photoresist. If the optical parameters of the stepper or the contrast of the photoresist remain constant, the printing bias may be characterized and corrected by making the mask bias. This type of mask bias that depends on geometry is typically used to maintain critical dimensions and critical dimension tolerances on actual wafers. Mask bias is defined as actual chrome width minus nominal (no bias) chrome width. Therefore, a positive bias means that the chrome is made larger.

[0107] In step 650, the machine learning network may perform identification of relevant trained features based on the pattern information extracted from the inspection image in step 640. The machine learning network may be automated to obtain trained feature information from an information file or to obtain trained feature information from a training image. The machine learning network may also be automated to extract relevant trained features from an information file or a training image. The extracted trained features (e.g., Figure 4 The trained features 430) can be stored in a storage module or a repository configured to store information. The extracted trained features can be temporarily stored or used for long-term purposes. The extracted trained features can be accessed by an image enhancement module or an image enhancer.

[0108] In step 660, the inspection image is enhanced based on the comparison results between the trained relevant features identified from the training image and the pattern information extracted from the inspection image. In step 660, image enhancement includes: generating an enhanced image (e.g., Figure 4 The enhanced image 420) or the enhanced inspection image. Then, the matching result and the trained features are used to enhance the inspection image. The matching result and the trained features can be stored in the storage module.

[0109] The enhanced image can be used in real time for inspection, defect identification and analysis, process verification, quality control, yield improvement analysis, etc., while maintaining the high throughput required during the inspection step in the wafer manufacturing process. The enhanced image can be displayed on multiple displays simultaneously. For example, an enhanced image representing an inspection image of a wafer after a photoresist stripping step but before the following metal deposition step can be reviewed by multiple reviewers or multiple users requesting information. In some embodiments, the enhanced image can be retrieved by a user prompt at a later time for review and in-depth analysis. The enhanced image can be stored in a suitable format (e.g., JPEG file, PNG file, PDF file, TIFF file, etc.).

[0110] Generally return reference Figure 3 . In some embodiments, the image enhancement system 300 may analyze the high-resolution image 310 to develop a deconvolution or benchmarking strategy to enhance the image quality of the (multiple) low-resolution inspection image 330. In some embodiments, the image enhancement system 300 may analyze the high-resolution image 310 to develop the deconvolution or benchmarking strategy in an offline mode. Alternatively or additionally, the image enhancement system 300 may analyze the high-resolution image 310 to develop the deconvolution or benchmarking strategy during a high-throughput inspection. It is contemplated that in addition to performing the image pattern-based enhancement process described above, the image enhancement system 300 may also utilize such a deconvolution or benchmarking strategy to enhance the image quality of the (multiple) low-resolution inspection image 330. Alternatively, the image enhancement system 300 may utilize such a deconvolution or benchmarking strategy to enhance the image quality of the (multiple) low-resolution inspection image 330, rather than performing the image pattern-based enhancement process described above. In some embodiments, the image enhancement system 300 may allow a user to specify (e.g., via a user interface) which enhancement process or processes to perform.

[0111] Figure 7 is a diagram that can be obtained by an image enhancement system (e.g. Figure 3 The image enhancement system 300 is executed to analyze the high-resolution image (e.g. Figure 4 high resolution images 310), developing deconvolution or benchmarking strategies and enhancing (multiple) low resolution inspection images (e.g. Figure 4 Flow chart of an exemplary image enhancement method for improving the image quality of a low-resolution image 330).

[0112] In step 710, one or more scanned images of the inspection area may be acquired. At least one of the images may include a high-resolution image, and at least one of the images may include a low-resolution inspection image. For illustrative purposes, the high-resolution image may be referred to as a first image (or a first SEM image), and the low-resolution inspection image may be referred to as a second image (or a second SEM image). It should be understood that the terms "first" and "second" do not suggest that the first image needs to be acquired before the second image. It is contemplated that the first image and the second image may be acquired simultaneously (e.g., a first image may be acquired using a coaxial beam of a multi-beam system, and a second image may be acquired using an off-axis beam of the multi-beam system) or sequentially (e.g., the first image may be acquired before the second image, or vice versa).

[0113] In step 720, the data representing the high-resolution image may be analyzed to obtain one or more spatial-spectral characteristics contained therein. The relevant spatial-spectral characteristics may include phase and amplitude characteristics. In step 730, the obtained spatial-spectral characteristics may be used to numerically compensate the data representing the low-resolution image, thereby producing an enhanced image in step 740. It should be noted that because wafer inspection is typically performed by scanning repetitive features with only small changes, performing such numerical compensation can effectively and efficiently enhance the image quality of the low-resolution inspection image.

[0114] Figure 88 is a diagram depicting such a numerical compensation process. Specifically, a SEM signal representing a first image (or a portion of a first image) acquired at a high resolution is depicted as high resolution data 810. A SEM signal representing a second image (or a portion of a second image) acquired at a low resolution is depicted as low resolution data 820. It should be noted that both the high resolution data 810 and the low resolution data 820 contain noise, which may complicate direct deconvolution processes (such as Wiener deconvolution processes, etc.) and render these deconvolution processes invalid. In addition, it should be noted that the deconvolution of SEM images is more complex than the deconvolution of optical images. For example, although many optical deconvolution strategies rely on the assumption that edges can be well approximated by step functions, this assumption cannot be extended to deconvoluted SEM signals because these deconvoluted SEM signals are often not well approximated by step functions. The SEM signal is largely determined by the amount of interaction of the electrons inside the target, making the SEM signal highly dependent on the properties of the target and the settings of the SEM. Therefore, when Wiener deconvolution is applied to deconvolute the low-resolution data 820, the resulting deconvolved data (not shown) fails to conform to the high-resolution data 810. The resulting deconvolved data also exhibits noticeable artifacts.

[0115] The deconvolution process configured according to an embodiment of the present disclosure improves the deconvolution results by taking into account the spatial-spectral characteristics obtained from the high-resolution data 810. The relevant spatial-spectral characteristics may include phase and amplitude characteristics. Utilizing such characteristics, the image enhancement system 300 can deconvolve the low-resolution data 820 by examining at which spatial frequency the noise dominates the amplitude spectrum, and replacing the amplitude and phase of these high spatial frequencies with the corresponding amplitude and phase obtained from the high-resolution data 810. The resulting deconvolved data (depicted as deconvolved data 830) exhibits noise comparable to the noise present in the high-resolution data 810. The deconvolved data 830 can also eliminate the obvious artifacts produced by Wiener deconvolution and better conform to the ideal high-resolution data.

[0116] It is contemplated that the vertical compensation process configured in accordance with embodiments of the present disclosure may be applied in the frequency domain. In some embodiments, the image enhancement system 300 may utilize a Fourier transform to transform the SEM signal in the time domain into the frequency domain (which may also be referred to as the Fourier domain) prior to applying numerical compensation. However, it should be understood that the image enhancement system 300 may also be configured to apply numerical compensation in the time domain. It should be understood that the specific implementation of the image enhancement system 300 may vary without departing from the scope and spirit of the present disclosure.

[0117] In some embodiments, the image enhancer 300 may also be configured to focus the inspection on certain features, such as edge locations, etc. In such embodiments, rather than attempting to enhance the image quality of the entire low-resolution inspection image, the image enhancement system 300 may focus its enhancement on a few features or regions of interest. It is contemplated that by focusing its enhancement on a few features or regions of interest, the image enhancement system 300 may also improve its accuracy and throughput.

[0118] Fig. 9 is a diagram that can be obtained by an image enhancement system (e.g. Figure 3 Flow chart of an exemplary image enhancement method performed by an image enhancement system 300) to help identify the presence and location of features or regions of interest.

[0119] In step 910, one or more scanned images of an examination region may be acquired. At least one of the images may include a high resolution image (e.g., Figure 4 310), and at least one of the images may include a low-resolution inspection image (e.g. Figure 4 330 of a low-resolution inspection image). For illustrative purposes, the high-resolution image may be referred to as a first image (or a first SEM image), and the low-resolution inspection image may be referred to as a second image (or a second SEM image). It should be understood that the terms "first" and "second" do not suggest that the first image needs to be acquired before the second image. It is contemplated that the first image and the second image may be acquired simultaneously (e.g., the first image may be acquired using an on-axis beam of a multi-beam system, and the second image may be acquired using an off-axis beam of the multi-beam system) or sequentially (e.g., the first image may be acquired before the second image, or vice versa).

[0120] In step 920, the image enhancement system may analyze the data representing the first image to identify one or more features contained therein. Fig.10 , the image enhancement system 300 may analyze the high-resolution data 1010 representing the first image (or a portion of the first image) to identify edge features contained in the first image. Alternatively or additionally, the image enhancement system 300 may feed the high-resolution data 1010 to the machine learning network 320, which may be configured to identify such features contained in the first image.

[0121] In step 930, the image enhancement system may numerically blur the data representing the first image to simulate data representing a blurred image acquired at a lower resolution. Fig.10, the image enhancement system 300 may numerically blur the high-resolution data 1010 representing the first image (or a portion of the first image) to produce simulated low-resolution data 1020. More specifically, to numerically blur the high-resolution data 1010, the image enhancement system 300 may convolve the high-resolution data 1010 with a beam profile similar to the beam used to acquire the low-resolution data 1030. For example, if it is known that the beam used to acquire the low-resolution data 1030 has a specific size, has a specific current level, or has a specific beam position, then this information may be used to create a beam profile, which may then be used to numerically blur the high-resolution data 1010 to simulate what would be acquired using a beam of that specific size, that specific current level, and that is positioned at that specific location. On the other hand, if the beam profile is not explicitly known, the image enhancement system 300 may utilize the machine learning network 320 to analyze the high-resolution data and the low-resolution data acquired in the past to help determine how blurring should be applied. It is expected that utilizing the machine learning network 320 in this manner may cause a slight delay, but the delay may be negligible compared to the increased throughput. In the case of a multi-beam system that uses coaxial beams to provide high-resolution images, the image enhancement system 300 may be able to verify or update the machine learning network 320 during the inspection process without incurring any noticeable delay.

[0122] Return to reference Fig. 9 In step 940, the image enhancement system may determine whether a portion of the simulated low-resolution data representing the blurred image fits a portion of the data representing the second image. Fig.10 , the image enhancement system 300 may compare the simulated low resolution data 1020 relative to the acquired low resolution data 1030 to determine if there is a fit. In some embodiments, a determination is made that the portion of the simulated low resolution data 1020 fits the portion of the low resolution data 1030 based on a cross-correlation between a portion of the simulated low resolution data 1020 and a portion of the low resolution data 1030. In some embodiments, a determination is made that the portion of the simulated low resolution data 1020 fits the portion of the low resolution data 1030 based on a root mean square of a portion of the simulated low resolution data 1020 and a root mean square of a portion of the low resolution data 1030. It is contemplated that other data fitting processes may also be utilized to make the determination without departing from the scope and spirit of the present disclosure.

[0123] Return to reference Fig. 9In response to a determination that a portion of the simulated low-resolution data representing the blurred image fits a portion of the data representing the second image, the image enhancement system may identify the portion of the data representing the second image as a region of interest because it contains an identified feature (e.g., an edge location in the example presented above) in step 950. Based on this identification, the image enhancement system may select to enhance the image quality of the second image for the region that has been identified as a region of interest in step 960. In this way, the image enhancement system 300 may focus its enhancement on a few features or regions of interest, thereby further improving its accuracy and yield.

[0124] The enhanced image can be used in real time for inspection, defect identification and analysis, process verification, quality control, yield improvement analysis, etc., while maintaining the high throughput required during the inspection step in the wafer manufacturing process. The enhanced image can be displayed on multiple displays simultaneously. For example, an enhanced image representing an inspection image of a wafer after a photoresist stripping step but before the following metal deposition step can be reviewed by multiple reviewers or multiple users requesting information. In some embodiments, the enhanced image can be acquired by a user prompt at a later time for review and in-depth analysis. The enhanced image can be stored in a suitable format (e.g., JPEG file, PNG file, PDF file, TIFF file, etc.).

[0125] Embodiments may also be described using the following first set of clauses, where references to other clauses are references to clauses in the first set of clauses:

[0126] 1. A method for enhancing an image, the method comprising:

[0127] acquiring a first image of the sample;

[0128] extracting pattern information of the first image;

[0129] Using the pattern information to identify trained features; and

[0130] An enhanced image is provided from the first image using the identified trained features.

[0131] 2. The method according to clause 1 further comprises obtaining training images from a user defined database, wherein the user defined database comprises a graphical database system.

[0132] 3. The method according to any one of clauses 1 and 2, further comprising extracting trained features from the training images via a machine learning network.

[0133] 4. The method according to any one of clauses 2 and 3, wherein acquiring training images and extracting trained features are performed in an offline mode.

[0134] 5. A method according to any one of clauses 1 to 4, wherein using the pattern information to identify the trained features comprises comparing the pattern information to the trained features.

[0135] 6. A method according to any one of clauses 1 to 5, wherein the first image comprises a low resolution electron beam inspection image.

[0136] 7. The method of clause 2, wherein the training images comprise high resolution electron beam images.

[0137] 8. A method for enhancing an image, the method comprising:

[0138] Get training images;

[0139] extracting trained features from the training images via a machine learning network;

[0140] acquiring a first image of the sample;

[0141] extracting pattern information of the first image;

[0142] Using the pattern information to identify trained features; and

[0143] An enhanced image is provided from the first image using the identified trained features.

[0144] 9. The method of clause 8, further comprising obtaining training images from a user defined database, wherein the user defined database comprises a graphical database system.

[0145] 10. The method according to any one of clauses 8 and 9, wherein acquiring training images and extracting trained features are performed in an offline mode.

[0146] 11. A method according to any of clauses 8 to 10, wherein using the pattern information to identify the trained features comprises comparing the pattern information to the trained features.

[0147] 12. A method according to any of clauses 8 to 11, wherein the first image comprises a low resolution electron beam inspection image.

[0148] 13. A method according to any one of clauses 8 to 11, wherein the training images comprise high resolution electron beam images.

[0149] 14. An inspection image enhancement system, comprising:

[0150] a memory storing an instruction set; and

[0151] A processor configured to execute a set of instructions to cause the inspection image enhancement system to:

[0152] obtaining an inspection image of the sample;

[0153] extracting pattern information from the inspection image;

[0154] Using the pattern information to identify trained features; and

[0155] Using the identified trained features, an enhanced image is provided from the inspection image.

[0156] 15. A system according to clause 14, wherein the instruction set further causes the inspection image enhancement system to:

[0157] obtaining training images; and

[0158] Through a machine learning network, trained features are extracted from training images.

[0159] 16. The system of clause 15, wherein the set of instructions further causes the inspection image enhancement system to acquire training images and extract trained features in an offline mode.

[0160] 17. A system according to any of clauses 15 and 16, wherein the set of instructions further causes the inspection image enhancement system to obtain training images from a user defined database.

[0161] 18. The system of any of clauses 14 to 17, wherein the set of instructions further causes the inspection image enhancement system to use the pattern information to identify the trained features by comparing the pattern information to the trained features.

[0162] 19. A system according to any of clauses 14 to 18, wherein the inspection image comprises a low resolution electron beam image.

[0163] 20. A system according to any one of clauses 15 to 18, wherein the training images comprise high resolution electron beam images.

[0164] 21. The system according to any one of clauses 14 to 20, further comprising a storage device configured to store data related to the trained features and pattern information.

[0165] 22. A non-transitory computer readable medium storing a set of instructions executable by at least one processor of a device to cause the device to perform a method comprising:

[0166] acquiring a first image of the sample;

[0167] extracting pattern information of the first image;

[0168] Using the pattern information to identify trained features; and

[0169] An enhanced image is provided from the first image using the identified trained features.

[0170] 23. The computer readable medium of clause 22, wherein the set of instructions further causes the apparatus to:

[0171] obtaining training images; and

[0172] Through a machine learning network, trained features are extracted from training images.

[0173] 24. The computer readable medium of clause 24, wherein the set of instructions further causes the apparatus to:

[0174] Acquire training images and extract trained features in offline mode.

[0175] 25. The computer readable medium of clause 22, wherein the set of instructions further causes the apparatus to:

[0176] The pattern information is used to identify the trained features by comparing the pattern information to the trained features.

[0177] Embodiments may also be described using the following second set of clauses, where references to other clauses are references to clauses in the second set of clauses:

[0178] 1. A method for enhancing an image, the method comprising:

[0179] acquiring a first scanning electron microscope (SEM) image at a first resolution;

[0180] acquiring a second SEM image at a second resolution; and

[0181] An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

[0182] 2. A method according to clause 1, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0183] 3. The method according to clause 1, further comprising:

[0184] Trained features are extracted from the first SEM image.

[0185] 4. The method according to clause 3, wherein the trained features are extracted from the first SEM image using a machine learning network.

[0186] 5. The method according to clause 4, further comprising:

[0187] acquiring at least one additional SEM image at the first resolution; and

[0188] Using the machine learning network, at least one additional trained feature is extracted from at least one additional SEM image.

[0189] 6. The method according to any one of clauses 2 to 5, further comprising:

[0190] extracting pattern information of the second SEM image;

[0191] Using the pattern information, determining that the trained features are identified on the second SEM image; and

[0192] In response to a determination that the trained feature is identified on the second SEM image, an enhanced image is provided by enhancing the second SEM image using the identified trained feature.

[0193] 7. The method according to any of clauses 2 to 6, wherein acquiring the first SEM image and extracting the trained features from the first SEM image are performed in an offline mode.

[0194] 8. The method according to any of clauses 2 to 6, wherein determining that the trained feature is identified on the second SEM image comprises: comparing the pattern information and the trained feature.

[0195] 9. The method according to clause 1, further comprising:

[0196] analyzing data representing the first SEM image to obtain one or more spatial-spectral characteristics; and

[0197] An enhanced image is provided by applying one or more numerical compensations to data representing the second SEM image based on the obtained one or more spatio-spectral characteristics.

[0198] 10. The method of clause 9, wherein the one or more spatial-spectral characteristics obtained include phase and amplitude characteristics.

[0199] 11. The method according to clause 9, wherein one or more numerical compensations are applied in the Fourier domain.

[0200] 12. The method according to clause 9, wherein the one or more numerical compensations deconvolute the second SEM image.

[0201] 13. The method according to clause 1, further comprising:

[0202] analyzing data representing the first SEM image to identify features in the first SEM image;

[0203] numerically blurring data representing the first SEM image to simulate data representing a blurred SEM image acquired at a second resolution;

[0204] Determining that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image; and

[0205] Responsive to a determination that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image, the portion of the data representing the second SEM image is identified as containing the location of the feature identified in the first SEM image.

[0206] 14. The method of clause 13, wherein the feature is an edge position.

[0207] 15. A method according to any one of clauses 13 and 14, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on a cross-correlation between the portion of the data representing the blurred SEM image and the portion of the data representing the second SEM image.

[0208] 16. A method according to any one of clauses 13 and 14, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on a root mean square of a portion of the data representing the blurred SEM image and a root mean square of a portion of the data representing the second SEM image.

[0209] 17. A method according to any of clauses 1 to 16, wherein the second resolution is lower than the first resolution.

[0210] 18. A method according to any one of clauses 1 to 17, wherein the first resolution and the second resolution correspond to at least one of: a signal averaging amount, a noise ratio of a SEM image frame, a pixel size, a SEM beam width of a coaxial beam of a multi-beam system, a SEM beam width of an off-axis beam of a multi-beam system, a SEM beam width of a single beam system, or a current supplied to the SEM beam.

[0211] 19. The method according to any of clauses 1 to 18, wherein the first SEM image is obtained using a coaxial beam of a multi-beam system and the second SEM image is obtained using an off-axis beam of the multi-beam system.

[0212] 20. The method according to any of clauses 1 to 18, wherein the first SEM image is obtained using a low current beam and the second SEM image is obtained using a high current beam.

[0213] 21. An inspection system comprising:

[0214] a memory storing an instruction set; and

[0215] A processor configured to execute a set of instructions to cause the inspection system to:

[0216] acquiring a first scanning electron microscope (SEM) image at a first resolution;

[0217] acquiring a second SEM image at a second resolution; and

[0218] An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

[0219] 22. The system according to clause 21, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0220] 23. A system according to clause 21, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0221] Trained features are extracted from the first SEM image.

[0222] 24. The system according to clause 23, wherein the trained features are extracted from the first SEM image using a machine learning network.

[0223] 25. A system according to clause 24, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0224] acquiring at least one additional SEM image at the first resolution; and

[0225] Using the machine learning network, at least one additional trained feature is extracted from at least one additional SEM image.

[0226] 26. A system according to any of clauses 22 to 25, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0227] extracting pattern information of the second SEM image;

[0228] Using the pattern information, determining that the trained features are identified on the second SEM image; and

[0229] In response to a determination that the trained feature is identified on the second SEM image, an enhanced image is provided by enhancing the second SEM image using the identified trained feature.

[0230] 27. The system according to any of clauses 22 to 26, wherein the processor is further configured to execute the set of instructions to cause the inspection system to acquire a first SEM image in an offline mode and to extract the trained features from the first SEM image.

[0231] 28. The system of clause 26, wherein the determination that the trained feature is identified on the second SEM image comprises a comparison of the pattern information and the trained feature.

[0232] 29. A system according to clause 21, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0233] analyzing data representing the first SEM image to obtain one or more spatio-spectral characteristics; and

[0234] An enhanced image is provided by applying one or more numerical compensations to data representing the second SEM image based on the obtained one or more spatio-spectral characteristics.

[0235] 30. A system according to clause 29, wherein the one or more spatial-spectral characteristics obtained include phase and amplitude characteristics.

[0236] 31. The system of clause 29, wherein the one or more numerical compensations are applied in the Fourier domain.

[0237] 32. The system of clause 29, wherein the one or more numerical compensations deconvolute the second SEM image.

[0238] 33. The system of clause 21, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0239] analyzing data representing the first SEM image to identify features in the first SEM image;

[0240] numerically blurring data representing the first SEM image to simulate data representing a blurred SEM image acquired at a second resolution;

[0241] Determining that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image; and

[0242] Responsive to a determination that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image, the portion of the data representing the second SEM image is identified as containing the location of the feature identified in the first SEM image.

[0243] 34. A system according to clause 33, wherein the feature is an edge position.

[0244] 35. A system according to any one of clauses 33 to 34, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on a cross-correlation between the portion of the data representing the blurred SEM image and the portion of the data representing the second SEM image.

[0245] 36. A system according to any one of clauses 33 to 34, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on the root mean square of the portion of the data representing the blurred SEM image and the root mean square of the portion of the data representing the second SEM image.

[0246] 37. A system according to any of clauses 21 to 36, wherein the second resolution is lower than the first resolution.

[0247] 38. A system according to any of clauses 21 to 37, wherein the first resolution and the second resolution correspond to at least one of: a signal averaging amount, a noise ratio of a SEM image frame, a pixel size, a SEM beam width of a coaxial beam of a multi-beam system, a SEM beam width of an off-axis beam of a multi-beam system, a SEM beam width of a single beam system, or a current supplied to the SEM beam.

[0248] 39. A system according to any of clauses 21 to 38, wherein the first SEM image is obtained using a coaxial beam of the multi-beam system and the second SEM image is obtained using an off-axis beam of the multi-beam system.

[0249] 40. The system according to any of clauses 21 to 38, wherein the first SEM image is obtained using a low current beam and the second SEM image is obtained using a high current beam.

[0250] 41. A non-transitory computer readable medium storing a set of instructions executable by at least one processor of a device to cause the device to perform a method comprising:

[0251] acquiring a first scanning electron microscope (SEM) image at a first resolution;

[0252] acquiring a second SEM image at a second resolution; and

[0253] An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

[0254] 42. The computer readable medium of clause 41, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0255] 43. The computer-readable medium of clause 41, wherein the set of instructions is executable by at least one processor of the device to cause the device to further:

[0256] Trained features are extracted from the first SEM image.

[0257] 44. The computer-readable medium of clause 43, wherein the trained features are extracted from the first SEM image using a machine learning network.

[0258] 45. The computer-readable medium of clause 44, wherein the set of instructions is executable by at least one processor of the device to cause the device to further:

[0259] acquiring at least one additional SEM image at the first resolution; and

[0260] Using the machine learning network, at least one additional trained feature is extracted from at least one additional SEM image.

[0261] 46. ​​A computer-readable medium according to any of clauses 42 to 45, wherein the set of instructions is executable by at least one processor of an apparatus to cause the apparatus to further:

[0262] extracting pattern information of the second SEM image;

[0263] Using the pattern information, determining that the trained features are identified on the second SEM image; and

[0264] In response to a determination that the trained feature is identified on the second SEM image, an enhanced image is provided by enhancing the second SEM image using the identified trained feature.

[0265] 47. The computer-readable medium according to any of clauses 42 to 46, wherein acquiring the first SEM image and extracting the trained features from the first SEM image are performed in an offline mode.

[0266] 48. The computer-readable medium of clause 46, wherein the determination that the trained feature is identified on the second SEM image comprises a comparison of the pattern information and the trained feature.

[0267] 49. The computer-readable medium of clause 41, wherein the set of instructions is executable by at least one processor of the device to cause the device to further:

[0268] analyzing data representing the first SEM image to obtain one or more spatio-spectral characteristics; and

[0269] An enhanced image is provided by applying one or more numerical compensations to data representing the second SEM image based on the obtained one or more spatio-spectral characteristics.

[0270] 50. The computer-readable medium of clause 49, wherein the one or more spatial-spectral characteristics obtained include phase and amplitude characteristics.

[0271] 51. The computer-readable medium of clause 49, wherein the one or more numerical compensations are applied in the Fourier domain.

[0272] 52. The computer-readable medium of clause 49, wherein the one or more numerical compensations deconvolute the second SEM image.

[0273] 53. The computer-readable medium of clause 41, wherein the set of instructions is executable by at least one processor of the device to cause the device to further:

[0274] analyzing data representing the first SEM image to identify features in the first SEM image;

[0275] numerically blurring data representing the first SEM image to simulate data representing a blurred SEM image acquired at a second resolution;

[0276] Determining that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image; and

[0277] Responsive to a determination that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image, the portion of the data representing the second SEM image is identified as containing the location of the feature identified in the first SEM image.

[0278] 54. A computer readable medium according to clause 53, wherein the feature is an edge position.

[0279] 55. A computer-readable medium according to any one of clauses 53 to 54, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on a cross-correlation between the portion of the data representing the blurred SEM image and the portion of the data representing the second SEM image.

[0280] 56. A computer-readable medium according to any one of clauses 53 to 54, wherein a determination is made that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image based on a root mean square of a portion of the data representing the blurred SEM image and a root mean square of a portion of the data representing the second SEM image.

[0281] 57. A computer-readable medium according to any of clauses 41 to 56, wherein the second resolution is lower than the first resolution.

[0282] 58. A computer-readable medium according to any one of clauses 41 to 57, wherein the first resolution and the second resolution correspond to at least one of: a signal averaging amount, a noise ratio of a SEM image frame, a pixel size, a SEM beam width of a coaxial beam of a multi-beam system, a SEM beam width of an off-axis beam of a multi-beam system, a SEM beam width of a single beam system, or a current supplied to the SEM beam.

[0283] 59. The computer readable medium of any of clauses 41 to 58, wherein the first SEM image is obtained using a coaxial beam of a multi-beam system and the second SEM image is obtained using an off-axis beam of the multi-beam system.

[0284] 60. The computer-readable medium of any of clauses 41 to 58, wherein the first SEM image is obtained using a low current beam and the second SEM image is obtained using a high current beam.

[0285] Embodiments may also be described using the following third set of clauses, where references to other clauses are references to clauses in the third set of clauses:

[0286] 1. A method for enhancing an image, the method comprising:

[0287] acquiring a first scanning electron microscope (SEM) image by using a coaxial beam of a multi-beam system;

[0288] acquiring a second SEM image using an off-axis beam of a multi-beam system; and

[0289] An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

[0290] 2. A method according to clause 1, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0291] 3. The method according to clause 1, further comprising:

[0292] Trained features are extracted from the first SEM image.

[0293] 4. An inspection system comprising:

[0294] a memory storing an instruction set; and

[0295] A processor configured to execute a set of instructions to cause the inspection system to:

[0296] acquiring a first scanning electron microscope (SEM) image by using a coaxial beam of a multi-beam system;

[0297] acquiring a second SEM image using an off-axis beam of a multi-beam system; and

[0298] An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

[0299] 5. A system according to clause 4, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first image, or by numerically enhancing the second SEM image using the first SEM image as a reference.

[0300] 6. A system according to clause 4, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0301] Trained features are extracted from the first SEM image.

[0302] 7. The system of clause 6, wherein the trained features are extracted from the first SEM image using a machine learning network.

[0303] 8. A system according to clause 7, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0304] acquiring at least one additional SEM image using a coaxial beam; and

[0305] Using the machine learning network, at least one additional trained feature is extracted from at least one additional SEM image.

[0306] 9. A system according to clause 4, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0307] extracting pattern information of the second SEM image;

[0308] Using the pattern information, determining that the trained features are identified on the second SEM image; and

[0309] In response to a determination that the trained feature is identified on the second SEM image, an enhanced image is provided by enhancing the second SEM image using the identified trained feature.

[0310] 10. The system of clause 4, wherein the processor is further configured to: execute the set of instructions to cause the inspection system to acquire the first SEM image in an offline mode and extract the trained features from the first SEM image.

[0311] 11. The system of clause 9, wherein the determination that the trained feature is identified on the second SEM image comprises: a comparison of the pattern information and the trained feature.

[0312] 12. A system according to clause 4, wherein the processor is further configured to execute the set of instructions to cause the inspection system to:

[0313] analyzing data representing the first SEM image to obtain one or more spatial-spectral characteristics; and

[0314] An enhanced image is provided by applying one or more numerical compensations to the data representing the second SEM image based on the obtained one or more spatio-spectral characteristics.

[0315] 13. A system according to clause 12, wherein the one or more spatial-spectral characteristics obtained include phase and amplitude characteristics.

[0316] 14. The system according to clause 12, wherein the one or more numerical compensations are applied in the Fourier domain.

[0317] 15. The system of clause 12, wherein the one or more numerical compensations deconvolute the second SEM image.

[0318] A non-transitory computer-readable medium may be provided that stores instructions for a processor (e.g., Figure 1 The processor of the controller 109 of the computer 100 performs image processing, data processing, database management, graphic display, operation of a charged particle beam device or other imaging equipment, etc. Common forms of non-transitory media include, for example, floppy disks, flexible magnetic disks, hard disks, solid-state drives, magnetic tapes or any other magnetic data storage media, CD-ROMs, any other optical data storage media, any physical media with a pattern of holes, RAM, PROM and EPROM, FLASH-EPROM or any other flash memory, NVRAM, cache, registers, any other memory chip or cassette tape, and networked versions thereof.

[0319] The block diagrams in the accompanying drawings illustrate the architecture, functionality and operation of possible implementations of the system, method and computer hardware or software product according to various exemplary embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, segment or part of a code, which includes one or more executable instructions for implementing a specified logical function. It should be understood that in some alternative implementations, the functions indicated in the box may not occur in the order shown in the accompanying drawings. For example, depending on the functionality involved, the two boxes shown in succession can be executed or implemented substantially in parallel, or the two boxes can sometimes be executed in reverse order. Some boxes can also be omitted. It should also be understood that each box in the block diagram and the combination of these boxes can be implemented by a system based on dedicated hardware that performs a specified function or action, or by a combination of dedicated hardware and computer instructions.

[0320] It will be appreciated that the embodiments of the present disclosure are not limited to the exact constructions that have been described above and illustrated in the accompanying drawings, and that various modifications and changes may be made without departing from the scope of the present disclosure.

Claims

1. An inspection system, include: Memory, storing instruction sets; as well as a processor configured to execute the set of instructions so that the inspection system: Acquire a first scanning electron microscope (SEM) image at a first resolution; acquiring a second SEM image at a second resolution, wherein the second resolution is lower than the first resolution; as well as An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

2. The system of claim 1, wherein the enhanced image is provided by enhancing the second SEM image using one or more features extracted from the first SEM image, or numerically enhancing the second SEM image using the first SEM image as a reference.

3. The system of claim 1 , wherein the processor is further configured to execute the set of instructions to cause the inspection system to: Training features are extracted from the first SEM image.

4. The system of claim 1 , wherein the processor is further configured to execute the set of instructions to cause the inspection system to: analyzing data representing the first SEM image to obtain one or more spatial-spectral characteristics; and The enhanced image is provided by applying one or more numerical compensations to data representing the second SEM image based on the obtained one or more spatio-spectral characteristics.

5. The system of claim 1 , wherein the processor is further configured to execute the set of instructions to cause the inspection system to: analyzing data representing the first SEM image to identify features in the first SEM image; numerically blurring data representing the first SEM image to simulate data representing a blurred SEM image acquired at the second resolution; determining that a portion of the data representing the blurred SEM image fits a portion of the data representing the second SEM image; and In response to determining that the portion of the data representing the blurred SEM image fits the portion of the data representing the second SEM image, the portion of the data representing the second SEM image is identified as a location containing a feature identified in the first SEM image.

6. The system of any one of claims 1 to 5, wherein the first resolution and the second resolution correspond to at least one of the following items: signal averaging, noise ratio of a SEM image frame, pixel size, SEM electron beam width of a coaxial electron beam of a multi-electron beam system, SEM electron beam width of an off-axis electron beam of a multi-electron beam system, SEM electron beam width of a single electron beam system, or current supplied to the SEM electron beam.

7. The system according to any one of claims 1 to 5, wherein the first SEM image is obtained using a coaxial electron beam of a multi-electron beam system, and the second SEM image is obtained using an off-axis electron beam of the multi-electron beam system.

8. The system according to any one of claims 1 to 5, wherein the first SEM image is obtained using a low current electron beam and the second SEM image is obtained using a high current electron beam.

9. A non-transitory computer-readable medium storing a set of instructions executable by at least one processor of a device to cause the device to perform a method, the method include: Acquire a first scanning electron microscope (SEM) image at a first resolution; acquiring a second SEM image at a second resolution, wherein the second resolution is lower than the first resolution; as well as An enhanced image is provided, the enhanced image being provided by enhancing the second SEM image using the first SEM image as a reference.

Images