A method for obtaining the actual pixel size of a charged particle beam scanning imaging device

Abstract

The present invention discloses a method for obtaining the actual pixel size of a charged particle beam scanning imaging device, comprising: acquiring a template image on a wafer, extracting a template from the template image; moving the wafer in a preset direction and acquiring a target image, performing template matching in the target image according to the template to obtain a matching position, obtaining a target displacement between the template and the matching position, using the target image as the template image, performing the next template extraction and template matching to obtain another target displacement, stopping image acquisition and template extraction when a preset stop condition is met; and obtaining the actual pixel size based on the accumulated target displacement and the actual wafer displacement. The present invention provides a method for obtaining the actual pixel size of a charged particle beam scanning imaging device, wherein multiple matching can be performed to improve the speed, accuracy, and reliability of obtaining the actual pixel size.

Description

Technical Field

[0001] The present invention relates to the technical field of semiconductor equipment, and in particular to a method for acquiring the actual pixel size of a charged particle beam scanning imaging device. Background Art

[0002] In the front-end manufacturing process of semiconductor large-scale integrated circuits, charged particle beam scanning imaging equipment is often required. For example, a scanning electron microscope (SEM) is used to perform defect detection or re-inspection on wafers. SEM equipment can be an initial inspection device (E-Beam Inspection, EBI) or a re-inspection device (E-Beam Review, EBR), and in principle, they are all electron beam scanning imaging systems. Taking the EBR device as an example, but not excluding the EBI device in certain applications, a quick and easy method is needed to obtain the actual pixel size of the device for wafer scanning imaging, which is used to estimate the actual physical size of the object to be tested (not the size in pixels), such as the defect size or critical dimension (Critical Dimension).

[0003] refer to Figure 1 The EBR device 100 usually includes a mechanical motion platform 110, which can move and rotate in the X, Y, and Z directions, and has an electrostatic tray (Electro-Static Chuck, E-Chuck) 120 on which the wafer can be placed, on which the wafer (Wafer) 111 can be placed. The EBR device usually also includes an optical microscope imaging system (Optical Microscope, OM) 130, which has a low magnification but a large field of view (FOV), and is usually used for auxiliary work such as primary wafer alignment. The EBR device also includes a core task component 140, which is an electron optical imaging system, that is, a scanning electron microscopy (SEM) system, including a lens barrel, which has functional components such as electron emission, focusing, beam limiting and scanning, as well as a collection and signal amplification processing circuit for electrons (mainly secondary electrons) emitted from the wafer surface. The EBR device also includes a computer 150, which can be used to process data and display images.

[0004] refer to Figure 2A and Figure 2BThe electron beam imaged by the SEM system in the EBR device is shown as 210, with the focal plane at 211 and the depth of focus at 212. The z-direction represents the density of the electron beam 210, and the x-direction represents the x-coordinate on the wafer. During operation, the SEM system scans back and forth across a defined area, sampling at intervals of Δx and Δy in the X and Y directions on the wafer surface (and within its focal depth / range). The system must dwell at each sampling point for a certain amount of time to accumulate sufficient emitted electrons (to achieve a certain signal-to-noise ratio). The resulting SEM image is, in theory, roughly equivalent to the convolution of the electron beam spot shape (approximately a Gaussian distribution) upon reaching the wafer surface and the wafer surface topography (including different materials and structures). Once the theoretical field of view (FOV) of an SEM system and the number of pixels in the image in the X and Y directions are determined, the pixel size in the X and Y directions, or the nominal pixel size (NPS), is also determined. The nominal pixel size is the theoretical FOV divided by the image size (for example, an image size of 1024 / 2048 pixels in width / height). Therefore, the nominal pixel size is Δx and Δy. In actual use, the operating parameters and conditions of SEM systems often drift. For example, this can be caused by 1) slight changes in the actual beam spot size or the scanning sampling interval determined by the device control circuitry; or 2) changes in the working distance between the SEM objective lens and the wafer surface (due to, for example, Z-axis drift of the mechanical motion stage or variations in wafer thickness, which alter the actual sampling interval on the wafer). This combination of factors can lead to significant discrepancies between the theoretical and actual pixel sizes. Therefore, a more accurate representation of the actual pixel size is often required in practical applications.

[0005] Please refer to Figure 3A and Figure 3BA commonly used method for obtaining the actual pixel size in the prior art is to image a wafer or sample of known size and then calculate the actual pixel size. For example, 1) a custom wafer is used with an object of known size on it, or 2) a custom sample 311 is placed on a mechanical motion platform 310, and the image 313 captured on sample 311 contains an object 312 of known size and a known length, such as 5μm. However, the disadvantages of method 1 are obvious. Custom wafers increase cost, and their thickness cannot be guaranteed to be close to that of wafers used in actual work, resulting in a difference between the distance from their surface to the lens barrel in the SEM system, i.e., the working distance, and the distance from the actual customer wafer surface to the lens barrel in the SEM system. Furthermore, the need to load and unload the wafer each time it is used is very inconvenient, making it difficult to implement on an actual production line. In addition, the custom sample 311 is prone to local material damage after multiple scans, and the working distance from the surface of the sample 311 (and the custom wafer) to the electron microscope may differ significantly from the working distance from the actual wafer surface to the electron microscope in actual applications. Therefore, in the current EBR equipment operation, there is also a method 2 to directly use the customer wafer (limited to patterned wafers) to obtain the actual pixel size. The process is roughly as follows:

[0006] Figure 4 The current method for measuring actual pixel size is shown by capturing images of the wafer before and after movement (two frames in total) and performing template matching (also known as image matching). (The diagram is not drawn to scale.) The coordinate axes of the mechanical motion platform coordinate system are the x-axis and y-axis, and the field of view 402 of the wafer 401 and the SEM system when capturing the image are shown. The position of field of view 402 is fixed. The principle is as follows: first, the first frame of the image is captured, i.e., the template image. An image within it is selected as template 403. Template 403 is located at (Xm, Ym). The displacement of the mechanical motion platform in the X and Y directions is dXs and dYs, respectively. dXs and dYs are labeled 404 and 405 in the diagram. The second frame of the image is captured, i.e., the target image, and template matching is performed to determine the matching position (Xm′, Ym′). Dividing dXs by (Xm′-Xm) and dYs by (Ym′-Ym) gives the actual pixel size.

[0007] Typically, the actual pixel size needs to be determined for different magnifications (corresponding to different fields of view and different pixels to be measured). For example, some EBR applications often operate in multiple magnification ranges, including low magnification (LM) and high magnification (HM), each with multiple magnification values. For example, LM has three different theoretical pixel sizes of 500nm, 400nm, and 50nm, corresponding to three different magnifications. HM has three different theoretical pixel sizes of 20nm, 10nm, and 2nm, corresponding to three different magnifications. These are all determined by the specific application.

[0008] by Figure 5A For example, method 2 usually includes:

[0009] 1) Determine a certain position on the wafer, capture a template image 510 , and select a template 511 from the template image 510 ;

[0010] 2) Move the mechanical motion stage, with the wafer and the mechanical motion stage moving synchronously, to another position to capture the target image 512 and record the relative displacement dXs, dYs of the mechanical motion stage; the movement position is limited to ensure that the target remains in the target image (that is, within the field of view of the SEM system) when the mechanical motion stage has a known error range;

[0011] 3) searching for a template in the target image, i.e., performing template matching, to obtain a matching position 513;

[0012] 4) Obtain the distances dXm, dYm between the template 511 and the matching position 513, i.e., dXm and dYm are the target displacements in the x-direction and y-direction, respectively, as follows:

[0013] dXm=Xm′-Xm

[0014] dYm=Ym′-Ym

[0015] Where (Xm, Ym) is the position of the template in the template image (initial position), and (Xm′, Ym′) is the position where the template is matched in the target image (matching position).

[0016] 5) Get the actual pixel sizes Px and Py, specifically:

[0017] Px=dXs / dXm

[0018] Py=dYs / dYm

[0019] The actual displacement of the mechanical motion platform in the x- and y-directions is dXs and dYs, meaning the coordinate position change of the mechanical motion platform is (dXs, dYs). For example, the precision barcode reader / laser interferometer typically included on the mechanical motion platform can provide a more precise indication of the actual arrival position of the mechanical motion platform. The actual arrival position may differ slightly from the position commanded by the system, typically with an accuracy of 0.5 μm, for example.

[0020] Please refer to Figure 5B If the LM has a higher magnification, the method is the same, which is to collect the template image 520, select the template 521 from it, and then move the wafer to collect the target image 522, perform template matching to obtain the matching position 523. Please refer to Figure 5C The method is the same for HM, which is to collect template image 530, select template 531 from it, and then move the wafer to collect target image 532, perform template matching to obtain matching position 533.

[0021] The above methods have the following problems: In method 1, a special wafer or sample is required to obtain the actual pixel size, which is time-consuming and costly (occupies valuable machine time of equipment on the IC production line), and the working distance between the wafer or sample and the electron optical lens tube is significantly different from the working distance between the customer wafer and the electron optical lens tube, resulting in errors, resulting in poor accuracy of the obtained actual pixel size and inconvenience in use; in method 2, on the one hand, only a single frame of target image is collected, a single template matching is performed, and the mechanical motion platform moves once, with a limited distance, so the relative error is large, resulting in a large relative error in the actual pixel size; on the other hand, the entire measurement only uses a single template matching, and since it is often affected by image quality (for example, brightness, contrast or noise), the matching of grayscale images including their gradient feature images is prone to failure or large errors.

[0022] In summary, in the prior art, the time cost is high, and the actual pixel size is obtained by only a single template matching, which results in a large error. Therefore, the calculated actual pixel size has poor accuracy and reliability. Summary of the Invention

[0023] The object of the present invention is to provide a method for acquiring the actual pixel size of a charged particle beam scanning imaging device, so as to improve the speed, accuracy and reliability of the calculated actual pixel size.

[0024] To achieve this object, the present invention adopts the following technical solutions:

[0025] A method for obtaining the actual pixel size of a charged particle beam scanning imaging device, comprising:

[0026] Acquiring a template image on a wafer, and extracting a template from the template image;

[0027] Moving the wafer in a preset direction and capturing a target image, performing template matching in the target image according to the template to obtain a matching position, obtaining a target displacement between the template and the matching position, using the target image as the template image, performing the next template extraction and template matching to obtain another target displacement, and stopping image acquisition and template extraction when a preset stop condition is met;

[0028] The actual pixel size is obtained according to the accumulated target displacement and the actual displacement of the wafer.

[0029] Beneficial effects of the present invention:

[0030] The present invention provides a method for obtaining the actual pixel size of a charged particle beam scanning imaging device. On the one hand, the method eliminates the need to prepare a special wafer or sample as in the prior art to obtain the actual pixel size, thereby saving time in obtaining the actual pixel size and improving the measurement accuracy of the actual pixel size. On the other hand, multiple matching operations are performed to obtain the actual pixel size, thereby reducing errors and improving the accuracy and reliability of the calculated actual pixel size. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the relevant parts of an EBR device in the prior art;

[0032] Figure 2A This is a schematic diagram of the density distribution of an electron beam in a SEM system in the prior art;

[0033] Figure 2B is a schematic diagram of the focal length and focus depth of an electron beam in the prior art;

[0034] Figure 3A It is a schematic diagram of a specially prepared sample placed on a mechanical motion platform in the prior art;

[0035] Figure 3B This is a schematic diagram of an SEM image obtained from a specially prepared sample in the prior art;

[0036] Figure 4 is a schematic diagram of a method for obtaining the actual pixel size of an image in the prior art;

[0037] Figure 5A It is a schematic diagram of performing template matching in a target image using a template at a LM magnification in the prior art;

[0038] Figure 5B This is a schematic diagram of another LM magnification in the prior art using a template to perform template matching in a target image, and the magnification is greater than Figure 5A The magnification is high;

[0039] Figure 5C It is a schematic diagram of performing template matching in a target image using a template at an HM magnification in the prior art;

[0040] Figure 6 is a flow chart of a method for obtaining actual pixel size provided by an embodiment of the present invention;

[0041] Figure 7 is a schematic diagram of relay template matching provided by an embodiment of the present invention;

[0042] Figure 8 is a schematic diagram of an image neutron template provided by an embodiment of the present invention;

[0043] Figure 9A is a schematic diagram of reverse matching provided by an embodiment of the present invention;

[0044] Figure 9B is a schematic diagram of a forward matching provided by an embodiment of the present invention;

[0045] Figure 9C Schematic diagram of the first m positions and the last n positions where the mechanical motion platform stays after the first round of forward template matching provided by an embodiment of the present invention;

[0046] Figure 10 It is a schematic diagram of pairing in a starting set and an ending set provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present invention are described clearly and completely below with reference to the accompanying drawings.

[0048] Please refer to Figure 6 , an embodiment of the present invention provides a method for obtaining the actual pixel size, comprising the following steps:

[0049] S1. Acquire a template image on a wafer, and extract a template from the template image;

[0050] In this embodiment, the wafer is placed on a mechanical motion platform, which is controlled to move in a certain step length to move the wafer. The displacement of the wafer is the actual displacement of the wafer, and the image acquisition system in the charged particle beam scanning imaging device (for example, an SEM system) is used to realize image acquisition of different image acquisition areas on the wafer.

[0051] In this embodiment, an image is collected in the image collection area of the wafer, and after processing, a normal grayscale image or an edge image extracted therefrom can be obtained. The edge image can be extracted by using methods such as Sobel, Canny, or convolution of the image with the gradient ΔG of the Gaussian function G. An area in the template image that contains many features (edges and / or corners), is clear, and is unique in the image can be selected as the template, such as Figure 5A In the template 511, generally speaking, an image usually contains many edge (gradient) features and corner features in the X and Y directions, especially in LM, which has richer features. The present invention does not limit whether the template is selected manually or automatically, because the principles of the two are the same.

[0052] Template selection algorithms include edge features extracted from a normal grayscale image or from it (the same method as above) and corner features (for example, using Harris, FAST, etc.). Additionally, templates can be obtained using methods available in commercial image processing / machine vision software, which will not be detailed here.

[0053] Image template matching typically calculates the similarity of corresponding regions. Common algorithms include Normalized Cross Correlation (NCC), which is used for grayscale images or their gradient images. Alternatively, image features such as Sobel, Canny, and FAST can be used to match feature vectors. The calculation can be performed using all or a portion of the pixels in the template, or using a combination of single or multiple feature points (with fixed distances and orientations between them). These features, such as edge features, include gradient values in the X and Y directions. FAST, for example, has its own specially defined descriptor features. These features, regardless of form, can all constitute feature vectors. When matching features, the dot product between the feature vectors of each feature point in the template image and the feature vectors of each feature point in the target image is averaged or weighted, normalized between 0 and 1. The maximum value corresponds to the best match. In practice, a threshold is typically imposed; for example, the successful match threshold can be 0.65.

[0054] In this embodiment, extracting a template from the template image includes:

[0055] Extracting a sub-template that meets a preset extraction condition from the template image, wherein the sub-template includes a sub-image and / or independent feature points;

[0056] A plurality of sub-templates are used as the template.

[0057] In this embodiment, the sub-template includes a sub-image and / or independent feature points. The sub-image can be a small grayscale image or its edge image area. The sub-image includes features (features include edges and / or corners). Generally, the features and the smaller image area around them are used as sub-images. The independent feature points come from a point (pixel) in the image and its surrounding neighborhood, such as corner points extracted by the FAST algorithm or feature points extracted by the SIFT algorithm. The sub-template can include both sub-images and independent feature points as long as their positions can be clearly defined, for example, in Figure 7 In FIG, sub-template 711 and sub-template 712 are sub-images, and sub-template 713 is an independent feature point.

[0058] Selecting the above-mentioned sub-templates has the following advantages: 1) It is easier to select a sub-template (smaller in size) in an image of limited scale than to select a larger template; 2) The search space of the sub-template with smaller size is relatively larger during matching, and the restriction on the single movement distance of the mechanical motion platform is smaller; 3) If multiple sub-templates are selected as templates, then there are strong constraints on the distance and orientation between the multiple sub-templates (after matching, the distance and orientation between the templates are checked, and those that change too much and exceed the established threshold are not considered to be correct matches). In this way, the image quality requirements can be appropriately reduced when matching multiple templates, that is, the similarity threshold can be significantly lower than that of matching a single template. For example, in a certain application, the similarity threshold of a single template match is 0.70, the similarity threshold of each sub-template match in a template composed of two sub-templates is 0.60, and the similarity threshold of each sub-template match in a template composed of three sub-templates is 0.50 or even lower, so it can better cope with poor quality images; 4) If multiple sub-templates are selected as templates, a single sub-template does not need to be unique in the template image; 5) When there are multiple sub-templates in the image, different templates can be found from the multiple sub-templates. The template selection is more flexible and is also conducive to the successful implementation of the long-distance relay template matching in the present invention (which will be described in detail later). On the one hand, if a larger template is used, the motion step length of the mechanical motion platform will be limited, and the selection of the motion step length will be restricted (worrying that the target is out of bounds, that is, part of the target runs out of the field of view), while the sub-template size is smaller and can be flexibly combined, so there is no such restriction when using sub-templates; on the other hand, please continue to refer to Figure 7 , using sub-templates 701, 702 and 703 as templates to perform template matching in the target image, if the matching position of sub-template 701 exceeds the range of the target image, but there is still a matching position 702′ (or 703′) of sub-template 702 (or 703), positioning can be achieved based on the matching position 702′ (or 703′), and the matching position can also be determined more accurately.

[0059] In one embodiment of the present invention, the adopting of the plurality of sub-templates as the template includes:

[0060] It is determined whether the number and / or type of the sub-templates meet the preset requirements. If so, several sub-templates that meet the preset requirements are used as the template.

[0061] Among them, judging the type of the sub-template means judging whether the sub-template includes only a sub-image, only independent feature points, or both a sub-image and independent feature points. In addition, the template may include sub-templates that have not been adopted, or may include some sub-templates that have been adopted.

[0062] In the embodiment of the present invention, the adopting of the plurality of sub-templates as the template specifically includes:

[0063] If the template image only includes the sub-image, the sub-image that reaches or exceeds the preset sub-image number threshold u is used as the template; if the template image only includes the independent feature points, the independent feature points that reach or exceed the preset independent feature point number threshold v are used as the template; if the template image includes both the sub-image and the independent feature points, the sub-template and the independent feature points when the sum of the number of the sub-template and the independent feature points reaches or exceeds the preset total number threshold w are used as the template, or the sub-image that reaches or exceeds the preset sub-image number threshold u is used as the template, or the independent feature points that reach or exceed the preset independent feature point number threshold v are used as the template.

[0064] A sub-image number threshold u, an independent feature point number threshold v, and a total number threshold w may be set, and u≥2 and v≥5, but the present invention is not limited thereto.

[0065] In addition, a threshold requirement may be set for the distance between each sub-template in the template, that is, the distance between each sub-template in the template is greater than or equal to a preset minimum distance, such as ≥25 pixels, and at the same time cannot be too far apart, such as ≤1 / 3 of the image width or height.

[0066] Compared to independent feature points, sub-images encompass a larger image area, which helps ensure template uniqueness. Therefore, a smaller sub-image number threshold u and a larger independent feature point number threshold v can be set. Accordingly, template uniqueness can be ensured with a smaller number of sub-images, while template uniqueness can be ensured with a larger number of independent feature points. Specifically, the sub-image number threshold u is less than the independent feature point number threshold v, and the total number threshold w is between the sub-image number threshold u and the independent feature point number threshold v (u≤w≤v).

[0067] Among them, if there is a single sub-image that meets the requirements of the preset extraction conditions (for example, the brightness, contrast and uniqueness all meet the corresponding threshold requirements) and its size exceeds the established threshold in the template image, for example, it accounts for 1 / 4, then the single sub-image can be used as a template. It should be noted that in actual applications, this situation should be avoided as much as possible. A template that is too large will have other side effects. For example, it can make the single stroke smaller, resulting in excessive image acquisition in the same graphic acquisition area of the wafer, or it may be detrimental to the relay template matching described later, because the continuous emergence of new sub-templates within a limited movement distance is conducive to completing the relay template matching.

[0068] Specifically, the following settings can be made: 1) If the template is entirely sub-images, the threshold for the number of sub-images is u≥2, preferably u≥3. Of course, in the case of insufficient sub-images, a sufficiently large sub-image as described above can also be used; 2) If the template is entirely independent feature points, the threshold for the number of independent feature points is v≥5, preferably v≥7. For example, the number of independent feature points should be one between 7 and 11, and they should be at a certain distance from each other; 3) If the template is mixed, including both sub-images and independent feature points, for example, when there is one sub-image, the number of independent feature points is required to be greater than or equal to 4. When there are two sub-images, two sub-images are also acceptable, but it is best to choose two sub-images and greater than or equal to 1 independent feature point.

[0069] S2, moving the wafer in a preset direction and capturing a target image, performing template matching in the target image according to the template to obtain a matching position, obtaining a target displacement between the template and the matching position, using the target image as the template image, performing the next template extraction and template matching (in the next target image) to obtain another target displacement (this template matching method is defined as relay template matching, or relay matching), and stopping image acquisition (stopping acquiring template images and target images) and template extraction when a preset stop condition is met;

[0070] Please continue to refer to Figure 7In one embodiment of the present invention, a first frame of image is captured in an image acquisition area, and a first template that meets a preset extraction condition is extracted from the first frame of image. The first template includes one or more (at least two) sub-templates that meet the preset extraction condition. For example, three sub-templates can be selected as the first template, and the three sub-templates are sub-templates 701, 702, and sub-template 703; the mechanical motion platform is moved in a preset direction to move the wafer, and a second frame of image is captured in another image acquisition area. The first template is used to perform template matching in the second frame of image to obtain a matching position, and the matching position includes matching positions 701′, 702′, and matching position 703′, and the target displacement between the first template and the matching position is obtained. For example, the point in the upper right corner of the first template (but not limited to this) can be used as the coordinate position (Xm1, Ym1) of the first template, and the point in the upper right corner of the matching position can be used as the coordinate position (Xm2, Ym2) of the matching position, then the target displacement can be obtained. The decomposition distances of the target displacement in the X direction and the Y direction are dXm1 and dYm1 respectively, and dXm1=Xm2-Xm1, dYm1=Ym2-Ym1; the second frame image is used as the template image, and the second template that meets the preset extraction conditions is extracted from the second frame image. The sub-templates in the second template can be obtained only from the sub-templates 711, 712 and sub-template 713 that have not been extracted, or can be obtained not only from the sub-templates 711, 712 and sub-template 713, but also from the extracted sub-templates 701′, 702′ and sub-template 703′. The mechanical motion platform and the wafer are moved in the preset direction, and the third frame image is collected. The second template is used to perform template matching in the third frame image, which will not be repeated here.

[0071] The preset stopping condition for relay template matching can be reaching a preset maximum number of matches (I), for example, a maximum of 19 matches; or reaching a preset value for the cumulative sum of the wafer's actual displacements, for example, when the total travel of the mechanical motion platform reaches a preset maximum; or failing to obtain a template that meets the preset extraction conditions in the current image, for example, when the preset extraction conditions can include brightness, contrast, and uniqueness all meeting corresponding threshold conditions; or failing to obtain a number of sub-templates that meet the threshold conditions in the current image, for example, when the number of sub-templates is insufficient or the number of unextracted sub-templates is zero. Of course, the preset stopping condition is not limited to this. For example, the preset stopping condition can also be a combination of at least two of the above conditions, with the preset stopping condition being considered satisfied as long as one of them is met. It should be noted that when the mechanical motion platform moves a small distance, it is possible to ensure that a matching position is found, but this should be within the allowable range of system error. Furthermore, the template can be corrected. If the template matching fails, the template is re-extracted from the template image and template matching is performed again until a successful match is obtained to obtain the matching position.

[0072] In yet another embodiment of the present invention, when there are multiple matchable templates, the one with the highest matching degree is used as much as possible.

[0073] Exemplarily, extracting a template from the template image, moving the wafer in a preset direction and capturing a target image, and performing template matching in the target image according to the template to obtain a matching position includes:

[0074] Extract multiple templates from the template image, move the wafer in a preset direction and capture a target image, perform template matching in the target image based on the multiple templates, and select the template with the highest matching degree to obtain the matching position.

[0075] Preferably, the motion step length of the mechanical motion platform is a distance that satisfies i) the template has the largest search space in the target image and ii) the mechanical motion platform has the largest single stroke, but is not limited thereto. The user can set the motion step length according to actual needs.

[0076] The template includes a sub-template that meets a preset extraction condition, and the sub-template includes a sub-image and / or independent feature points, which have been introduced above and will not be repeated here.

[0077] by Figure 8 Taking the template image 810 and target image 820 given in as an example, there are many sub-templates (sub-templates a, b, c, d, e, f, g and h) in the template image 810, which can be composed of multiple templates. For example, at least two sub-templates are required to form a template, and there are templates ab, template abc, template abd, template ce, template cef, template ef and template fgh to choose from, all of which can be used for template matching. The template with the best match is used as the template of the current template image 810, and the target image 820 also has sub-templates o, p, q, r, s and t that have not been extracted, which can be composed of multiple templates. Similarly, the template used for the next template matching can be obtained.

[0078] Please continue to refer to Figure 8 If the template extracted from template image 810 fails to match the target image 820, a new template is selected from template image 810. Flexible combinations of more sub-templates, such as forming a template based on sub-templates bcdef, combined with a smaller motion step size for the mechanical motion platform, effectively ensure successful matching, essentially achieving a preset maximum number of matches I, for example, 19. In actual applications, even if the preset maximum number of matches I is not reached, for example, 15 successful template matches are completed and the matching process stops due to a preset stop condition being triggered, the accuracy of the calculated actual pixel size can still be significantly improved compared to existing technologies.

[0079] S3. Acquire the actual pixel size according to the accumulated target displacement and the actual displacement of the wafer.

[0080] It should be noted that by selecting an appropriate template and motion step size, the success of the first and second template matching can be ensured, so at least two actual wafer displacements and two target displacement values can be accumulated. In fact, the image collected during LM has more features, which can provide more sub-templates to choose from, while the image during HM still has more fragmented features (including edge segments, corner points, or specialized feature points like FAST and SIFT). Therefore, it is easy to find sufficient sub-templates in the image. Combined with an appropriate motion step size, the continuation of template matching can be guaranteed, and multiple matches (generally at least three times, but not excluding two times) can be completed, which can reduce errors and improve the accuracy and reliability of the calculated actual pixel size.

[0081] It should be noted that the actual pixel size can be obtained based on the accumulated actual displacement of part or all of the wafer and the target displacement. The formula for calculating the actual pixel size using relay template matching in this embodiment is given below, which takes the actual pixel size obtained by accumulating all the actual displacement of the wafer and the target displacement as an example.

[0082] The actual pixel size is obtained by establishing a formula based on the accumulated actual displacement and target displacement of the wafer. For example, if n+1 frames of images are collected and n template matchings are successfully completed, after the i-th template matching is successful, the coordinate position (Xs i ,Ys i ) moves to the coordinate position (Xs) of the mechanical motion platform corresponding to the i+1 frame image i+1 ,Ys i+1 ), the coordinate position change of the mechanical motion platform is (dXs i ,dYs i ), dXs i and dYs i are the actual displacements of the wafer corresponding to the i-th template matching in the x-direction and y-direction respectively, specifically:

[0083] dX i =Xs i+1 -Xs i

[0084] dY i =Ys i+1 -Ys i

[0085] The coordinate position of the template extracted from the i-th frame image is (Xm i ,Ym i), the coordinate position of the matching position obtained in the i+1 frame image is (Xm i+1 ,Ym i+1 ), we can get the target displacement between the template in the i-th frame image and the matching position in the i+1-th frame image (which can be decomposed into the distance in the X direction and the distance in the Y direction, and the distance is the change in coordinate position). Without a doubt, we can know that the corresponding coordinate position change is (dXm i ,dYm i ), dXm i and dYm i are the target displacement corresponding to the i-th template matching in the x-direction and y-direction respectively, specifically:

[0086] dX i =Xm i+1 -Xm i

[0087] dY i =Ym i+1 -Ym i

[0088] Therefore, the actual pixel size obtained after completing the relay template matching is:

[0089]

[0090] After completing the relay template matching, the coordinate position of the starting position of the mechanical motion platform (Xs1, Ys1) and the coordinate position of the end position (Xs n+1 ,Ys n+1 ), where the coordinate position of the starting position (Xs1, Ys1) is the coordinate position of the mechanical motion platform corresponding to the first frame image, and the coordinate position of the end position (Xs n+1 , Ys n+1 ) is the coordinate position of the mechanical motion platform corresponding to the n+1th frame image, so the above formula is further simplified to:

[0091]

[0092] Wherein, 1≤i≤n, n is a positive integer, and n≥2.

[0093] In order to demonstrate that the above relay template matching method can reduce errors and improve the accuracy and reliability of the calculated actual pixel size, the error analysis process will be given below.

[0094] The above formula (1) for calculating the actual pixel size Px and Py is given. It is an ideal formula and does not take into account the influence of errors. In actual measurement, there are errors. In the following, it is assumed that the maximum error of the motion position of the mechanical motion platform (hereinafter referred to as error) is δS (usually around 0.5 μm), and the maximum error of single image template matching (hereinafter referred to as error) δ M (usually between 0.025 pixels and 0.25 pixels, determined by many factors.) To better illustrate the problem, let’s use the larger 0.25 pixels as δ M When the FOV is 10 μm and the image width and height are both 1024 pixels, the theoretical pixel value is about 10 nm, δ M The value is about 2.5nm. The actual pixel size of the entire measurement is its actual value (Px, Py) plus the error (maximum error) term (δPx, δPy), specifically:

[0095]

[0096] Expand it with

[0097]

[0098] Since the number of template matching times i actually completed is limited, it is less than or equal to the preset maximum matching number I, for example, I = 19. As mentioned above, δ M About 0.25 pixels, compared to the position of each match (dXm i ,dYm i ) (average 250 to 500 pixels in a 1024 × 1024 image) is about three orders of magnitude different, nδ M (2.5 pixels) and the sum of all displacements in the image or The difference is also about 3 orders of magnitude, so nδ can be discarded in the denominator of formula (2) M Term, so we have the following high-precision approximation of the actual pixel size, that is, the actual pixel sizes Px and Py are close to the ideal value formula (1) without considering the error above, specifically:

[0099]

[0100] The error estimate at this time is:

[0101]

[0102] As above, nδ in the denominator can be ignored M Term, so the error estimate is:

[0103]

[0104]

[0105] Since the error of the mechanical motion platform here is fixed, its error only comes from the position readings of the starting position and the end position. Obviously, the longer the total stroke of the mechanical motion platform, the smaller the error. If a more rigorous derivative error analysis method is used, the same conclusion will be reached. In most LM cases, due to the rich features in the template image, there are usually many templates to choose from. The present invention can support the selection of more than one for matching and finally use the best match; while in the HM case, the features in the template image are sparse, but the present invention is not limited to using sub-images, but can use curve features and hybrid templates (sub-images and independent feature points). Therefore, a template image can also support the selection of more than one for matching and finally use the best match, which can ensure that a predetermined number of matches are achieved. In addition, the template can be reselected when the matching fails, so that the mechanical motion platform can always have a longer moving distance, which can reduce the relative error and improve the accuracy and reliability of the calculated actual pixel size.

[0106] When this method is actually used, in order to complete a round of template matching (relay template matching or non-relay template matching mentioned later), the movement step of the mechanical motion platform should be large, and its total stroke should also be large. On the one hand, the relative error is small at this time, and on the other hand, it can also avoid excessive collection of SEM images at similar positions (when the movement step of the mechanical motion platform is small, the position where the mechanical motion platform stays is similar), especially when HM, to avoid damaging the image acquisition area of the wafer due to excessive scanning (for example, local polarization produces local potential).

[0107] Considering that an actual pixel size with higher accuracy and reliability can be obtained when the total stroke of the mechanical motion platform is relatively long, as a variation of this embodiment, position pairing can be performed. A pixel size measurement value (the measurement value of the actual pixel size, referred to as the pixel size measurement value for short) can be obtained based on any pairing, and then the actual pixel size is obtained based on the pixel size measurement value. No additional stroke of the mechanical motion platform and pauses are added for image collection, so as to fully utilize the first round of forward template matching (which does not involve more mechanical motion) to obtain more pixel size measurement values, which is conducive to obtaining an actual pixel size with higher statistical accuracy.

[0108] Specifically, please refer to Figure 10 , step S3 includes:

[0109] Establishing a starting set comprising first k positions and an ending set comprising last k positions such that a distance between two closest positions in the starting set and the ending set reaches or exceeds a preset distance threshold, wherein the wafer is imaged at the positions;

[0110] Establishing a one-to-one pairing relationship between the starting set and the ending set to form k pairs, so that any position in the starting set and the ending set is paired only once;

[0111] Obtaining a starting calculation position in the starting set, obtaining a termination calculation position paired with the starting calculation position in the termination set, and obtaining the target displacement, the actual wafer displacement, and a corresponding pixel size measurement value accumulated between any pair of the starting calculation position and the termination calculation position;

[0112] Obtaining the actual pixel size according to the statistical results of k pixel size measurement values;

[0113] Wherein, k≥1, the starting set includes the starting position and k-1 positions around it, and the ending set includes the ending position and k-1 positions around it.

[0114] Taking the example of a mechanical motion platform having 20 positions and k=5 after completing the first round of template matching, the first position and the 20th position are the starting position and the ending position, respectively. The total stroke of the mechanical motion platform (the distance between the first position and the 20th position) can be obtained by pairing only the first position and the 20th position, and respectively serving as the starting calculation position and the ending calculation position, and then substituted into formula (1) to obtain a pixel size measurement value; the second position and the 19th position can also be paired, and respectively serving as the starting calculation position and the ending calculation position, and then the total stroke of the mechanical motion platform (the distance between the second position and the 19th position) can be obtained, and then combined with the corresponding accumulated target displacement and the revised formula (1) to obtain another pixel size measurement value, wherein the numerator in the revised formula (1) is the distance between the second position and the 19th position (movement of the mechanical motion platform), and the denominator comes from the accumulated target displacement between the second position and the 19th position. Of course, there are other pairing situations (for example, in another embodiment, the second position can be paired with the 20th position), which will not be repeated here.

[0115] It should be noted that obtaining the actual pixel size based on k pixel size measurement values is not limited to obtaining the actual pixel size based only on these k pixel size measurement values. In the following embodiments, other pixel size measurement values can also be obtained, and the actual pixel size can be obtained based on the statistical results of all these pixel size measurement values.

[0116] The above pairing method can obtain k pixel size measurements, and the actual pixel size can be obtained based on these k pixel size measurements, which is conducive to obtaining the actual pixel size with high statistical accuracy and high reliability. In addition, due to the limitation of the distance threshold, the actual pixel size can be further guaranteed to be obtained with high accuracy and reliability.

[0117] As mentioned above, multiple matches are performed, and by selecting a suitable template and a suitable motion step size, it is possible to ensure that multiple matches are successful, and to obtain an actual pixel size with higher accuracy and reliability, with a smaller error. However, there is still a statistical error, and it is necessary to further improve the accuracy and reliability of the actual pixel size.

[0118] Based on this, in this embodiment, a round-trip matching is performed to obtain the actual pixel size, wherein the round-trip matching includes the first round of template matching. The round-trip matching can reduce statistical errors and improve statistical accuracy and reliability.

[0119] In this embodiment, the preset direction includes a forward direction and a reverse direction, and performing round-trip matching to obtain the actual pixel size includes:

[0120] First, a first round of forward template matching (relay template matching) is performed to obtain at least one pixel size measurement value (in this embodiment, as described above, k pixel size measurements can be obtained), and then a round of reverse matching is performed to obtain pixel size measurements for the corresponding round to complete a set of round-trip matching. Then, the actual pixel size is obtained based on the statistical results of the pixel size measurements of each round. The first round of forward template matching includes multiple template matchings, each round being all rounds. However, when performing statistics, the actual pixel size can be obtained based on the average value of the pixel size measurements of all rounds, or the actual pixel size can be obtained by averaging the remaining pixel size measurements after eliminating some data (with significantly larger deviations from the mean). Both methods can reduce statistical errors and improve the accuracy and reliability of the actual pixel size. Reverse matching is not limited to relay template matching, but can also be non-relay template matching, which will be described in detail later.

[0121] Furthermore, after completing a set of round-trip matching, other rounds of forward matching and / or reverse matching may be performed to obtain pixel size measurements for the corresponding rounds, and then the actual pixel size may be obtained based on the statistical results of the pixel size measurements for each round. Similarly, the actual pixel size may be obtained based on the statistical results of the pixel size measurements for all rounds, but the present invention is not limited thereto.

[0122] The number of other rounds can be odd or even. For example, after completing one set of round-trip matching, a second set (or more sets) of round-trip matching can be performed. The second set (or more sets) of round-trip matching includes one round of forward matching and one round of reverse matching, thus completing an even number of rounds of matching. Alternatively, after completing the first set of round-trip matching, a forward matching round can be performed to complete an odd number of rounds of matching.

[0123] It should be noted that forward and reverse are two opposite directions of movement. Forward matching refers to moving in the same direction as the first round of forward matching and performing template matching (not limited to relay template matching). Specifically, the wafer is moved forward by a mechanical motion platform and template matching is performed. Reverse matching refers to moving in the opposite direction to the first round of forward matching and performing template matching (not limited to relay template matching). Specifically, the wafer is moved in the reverse direction by a mechanical motion platform and template matching is performed. For example, when performing reverse matching, the mechanical motion platform can be moved in reverse from the current position (for example, the last position where the mechanical motion platform stayed during the first round of forward relay template matching, i.e., the end position) to another position (for example, returning to the first position during the first round of forward template matching, i.e., the starting position) as the starting position, and the first frame of image is captured at the starting position.

[0124] Among them, the forward matching (or reverse matching) of other rounds can be relay template matching or non-relay template matching. In this embodiment, after the first round of forward relay template matching, the templates of the mechanical motion platform at each position (the positions where the mechanical motion platform stopped during the first round of forward relay template matching) are already known. Therefore, the known templates at each position can be used to move the mechanical motion platform to that position, and the corresponding forward matching or reverse matching can be performed using the known templates, that is, non-relay template matching can be performed.

[0125] It should be noted that the templates (part or all of the templates) used in the first round of forward template matching are saved to facilitate the implementation of the forward matching or reverse matching solution in this embodiment.

[0126] In this embodiment, performing the reverse matching to obtain the pixel size measurement value of the corresponding wheel includes:

[0127] Recapture the target image in reverse direction to at least one position identical to that in the first forward round, perform template matching on the corresponding recaptured target image using the same template as that in the first forward round to obtain an updated matching position and an updated target displacement, and replace the corresponding target displacement during the template matching in the first forward round according to the updated target displacement to obtain a pixel size measurement value of the corresponding round.

[0128] It should be noted that in order to recapture the reverse image in at least one of the same positions as in the first round of forward (relay template matching), the mechanical motion platform is returned to the position where it stopped during the first round of forward relay template matching (of course, with inherent errors as described above), and then recapture the target image. It is not necessary to traverse all positions, but rather to stop at one or more positions where templates were extracted and successfully matched during the first round of forward relay template matching.

[0129] Considering the large distance between the first position (starting position) and the last position (ending position), and the large distance between positions near the starting position and those near the ending position, the m positions at the starting position can be considered as one set. Similarly, the n positions at the ending position can be considered as another set. Each position and corresponding image in the two sets can be arbitrarily paired. Since the travel of the mechanical motion platform is relatively long at this point, the relative error in calculating the actual pixel size can still be significantly reduced. In this embodiment, after completing a pairing, a pixel size measurement value can be further calculated. For example, the starting position and the ending position can be paired to achieve reverse matching to obtain a pixel size measurement value. This can significantly reduce the relative error in calculating the actual pixel size.

[0130] In this embodiment, preferably, the first set of round trips (including: the first round of forward template matching and the second round of reverse matching) is between the starting position and the end position, that is, between the positions farthest apart. Similarly, it is preferred to perform other rounds of reverse matching and / or forward matching (forward matching will be introduced later) between the starting position and the end position to achieve long-distance movement and reduce the relative error in calculating the actual pixel size.

[0131] Specifically, please refer to Figure 9C , and recapture the target image at any of the first m positions (the first position to the mth position) that are the same as the first round forward direction, where m is a positive integer less than or equal to the first preset value. For example, the target image can be recaptured at the first position (i.e., the starting position), or the target image can be recaptured at the second position in the reverse direction. If the target image is recaptured at the second position, the template in the current image corresponding to the second position can be used for template matching to replace the target displacement during the template matching in the first round forward direction (the second position is used as the starting position, the target displacement obtained by performing template matching based on the target images of the first position and the second position is discarded, and the target displacement obtained by performing template matching based on the second position and the known third position is replaced, that is, the target displacements of each template matching before the reverse movement position in formula (1) are discarded, and the remaining target displacements are replaced according to the result of template matching based on the reverse movement position and the next position) to obtain the pixel size measurement value of the corresponding round.

[0132] Please refer to Figure 9A (Additional note, although Figure 9A(The movement in the X direction is large and obvious, but in actual applications, both X and Y directions move significantly.) For example, after completing the first round of forward relay template matching, the mechanical motion platform has twenty positions (the first position to the twentieth position, the first position can also be called the starting position, and the twentieth position can also be called the ending position), and the actual number of matches completed i and the preset maximum number of matches I are both 19. Images are captured at different image acquisition areas on the wafer, respectively, to obtain the first frame image 901, the second frame image 902, the third frame image 903, up to the nineteenth frame image 919, and the twentieth frame image 920. At this point, the first template 1 (obtained from the first frame image 901, including sub-templates a and b), the second template 2 (obtained from the second frame image 902, including sub-template c), up to the nineteenth template 19 (obtained from the nineteenth frame image 919, including sub-templates s and t) are known, and the current position of the mechanical motion platform is the twentieth position. When performing forward matching and / or reverse matching in other rounds, taking reverse matching as an example, the mechanical motion platform can be moved from the 20th position to either the 1st position or the 19th position. Taking moving it to the first position as an example, image 901" is recaptured and template matching is performed using the first template 1 in the recaptured image 901", i.e., reverse matching is performed, resulting in an updated matching position 1" and an updated target displacement.

[0133] Although the equipment system instructs the mechanical motion platform to return from the twentieth position to the first position, due to the existence of errors, the actual coordinate position (Xs1", Ys1") of the mechanical motion platform often deviates from the first position (Xs1, Ys1). The first template 1 (including sub-templates a and b) is difficult to match the sub-templates a and b at the same position in the re-captured image 910", but is matched to the updated matching position 1" (including sub-templates a" and b"), resulting in a mismatch between the coordinate position (Xm1, Ym1) of the first template 1 and the updated matching position 1". The target position (Xm1", Ym1") is different. The target displacement can be decomposed into dXm1" and dYm1", dXm1" = Xm1" - Xm1, dYm1" = Ym1" - Ym1, replacing dXm1, dYm1, Xs1 and Ys1 in formula (1) when calculating the pixel size measurement value. Specifically, dXm1" is used to replace dXm1, dYm1" is used to replace dYm1, Xs1" is used to replace Xs1, and Ys1" is used to replace Ys1. Other contents remain unchanged, and then formula (1) can be used to calculate another pixel size measurement value.

[0134] Among them, when performing reverse matching, the template in the current image corresponding to the position moved in reverse is used for template matching. For example, if returning to the first position, the first template 1 in the first frame image 901 is used for template matching. If returning to the second position, the second template 2 in the second frame image 902 is used for template matching.

[0135] It should be noted that the above example uses a single movement of the mechanical motion platform during reverse matching, but is not limited to this. A single reverse matching round can include one or more reverse matching steps. For example, the mechanical motion platform can reversely move from the 20th position to the third position, capture a frame of image, and perform a template matching operation. The mechanical motion platform can then move from the third position to the first position, capture another frame of image, and perform another template matching operation.

[0136] In this embodiment, when performing reverse matching, the mechanical platform is returned to its first position (the starting point). At this point, the mechanical platform's total travel is the longest (because reverse matching in this embodiment utilizes the template from the first round of forward movement, the distance between the starting and ending positions is the longest during reverse matching), which is more conducive to reducing relative error. To distinguish it from relay-based template matching, this non-relay template matching method is referred to as end-to-end template matching.

[0137] In summary, when performing reverse matching, it is not limited to relay template matching, and non-relay template matching (for example, end-to-end template matching) can also be performed. When performing non-relay template matching, the mechanical motion platform is controlled to move to a known position (Xs i , Ys i ) and re-acquire the image, using this known position (Xs i , Ys i ) is matched with the corresponding template in the re-collected image. After the image matching is completed, the actual motion position (Xs i 『,Ys i ") and the decomposition value of the target displacement obtained by template matching (dXm i ', dYm i ') Update the (Xs in formula (1) when calculating the pixel size measurement value i , Ys i ) and (dXm i , dYm i ), an updated pixel size measurement value can be obtained. If multiple rounds of reverse non-relay template matching are required, the pixel size measurement value can be calculated once after each round, or the pixel size measurement value can be obtained by simultaneously replacing the corresponding target displacement in formula (1) after completing multiple rounds according to the situation.

[0138] Similarly, forward matching can be relay template matching or non-relay template matching. In one embodiment of the present invention, forward matching can be non-relay template matching. Specifically, the template (partial or complete template) used in the first round of forward template matching is saved, wherein the forward matching to obtain the pixel size measurement value of the corresponding round includes:

[0139] The target image is recaptured in the forward direction to at least one position (the image acquisition area of the wafer) that is the same as that in the first round of forward direction, and template matching is performed on the corresponding recaptured target image using the same template as that in the first round of forward direction to obtain an updated matching position and an updated target displacement, and the corresponding target displacement during the template matching in the first round of forward direction is replaced according to the updated target displacement to obtain the pixel size measurement value of the corresponding round.

[0140] Similarly, the first m positions (including the starting position) are taken as one set, and the last n positions (including the ending position) are taken as another set. The positions and corresponding images in the two sets can be paired arbitrarily. For details, please refer to Figure 9C , recapture the target image by forwarding to any of the next n positions (e.g., the last position) that are the same as those in the first round of forward matching, where n is a positive integer greater than or equal to a second preset value. In this embodiment, after completing one pairing, a pixel size measurement value can be further calculated. For example, the starting position and the ending position can be used for pairing to achieve reverse matching and forward matching. This can significantly reduce the relative error in calculating the actual pixel size. One round of reverse matching can obtain one pixel size measurement value, and one round of forward matching can also obtain one pixel size measurement value. That is, in this embodiment, a set of round-trip template matching including forward matching and reverse matching can obtain two pixel size measurements, thereby obtaining an actual pixel size with higher statistical accuracy.

[0141] When forward matching is performed, the template in the previous frame of the current image corresponding to the position to which the forward movement is to be performed is used for template matching. For example, if the movement is to the last position (the 20th position, i.e., the end position), the template 19 in the 19th frame image 919 is used for template matching. If the movement is to the third position, the second template 2 in the second frame image 902 is used for template matching. Please refer to Figure 9B, taking the case where the mechanical motion platform moves to the twentieth position as an example, a new twentieth frame image 920' is collected, and the known nineteenth template 19 is used to perform template matching on the twentieth frame image 920' to obtain the matching position 19'. The matching position 19' is different from the position of the nineteenth template 19, and the target displacement corresponding to the nineteenth template matching in formula (1) can be replaced accordingly. Similarly, if the motion reaches the third position, the second template 2 can be used for template matching to replace the corresponding target displacement of the template matching in the first round of forward direction (the third position is taken as the last position, and the target displacement obtained by each template matching after the third position is discarded, and the target displacement obtained by template matching the third position with the known second position is used to replace it, that is, the target displacement of each template matching after the forward motion position in formula (1) is discarded, and the remaining corresponding target displacement is replaced according to the result of template matching based on the forward motion position and the previous position) to obtain the pixel size measurement value of the corresponding wheel, which will not be repeated here. It can be seen that the mechanical motion platform can obtain min(m,n) actual pixel measurement values by moving end-to-end in one direction, which is efficient.

[0142] In summary, in this embodiment, after performing the first round of forward template matching (relay template matching), k pixel size measurement values can be obtained, and then reverse matching and / or forward matching are performed to obtain another j pixel size measurement values. The total pixel size measurement values are (k+j). The actual pixel size can be obtained based on the statistical results of the (k+j) pixel size measurement values, and its accuracy is necessarily high.

[0143] It should be noted that the above-described method for obtaining the actual pixel size can be used to obtain the actual pixel size of images at different magnifications (LM or HM). Typically, the actual pixel size at a lower magnification is calculated first, and then the actual pixel size at a higher magnification is calculated. LM or HM can have multiple magnification levels. Examples of LM and HM are provided in the background section and will not be repeated here.

[0144] In summary, in the embodiments of the present invention, after performing the first round of forward template matching (relay template matching), non-first round reverse matching and / or forward matching is also performed. That is, based on the first round of forward relay template matching, at least one round of reverse matching is performed to form at least one set of round-trip matching. This can improve measurement efficiency, reduce errors, and obtain an actual pixel size with higher statistical accuracy and reliability, thereby significantly surpassing existing technologies in terms of speed, accuracy, and reliability.

[0145] In this embodiment, the above-mentioned method for obtaining the actual pixel size is applied to a charged particle beam scanning imaging device, which includes, for example, a SEM device (e.g., an EBI device or an EBR device), and may also include a FIB (focused ion beam) device. The above-mentioned method can be used to obtain the actual pixel size of the charged particle beam scanning imaging device for wafer scanning imaging, which can then be used to estimate the actual physical size of the object to be measured, such as the defect size or critical size. The device is mainly used for detection and measurement applications related to semiconductor wafers.

[0146] The above embodiments only illustrate the basic principles and characteristics of the present invention. The present invention is not limited to the above embodiments. Without departing from the spirit and scope of the present invention, all other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative work are within the scope of protection of the present invention.

Claims

1. A method for obtaining the actual pixel size of a charged particle beam scanning imaging device, characterized in that: include: Acquiring a template image on a wafer, and extracting a template from the template image; Moving the wafer in a preset direction and capturing a target image, performing template matching in the target image according to the template to obtain a matching position, obtaining a target displacement between the template and the matching position, using the target image as the template image, performing the next template extraction and template matching to obtain another target displacement, and stopping image acquisition and template extraction when a preset stop condition is met; The actual pixel size is obtained according to the accumulated part or all of the target displacement and the actual displacement of the wafer.

2. The method according to claim 1, characterized in that The preset direction includes a forward direction and a reverse direction, and performing a round-trip matching to obtain the actual pixel size includes: A first round of forward template matching is performed to obtain at least one pixel size measurement value, and then a round of reverse matching is performed to obtain pixel size measurement values of a corresponding round to complete a set of round-trip matching. Then, the actual pixel size is obtained based on the statistical results of the pixel size measurement values of each round, wherein the first round of forward template matching includes multiple template matchings.

3. The method according to claim 2, characterized in that After completing a set of round-trip matching, other rounds of forward matching and / or reverse matching are performed to obtain pixel size measurement values of the corresponding rounds, and then the actual pixel size is obtained based on the statistical results of the pixel size measurement values of each round.

4. The method according to claim 2 or 3, characterized in that Saving the template used in the first round of forward template matching, wherein performing the reverse matching to obtain the pixel size measurement value of the corresponding round includes: Recapture the target image in reverse direction to at least one position identical to that in the first forward round, perform template matching on the corresponding recaptured target image using the same template as that in the first forward round to obtain an updated matching position and an updated target displacement, and replace the corresponding target displacement during the template matching in the first forward round according to the updated target displacement to obtain a pixel size measurement value of the corresponding round.

5. The method according to claim 3, characterized in that Saving a template used in the first round of forward template matching, wherein performing the forward matching to obtain pixel size measurements of the corresponding round includes: Recapture the target image in a forward direction to at least one position that is the same as that in the first forward round, perform template matching on the corresponding recaptured target image using the same template as that in the first forward round to obtain an updated matching position and an updated target displacement, and replace the corresponding target displacement in the template matching in the first forward round according to the updated target displacement to obtain a pixel size measurement value of the corresponding round.

6. The method according to claim 1, characterized in that Extracting the template from the template image includes: Extracting a sub-template that meets a preset extraction condition from the template image, wherein the sub-template includes a sub-image and / or independent feature points; A plurality of sub-templates are used as the template.

7. The method according to claim 6, characterized in that The adopting of a plurality of sub-templates as the template includes: It is determined whether the number and / or type of the sub-templates meet the preset requirements. If so, several sub-templates that meet the preset requirements are used as the template.

8. The method according to claim 7, characterized in that The adopting of a plurality of sub-templates as the template includes: If the template image only includes the sub-image, the sub-image that reaches or exceeds the preset sub-image number threshold u is used as the template; if the template image only includes the independent feature points, the independent feature points that reach or exceed the preset independent feature point number threshold v are used as the template; if the template image includes both the sub-image and the independent feature points, the sub-template and the independent feature points when the sum of the number of the sub-template and the independent feature points reaches or exceeds the preset total number threshold w are used as the template, or the sub-image that reaches or exceeds the preset sub-image number threshold u is used as the template, or the independent feature points that reach or exceed the preset independent feature point number threshold v are used as the template.

9. The method according to claim 8, characterized in that The sub-image quantity threshold u is smaller than the independent feature point quantity threshold v, and the total quantity threshold w is between the sub-image quantity threshold u and the independent feature point quantity threshold v.

10. The method according to claim 1, characterized in that Extracting a template from the template image, moving the wafer in a preset direction and acquiring a target image, and performing template matching in the target image according to the template to obtain a matching position includes: Extract multiple templates from the template image, move the wafer in a preset direction and capture a target image, perform template matching in the target image based on the multiple templates, and select the template with the highest matching degree to obtain the matching position.

11. The method according to claim 1, characterized in that Acquiring the actual pixel size according to the accumulated part or all of the target displacement and the actual displacement of the wafer includes: Establishing a starting set comprising first k positions and an ending set comprising last k positions such that a distance between two closest positions in the starting set and the ending set reaches or exceeds a preset distance threshold, wherein the wafer is imaged at the positions; Establishing a one-to-one pairing relationship between the starting set and the ending set to form k pairs, so that any position in the starting set and the ending set is paired only once; Obtaining a starting calculation position in the starting set, obtaining a termination calculation position paired with the starting calculation position in the termination set, and obtaining the target displacement, the actual wafer displacement, and a corresponding pixel size measurement value accumulated between any pair of the starting calculation position and the termination calculation position; The actual pixel size is obtained according to the statistical result of k pixel size measurement values.

12. The method according to claim 1, characterized in that The obtaining of the actual pixel size according to the accumulated part or all of the target displacement and the actual displacement of the wafer comprises: A formula is established based on the accumulated target displacement and the actual displacement of the wafer to obtain the actual pixel size, and the formula is: in, dXs i =Xs i+1 -Xs i day i =Yes i+1 -Yes i dXm i =Xm i+1 -Xm i dYm i =No i+1 -No i Among them, n+1 frames of images are collected and n template matching is completed, n≥2, and the coordinate position of the mechanical motion platform corresponding to the i-th frame image is (Xs i ,Ys i ), the coordinate position of the mechanical motion platform corresponding to the i+1 frame image (Xs i+1 ,Ys i+1 ), 1≤i≤n, dXs i and dYs i are the actual displacements of the wafer corresponding to the i-th template matching in the x-direction and y-direction respectively, and the coordinate position of the template extracted in the i-th frame image is (Xm i ,Ym i ), the coordinate position of the matching position obtained in the i+1 frame image is (Xm i+1 ,Ym i+1 ),dXm i and dYm i are the target displacements corresponding to the i-th template matching in the x-direction and the y-direction respectively, and Px and Py are the actual pixel sizes in the x-direction and the y-direction respectively.

Images