Photoetching field domain image automatic focusing method
By using the average edge gradient rate as the focusing index and the least squares method to obtain the optimal focus position in the lithography field image alignment device, the shortcomings in clarity and focus of the lithographic field image alignment device are solved, and higher accuracy and stability are achieved.
Patent Information
- Application Number
- CN202411731597.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-28
- Publication Date
- 2025-03-21
AI Technical Summary
The field image alignment devices in existing lithography machines have shortcomings in the clarity and focus of the alignment marks, which leads to the accuracy and stability of the alignment results being unable to meet the high-precision requirements of the lithography machine.
Improve the accuracy and stability of autofocus in the photolithographic field image by using the average edge gradient rate as a quantization index of the image focus value and fitting the curve using the least squares method.
The accuracy and stability of the photolithographic airport domain image autofocus is improved, ensuring that the alignment marks are clearly presented in the image, thereby improving the accuracy of the output results of the field image alignment device.
Smart Images
Figure CN119668044A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of photolithography machines, and more specifically, to an automatic focusing method for a photolithography machine field image. Background Art
[0002] The photolithography machine is an important equipment in the semiconductor manufacturing process. Its function is to accurately transfer the circuit pattern in the mask plate (also known as the mask plate) to a silicon wafer or other substrate according to the predetermined size and position through the exposure process.
[0003] The field image alignment (FIA) device is an important tool for solving the mask alignment and silicon wafer alignment problems in the lithography machine. Its core principle is to use a camera to capture a specific alignment mark pattern engraved on the silicon stage or wafer, obtain the corresponding field image, and then perform image signal processing on the field image to locate the edge of the alignment mark in the image. The symmetry of the alignment mark is used to accurately calculate the position of the center of the alignment mark in the image, and then the image position is converted into a real physical position according to the existing coordinate transformation relationship, thereby completing the alignment work.
[0004] Therefore, in order to ensure the accuracy of the alignment results, the field image alignment device needs to be maintained and adjusted before it is officially used for alignment work to ensure that the field image containing the alignment mark captured by its camera is focused well enough to clearly and completely capture the alignment mark, thereby providing high-quality and accurate input data for the subsequent image signal processing algorithm and ensuring the accuracy of the output results of the entire alignment device. Considering that the position of the entire device is basically fixed after the tooling is completed, the parameter that can be adjusted during focusing is the height position of the silicon wafer stage. By changing the object distance between the camera lens and the alignment mark pattern engraved on the silicon wafer stage or the silicon wafer, the field image can finally achieve the best focusing effect, so that the alignment mark can be clearly presented in the field image.
[0005] In summary, in order to ensure the accuracy of the positioning results output by the field image alignment device, it is necessary to ensure that the alignment mark is imaged clearly enough in the image. Therefore, the device needs to design a set of accurate image autofocus algorithms to accurately calculate the optimal focus height position of the silicon wafer stage under the premise that the camera tooling position is fixed.
[0006] Therefore, it is desirable to provide an improved method for automatically focusing images in the lithography field. Summary of the invention
[0007] An embodiment of the present application provides a method for automatically focusing an image in a lithography field, which can improve the accuracy and stability of automatic focusing of an image in a lithography field by using the average edge gradient rate as a quantitative indicator of the image focus value and using the least squares method to fit the curve to obtain the optimal focus position.
[0008] According to one aspect of the present application, a method for automatically focusing a field image of a lithography machine is provided, comprising: acquiring an initial field image, wherein, for the initial field image, the center of an alignment mark of a wafer stage is near the center of a field of view of a camera, and the object distance between the wafer stage and the lens of the camera is greater than a predetermined threshold; calculating an average edge gradient rate of the initial field image as an image focus value for automatically focusing the field image; and moving the wafer stage to obtain a plurality of sampled field images, and fitting a focus position and a focus value function curve using a least squares method based on the image focus values of the plurality of sampled field images to obtain an optimal focus position corresponding to a vertex.
[0009] In the above-mentioned lithography field image autofocus method, obtaining the initial field image includes: calculating the coordinate difference between the mechanical coordinate system coordinate of the center of the alignment mark of the wafer stage and the mechanical coordinate system coordinate corresponding to the field of view center of the camera and driving the height control motor of the wafer stage in the z direction to move the wafer stage to obtain the initial field image.
[0010] In the above-mentioned lithography field image auto-focusing method, the average edge gradient rate is expressed as:
[0011]
[0012] Where n is the number of edges detected by the initial field image through the average projection method according to the amplitude screening, is the grayscale of the image before the grayscale changes in the i-th edge, is the grayscale of the image after the grayscale changes in the i-th edge, is the X coordinate corresponding to the moment when the grayscale of the image in the i-th edge changes, and It is the X coordinate corresponding to the grayscale change of the image in the i-th edge.
[0013] In the above-mentioned lithography field image auto-focusing method, the average edge gradient rate is an average value of multiple single average edge gradient rates obtained by measuring and calculating the average edge gradient rate of the initial field image for multiple times.
[0014] In the above-mentioned lithography field image automatic focusing method, moving the wafer stage to obtain a plurality of sampling field images, and fitting the focus position and focus value function curve using the least square method based on the image focus values of the plurality of sampling field images to obtain the best focus position corresponding to the vertex includes:
[0015] Determine a plurality of focus position points x corresponding to the plurality of sampling field images i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as:
[0016]
[0017] Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and
[0018] Determine the best focus position corresponding to the vertex where x=μ.
[0019] In the above-mentioned lithography field image auto-focusing method, fitting the focus position and focus value function curve using the least square method based on the image focus value to obtain the optimal focus position corresponding to the vertex includes:
[0020] Using a hill climbing search method starting from the focus position corresponding to the initial field image, and determining the focus position when the image focus value first decreases during the search process as the right boundary of the fine sampling area, and the focus position where the hill climbing search method is retracted by a predetermined number of steps as the left boundary of the fine sampling area;
[0021] Determine multiple focus positions x within the fine sampling area i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as:
[0022]
[0023] Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and
[0024] Determine the best focus position corresponding to the vertex where x=μ.
[0025] In the above-mentioned lithography field area image autofocus method, the hill climbing search method corresponds to a first search step, the sampling in the fine sampling area corresponds to a second search step, and the first search step is larger than the second search step.
[0026] In the above-mentioned lithography field image automatic focusing method, a plurality of focus position points x in the fine sampling area are determined. i The image focus value y under i The method comprises: dividing the coordinate difference of the focus position corresponding to the right boundary of the fine sampling area and the left boundary of the fine sampling area by the total number of steps of the least square fitting method to obtain a sampling step length; and starting from the focus position corresponding to the right boundary of the fine sampling area, backtracking by the sampling step length to obtain the multiple focus position points x i The image focus value y under i .
[0027] In the above-mentioned lithography field area image automatic focusing method, the above-mentioned steps are repeatedly performed to obtain multiple optimal focus positions, and the average value of the multiple optimal focus positions is output as the final focus position.
[0028] The automatic focusing method for lithography field area images provided in the embodiment of the present application can improve the accuracy and stability of automatic focusing of lithography field area images by taking the average edge gradient rate as a quantitative indicator of the image focus value and using the least squares method to fit the curve to obtain the optimal focus position. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] By reading the detailed description of the preferred specific embodiments below, various other advantages and benefits of the present application will become clear to those of ordinary skill in the art. The drawings in the specification are only used for the purpose of illustrating the preferred embodiments and are not considered to be limitations of the present application. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work. Moreover, the same reference numerals are used to represent the same components throughout the drawings.
[0030] Figure 1 The figure shows a flowchart of the hill climbing search method.
[0031] Figure 2 A schematic diagram showing how the hill climbing search method falls into a local optimum.
[0032] Figure 3 The figure illustrates a schematic flow chart of a method for automatically focusing a lithography field area image according to an embodiment of the present application.
[0033] Figure 4 A schematic diagram of an edge positioning method of an alignment mark in a field image alignment device of a lithography machine is shown.
[0034] Figure 5 A schematic diagram of a cascade search process in a lithography field area image auto-focusing method according to an embodiment of the present application is illustrated.
[0035] Figure 6 A schematic diagram illustrating an application example of the method for automatically focusing an image in a lithography field according to an embodiment of the present application is shown. DETAILED DESCRIPTION
[0036] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0037] Application Overview
[0038] Generally speaking, for camera autofocus algorithms, there are two parts that need to be focused on and solved: 1) how to evaluate and quantify the focus level of the current image; 2) how to search and locate the best focus position.
[0039] In the current mainstream camera autofocus method, Contrast Detection Autofocus (CDAF), the contrast of the image is set as a quantitative indicator of the image focus degree. The contrast is the difference between the maximum and minimum values of the image grayscale. When the focus is poor and the image is blurred, the contrast is low. When the focus is good and the image is clear, the contrast between light and dark is more obvious, and the contrast will increase. Therefore, the CDAF algorithm selects contrast as the image focus quantitative indicator.
[0040] As for the search and positioning of the best focus position, the CDAF algorithm uses the hill climbing search method. As the name implies, the algorithm regards the shape of the functional relationship between the focus position and the focus value in the coordinate system as a mountain, and the process of finding the best focus position is regarded as mountain climbing. The core idea of the algorithm is to gradually climb from the bottom of the mountain to the top of the mountain to determine the best focus position. Specifically, the CDAF algorithm will first select a position with a lower focus value (large or small object distance) as the starting point, and then move forward in a certain step length to the middle direction, and continue to detect the focus value. If the focus value is found to be larger, it means that it is still in the uphill stage and continues to move forward. If the focus value is found to be smaller, it means that it has entered the downhill stage. At this time, first determine whether the decrease in the focus value is within the acceptable error range. If so, take a step back and assume that the vertex position has been reached at this time; if not, reverse the forward direction and reduce the forward step length. Continue to execute this process until the focus value decreases and the amplitude is within the acceptable error range. Finally, it is ensured that it can be positioned near the vertex position. The overall process is as follows Figure 1 Here, Figure 1 The figure shows a schematic diagram of the process of hill climbing search.
[0041] However, for the image autofocus task of the field image alignment device of the lithography machine, the CDAF algorithm cannot meet the accuracy requirements of the lithography machine business process, and its stability in some complex scenarios is not good enough. More importantly, this autofocus algorithm is not specifically adapted to the functional characteristics and actual business needs of the field image alignment device. Therefore, it is necessary to further analyze the business process and principles of the field image alignment device of the lithography machine, and design an image autofocus algorithm that can truly meet the actual needs of the field image alignment device, so as to better serve the overall business process of the lithography machine.
[0042] First, in mainstream autofocus algorithms such as CDAF, image contrast is used as a quantitative indicator of image focus. However, for field image alignment of lithography machines, using contrast as a quantitative indicator of focus has the following two problems:
[0043] 1) In the CDAF algorithm, image contrast is defined as the contrast between the maximum and minimum values of the image grayscale. However, in the internal environment of the lithography machine, an increase in the contrast of the field image does not necessarily mean that the field image is clearer. It may also be that the lighting environment has changed or fluctuated, causing the local area to become brighter or darker, which will also increase the contrast, but the image content is not presented more clearly at this time;
[0044] 2) Analyze the demand for automatic image focusing of the field image alignment device. As mentioned above, the core essence of this demand is to hope that the alignment mark pattern can be presented more clearly in the field image. Specifically, it is hoped that the edge of the alignment mark can be imaged more clearly, so as to ensure that the field image alignment device can more accurately locate the edge position of the alignment mark, and then more accurately locate the entire alignment mark based on the edge position to complete the alignment work. Therefore, when designing the focus index, more attention should be paid to whether the focus index can intuitively reflect the ease of positioning of the edge of the alignment mark. However, the contrast index is more inclined to reflect the global clarity of the entire image, and its measurement object is not necessarily the edge of the alignment mark. Specifically, the increase in contrast may not be caused by the clearer edge of the alignment mark, but may also be caused by changes in the brightness or clarity of other areas of the image. If it is the latter, then even if the silicon wafer stage is finally adjusted to the maximum contrast position according to the algorithm, the edge of the alignment mark in the image may still not be clear enough. In this case, the "best focus position" cannot meet the actual business needs. Therefore, the contrast index does not directly meet the actual business needs of the field image alignment device, and this uncertainty may interfere with the measurement accuracy of the field image alignment device.
[0045] In addition, as mentioned above, for the search and positioning of the best focus position, mainstream autofocus algorithms such as CDAF use the hill climbing search method for positioning. However, for the image autofocus of the field image alignment device of the lithography machine, the hill climbing search method has the following two problems:
[0046] 1) Although the hill climbing search method can ensure that the optimal focus position of the final positioning is near the actual optimal focus position, it cannot give the theoretical optimal focus position, because the termination condition of the algorithm is that the focus value decreases and the decrease is less than the set error acceptable threshold. If the error threshold is set too large, it will not meet the overall accuracy requirements of the lithography machine. If it is set too small, it may cause the overall search and positioning to take too long, reducing business efficiency.
[0047] 2) The core of the hill climbing search method is to compare the difference between the current focus value and the previous focus value to determine whether it is currently in the climbing stage or the descending stage, and then continuously narrow the local range. However, the internal environment of the lithography machine is not necessarily completely stable. During the search and positioning process, the calculation of the focus value may fluctuate or error, which will cause the actual focus value and the focus position curve to not be a strict single-peak function, and there may be a local optimal point. In this case, the hill climbing search method, which only considers the local difference calculation method, may fall into the local optimal point and cannot give a theoretical optimal position. The schematic process is as follows Figure 2 shown. Figure 2 A schematic diagram showing how the hill climbing search method falls into a local optimum.
[0048] Schematic method
[0049] Figure 3 The figure illustrates a schematic flow chart of a method for automatically focusing a lithography field area image according to an embodiment of the present application.
[0050] like Figure 3 As shown, the automatic focusing method of the lithography field image according to the embodiment of the present application includes the following steps.
[0051] Step S110, acquiring an initial field image, wherein for the initial field image, the center of the alignment mark of the wafer stage is near the center of the field of view of the camera, and the object distance between the wafer stage and the lens of the camera is greater than a predetermined threshold.
[0052] That is, in the autofocus scenario, it is first necessary to determine the starting point of the autofocus, that is, it is necessary to control the wafer stage to move the center of the alignment mark to near the center of the camera's field of view to ensure the accuracy of subsequent measurements and ensure that the object distance between the wafer stage and the camera lens is large enough, that is, the focus value of the camera's image is at a relatively small level, which serves as the search starting point of the autofocus process.
[0053] For example, the mechanical coordinate system coordinates of the center of the alignment mark of the wafer stage can be expressed as (C x ,C y ), and the mechanical coordinate system coordinates corresponding to the camera's field of view center, for example, recorded as (V x ,V y ), to calculate the coordinate difference between the two, so as to move the center of the wafer stage alignment mark to the vicinity of the camera's field of view. In addition, by driving the wafer stage z-direction height control motor, the wafer stage is moved to the bottom to ensure that the object distance between the wafer stage and the camera lens is large enough.
[0054] Therefore, in the lithography field image autofocus method according to an embodiment of the present application, obtaining the initial field image includes: calculating the coordinate difference between the mechanical coordinate system coordinate of the center of the alignment mark of the silicon wafer stage and the mechanical coordinate system coordinate corresponding to the field of view center of the camera and driving the height control motor of the silicon wafer stage in the z direction to move the silicon wafer stage to obtain the initial field image.
[0055] Step S120, calculating the average edge gradient rate of the initial field image as the image focus value for field image autofocus. That is, in view of the problem of using contrast as a focus quantification index as described above, in the embodiment of the present application, the image average edge gradient rate is used as the image focus value.
[0056] Specifically, as mentioned above, in the task of automatic focusing of the field image alignment device, the fundamental purpose of focusing is to ensure that the subsequent algorithm can more accurately locate the edge position of the alignment mark in the image. Therefore, when designing the focusing index, it should intuitively reflect the ease of locating the edge part of the alignment mark. Then, it is necessary to first clarify how the subsequent algorithm locates the edge position of the alignment mark in the image, such as Figure 4 shown. Figure 4 A schematic diagram of an edge positioning method of an alignment mark in a field image alignment device of a lithography machine is shown.
[0057] like Figure 4 As shown, the alignment mark used by the field image alignment device is generally composed of a plurality of black strip-shaped patterns, such as Figure 4 As shown in (a), for example, it is an alignment mark pattern with two black strips, and since the background of the field image is generally white, the edge positioning method of the alignment mark calculates the edge position of the mark by averaging the projection. For example, if you want to obtain Figure 4In (a), the four edge X coordinate positions of the alignment marks are first averaged in the X direction. Since the color of the alignment marks is very different from the background color, the grayscale average projection will drop sharply near the X coordinate position of the edge, as shown in Figure 1. Figure 4 As shown in (b). By detecting the abrupt change of the average grayscale, the approximate position of each edge can be obtained. The reason why the accurate position cannot be obtained is that the grayscale abrupt change is not an ideal step-down in the actual image. Because of blurring caused by unclear focus or some other external factors, this abrupt change is often a large gradient drop. In other words, the X position where the abrupt change begins and ends is an interval. Therefore, in order to define the exact X position of the edge part, the grayscale value before the abrupt change can be regarded as 0%, and the grayscale value after the abrupt change can be regarded as 100%. Then, the X position corresponding to 50% grayscale is taken as the exact X position of this edge. By drawing the 50% grayscale line, we can get Figure 4 The exact X-direction coordinate positions of the four edges of the alignment mark in (a) are as follows: Figure 4 As shown in (c) and (d).
[0058] It can be seen that when the focus is good, the edge gradient is larger. To some extent, the gradient here is similar to the image contrast, both of which represent the maximum amplitude of the image grayscale change. However, it should be noted that when the subsequent algorithm locates the exact position of the edge, it uses the 50% line method. Therefore, only the edge with a faster gradient change will have a smaller error in the defect position located by the line. Visually, the alignment mark edge in this image is clearer. Figure 4 (c) and Figure 4 In (d), we can see that the image with better edge focus, the exact edge position obtained by drawing lines obviously has a smaller error range and is more accurate. Although the grayscale amplitude is also larger, this amplitude only helps the algorithm to roughly locate the edge position, and is not the only indicator that the algorithm cares about. Therefore, the amplitude (contrast) and gradient change range can be included in the definition and calculation of the focus index. In the embodiment of the present application, the concept of image edge gradient rate S is proposed to describe the speed of edge gradient change. Taking the focus position calculation in the X-axis direction as an example, it is defined as follows:
[0059]
[0060] Among them G 0％ is the grayscale of the image before the grayscale changes, X 0％ is the X coordinate corresponding to the moment when the grayscale of the image changes, G 100％ is the grayscale of the image after the grayscale changes, X 100％ It is the X coordinate corresponding to when the image grayscale has just changed.
[0061] Furthermore, considering that the alignment mark pattern in the image contains multiple edge positions, in the embodiment of the present application, the average edge gradient rate of the image can be used as a focus quantification indicator, which is specifically defined as:
[0062]
[0063] Where n represents the n edges detected by amplitude screening in the average projection method, G and S represent the grayscale and position information corresponding to the edge. That is, the average edge gradient rate of the image, which is used as an indicator to quantify the degree of focus. It can be seen that this indicator not only hopes that the contrast of the image edge is greater, but also hopes that the rate of change of the image gradient is faster, avoiding the contrast fluctuation problem caused by changes in the local lighting environment. More importantly, it is directly consistent with the algorithm for locating the edge position of the subsequent algorithm, ensuring the accuracy of the output results of the field image alignment device, and fundamentally solving the problem of incomplete matching between the contrast indicator and the overall business goal.
[0064] Therefore, in the lithography field area image auto-focusing method according to the embodiment of the present application, the average edge gradient rate is expressed as:
[0065]
[0066] Where n is the number of edges detected by the initial field image through the average projection method according to the amplitude screening, is the grayscale of the image before the grayscale changes in the i-th edge, is the grayscale of the image after the grayscale changes in the i-th edge, is the X coordinate corresponding to the moment when the grayscale of the image in the i-th edge changes, and It is the X coordinate corresponding to the grayscale change of the image in the i-th edge.
[0067] In addition, in actual application, the average edge gradient rate of the initial field image, that is, the image focus value, can be measured and calculated multiple times, for example, F times, and the average value is taken as the image focus value at this position, that is, the height of the silicon wafer stage.
[0068] That is, in the lithography field image autofocus method according to the embodiment of the present application, the average edge gradient rate is the average value of multiple single average edge gradient rates obtained by measuring and calculating the average edge gradient rate of the initial field image multiple times.
[0069] Step S130, moving the wafer stage to obtain a plurality of sampled field images, and fitting the focus position and focus value function curve using the least square method based on the image focus values of the plurality of sampled field images to obtain the best focus position corresponding to the vertex. That is, in view of the problem that the hill climbing search method as described above cannot give an accurate best focus theoretical position, and the problem that the search may fall into a local optimum due to environmental fluctuations or measurement errors, in the embodiment of the present application, a search and positioning method is proposed that uses a robust least square method to fit the focus position and focus value function curve, and then obtains the best focus position corresponding to the vertex based on the fit.
[0070] Since it is known here that the functional relationship between the focus position and the focus value is an approximately middle-symmetrical function curve, in order to calculate the theoretical vertex position of the curve, the function equation corresponding to the curve can be estimated first. Therefore, in an embodiment of the present application, it is proposed to use the least squares method to fit the function curve equation of the focus position and the focus value.
[0071] Specifically, firstly, the intermediate symmetry of the function curve is considered. In order to ensure the generalization ability of the fitting function, the functional relationship between the adjustment position and the focus value can be set as a generalized Gaussian-like distribution function:
[0072]
[0073] Among them, a represents the ratio coefficient of the curve, μ represents the axis of symmetry of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve. These are all unknown parameters and need to be solved by least squares fitting. Then, the corresponding image focus values can be measured at many evenly distributed focus positions as curve sampling data, and finally the robust least squares method is used for curve fitting to obtain the unknown parameters of the specific curve function, and finally x = μ can be obtained as the theoretical optimal focus position.
[0074] It can be seen that since curve fitting is no longer limited to local focus value changes, but is fitted to the overall data, the theoretical optimal focus position can be directly obtained when the sampling points are sufficient and uniform, and the robust least squares method has a certain robustness for curve fitting of sampling data with certain outliers due to local errors or numerical fluctuations (the weight of the outliers will be smaller), so the best focus position finally calculated will be more accurate.
[0075] That is, in the lithography field image auto-focusing method according to the embodiment of the present application, moving the wafer stage to obtain a plurality of sample field images, and fitting the focus position and focus value function curve using the least square method based on the image focus values of the plurality of sample field images to obtain the best focus position corresponding to the vertex includes:
[0076] Determine a plurality of focus position points x corresponding to the plurality of sampling field images i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as:
[0077]
[0078] Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and
[0079] Determine the best focus position corresponding to the vertex where x=μ.
[0080] In addition, if you want to ensure the accuracy of the curve fitting results, you must ensure that the number of sampling points is sufficient and uniform. However, if there are too many sampling points, the overall fitting calculation time will be too long. In order to improve the efficiency of the lithography field image autofocus method according to the embodiment of the present application, a cascade search and positioning combining hill climbing search and curve fitting can be performed.
[0081] Specifically, considering that the ultimate goal of using curve fitting is to obtain the focus position corresponding to the vertex of the curve, it is only necessary to ensure that the fitting of the curve near the vertex is accurate enough. Therefore, it is only necessary to perform dense sampling on the area near the vertex. Therefore, it is possible to consider combining the hill climbing search and the cascade search method of curve fitting to improve the operation efficiency. The specific process is as follows:
[0082] First, the hill climbing search method is used to locate the area near the coarse-grained vertex position. When the image focus value drops for the first time during the search process, the current focus position is used as the right boundary of the subsequent sampling area, and the current focus position is used as the left boundary of the subsequent sampling area after backing off for several steps, for example, backing off for two steps to determine the specific sampling area.
[0083] Then, uniformly and densely sampling data is performed on the specific sampling area with a smaller step spacing to obtain sampling data of multiple focus positions and image focus values in the area;
[0084] Use the weighted least squares method to fit the curve of all sampled data and solve the function equation
[0085]
[0086] Finally, the optimal focusing position x=μ corresponding to the vertex position is obtained.
[0087] It can be seen that this cascade search process framework combines the fast positioning characteristics of hill climbing search and the precise positioning characteristics of curve fitting, which ensures the accuracy of vertex positions solved by curve fitting while further improving the execution efficiency of the overall process. The overall process is as follows Figure 5 shown. Figure 5 A schematic diagram of a cascade search process in a lithography field area image auto-focusing method according to an embodiment of the present application is illustrated.
[0088] Therefore, in the lithography field image auto-focusing method according to the embodiment of the present application, fitting the focus position and focus value function curve using the least square method based on the image focus value to obtain the optimal focus position corresponding to the vertex includes:
[0089] Using a hill climbing search method starting from the focus position corresponding to the initial field image, and determining the focus position when the image focus value first decreases during the search process as the right boundary of the fine sampling area, and the focus position where the hill climbing search method is retracted by a predetermined number of steps as the left boundary of the fine sampling area;
[0090] Determine multiple focus positions x within the fine sampling area i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as:
[0091]
[0092] Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and
[0093] Determine the best focus position corresponding to the vertex where x=μ.
[0094] Furthermore, in the above-mentioned lithography field area image autofocus method, the hill climbing search method corresponds to a first search step, the sampling within the fine sampling area corresponds to a second search step, and the first search step is greater than the second search step.
[0095] For example, in the curve fitting stage, the sampling step length is obtained by dividing the coordinate difference between the focus position corresponding to the right boundary of the fine sampling area and the left boundary of the fine sampling area by the total number of steps of the least squares fitting method, and then retreating the sampling step length from the focus position corresponding to the right boundary of the fine sampling area to obtain the multiple focus position points x i The image focus value y under i .
[0096] That is, in the above-mentioned lithography field image automatic focusing method, a plurality of focus position points x in the fine sampling area are determined. i The image focus value y under i include:
[0097] The sampling step length is obtained by dividing the coordinate difference of the focus position corresponding to the right boundary of the fine sampling area and the left boundary of the fine sampling area by the total number of steps of the least square fitting method; and
[0098] Starting from the focus position corresponding to the right boundary of the fine sampling area, backtracking by the sampling step length obtains the plurality of focus position points x i The image focus value y under i .
[0099] In addition, in the embodiment of the present application, the above steps S110 to S130 may be repeatedly performed multiple times, and an average value of the best focus positions obtained multiple times is taken as the final focus position output.
[0100] That is, in the above-mentioned lithography field area image automatic focusing method, the above-mentioned steps are repeatedly performed to obtain multiple optimal focus positions, and the average value of the multiple optimal focus positions is output as the final focus position.
[0101] Therefore, the lithography machine field image autofocus method according to the embodiment of the present application improves the quantitative index of the image focus value into the image average edge gradient rate, which is more in line with the actual needs of the field image alignment of the lithography machine. That is, firstly, the image focus value will not fluctuate significantly due to the fluctuation of the internal lighting environment of the lithography machine. Secondly, the image focus value is directly related to the positioning of the alignment mark edge in the alignment algorithm of the field image alignment device. When the image focus value is higher, the output positioning result accuracy of the field image alignment device is also higher. Therefore, this quantitative index can better serve the actual business process of the entire device.
[0102] Moreover, according to the automatic focusing method of the lithography field image in the embodiment of the present application, the search method for the optimal focus position is improved into a curve fitting method based on robust least squares, which is more in line with the performance requirements of the lithography machine for accuracy and stability. That is, firstly, when the sampling points are sufficient and uniform, the curve fitting method can directly calculate the theoretical optimal focus position, which solves the problem of difficulty in setting the error threshold in the hill climbing search positioning method and improves the calculation accuracy. Secondly, the robust least squares method can better adapt to the influence of individual abnormal values caused by environmental fluctuations or measurement errors on the fitting results, thereby avoiding the search positioning from entering the local optimum or fitting errors.
[0103] In addition, according to the automatic focusing method of the lithography field image in the embodiment of the present application, a cascade search and positioning framework combining hill climbing search and curve sampling and fitting is proposed, which can ensure the search and positioning accuracy while effectively improving the execution calculation efficiency of the overall process. That is, first, in the early rough positioning stage, hill climbing search is used to perform coarse-grained vertex area positioning to avoid excessive sampling in the distant focusing position area that is not of concern and increase the overall calculation time. Secondly, after locating the area near the vertex, the fine positioning stage is entered, fine-grained uniform dense sampling is adopted, and then robust least squares is used for local curve fitting, which accurately calculates the optimal focusing position corresponding to the vertex while improving the overall calculation efficiency.
[0104] Application Examples
[0105] Figure 6 A schematic diagram illustrating an application example of the method for automatically focusing an image in a lithography field according to an embodiment of the present application is shown.
[0106] like Figure 6 As shown, the input of the lithography field area image auto-focusing method based on the image average edge gradient rate and cascade search positioning includes:
[0107] The number of cycles of the focus measurement algorithm N, the average number of frames when calculating the focus value F, the length of the step in the hill climbing search phase D, the total number of steps T on the z-axis of the curve fitting algorithm (T is set by the user. Generally speaking, the larger T is, the denser the sampling is, the higher the fitting accuracy is, but the overall algorithm takes more time), the mechanical coordinate system coordinates of the alignment mark on the wafer stage (C x ,C y ), the mechanical coordinate system coordinates corresponding to the center of the camera field of view (V x ,V y ).
[0108] The control items are the position control motors in the x and y directions of the horizontal plane of the wafer stage, and the height control motor in the z direction of the vertical plane of the wafer stage.
[0109] The initial output is the best focus position of the field image alignment device, for example, denoted as Z p .
[0110] 1) According to the mechanical coordinate system coordinates of the alignment mark on the silicon wafer stage (C x ,C y ), the mechanical coordinate system coordinates corresponding to the center of the camera field of view (V x ,V y ) information, calculate the coordinate difference between the two, and use this to control the wafer stage to move the center of the alignment mark to the center of the camera field of view to ensure the accuracy of subsequent measurements;
[0111] 2) Drive the height control motor of the wafer stage in the z direction to move the wafer stage to the bottom, ensure that the object distance between the wafer stage and the camera lens is large enough, and the focus value of the camera image is at a relatively small level, which serves as the search starting point for the hill climbing search phase;
[0112] 3) Calculate and measure the average edge gradient rate of the image at this time, that is, the image focus value, measure F times, and take the average focus value as the image focus value at this height;
[0113] 4) According to the value of the step length D of the climbing stage set by the algorithm input, drive the height control motor of the wafer stage in the z direction to move the wafer stage upward by a distance D, and refer to step 3 to measure the average image focus value at this height;
[0114] 5) Compare and determine the average focus value of the image at the current height and the previous height. If the focus value does not decrease, it is considered that the algorithm is in the uphill phase, and steps 4-5 are repeated; if the focus value decreases, it is considered that the algorithm is in the downhill phase, the climbing search phase ends, and go to step 6;
[0115] 6) The curve fitting phase begins, with the current height as the right boundary Z of the curve fitting sampling T , take the previous height as the left boundary Z0 of the curve fitting sampling, and according to the total number of steps T of the curve fitting stage set in the algorithm input, the sampling step length of the curve fitting stage can be calculated.
[0116]
[0117] 7) Drive the height control motor of the wafer stage in the z direction to move the wafer stage downward by a distance D′. Let the height at this time be Z n , refer to step 3 and measure the average image focus value at this height That is, get the current sampling point data
[0118] 8) Keeping other external conditions unchanged, continue to move the wafer stage downward by a distance D′, move the Z axis to the height of the next sampling point, and refer to step 7 to obtain new sampling point data until the last sampling height is measured. At this point, a sampling data set consisting of T+1 points is accumulated.
[0119] 9) Based on the fitting objective function, the obtained sampling data set is obtained using the robust least squares method Perform curve fitting to obtain the curve equation of focus position and focus value:
[0120]
[0121] At this point, we can know the best focus height position Z obtained by this sampling fitting. p =μ, focusing value Q p =a+d;
[0122] 10) Re-drive the height control motor in the z direction of the wafer stage, and execute steps 2-9 again N-1 times, and set the best focus height position Z calculated N times p Take the average as the final output.
[0123] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.
[0124] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.
[0125] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.
[0126] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
[0127] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.
Claims
1. A method for automatically focusing an image in a photolithography field, comprising: Acquire an initial field image, wherein, for the initial field image, a center of an alignment mark of the wafer stage is near a center of a field of view of the camera, and an object distance between the wafer stage and a lens of the camera is greater than a predetermined threshold; Calculating an average edge gradient rate of the initial field image as an image focus value for automatic focusing of the field image; and The wafer stage is moved to obtain a plurality of sampling field images, and a focusing position and focusing value function curve is fitted using a least square method based on image focusing values of the plurality of sampling field images to obtain an optimal focusing position corresponding to a vertex.
2. The method for automatically focusing a lithography field image according to claim 1, wherein: Acquiring the initial field image includes: The wafer stage is moved to obtain the initial field image by calculating the coordinate difference between the mechanical coordinate system coordinate of the center of the alignment mark of the wafer stage and the mechanical coordinate system coordinate corresponding to the field of view center of the camera and driving the height control motor of the wafer stage in the z direction.
3. The method for automatically focusing a lithography field image according to claim 1, wherein: The average edge gradient rate is expressed as: Where n is the number of edges detected by the initial field image through the average projection method according to the amplitude screening, is the grayscale of the image before the grayscale changes in the i-th edge, is the grayscale of the image after the grayscale changes in the i-th edge, is the X coordinate corresponding to the moment when the grayscale of the image in the i-th edge changes, and It is the X coordinate corresponding to the grayscale change of the image in the i-th edge.
4. The method for automatically focusing a lithography field image according to claim 1, wherein: The average edge gradient rate is an average value of multiple single average edge gradient rates obtained by measuring and calculating the average edge gradient rate of the initial field image for multiple times.
5. The method for automatically focusing a lithography field image according to claim 1, wherein: Moving the wafer stage to obtain a plurality of sampling field images, and fitting the focus position and focus value function curve using the least square method based on the image focus values of the plurality of sampling field images to obtain the best focus position corresponding to the vertex includes: Determine a plurality of focus position points x corresponding to the plurality of sampling field images i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as: Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and Determine the best focus position corresponding to the vertex where x=μ.
6. The method for automatically focusing a lithography field image according to claim 1, wherein: Fitting the focus position and focus value function curve using the least square method based on the image focus value to obtain the optimal focus position corresponding to the vertex includes: Using a hill climbing search method starting from the focus position corresponding to the initial field image, and determining the focus position when the image focus value first decreases during the search process as the right boundary of the fine sampling area, and the focus position where the hill climbing search method is retracted by a predetermined number of steps as the left boundary of the fine sampling area; Determine multiple focus positions x within the fine sampling area i The image focus value y under i , and the least squares fitting solution is performed based on the general Gaussian-like distribution function, and the general Gaussian-like distribution function is expressed as: Wherein, a represents the ratio coefficient of the curve, μ represents the symmetry axis of the curve, σ represents the standard deviation of the curve, and d represents the longitudinal offset of the curve; and Determine the best focus position corresponding to the vertex where x=μ.
7. The method for automatically focusing a lithography field image according to claim 6, wherein: The hill climbing search method corresponds to a first search step length, the sampling within the fine sampling area corresponds to a second search step length, and the first search step length is greater than the second search step length.
8. The method for automatically focusing a lithography field image according to claim 6, wherein: Determine multiple focus positions x within the fine sampling area i The image focus value y under i include: The sampling step length is obtained by dividing the coordinate difference of the focus position corresponding to the right boundary of the fine sampling area and the left boundary of the fine sampling area by the total number of steps of the least square fitting method; and Starting from the focus position corresponding to the right boundary of the fine sampling area, backtracking by the sampling step length obtains the plurality of focus position points x i The image focus value y under i .
9. The method for automatically focusing a lithography field image according to claim 1, wherein: Repeat the steps as claimed in claim 1 to obtain a plurality of optimal focus positions, and output an average value of the plurality of optimal focus positions as a final focus position.