A positioning method and a measurement method of a CD-SEM device

Abstract

The application discloses a positioning method and a measuring method of a CD-SEM device, which are used for positioning a measuring point on a wafer, and the positioning method comprises the following steps: when a work menu is created, the positions of all preset measuring points are determined by using a work image, and the positions of corresponding positioning points are determined, a first template image is collected at the positions of the positioning points with a first magnification, and a first template is selected, and a second template image is also collected with a second magnification; when the work menu is executed, each measuring point is traversed, a target image is collected at the position of the corresponding positioning point with the first magnification, the first template is matched with the target image, when the matching succeeds, the position of the measuring point is determined, when the matching fails, a down-sampling image with the second magnification is obtained by down-sampling the target image, the down-sampling image is matched with the second template image, and the position of the measuring point is determined. The positioning method and the measuring method have high reliability, can save time, and can improve the throughput of the device.

Description

Technical Field

[0001] This invention relates to the field of scanning electron microscopy, and more particularly to a positioning and measurement method for a CD-SEM device. Background Technology

[0002] In the fabrication of semiconductor integrated circuits (ICs), critical dimension (CD) measurement equipment based on scanning electron microscopy (SEM), namely CD-SEM equipment, plays a crucial role. Current CD-SEM equipment is largely similar, such as... Figure 1A As shown, its components include a core scanning electron microscope (SEM, an electron optical system), which contains an electron optical column 111 and a parallel optical microscopy system (OM) 112. The OM 112 is used for primary wafer alignment (WA) and can have multiple objectives with different magnifications. The device 100 also includes a mechanical motion platform 113 (located within the main cavity, which maintains a high vacuum) for placing wafers 114. The device 100 also includes a front-end module (EFEM) 115, which contains a robot 1151 and a pre-aligner 1152, and a wafer cassette 1153 can be placed outside it. Between the EFEM and the main cavity including the mechanical motion platform 113, there is a vacuum transition chamber 116. The main cavity also contains a robot 117 for wafer handling between the vacuum transition chamber 116 and the mechanical motion platform 113. After wafer 114 is mounted, it is placed on an electrostatic tray (E-chuck) on a mechanical motion platform 113. The electrostatic tray may also have a height fine-tuning mechanism. It also includes a multi-stage vacuum pump and a vacuum measuring device (not shown in the figure). Additionally, it has an industrial computer 118 that runs software 119, including graphical user interface (GUI) software, system software, database, algorithm software, and low-level hardware drivers and communication software. (Reference) Figure 1BThis is a schematic diagram of the structure of a scanning electron microscope (SEM) in one example of the described device. Currently, the structures of various SEMs in the prior art are largely similar. Their main components are integrated, responsible for the generation, acceleration, focusing, and scanning control of the electron beam, as well as the detection of secondary electrons and other electrons, including backscatted electrons (BSE), generated from the sample. After signal amplification, analog-to-digital conversion, and signal processing, an SEM image is finally generated, from which the algorithm software extracts the CD information of a given object. There are many other details, but these are all prior art and will not be elaborated upon further.

[0003] Preparation

[0004] Like most semiconductor devices, when using a CD-SEM device to measure the CD of a wafer, it is usually necessary to first create a working menu (Recipe) for the device, and then the device executes the working menu when it is in operation, including executing the contents of the working menu inline.

[0005] The preparation work includes wafer alignment (WA) performed after wafer mounting. CD-SEM equipment involves patterned wafers (WA). (Reference) Figure 2A Typically, when creating a WA recipe, start with the most basic level, which has the largest field of view (FOV), such as an OM WA using an OM image. Acquire a template image at position 201, near the wafer center. Select a template that meets the template matching (PM) requirements (brightness, contrast, uniqueness, etc.). Use this template to acquire target images on the wafer at positions an integer number of grain cycles away from position 201, such as positions 202, 203, 204, and 205, and perform template matching to determine the wafer orientation angle θ (positions 201, 206, and 207 can also be used to determine the orientation angle θ). Then, use the successful matching results that meet a predetermined threshold (e.g., reference...) Figure 2B If template 212 in template image 211 is successfully matched to position 2121 in a target image 2111, it is corrected (including translation and rotation of the mechanical motion platform to achieve correction, or obtaining / updating the transformation matrix relative to the coordinate system of the mechanical motion platform), and then the same method is used (e.g., referring to...). Figure 2CImages with higher magnification, such as SEM image 221, contain template 222. Template 222 is used to successfully match a position 2221 in a target image 2211, until each predetermined WA (Wait and Measure) is completed. The template images, templates, and target image acquisition positions (i.e., matching positions) for each level of WA are saved to the working menu. When the Recipe is executed, each level of WA matches the target image acquired at the aforementioned matching position based on its template, thus completing the WA. WA is a necessary prerequisite for the core function of the device, namely CD measurement, providing a fundamental guarantee for the subsequent positioning of measurement points by the CD-SEM device. After completing the WA, the reference point of the wafer coordinate system can be determined, thus defining the wafer coordinate system. It has a fixed transformation relationship relative to the mechanical motion platform coordinate system, which is existing technology and will not be elaborated further.

[0006] There are many commonly used template matching algorithms in existing technologies, including image similarity algorithms, such as the Normalized Cross Correlation (NCC) algorithm, which can be directly applied to grayscale images or their gradient images, as well as feature-based template matching methods, all of which can achieve sub-pixel accuracy. Template matching algorithms are also used in CD-SEM equipment for locating measurement points.

[0007] Besides WA, another crucial preparatory step is Quick Auto Focus (QAF), which is also existing technology and will not be elaborated upon here.

[0008] Measurement points

[0009] After completing the WA (Wait and Measure), the Measurement Point (MP) can be determined. When creating the Recipe, the die to be measured on the wafer is identified, such as... Figure 3 The grain 301 in the test includes at least one structure to be tested (such as a line, circle or arc). For each structure to be tested, a corresponding measurement point needs to be configured (i.e. set).

[0010] The measurement point MP is broadly defined, encompassing more information. First, it includes a location on the wafer, i.e., coordinates, representing the image acquisition location of the measurement image (i.e., the SEM image acquired by the CD-SEM device during CD measurement). Second, it also includes an image acquisition area of ​​the measurement image, i.e., the field of view (FOV), centered on this location (this image acquisition area is typically determined through trial and error in CD-SEM applications). Additionally, it includes at least one measurement object and its measurement configuration. The measurement object belongs to a structure under test on the wafer (such as a line, circle, or arc), and can be all of it (e.g., the entire circle) or a portion of it (e.g., a segment of the line under test). The measurement object typically occupies the main part of the measurement image, and different measurement objects have different measurement configurations (examples will be provided later).

[0011] In addition, there may be dedicated dies on the wafer for setting the structure under test, such as die 302 (or one of the dies 301), which can be referred to as setting dies.

[0012] The CD-SEM equipment measures either a local or all area of ​​the structure under test in order to calculate the dimensions of the structure, such as line width, line width roughness, and line edge roughness.

[0013] In practical applications, for example, for a certain object being measured, a location on the object, such as near the center, is taken as the location of the measurement point. Although in most cases, one measurement point corresponds to one object being measured, and the object being measured occupies the main position in the measurement image, it is not limited to this. For example, for three objects being measured, the center position of the three objects is taken as the location of the common measurement point for the three, without setting the location of three different measurement points. Therefore, there is only one measurement point here.

[0014] Taking the structure under test as a line as an example, the cross-section of a single line (i.e., a line structure) of 400 is as follows: Figure 4A The cross section 401 in the figure is shown (taking the X-direction measurement as an example), and its corresponding SEM image is as follows. Figure 4B As shown in image 402, which is a measurement image used for CD measurement, the bright band 403 indicates that the edge of line 400 contributes significantly to the electron emission. The one-dimensional sampled value in the X-direction of image 402 is called the profile (taking the X-direction as an example), as shown... Figure 4C Middle outline 405, in Figure 4BIn the image 402 shown, a measurement object 404 is divided into two parts (the measurement point corresponding to the measurement object 404 is located at the center between the two parts, which is also the center of the image 402). Alternatively, a rectangle 4041 can be used to represent the measurement object. The profile can be a single-row sample of the measurement object at y = y0 (y0 is a preset position in the Y direction), and the normalized grayscale values ​​of the pixels between positions (x1, y0) and (x2, y0) in the X direction (see reference). Figure 4C The CD measurement results for each profile are obtained by averaging the samples from multiple adjacent rows (from y0 to y0+ny-1, e.g., ny=10) using the ordinate of the line (the vertical axis of the line). These results are then integrated into the final CD value, and relevant statistical results are obtained, such as the mean and variance of each CD measurement. Furthermore, some lines are not rectangular but are isosceles trapezoidal in structure, requiring measurements of the top CD and bottom CD. Sometimes, there are many lines parallel in the X or Y direction, necessitating measurements of the period (left or right period) and the space between lines (e.g., the distance between the lines). Figure 4D Line spacing (406) etc. Other common structures to be tested include circles, ellipses, or partially circular arcs.

[0015] refer to Figure 4E One of the ring-shaped structures has a cross-section as shown in section 407, and the corresponding SEM image is shown below. Figure 4F As shown, the measurement objects can be set as, for example, a pair of measurement objects 408 (shown as rectangles) on opposite sides in the X direction, or a measurement object 409 in the form of a partial annulus (shown as a sector), or the entire annulus. Additionally, some measurement objects are represented by a single rectangle instead of the pair shown above, for example... Figure 4G As shown, SEM image 410 is illustrated. The measurement object used to calculate the line width of the line to be measured is represented by a single rectangle 411, and the measurement object used to calculate the line edge roughness of the line to be measured is represented by a single rectangle 412. These are common in the prior art and will not be listed one by one.

[0016] Measurement configurations typically include parameters (e.g., for line structures, multiple sampling point parameters are configured on the measurement object to measure the line edge roughness), measurement algorithms (e.g., line roughness calculation methods that calculate line edge roughness based on multiple sampling point parameters), and the measurement object. The measurement configurations vary depending on the specific measurement object. For example... Figure 4B The measurement of line width shown requires defining one or a pair of rectangles (the object to be measured is represented by a rectangle) in the drawing, specifying their position, size, spacing, etc., as well as algorithm configuration parameters, such as how many segments to divide the Y direction for averaging, how to average the X direction to obtain the profile, and then configuring the algorithm to use (direct measurement or derivative calculation), among many other settings. For example... Figure 4FThe linewidth measurement of the circle shown requires defining whether to measure the entire circle or a portion of it, the angular averaging parameters, and various configurations such as the starting point (inner diameter) and ending point (outer diameter) for searching the diameter. These are all existing technologies and will not be elaborated further.

[0017] Location point

[0018] The SEM image used to measure the measurement point MP is called the MP image, or the measurement image mentioned earlier. The MP image requires the structure under test (e.g., a line segment and sufficient surrounding background) to occupy as much of the space as possible within the SEM image. To facilitate the location of the measurement point MP, there is usually a corresponding addressing point (AP) nearby (avoiding its exact location). The SEM image acquired for the addressing point AP is called the AP image. The field of view (FOV) of the AP image is larger than that of the MP image, making it easier to search.

[0019] To ensure the device can locate the target under test (DUT) at a predetermined position within a given die on the wafer during operation, it is necessary to position each DUT after the WA (Wait and Measure) step (using the positioning point AP mentioned above). This is crucial; failure to do so or improperly will prevent the device from accessing the predetermined DUT MP position on the wafer during Recipe execution, or the measurement point MP may be entirely or partially outside the field of view, resulting in critical dimension measurement failure. (Reference) Figure 5 The SEM image 500 used to select the positioning point AP can be called the positioning working image or simply the working image. It includes the positioning point AP and measurement points MP, such as ring structures, line structures, and circular structures. A ring structure has a measurement point 501, whose measurement object is a ring. Measurement point 502 (the distance between the edges of the line structure within a pair of rectangular frames is the line width; the position of measurement point MP is the geometric center between the pair of rectangular frames) has a measurement object related to the line width of the line structure. Figure 5 (Illustrated by a pair of rectangles in the diagram), measurement point 503 is a measurement object related to the linear period, measurement point 504 is a measurement object related to a circular structure, and measurement point 505 is also a measurement object related to the diameter of a circular structure.

[0020] Figure 5 There is a positioning point AP 507 for positioning. Typically, there can be more than one positioning point AP; for example, there can also be a positioning point AP 508. Each positioning point AP can be shared by one or more nearby measurement points MP, and vice versa. These are all existing technologies and will not be elaborated further.

[0021] Current problem

[0022] The aforementioned prior art has several serious problems, including:

[0023] Problem 1: When executing the Recipe, the location of the positioning point AP is prone to failure. Specifically, the template matching (PM) fails when locating the positioning point AP (if there are multiple levels, the first level PM fails). Then, a very time-consuming spiral search must be adopted, that is, moving the mechanical motion platform to collect nine-square grid images of the wafer at the periphery of the predetermined positioning point AP in the Recipe to search for the positioning point AP. The mechanical motion platform moves a lot and the number of image acquisitions is also high, resulting in low reliability of the positioning method, high time consumption, and affecting the throughput of the CD-SEM equipment.

[0024] For example, in Figure 6 In this process, position O1 is the image acquisition position of the AP image. If the template used to locate the AP (called the AP template) fails to match the target image 601 acquired at position O1, the mechanical motion platform moves in a predetermined order to acquire target images at other positions on the wafer for template matching until successful, for example, from positions O2 to O9 (the order may be uncertain, for example, after position O1, images can be acquired at position O3 to obtain the target image 603). With so many mechanical motion platforms moving, this not only seriously increases the time consumption and thus affects the throughput of the equipment, but is also not very reliable. For example, sometimes there are similar patterns in the AP image, that is, the AP template is not unique in one AP image within the nine-grid range, resulting in matching failure, or the matching position happens to be between 2 or 3 frames of the nine-grid image, and the overlapping area between the images in the nine-grid image is insufficient, which can also easily lead to matching failure. The reasons for this situation are not singular and may include: a) improper field of view (FOV) selection when the user creates the recipe, such as an FOV that is too small (the distance from the positioning point AP to the measurement point MP relative to the maximum possible error of the device in the XY direction); b) insufficient allowance for overlapping areas between the nine-grid images, which is sometimes unavoidable, otherwise more images are needed, such as using 25-grid images to cover uncertain areas; c) poor AP template selection when the user creates the recipe (e.g., insufficient brightness, contrast, and X and Y direction features); coupled with other inherent factors of the device or wafer, including local polarization around the positioning point AP, variations in wafer surface film height, image noise, and insufficient wafer alignment accuracy, the image quality and template matching are ultimately affected.

[0025] Problem 2: Improper configuration between measurement point MP and positioning point AP (users have some arbitrariness in selecting the location and number of measurement points MP and positioning points AP) can also lead to low reliability and long time consumption in the positioning method.

[0026] Specifically, in the existing technology, the positioning points (APs) often become redundant. Because a large number of positioning points need to be located, the mechanical motion platform moves a lot and the number of image acquisitions is also high, resulting in low reliability and high time consumption of the positioning method, which affects the throughput of the CD-SEM device. Summary of the Invention

[0027] The purpose of this invention is to provide a positioning and measurement method for CD-SEM equipment to solve the above-mentioned problem 1.

[0028] To achieve this objective, the present invention adopts the following technical solution:

[0029] A positioning method for a CD-SEM device, used to locate measurement points MP on a wafer, includes:

[0030] When creating the working menu of the device, all preset measurement points MP on the wafer are acquired. Working images are acquired on the wafer to determine the positions of all measurement points MP and the corresponding positioning points AP. The working images are used to determine the position of each measurement point MP and the position of the positioning point AP required for each measurement point MP. A first template image is acquired at the position of the positioning point AP with a first magnification LM1 and a first template is selected therein. A second magnification LM2, which is less than the first magnification LM1, is set. A second template image is also acquired at the position of the positioning point AP with the second magnification LM2. The positions of the measurement points MP, the positions of the positioning points AP, the first template image, the first template and the second template image are saved to the working menu.

[0031] When executing the device's working menu, each measurement point MP is traversed, and a first target image is acquired at the position of the positioning point AP corresponding to each measurement point MP at a first magnification ratio LM1. Template matching is performed between the first template and the first target image. When the matching is successful, the position of the positioning point AP and the position of its corresponding measurement point MP are determined. When the matching fails, the first target image is downsampled to obtain a downsampled image, such that the magnification ratio of the downsampled image is the second magnification ratio LM2. The downsampled image is used as a template and template matching is performed with a second template image to determine the position of the positioning point AP and the position of its corresponding measurement point MP.

[0032] A measurement method for a CD-SEM device includes: locating each measurement point MP using the positioning method; acquiring measurement images for each measurement point MP based on a preset field of view; and measuring the key dimensions of a preset measurement object based on the measurement images of each measurement point MP and a preset measurement configuration to obtain the key dimensions of each measurement point MP. The measurement point MP includes the location, the field of view, the measurement object, and the measurement configuration of the measurement object.

[0033] The beneficial effects of this invention are:

[0034] This invention provides a positioning method and a measurement method for a CD-SEM device. In this positioning method, when matching for a positioning point AP fails, a downsampled image is obtained by downsampling a first target image, and template matching is performed based on the downsampled image and a second template image to achieve the positioning of the positioning point AP and the measurement point MP. This solves the aforementioned problem 1, improves the reliability of the positioning method, saves time, and increases the throughput of the device. Furthermore, since the measurement method includes the above-mentioned positioning method, all the advantages of this positioning method are obtained. Attached Figure Description

[0035] Figure 1A This is a schematic diagram of the structure of a CD-SEM device in the prior art;

[0036] Figure 1B This is a schematic diagram of the electron optical system on a CD-SEM device in the prior art;

[0037] Figure 2A This is a schematic diagram of wafer alignment on a CD-SEM device in the prior art;

[0038] Figure 2B This is a schematic diagram of template matching used in wafer alignment on CD-SEM equipment in the prior art;

[0039] Figure 2C This is a schematic diagram of another template matching used in wafer alignment on CD-SEM equipment in the prior art;

[0040] Figure 3 This is a schematic diagram of the die to be tested and the die settings on a CD-SEM device in the prior art;

[0041] Figure 4A This is a schematic diagram of the cross-section of a single test line structure in the prior art;

[0042] Figure 4B This is a schematic diagram of a CD-SEM image of a single line under test in the prior art, wherein the measurement object related to the line width is represented by a pair of rectangles;

[0043] Figure 4C This is a schematic diagram of the X-direction sampling results in a CD-SEM image of a single line structure under test in the prior art;

[0044] Figure 4D This is a schematic diagram illustrating the distance measurement between multiple test lines in existing technology.

[0045] Figure 4E This is a schematic diagram of the cross-section of the ring structure to be tested in the prior art;

[0046] Figure 4F This is a schematic diagram of a CD-SEM image of a ring structure under test in the prior art. The object under test with respect to the line width of the ring is represented by a pair of rectangular boxes, and the object under test with respect to the line width of an arc is represented by a sector box.

[0047] Figure 4G This is a schematic diagram of a CD-SEM image with a small field of view for a single line under test in the prior art. The measurement object related to the line width is represented by a rectangle, and the measurement object related to the line edge is represented by another rectangle.

[0048] Figure 5 This is a schematic diagram of a working image in the prior art, which illustrates multiple measurement points and multiple positioning points;

[0049] Figure 6 This is a schematic diagram of how a CD-SEM device in the prior art acquires SEM images in a spiral manner around a predetermined positioning point after template matching fails.

[0050] Figure 7 This is a schematic diagram of the positioning method in an embodiment of the present invention;

[0051] Figure 8A This is a first template image with a first magnification LM1 used for positioning in a CD-SEM device according to an embodiment of the present invention;

[0052] Figure 8B This is a second template image with a second magnification LM2 used for positioning in a CD-SEM device according to an embodiment of the present invention;

[0053] Figure 8C This is another example of a first template image with a first magnification LM1 used for positioning in a CD-SEM device according to an embodiment of the present invention, and the first template therein;

[0054] Figure 8D This is a first target image with a first magnification LM1 used for positioning in a CD-SEM device according to an embodiment of the present invention, wherein there are successfully matched locations;

[0055] Figure 8EThis is a first target image with a first magnification LM1 used for positioning in a CD-SEM device in an embodiment of the present invention, wherein there is no successfully matched location;

[0056] Figure 8F The first target image with a first magnification LM1 used for positioning in the use of a CD-SEM device in this embodiment of the invention, and the downsampled image obtained by downsampling the target image;

[0057] Figure 8G This is a second template image with a second magnification LM2 used for positioning in the use of a CD-SEM device in an embodiment of the present invention, wherein there are positions where a downsampled image is used as a template and successfully matched with the second template image;

[0058] Figure 8H This is a second template image with a second magnification LM2 used for positioning in a CD-SEM device in an embodiment of the present invention, wherein a large portion of the downsampled image is used as a template and the position is successfully matched with the second template image;

[0059] Figure 8I This is a schematic diagram illustrating template matching from a downsampled image to a second template image in an embodiment of the present invention;

[0060] Figure 8J This is a schematic diagram illustrating another method of template matching from a downsampled image to a second template image in an embodiment of the present invention;

[0061] Figure 9A This is a schematic diagram of the integration of positioning points in an example positioning point optimization method according to an embodiment of the present invention;

[0062] Figure 9B This is a schematic diagram of the integration of positioning points in another example of a positioning point optimization method in this embodiment of the invention. Detailed Implementation

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0064] It should be noted that semiconductor equipment operation includes creating the equipment's recipe and executing the recipe. The equipment's recipe includes recipes for preparatory work (such as wafer alignment, fast autofocus, etc.) and recipes for its primary function (such as CD measurement).

[0065] like Figure 7 As shown, this embodiment of the invention provides a positioning method for a CD-SEM device, used to locate measurement points MP on a wafer, including:

[0066] When creating the device's working menu, all preset measurement points MP on the wafer are acquired. Working images are acquired on the wafer to determine the positions of all measurement points MP and the corresponding positioning points AP. The working images are used to determine the positions of measurement points MP and the positions of positioning points AP required for each measurement point MP. A first template image is acquired at the position of the positioning point AP with a first magnification LM1 and a first template is selected therein. A second magnification LM2, which is less than the first magnification LM1, is set. A second template image is also acquired at the position of the positioning point AP with the second magnification LM2. The positions of measurement points MP, positioning points AP, the first template image, the first template, and the second template image are saved to the working menu.

[0067] When executing the device's working menu, each measurement point MP is traversed, and a first target image is acquired at the position of the positioning point AP corresponding to each measurement point MP at a first magnification of LM1. Template matching is performed between the first template and the first target image. When the matching is successful, the position of the positioning point AP and the position of its corresponding measurement point MP are determined. When the matching fails, the first target image is downsampled to obtain a downsampled image, such that the magnification of the downsampled image is the second magnification of LM2. The downsampled image is used as a template and template matching is performed with a second template image to determine the position of the positioning point AP and the position of its corresponding measurement point MP.

[0068] In this positioning method, when the matching of positioning point AP fails, a downsampled image is obtained by downsampling the first target image, and template matching is performed based on the downsampled image and the second template image to realize the positioning of positioning point AP and measurement point MP (i.e., determine the position). This can solve the above-mentioned problem 1, improve the reliability of the positioning method, save time, and increase the throughput of the equipment.

[0069] Those skilled in the art will know that when creating a device's working menu, the working menu is created based on a wafer, and when executing the menu, the working menu is executed for wafers of the same type as the original wafer. This is prior art and will not be described in detail here.

[0070] In this embodiment of the invention, obtaining all preset measurement points MP on the wafer refers to obtaining the positions of all preset measurement points MP on the wafer according to measurement requirements. Because the creation of the working menu involves the wafer mounting and wafer alignment WA process, there are still residual errors. In addition, the errors caused by the local environment (including local potential) on the wafer will be relatively amplified in the image (which is the SEM image acquired by the CD-SEM device) corresponding to the measurement point MP at a magnification much higher than that used by WA. This causes the position of the measurement point MP to deviate from the preset position. Therefore, it is necessary to acquire a working image. The purpose of acquiring the working image is to determine the position of all measurement points MP and the position of the corresponding positioning point AP.

[0071] In this embodiment of the invention, for the above-described positioning method, preferably, when creating the device's working menu, the working image is also used to determine the field of view, the measurement object, and the measurement configuration of the measurement object for each measurement point MP. However, this is not limited to this; the field of view, the measurement object, and the measurement configuration of the measurement object for each measurement point MP can also be determined after the positioning method is completed. The relevant content regarding the field of view, the measurement object, and the measurement position is prior art; please refer to the relevant content in the background art, and it will not be repeated here.

[0072] In this embodiment of the invention, when creating the Recipe, for a given measurement point MP on the wafer, the position of the corresponding positioning point AP is determined within a preset maximum distance threshold Rt (an empirical value based on the positional accuracy of the device components, which can be obtained experimentally for a given device). That is, in the working image, the corresponding positioning point AP is selected within the area whose distance from the measurement point MP is within the maximum distance threshold Rt, and the position of the positioning point AP is determined. The method for selecting the positioning point AP is prior art and will not be elaborated here.

[0073] When creating this recipe, it is necessary to first acquire a positioning working image (referred to as a working image). The corresponding SEM has a large field of view (FOV). The purpose of acquiring the working image is to include both the measurement point MP and the positioning point AP that is close to the measurement point MP in the working image, so as to select the positioning point AP required for the measurement point MP, and thus determine the position of the positioning point AP.

[0074] In this embodiment of the invention, the number of working images is one frame or more, depending on the needs, as long as the working images can cover the required measurement points MP. For example, if the number of all measurement points MP is small and they are close together, only one working image can be acquired to cover all measurement points MP, and the positions of all measurement points MP and their corresponding positioning points AP can be determined in this working image. For example, if the number of measurement points MP is large and their distribution is relatively scattered, and it is impossible to cover all measurement points MP with one working image, then working images are acquired at different measurement point MP positions to obtain multiple working images. For example, it is necessary to use the SEM in the device to acquire images of different positions on the wafer as the mechanical motion platform moves to obtain multiple working images, so as to cover all measurement points MP and facilitate the determination of the corresponding positioning points AP for all measurement points MP. In this embodiment of the invention, the wafer is typically moved to the position of a measurement point MP or near it at a preset magnification. Working images are then acquired at or near the measurement point MP. The acquisition of working images is also carried out while moving (the CD-SEM device operates in real-time acquisition mode, Live mode). The image acquisition position or magnification can be continuously adjusted to finally determine at least one working image. Each working image contains at least one measurement point MP and at least one corresponding positioning point AP.

[0075] In one embodiment, after determining the position of a measurement point MP in the working image, the position of its corresponding positioning point AP can be determined in the working image; in another embodiment, after determining the positions of all measurement points MP in the working image, the position of the corresponding positioning point AP can be determined for each measurement point MP based on the working image.

[0076] In addition, the location and range of the measurement image of the measurement point MP can be determined by acquiring SEM images at the measurement point MP location more than once, and the working image does not need to be saved to the working menu. These are all existing technologies.

[0077] In embodiments of the present invention, such as Figure 5 As shown in the working image 500, there are measurement points MP 501 and 502. Measurement point 501 is indicated by a rectangle in the image (the same applies to measurement point 502; however, since measurement point 502 has two measurement objects indicated by rectangles, its position can be defined as the geometric center between the two rectangles). In this case, the field of view of the measurement image is a rectangular area containing the two rectangles, and it can have a certain margin, such as... Figure 5As shown in the rectangular area 5021, the size of the field of view of the measurement image corresponding to all measurement points in this embodiment can be understood as follows. It is also the range (field of view) of the measurement image acquired when CD measurement is performed on the measurement point MP. There are also positioning points AP 507 and 508 that serve the positioning of measurement points 501 and 502 (for example, positioning point AP 507 is for positioning measurement point MP 501, and positioning point AP 508 is for positioning measurement point MP 502), and ranges 5071 and 5081 including these positioning points AP. Ranges 5071 and 5081 are respectively the image acquisition areas of the AP template image (i.e., the first template image) (the actual size of the acquired SEM image usually has a fixed width and height, so here we only say "the image acquisition area of ​​the template image" rather than the template image). The magnification of the AP template image (i.e., the first template image) is the first magnification LM1, which will be used later. It should be noted that this working image 500 is the image required when determining the measurement point MP and the relevant positioning point AP during Recipe creation. It is used to help determine the measurement point MP and the positioning point AP, and is no longer needed when executing the Recipe later, because the positioning point AP will then have a first template image with a first magnification of LM1 (in this embodiment, there is also a second template image with a second magnification of LM2, which has a larger field of view). In this embodiment, it is further explained that the distance between the measurement point MP and the positioning point AP is d. Usually, the shortest distance between the bounding rectangles of the measurement point MP and the positioning point AP can be used as the distance d. Roughly, the center distance between the two can also be used as the distance d.

[0078] To address problem 1 in the prior art, the first approach that comes to mind is to collect template images around the location point AP during recipe creation. These images are taken at the first magnification (LM1) and the selected AP template (referred to as the main template). This creates a 3×3 grid (unlike the existing technique where a 3×3 grid is collected when AP matching fails, thus having minimal impact on device throughput). Templates selected from these surrounding images are called peripheral templates. When the main template fails to match the target image, peripheral templates can be used for matching, saving more time than existing techniques. However, this approach has significant drawbacks: 1) the coverage area of ​​the peripheral template images may be insufficient to accurately represent the actual position of the AP template (i.e., the main template); 2) some template images may not meet the required conditions (brightness, contrast, feature content, etc., including edges and corners). Therefore, this embodiment uses a more reliable and time-saving positioning method.

[0079] In this embodiment of the invention, when creating a Recipe, the positions of all measurement points MP are traversed. To determine the position of the corresponding positioning point AP, firstly, the working image, magnified at an appropriate magnification, reaches the position of measurement point MP. The working image has a large field of view, which can cover the positions of measurement point MP and positioning point AP. This is the same as the steps in the prior art, as described above. Figure 5 As shown, the location of the positioning point AP required for each measurement point MP can be determined in the working image. Then, refer to Figure 8A An AP template image (i.e., the first template image, with a magnification of LM1) 801 is acquired from the acquisition area of ​​the template image, containing the AP template (i.e., the first template) 8011. The field of view of the working image is larger than the field of view corresponding to the AP template image, and even larger than the field of view corresponding to the MP image acquired at the measurement point during CD measurement, which is the same as or similar to that in the prior art. The AP template meets the requirements of the predetermined template matching algorithm and is unique within the field of view of the SEM (corresponding to the first magnification LM1).

[0080] In this embodiment of the invention, when creating the Recipe, a second template image is also acquired at the location of the positioning point AP (SEM in the device) with a larger field of view (FOV), that is, with a lower magnification (denoted by the second magnification LM2, LM2 < LM1). For example, refer to Figure 8A This illustrates an AP template image with a first magnification LM1, i.e., the first template image 801. The image may contain one or more AP templates (i.e., the first templates), such as template 8011. (Reference) Figure 8B The diagram illustrates a second template image 802 with a second magnification LM2 (since the second template image 802 can also be used to locate the positioning point AP, the second template image 802 is also an AP template image). Region 803 in the diagram is equivalent to the entire contents of the first template image 801 mentioned above. Region 803 and the second template image 802 have the same center position, while template region 8021 in the diagram corresponds to template 8011 in the first template image 801.

[0081] In this embodiment of the invention, the first magnification LM1 of the first template image at each measurement point MP does not need to be the same, but it should and can be easily made the same or close (e.g., the variation range does not exceed 10%), so as not to affect the positioning accuracy at the positioning point AP. Similarly, the second magnification LM2 of the second template image at each measurement point MP does not need to be exactly the same, but it should and can be easily made the same or close (e.g., the variation range does not exceed 10%). In this embodiment of the invention, the number of measurement points MP is multiple, and the second magnification LM2 corresponding to all measurement points MP is the same, or at least one of the second magnification LM2 corresponding to all measurement points MP is different from the rest of the second magnification LM2.

[0082] When executing the Recipe, if the first target image acquired from the first template to the location of the location point AP is successfully matched, the location of the location point AP is considered complete. The location of the corresponding measurement point MP can be determined without a doubt based on the location of the location point AP (this is because the location of the measurement point MP and the corresponding location point AP are saved when the working menu is created; therefore, when the working menu is executed, the relative position between the two is obtained based on the saved location of the measurement point MP and the corresponding location point AP, and the location of the corresponding measurement point MP is obtained based on this relative position and the location of the location point AP obtained through template matching). If the matching fails in the first target image acquired from the first template to the location point AP (with the same magnification LM1 and the same field of view, magnification, and image size as the first template image), the following method can be used.

[0083] In this embodiment of the invention, the method can be called a downsampling matching method. It involves downsampling the first target image of the current location point AP to obtain a downsampling image with a second magnification ratio LM2 (the downsampling image is smaller in size compared to the first target image), and then performing template matching using the entire downsampling image or most of the downsampling image with a second template image. In other words, in this embodiment of the invention, using the downsampling image as a template and performing template matching with the second template image includes: using a region with a preset area in the downsampling image as a template and performing template matching with the second template image, wherein the ratio of the preset area to the area of ​​the downsampling image is a number between 80% and 100%, for example, an area ratio of 80% or 100% (the entire area). In this embodiment of the invention, the size ratio of the first target image and the downsampling image is the same as the ratio of the first magnification ratio LM1 to the second magnification ratio LM2.

[0084] The template matching algorithm may include image grayscale-based algorithms such as normalized cross correlation (NCC) or feature-based algorithms. The specific shape of the downsampled image is not required, and as mentioned above, it is not necessary to use all the pixels in it. Parts with no or almost no features in the downsampled image (the features mainly refer to the edges and corners in the image) can be excluded. Then, the search is performed in the second template image of the location point AP based on the downsampled image.

[0085] In this embodiment of the invention, the current first target image is downsampled to obtain a downsampled image. Then, the entirety or most of the downsampled image is used to perform a matching search in the second template image, i.e., template matching. Since the size of the second template image is larger than the size of the first template image, and the size of the downsampled image is smaller than the size of the first target image (the size of the first target image is the same as the size of the first template image), the size of the second template image is larger than the size of the downsampled image. Furthermore, the downsampled image is obtained by scaling the first target image to a second magnification factor LM2. The downsampled image typically includes more features, making template matching from the downsampled image to the second template image easier to succeed.

[0086] In this context, the second template image is the image to be searched, serving as the target image at this point, while the entire downsampled image, or most of it, serves as the template. Such a large template will inevitably meet the requirements for template matching (including feature content, contrast, and uniqueness). The following section will use... Figure 8C , Figure 8D , Figure 8E , Figure 8F , Figure 8G and Figure 8H Let's take an example to explain in detail.

[0087] refer to Figure 8C This illustrates an example of the first template image 810 in an embodiment of the present invention, which includes the first template 811. As mentioned above, both are determined when the Recipe is created. (See reference...) Figure 8D This is an example of a successful match between the first template 811 and the first target image 812. The successful match between the template 811 and the first target image 812 yields a matching position (also called a matching region) 813. As described above, this completes the localization of the location point AP. (Reference) Figure 8EThis is an example of a failed match between the first template 811 and the first target image 814, where only a corresponding portion 815 of the first template 811 is present in the first target image 814. Therefore, following the method described in the above embodiment of the invention, the current first target image 814 is downsampled to the same magnification ratio as the second template image, i.e., the second magnification ratio LM2, to obtain a downsampled image 8141. Figure 8F The center point of the downsampled image is indicated by a cross shape. Figure 8G The center point of the matching position 816 is also indicated by a cross shape. Figure 8H Similarly, without further explanation, the entire downsampled image 8141 is used as a template, and then the second template image 819 (see reference) is applied. Figure 8G Template matching is performed on the second template image 819 (which is then used as the target image). As mentioned earlier, the size and content of the downsampled image 8141 ensure successful matching to obtain the matching position 816. The matching position of the original first template 811 at this time can be calculated, thus completing the positioning of the location point AP. Alternatively, as mentioned earlier, not all of the downsampled image 8141 can be used, but rather the vast majority (for example, the ratio of the area of ​​the downsampled image to the area of ​​the first target image is between 80% and 100%, but not 100%). Figure 8H Image 817 in the image is the majority of the downsampled image 8141, that is, the part remaining after discarding the parts with low surrounding feature content (for example, image 817 obtained after discarding the lower left edge of the downsampled image 8141). Image 817 is used as a template to match the second template image 819, and a successful matching position 818 is obtained. At this time, the similarity value used for template matching may be slightly higher than the value when the entire downsampled image 8141 is used as the template.

[0088] In this embodiment of the invention, when performing the downsampling matching, the similarity threshold corresponding to the template matching (a successful match is defined as a matching score greater than or equal to this similarity threshold) should also be relatively reduced, for example, to 95% of the similarity threshold corresponding to the original AP template (i.e., the first template) at the location point AP. Since the positions of the first template in the first template image and the second template image are fixed, when the downsampled image is successfully matched in the second template image, the relative position between the first template and the first target image can be indirectly determined, and thus the position of the location point AP can be determined. Since there is a correspondence between the location point AP and the measurement point MP, that is, the relative relationship between their positions is known, the position of the related measurement point MP can be determined from the location point AP.

[0089] In this embodiment of the invention, using the downsampled image as a template and performing template matching with a second template image to determine the position of the location point AP includes: using the downsampled image as a template to perform template matching in the second template image to obtain a matching region and its position; obtaining a correction amount for the position of the location point AP based on the position of the matching region and the center position of the second template image; and determining the position of the location point AP by adding the correction amount to the position of the location point AP stored in the working menu. For example, refer to... Figure 8G In the second template image 819, the first template image corresponds to the central region 8101 of the second template image 819 (the size of the central region 8101 is the same as the size of the first template image). The center position of the central region 8101 (which is also the center position of the second template image 819) is coordinate (xc, yc). The downsampled image 8141 successfully matches the matching region 816 in the second template image 819. The center position of the matching region 816 is coordinate (xp, yp). The difference between the two is (dx, dy) = (xp-xc, yp-yc). Therefore, the correction amount for the position of the positioning point AP on the wafer is (-fr×dx, -fr×dy). The position of the positioning point AP is determined by adding the correction amount to the position of the positioning point AP saved in the working menu (which is the image acquisition position of the first target image that failed to match). The determined position of the positioning point AP is applicable to all measurement points MP that are located with reference to the positioning point AP. Where fr is the ratio of the first magnification LM1 to the second magnification LM2 (that is, the ratio of the field size corresponding to the second magnification LM2 to the field size corresponding to the first magnification LM1), and it is assumed that the values ​​of fr in the X and Y directions are the same, which are fr = LM1 / LM2.

[0090] The accuracy of the matching result at this point will be slightly different from the accuracy of the matching result from the first template to the first target image corresponding to the original positioning point AP. For example, if the original accuracy was δ pixels (δ < 1.0), the current accuracy will be δ. ′ Pixels

[0091]

[0092] Wherein, FOV2 and FOV1 are the field of view (FOV) sizes of the second and first template images corresponding to the positioning point AP, respectively. Typically, if δ is 0.25 pixels (this accuracy can be achieved without complex algorithms and time consumption in existing technologies; 0.1 pixels is also common), and FOV2 is twice the size of FOV1, then δ... ′ That is, 0.5 pixels, and a difference of a few pixels is not enough to affect the measurement of MP at the measurement point.

[0093] Alternatively, when creating the Recipe, multiple alternative templates can be determined around the region 8021 corresponding to the first template in the second template image (e.g., template 8011 in the first template image 801) in the second template image. For example, m alternative templates, where 4 ≤ m ≤ 12. Figure 8B The image uses backup templates 804, 805, 806, 807, and 808. Then, one or more of these backup templates 804, 805, 806, 807, and 808 are upsampled, and the upsampled backup templates are used to search the current first target image (excluding region 8021, which is assumed to be used only if the matching of its corresponding template 8011 fails). There are many image upsampling methods, and these are existing image processing techniques. Although this method is not excluded in this invention and is an optional method, it is used in this embodiment because its work in creating a Recipe is relatively cumbersome, and its success rate and computation speed are theoretically and practically inferior to the downsampling matching method described above.

[0094] In this embodiment of the invention, the second magnification LM2 is located between a preset lower limit and a preset upper limit. The second magnification LM2 can be either the preset lower limit or the preset upper limit.

[0095] In this embodiment of the invention, the method for determining the specific range of the second magnification factor LM2 of the second template image is described below.

[0096] In this embodiment of the invention, the preset upper limit of the second magnification LM2 corresponds to the lower limit of the field of view (FOV2) corresponding to the second magnification LM2, which is p times the first magnification LM1 of the first template image. That is, the preset lower limit is p×LM1, where p is determined empirically and is usually a number between 65% and 85% (for example, it can be 65% or 85%), because a magnification that is too close to the first magnification LM1 will not help much in terms of successful matching.

[0097] In practical applications, the second magnification LM2 is selected between the preset upper limit and the preset lower limit.

[0098] Regarding the preset lower limit of the second magnification LM2 of the aforementioned second template image, which corresponds to the upper limit of the field of view (FOV) value corresponding to the second magnification LM2, this embodiment of the invention provides an experimentally based method for determining this preset lower limit, which includes: performing the following steps on the wafer before creating the Recipe to determine the preset lower limit of the second magnification LM2:

[0099] a. Obtain the maximum value LM1max among all preset first magnifications LM1 (corresponding to all the positioning points AP), and obtain q times the minimum value among all preset second magnifications LM2 as LM2min (at this time, it is an initial value, and in subsequent step e, the value of LM2min can be increased by a predetermined step size), where 0 <q<1;

[0100] As mentioned above, in the embodiments of the present invention, each first magnification LM1 is actually the same or close to the same. Therefore, the difference between LM1max and LM1 at each measurement point MP position is not significant. Thus, it can be considered that the ratio of LM2min to any first magnification LM1 is approximately the same as the ratio of LM2min to LM1max. Therefore, it is only necessary to consider where LM2min is.

[0101] b. Select N regions on the wafer, acquire images in the N regions respectively with the maximum value LM1max to obtain N images I, and acquire images in the N regions respectively with the maximum value LM2min to obtain N images G. Downsample each image I to obtain N images D, such that each image D has the same magnification as image G.

[0102] Each of the N regions meets the contrast and brightness requirements of the template matching requirements, and has sufficient features (e.g., edges) in the X and Y directions (or orthogonal methods), but avoids using the position of the positioning point AP and the position of the measurement point MP as much as possible.

[0103] c. For each of the regions, the overlapping region when the corresponding image D and image G are matched is obtained to obtain N overlapping regions. The overlapping region is the overlapping region of image D and image G when the matching region corresponding to image D in image D and image G are overlapped.

[0104] In this embodiment of the invention, image D is used as a template, and image G is used as the image to be searched, i.e., the target image. A template matching method is used to align images D and G to obtain their relative displacement. Furthermore, it is possible to obtain the overlapping region when the matching regions in images D and G that match image D are overlapped. For example, refer to... Figure 8I Images 821 and 822 are images D and G, respectively. Matching and aligning them (often requiring sub-pixel interpolation, where the matching regions in images D and G have already been successfully matched) yields the overlapping region 823; for example, in... Figure 8J In the image, images 821 and 822 are images D and G, respectively. Matching and aligning the two will also yield the overlapping area. In this case, the entire image 821 is the overlapping area.

[0105] d. Calculate the average grayscale error AE of the N overlapping regions based on the following formula:

[0106]

[0107] Where D[i,j] is the gray value of the pixel position (i,j) of image D in the overlapping region, G[i,j] is the gray value of the pixel position (i,j) of image G in the overlapping region, Wk and Hk are the width and height of the overlapping region after alignment of image D and image G corresponding to the kth region, respectively, 1≤k≤N, 6≤N≤10;

[0108] e. Compare the average value AE with the predetermined threshold Th2. If it is less than the predetermined threshold Th2, then LM2min is the preset upper limit. If it is greater than or equal to the predetermined threshold Th2, then increase the value of LM2min by a predetermined step size and repeat step ae until the average value AE is less than the predetermined threshold Th2. Then, use the current LM2min as the preset lower limit.

[0109] In this embodiment of the invention, the threshold Th2 is based on experience and is lower than the threshold Th originally used for template matching with the first template image, for example, 90% of it, but the two have no further correlation.

[0110] The device's computer can be equipped with a dedicated function including a graphical user interface (GUI) to support the determination of the preset lower limit for the second magnification LM2. Its functions include assisting the user in selecting the N locations and displaying the experimental results.

[0111] In this embodiment of the invention, as mentioned above, the preset lower limit of the second magnification LM2 can typically be determined before creating the Recipe, if it has never been done before; otherwise, the previously saved result of LM2min is used, which can be used at least once for the current Recipe. This also applies to similar wafers under test (with the same surface structure but different Recipes), considering the limited time cost and without affecting the throughput of the CD-SEM equipment—that is, when executing the Recipe.

[0112] It should also be noted that, in another embodiment of the present invention, the calculation of the average grayscale error AE of the N overlapping regions in step d above can be replaced by the Normalized CrossCorrelation (NCC) method for the N overlapping regions, which is also perfectly feasible. The template matching used when acquiring the overlapping regions employs the NCC method. The main difference between the two is that the NCC calculation normalizes the overall grayscale of the image, relying primarily on features such as edges and corners, while the calculation of the average AE also emphasizes the grayscale difference between image D and image G, which has its advantages. If the NCC method is used directly, there is also a corresponding threshold Th2' (between 0 and 1) used to determine whether the standard is met. Since the NCC method itself is existing technology, it will not be elaborated upon.

[0113] This concludes the description of the positioning method in this invention. In short, the above positioning method is obviously much faster and more reliable than the methods in the prior art.

[0114] In one embodiment of the present invention, when creating a working menu, the location point AP is determined such that no location point AP is shared by two or more measurement points. That is, when creating the working menu, the position of the location point AP corresponding to any measurement point MP is different from the positions of the location point APs corresponding to the other measurement points MP. This method is inflexible, as the measurement points MP cannot share their corresponding location point APs, resulting in relatively time-consuming execution of the Recipe and impacting device throughput.

[0115] In existing technologies, there is a problem of improper configuration between measurement points (MP) and positioning points (AP), which is referred to as Problem 2 above. This problem is quite common, and it often stems from users' arbitrary selection of the location and number of measurement points (MP) and positioning points (AP). For example, when the number of measurement points (MP) is large, the user often selects redundant positioning points (AP). To solve Problem 2, when creating the work menu, after determining all positioning points (AP), the configuration between measurement points (MP) and positioning points (AP) is optimized.

[0116] This can be done after each measurement point MP and each positioning point AP has been determined, but before the first template image and the second template image have been acquired (of course, it is not impossible to do it afterward, but doing so would be slightly wasteful, as the first and second template images corresponding to the positioning point AP that were discarded by the method below would be wasted), which is more economical.

[0117] In this embodiment of the invention, when creating the working menu, after determining the location of the positioning point AP required for each measurement point MP, the configuration of all measurement points MP and positioning points AP is optimized. The optimization process includes the following sub-steps:

[0118] S11. Select the positioning point AP corresponding to each measurement point MP on the wafer according to the preset maximum distance threshold Rt between the measurement point MP and the positioning point AP, and then initialize the measurement point MP and the positioning point AP so that the state of each measurement point MP is no positioning point NA, the integral S of each positioning point AP is 0 and the record of whether it has been used is no.

[0119] In this embodiment of the invention, when a user creates a Recipe, the graphical user interface (GUI) of the computer in the device determines all measurement points (MP) and all location points (AP) on the wafer according to the maximum distance threshold Rt (at this time, the configuration of all measurement points (MP) and location points (AP) is considered globally on the wafer). Then, the measurement point (MP) and location point (AP) records are initialized (in the device software, there is a corresponding recorder for each MP and each location point (AP)). This includes setting the state of each measurement point (MP) to No Addressing (NA), setting the integral (S) of each location point (AP) to 0, and recording whether it has been used (the initial state is No, and it becomes Yes once it is associated with at least one measurement point (MP).

[0120] S12. Iterate through each measurement point MP, and for each location point AP corresponding to it, increment the integral S by 1 to obtain the integral S of all the location points AP.

[0121] In an embodiment of the present invention, reference is made to Figure 9A (Note: Figure 9A (Note: 9B is merely a schematic diagram and not drawn to strict scale.) Figure 9ATaking image 900 on the left (which is the working image) as an example, image 900 contains measurement points MP 901, 902, 903, 904, 905 and positioning points AP 906, 907. According to the above rules, the integral S of positioning point AP 906 is 3, and the integral of positioning point AP 907 is 4, because 4 of the measurement points MP in the image are within the maximum distance threshold Rt of positioning point AP 907, hence the integral S is 4. However, the distance between positioning point AP 906 and measurement points MP 903, 904 exceeds the maximum distance threshold Rt, therefore they are not associated; AP 906 is only associated with measurement points MP 901, 902, 905, hence the integral S is 3. After this step is completed, each measurement point MP has a list of relationships with positioning points AP (within the maximum distance threshold Rt) associated with it. Note that, as mentioned above, access to the measurement point MP and positioning point AP is achieved through a mobile mechanical motion platform, which acquires working images. Rectangular frames at the locations of each measurement point MP and positioning point AP can be displayed in the working images. Furthermore, the positioning point AP can be marked on the computer's GUI to indicate whether it has integration or not; for example, the rectangle is one color when there is integration and another color when there is no integration. Figure 9A The text uses two colors, black and white, to distinguish between the two (but in practical applications, it's not limited to any particular color; it can be colored as long as it can differentiate between the two). Similarly, for the measurement point MP, it uses one color when the state is NA, and another color otherwise (when the state is HA). This is easily achieved in the GUI.

[0122] S13. Starting from the positioning point AP corresponding to the highest score among all the integrals S, clear the integrals S one by one in descending order of integral. Change the state of each measurement point MP that is within the maximum distance threshold Rt of the positioning point AP to have a positioning point HA. For each measurement point MP with a state of HA, subtract 1 from the integral S of the other positioning points AP that are within the maximum distance threshold Rt of the measurement point MP. Change the record of whether the positioning point AP with the highest score has been used from no to yes. Repeat sub-steps S11-S13 for the current positioning point AP with the highest score until the state of all the measurement points MP is HA. Then discard the positioning points AP with the record of no.

[0123] In this embodiment of the invention, starting with the location point AP with the highest integral, all integrals of each location point AP are cleared one by one (S=0). Simultaneously, the state of measurement points MP associated with location points AP (i.e., those within the maximum distance threshold Rt) that were previously NA are changed to HA (Has Addressing). Location points AP associated with measurement points MP that have changed to HA are also affected; for every measurement point MP that changes to HA, one point is deducted. (Continue to refer to...) Figure 9AIn the right figure, following the method described above, the integral of positioning point AP 907 after the initial zeroing is 0 ( Figure 9A (The score is not shown in the text). The integral of the relevant positioning point AP 906 is 1, leaving only the measurement point MP 901. Continue as follows (for the remaining positioning points AP) until all measurement points MP have a corresponding positioning point AP, i.e., all are in HA state. Positioning points AP that have never been used, i.e., have no corresponding measurement point MP (their state is still No), are considered redundant and can be discarded (simplified). Similarly, another example can be found in [reference missing]. Figure 9B Image 911 (i.e., the working image as described above) contains positioning point AP 912 with an integral of 2 and positioning point AP 913 with an integral of 5. First, the positioning point AP 913 with the higher integral is cleared. Since the measurement point MP involved in positioning point AP 912 is also involved in positioning point AP 913, after clearing positioning point AP 912, positioning point AP 913 will no longer correspond to any measurement point MP in state NA due to the aforementioned influence. Therefore, positioning point AP 912 has never been used and its state is still No, which is so-called redundant, and it is discarded.

[0124] Additional notes: Optionally, in some applications where each measurement point MP requires two positioning points AP, the remaining positioning points AP from the first round can be used. However, it's crucial to ensure that each measurement point MP requiring two positioning points AP has at least one additional positioning point AP. If not, supplement it first, then perform another round using the method described above to obtain the positioning points AP for the second round. Any excess APs should be discarded. The same method can support the association of positioning points AP for a third or more rounds, although this is usually unnecessary in practical applications. It should be noted that many similar variations based on the above approach exist, which will not be elaborated upon further.

[0125] As mentioned above, steps S11 to S13 are performed when the Recipe is created, so they do not add any extra burden (since these measurement points MP and positioning points AP need to be determined anyway), and in particular, they do not require additional movement of the mechanical motion platform and image acquisition.

[0126] Furthermore, in this embodiment of the invention, multi-level positioning point (AP) positioning is generally not required, although there is no limitation, because in practice, if the initial wafer positioning (WA) is done well, multi-level AP positioning is unnecessary. WA only needs to be performed once for a single wafer, which is much more efficient than performing multi-level positioning in multiple locations.

[0127] Therefore, the positioning method in the above embodiments of the present invention, compared with the prior art, not only greatly reduces the time consumption for positioning the positioning point AP and the measurement point MP, and improves the positioning speed, thereby greatly increasing the throughput of the CD-SEM device, but also greatly improves the reliability of the positioning method, thus solving problem 1. Furthermore, the embodiments of the present invention also solve the problem of improper configuration between the positioning point AP and the measurement point MP, i.e., problem 2.

[0128] The following describes the measurement method of the CD-SEM device in the embodiments of the present invention, including: using the positioning method to locate each measurement point MP; for any measurement point MP, acquiring a measurement image based on a preset field of view; and performing key dimension measurement on a preset measurement object based on the measurement image of each measurement point MP and a preset measurement configuration to obtain the key dimensions of each measurement point MP, wherein the measurement point MP includes the position, the field of view, the measurement object, and the measurement configuration of the measurement object.

[0129] In some embodiments of the present invention, in the above measurement method, after the positioning of each measurement point MP is completed, the acquisition and measurement of the measurement image and key dimensions are performed on the measurement point MP, and finally the key dimensions of all measurement points MP are obtained.

[0130] In this embodiment of the invention, as described above, whether during the implementation of the positioning method or after the positioning method is completed, the field of view, the measurement object, and the measurement configuration of each measurement point MP can be determined, so that these contents are preset contents relative to the measurement method, so as to be used when implementing the measurement method.

[0131] In this embodiment of the invention, the measurement preparation work for the CD-SEM equipment is the same as that in the prior art, including wafer alignment after loading and rapid autofocus. The measurement methods themselves are also the same as in the prior art. (See reference...) Figures 4A to 4F For example, it can measure line width and line width roughness, line spacing, line edge roughness, the inner and outer diameters of circles, arcs, ellipses or annulus, the angle of corners, or the distance between any pair of structures. Unlike existing technologies, the measurement methods of the CD-SEM device all use the positioning methods in the above embodiments of the present invention, thus obtaining all its advantages. Compared with existing technologies, the measurement method greatly improves reliability, saves time, and thus greatly improves the throughput of the device.

[0132] 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 those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

Claims

1. A positioning method for a CD-SEM device, used to locate measurement points MP on a wafer, characterized in that, include: When creating the working menu of the device, all preset measurement points MP on the wafer are acquired. Working images are acquired on the wafer to determine the positions of all measurement points MP and the corresponding positioning points AP. The working images are used to determine the position of each measurement point MP and the position of the positioning point AP required for each measurement point MP. A first template image is acquired at the position of the positioning point AP with a first magnification LM1 and a first template is selected therein. A second magnification LM2, which is less than the first magnification LM1, is set. A second template image is also acquired at the position of the positioning point AP with the second magnification LM2. The positions of the measurement points MP, the positions of the positioning points AP, the first template image, the first template and the second template image are saved to the working menu. When executing the device's working menu, each measurement point MP is traversed, and a first target image is acquired at the position of the positioning point AP corresponding to each measurement point MP at a first magnification ratio LM1. Template matching is performed between the first template and the first target image. When the matching is successful, the position of the positioning point AP and the position of its corresponding measurement point MP are determined. When the matching fails, the first target image is downsampled to obtain a downsampled image, such that the magnification ratio of the downsampled image is the second magnification ratio LM2. The downsampled image is used as a template and template matching is performed with a second template image to determine the position of the positioning point AP and the position of its corresponding measurement point MP.

2. The method according to claim 1, characterized in that, Using the downsampled image as a template and performing template matching with a second template image includes: using a region with a preset area in the downsampled image as a template and performing template matching with the second template image, wherein the ratio of the preset area to the area of ​​the downsampled image is a number between 80% and 100%.

3. The method according to claim 1, characterized in that, The number of measurement points MP is multiple, and the second magnification LM2 corresponding to all measurement points MP is the same, or at least one of the second magnification LM2 corresponding to all measurement points MP is different from the other second magnification LM2.

4. The method according to claim 1, characterized in that, The second magnification LM2 is located between a preset lower limit and a preset upper limit.

5. The method according to claim 4, characterized in that, The preset upper limit is p×LM1, where p is a number between 65% and 85%.

6. The method according to claim 4, characterized in that, Before creating the working menu for the device, the following steps are performed on the wafer to determine a preset lower limit for the second magnification LM2: a. Obtain the maximum value LM1max among all preset first magnifications LM1, and obtain q times the minimum value among all preset second magnifications LM2 as LM2min, where 0 <q<1; b. Select N regions on the wafer, acquire images in the N regions respectively with the maximum value LM1max to obtain N images I, and acquire images in the N regions respectively with the maximum value LM2min to obtain N images G. Downsample each image I to obtain N images D, such that each image D has the same magnification as image G. c. For each of the regions, the overlapping region when the corresponding image D and image G are matched is obtained to obtain N overlapping regions. The overlapping region is the overlapping region of image D and image G when the matching region corresponding to image D in image D and image G are overlapped. d. Calculate the average grayscale error AE of the N overlapping regions based on the following formula: Where D[i,j] is the gray value of the pixel position (i,j) of the image D in the overlapping region, G[i,j] is the gray value of the pixel position (i,j) of the image G in the overlapping region, Wk and Hk are the width and height of the overlapping region after the image D and image G are aligned, respectively, 1≤k≤N, 6≤N≤10; e. Compare the average value AE with a predetermined threshold Th2. If AE is less than the predetermined threshold Th2, then LM2min is the preset lower limit. If LM2min is greater than or equal to the predetermined threshold Th2, then the value of LM2min is increased by a predetermined step size, and step ae is repeated until the average value AE is reached. If it is less than the predetermined threshold Th2, then the current LM2min is used as the preset lower limit.

7. The method according to claim 1, characterized in that, Using the downsampled image as a template and performing template matching with a second template image to determine the position of the location point AP includes: using the downsampled image as a template to perform template matching in the second template image to obtain a matching region and its position; obtaining a correction amount for the position of the location point AP based on the position of the matching region and the center position of the second template image; and determining the position of the location point AP by adding the correction amount to the position of the location point AP saved in the working menu.

8. The method according to claim 1, characterized in that, When creating the working menu, the position of the positioning point AP corresponding to any of the measurement points MP is different from the position of the positioning point AP corresponding to the other measurement points MP.

9. The method according to claim 1, characterized in that, When creating the working menu, after determining the location of the positioning point AP required for each measurement point MP, the configuration of all measurement points MP and positioning points AP is optimized. The optimization process includes the following sub-steps: S11. Select the positioning point AP corresponding to each measurement point MP on the wafer according to the preset maximum distance threshold Rt between the measurement point MP and the positioning point AP, and then initialize the measurement point MP and the positioning point AP so that the state of each measurement point MP is no positioning point NA, the integral S of each positioning point AP is 0 and the record of whether it has been used is no. S12. Traverse each measurement point MP, and for each location point AP corresponding to it, increment the integral S by 1 to obtain the integral S of all location points AP; S13. Starting from the positioning point AP corresponding to the highest score among all the integrals S, clear the integrals S one by one in descending order of integral. Change the state of each measurement point MP that is within the maximum distance threshold Rt of the positioning point AP to have a positioning point HA. For each measurement point MP with a state of HA, subtract 1 from the integral S of the other positioning points AP that are within the maximum distance threshold Rt of the measurement point MP. Change the record of whether the positioning point AP with the highest score has been used from no to yes. Repeat sub-steps S11-S13 for the current positioning point AP with the highest score until the state of all the measurement points MP is HA. Then discard the positioning points AP with the record of no.

10. A measurement method for a CD-SEM device, characterized in that, include: The positioning method according to any one of claims 1-9 is used to locate each of the measurement points MP. For any measurement point MP, a measurement image is acquired based on a preset field of view, and a preset measurement object is measured based on the measurement image of each measurement point MP and a preset measurement configuration to obtain the key dimensions of each measurement point MP. The measurement point MP includes the position, the field of view, the measurement object, and the measurement configuration of the measurement object.

Images