A wafer residue detection method for cleaning etching machine
The etching and cleaning residue coefficient is constructed through image acquisition and edge detection algorithms, which solves the problem of difficulty in detecting wafer surface residues after cleaning by the etching machine, and improves detection accuracy and accuracy.
Patent Information
- Application Number
- CN202411132431.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing etching machine cleaning detection methods are difficult to accurately detect the etching film remaining on the wafer surface, resulting in misjudgment of the cleaning effect and unable to effectively monitor the surface residue after cleaning of the etching machine.
Using image acquisition and edge detection algorithms, the residual confidence and correction coefficient of etching are calculated by constructing the etching cleaning residue coefficient, including gloss disturbance index, cleaning effect evaluation parameters, gloss change law coefficient and local gloss consistency index, and the crystal is judged. Whether there is any residue on the circular surface.
The accuracy and accuracy of wafer surface etchant residue detection can be improved, and the location where etchant residues can be more accurately identified, ensuring the accuracy of the cleaning effect.
Smart Images

Figure CN118658801B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of cleaning and detection of etching machines, and in particular to a wafer residue detection method for cleaning an etching machine. Background Art
[0002] Cleaning the etcher is a key step to ensure etching quality and improve product performance. Incomplete cleaning may lead to problems such as uneven etching, circuit short circuit, and product performance degradation. Therefore, the etching cleaning process needs to be strictly controlled to ensure that each step achieves the expected effect and avoid safety hazards on the chip caused by surface residues due to poor cleaning effect of the etcher.
[0003] Taking into account the special nature of the etching glue, it will firmly adhere to the surface of the wafer during the etching process. During the cleaning process, a thin film of etching glue may remain on the wafer surface due to insufficient debonding strength and cleaning strength. When a thin film of etching glue remains on the wafer surface, it has little effect on the refraction change of the light source on the wafer surface. It is difficult to detect the residual etching glue using traditional visual inspection methods, resulting in misjudgment of the cleaning effect of the wafer, and inability to accurately detect the residual situation on the surface after cleaning by the etcher. Summary of the invention
[0004] In order to solve the above technical problems, the present application provides a wafer residue detection method for a cleaning and etching machine to solve the existing problems.
[0005] A wafer residue detection method for cleaning an etching machine of the present application adopts the following technical solution:
[0006] An embodiment of the present application provides a wafer residue detection method for cleaning an etching machine, the method comprising the following steps:
[0007] Collect wafer surface images and extract the edges of each circuit using edge detection algorithms;
[0008] The etching cleaning residual coefficient is constructed based on the situation of each circuit edge and the difference between each pixel point. The construction process of the etching cleaning residual coefficient includes:
[0009] Determine the gloss disturbance index of each circuit edge based on the grayscale information of each circuit edge and the difference relationship between the edge and the adjacent circuit edge; determine the wafer cleaning effect evaluation parameter based on the gloss disturbance index of all circuit edges;
[0010] Determine the gloss variation coefficient of each pixel based on the grayscale change of the pixel in the direction of the line connecting each pixel and the projection point of the light source, wherein the light source is the projection point of the center point of the light source for image acquisition in the image; determine the local gloss consistency index of each pixel based on the grayscale difference and distribution in the local neighborhood of each pixel; determine the etching glue residue confidence of each pixel based on the gloss variation coefficient and the local gloss consistency index, and determine the etching glue residue correction coefficient based on the etching glue residue confidence of all pixels;
[0011] The etching cleaning residual coefficient is determined based on the wafer cleaning effect evaluation parameters and the etching glue residual correction coefficient; when the etching cleaning residual coefficient is greater than the preset threshold, there is residue on the wafer surface, and the wafer residue detection is completed.
[0012] Preferably, the gloss disturbance index is determined by:
[0013] The grayscale values of all pixels on each circuit edge constitute the circuit glossiness sequence of each circuit edge; all detected circuit edges are divided into two categories: horizontal edges and vertical edges. For horizontal edges, the grayscale values of all pixels on the line connecting the two pixels with the largest and smallest horizontal coordinates constitute a direction connection sequence; for vertical edges, the grayscale values of all pixels on the line connecting the two pixels with the largest and smallest vertical coordinates constitute a direction connection sequence;
[0014] A perpendicular line is drawn through each pixel point on the circuit edge, and the angle between the tangent line of each pixel point on the circuit edge and the corresponding perpendicular line is counted; the Euclidean distance between the pixel points at the corresponding position in the circuit gloss sequence of each circuit edge and the circuit edge closest to it is calculated as the adjacent offset distance of each pixel point on each circuit edge, and then the parallel variation coefficient of each circuit edge is calculated, and the parallel variation coefficient B of the i-th circuit edge is i The corresponding calculation formula is:
[0015] ;
[0016] Where P is the total number of pixels at the edge of the i-th circuit, λ i,a ,λ g,a are the angles between the ith circuit edge and the ath pixel in the nearest circuit edge g, respectively. i,a is the adjacent offset distance of the ath pixel point in the i-th circuit edge, is the mean of the adjacent offset distances of the i-th circuit edge;
[0017] The Euclidean distance between the circuit glossiness sequence of each circuit edge and the pixel point at the corresponding position in the direction connection sequence is used as the offset distance of each pixel point on each circuit edge, and the directional offset of each circuit edge is determined based on the distribution of the offset distance of each circuit edge and the cosine similarity between the circuit glossiness sequence and the direction connection sequence;
[0018] The gloss disturbance index of each circuit edge is determined based on the parallel variation coefficient and the directional offset.
[0019] Preferably, the process of determining the direction offset is:
[0020] Calculate the cosine similarity between the circuit glossiness sequence and the directional connection sequence of the circuit edge, obtain the mean offset distance of all pixel points on the circuit edge, calculate the cumulative result of the absolute value of the difference between the offset distance of all pixel points on the circuit edge and the mean offset distance, use the negative number of the ratio of the cosine similarity to the cumulative result as the exponent of an exponential function with a natural constant as the base, and use the calculation result of the exponential function as the directional offset of the circuit edge.
[0021] Preferably, the gloss disturbance index is the product of the parallel variation coefficient and the directional offset.
[0022] Preferably, the method for determining the wafer cleaning effect evaluation parameters is:
[0023] Cluster the gloss disturbance indexes of all circuit edges, and take the circuit edge corresponding to the gloss disturbance index in the cluster with the largest mean value of the gloss disturbance index as the etching glue interference edge, and the circuit edge corresponding to the gloss disturbance index in the cluster with the smallest mean value as the circuit stripe clean edge; calculate the sum of the number of etching glue interference edges and the preset value;
[0024] The number of clean edges of the circuit stripes is divided by the sum result to obtain a wafer cleaning effect evaluation parameter.
[0025] Preferably, the method for determining the gloss gradient coefficient is:
[0026] Obtaining a line connecting each pixel point and the projection point of the light source, forming a gloss disturbance judgment sequence of each pixel point with a plurality of points on the line with the closest Euclidean distance to each pixel point, and calculating the cumulative sum of the absolute values of the gray value differences between any two adjacent pixel points in the sequence;
[0027] The ApEn algorithm is used to calculate the ApEn value of the sequence, and the result of adding the reciprocal of the sum of the ApEn value and the preset value to the accumulated sum is used as the gloss gradient coefficient of each pixel point.
[0028] Preferably, the local gloss consistency index is determined by:
[0029] With each pixel as the center, a gloss distribution judgment window with a preset side length is constructed, and the local gloss consistency index of the Lth pixel is recorded as μ L , the corresponding calculation formula is:
[0030]
[0031] In the formula, exp( ) is an exponential function with a natural constant as the base, is the total number of pixels in the gloss distribution judgment window, b L is the gray value of the Lth pixel, b ρ(ω) is the gray value of the ω-th pixel in the gloss distribution judgment window, σ is the total number of pixels in the image, , are the grayscale means of the gloss distribution judgment windows of the Lth and τth pixels respectively.
[0032] Preferably, the etching glue residue confidence is the reciprocal of the sum of the gloss gradient law coefficient and the local gloss consistency index.
[0033] Preferably, the etching glue residue correction coefficient is the average of the etching glue residue confidence levels of all pixels in the wafer surface image.
[0034] Preferably, the etching cleaning residue coefficient is a normalized value of the product of the inverse of the wafer cleaning effect evaluation parameter and the etching glue residue correction coefficient.
[0035] This application has at least the following beneficial effects:
[0036] The present application aims at monitoring the etching glue film existing after the wafer cleaning is completed, analyzing the cleaning effect of the wafer surface, and calculating the cleaning effect evaluation parameters, which are used to preliminarily determine the residual cleaning on the wafer surface and improve the subsequent wafer residue detection accuracy; at the same time, the etching glue residue confidence of each pixel point in the image is constructed to characterize the situation where the etching glue film may be attached to the wafer, and more accurately distinguish the location where the etching glue residue exists, thereby improving the detection accuracy of the etching glue on the wafer surface; further constructing the etching cleaning residual coefficient, and then judging the etching glue residue on the wafer surface, solves the problem of difficulty in detection when the etching glue film exists on the wafer surface, and further improves the detection accuracy of the residue on the wafer cleaning surface. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0038] Figure 1 A flowchart of the steps of a method for detecting wafer residues in a cleaning etcher provided in the present application;
[0039] Figure 2 Schematic diagram for extracting the etching cleaning residue coefficient. DETAILED DESCRIPTION
[0040] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following is a detailed description of the wafer residue detection method for cleaning and etching machines proposed in the present application, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0041] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
[0042] The following is a detailed description of a specific solution of a wafer residue detection method for a cleaning etcher provided by the present application in conjunction with the accompanying drawings.
[0043] An embodiment of the present application provides a method for detecting wafer residues for cleaning an etching machine. Specifically, the following method for detecting wafer residues for cleaning an etching machine is provided. Figure 1 , the method comprises the following steps:
[0044] Step S001, collecting wafer surface image data and preprocessing it as basic data for wafer surface residue detection.
[0045] First, an industrial high-resolution camera is used to collect image data of the same batch of wafers after cleaning under sufficient lighting conditions, and the image data is converted into a grayscale image using the grayscale value averaging method, and the collected image data is denoised using the wavelet denoising algorithm, and then the image data is sharpened using the Laplace operator. The grayscale value averaging method, the wavelet denoising algorithm, and the Laplace operator are all well-known technologies, and their processes are not repeated in this embodiment. Finally, a high-definition grayscale image of the wafer surface after cleaning is obtained, and it is recorded as images A1, A2, ..., A n . Among them A n The meaning is a high-definition grayscale image of the surface of the nth wafer, and the grayscale values of all pixels therein are used as its glossiness value.
[0046] Step S002, analyze the wafer surface cleaning evaluation effect, construct cleaning effect evaluation parameters, and at the same time, construct the etching glue residue confidence according to the gloss change characteristics formed by the etching glue residue on the wafer surface, and further construct the etching cleaning residue coefficient in combination with the cleaning effect evaluation parameters to characterize the etching glue residue on the wafer surface.
[0047] First, take image A n For example, image A n As input, using the Canny edge detection operator and the expansion corrosion algorithm, the output is all the edges in the image A. The Canny edge detection algorithm and the expansion corrosion algorithm are both well-known technologies, and the process is not repeated in this embodiment. Since the circuit design on the surface of the wafer is usually horizontal and vertical, and the edges on the surface of the wafer are all circuit edges, and usually adjacent circuit edges are parallel, when the etching liquid remains on the surface of the wafer, it may have a certain impact on the originally relatively straight circuit edges, causing them to be distorted and offset, and also causing their spacing to change.
[0048] The circuit edge is divided into horizontal edge and vertical edge. First, obtain the pixel points with the largest and smallest horizontal coordinates among all detected edges, connect the two pixel points, calculate the Euclidean distance between the two points, and record it as dh; obtain the pixel points with the largest and smallest vertical coordinates, connect the two pixel points, calculate the Euclidean distance between the two points, and record it as dz. Taking the i-th edge as an example, when the dh of the edge is greater than dz, the edge is divided into a horizontal edge. Conversely, when the dz of the edge is greater than or equal to dh, the edge is divided into a vertical edge. It should be noted that the implementer can also use other methods to divide the circuit edge, such as using a neural network model to classify the circuit edge, artificially making label data, marking the horizontal edge as 1 and the vertical edge as 0, and training the classification neural network through the label data to realize the classification and identification of the horizontal edge and the vertical edge. The specific method is not particularly limited in this embodiment, and the implementer can choose the division method at will.
[0049] For a horizontal edge, all the pixels on the line connecting the two pixels with the largest and smallest horizontal coordinates are used to construct the direction line of the edge. The glossiness, i.e., the grayscale value, of all the pixels contained therein is used to construct a direction line sequence, which is recorded as , the glossiness of all pixels on the edge is used to construct the circuit glossiness sequence of the edge. For a longitudinal edge, the glossiness of all pixels on the line connecting the two pixels with the largest and smallest ordinates is constructed. Correspondingly, the circuit glossiness sequence is also constructed from the glossiness of all pixels on the edge.
[0050] Taking the ith circuit edge as an example, starting from the first pixel point in the circuit gloss sequence of the circuit edge, calculate the Euclidean distance between each pixel point and each pixel point at the corresponding position in the direction line sequence one by one, and record it as the offset distance d. If the total number of pixels is different, the total number of pixels of the ith circuit edge shall be used as the basis for calculation, that is, the number of pixels in the circuit gloss sequence shall be used as the basis. When the number of pixels in the circuit gloss sequence is greater than the number of pixels in the direction line sequence, the grayscale value of the pixel point at the insufficient position in the direction line sequence shall be filled with 0 to ensure data integrity. It should be noted that the implementer may also use the mean filling method, which is not particularly limited in this embodiment. This embodiment adopts the zero filling method. Then calculate the adjacent offset distance between the ith circuit edge and each pixel point on the circuit edge closest to it in the same way, that is, calculate the Euclidean distance between each circuit edge and the pixel point at the corresponding position in the circuit gloss sequence of the circuit edge closest to it as the adjacent offset distance of each pixel point on each circuit edge, and record it as f.
[0051] A perpendicular line is drawn through each pixel point on the edge of the i-th circuit to connect the direction of the circuit edge. The tangent line of each pixel point on the i-th circuit edge will form an angle with the corresponding perpendicular line, and the size of the angle is calculated using the function in the OpenCV library. The implementer may also use other methods to count the size of the above angle, where the OpenCV library is a well-known technology and will not be described in detail in this embodiment. Based on the above process, this embodiment will construct the gloss disturbance index (F) of each circuit edge:
[0052] ;
[0053] ;
[0054] ;
[0055] In the formula, B i is the parallel variation coefficient of the edge of the ith circuit, P is the total number of pixels on the edge of the ith circuit, and λ i,a ,λ g,a are the angles between the ith circuit edge and the ath pixel in the nearest circuit edge g, respectively. i,a is the adjacent offset distance of the ath pixel point in the i-th circuit edge, is the mean of the adjacent offset distances of the i-th circuit edge. i is the direction offset of the i-th circuit edge, D i is the circuit glossiness sequence of the i-th circuit edge, is the direction connection sequence of the i-th circuit edge, cos( ) is the cosine similarity, is the cosine similarity between the circuit glossiness sequence and the direction connection sequence of the i-th circuit edge, d i,a is the offset distance of the ath pixel point on the edge of the ith circuit, is the mean offset distance of the edge of the ith circuit, P is the total number of pixels on the edge of the ith circuit, α is a parameter to prevent the denominator from being zero, α=0.0001, exp( ) is an exponential function with e as the base, F i is the gloss disturbance index of the i-th circuit edge.
[0056] When the difference in the angle between the i-th circuit edge and the nearest circuit edge is greater, and the difference in the adjacent offset distance of each pixel is greater, it means that there may be multiple distorted positions on the i-th circuit edge, and its parallel situation has changed. When the cosine similarity between the circuit gloss sequence and the direction connection sequence of the i-th circuit edge is greater, and the difference in the offset distance change on the circuit edge is smaller, that is, E iThe smaller it is, the less the offset of the circuit edge is, and the more likely it is that the circuit edge is straighter, and the less affected by the etching glue. When the glossiness of the i-th circuit edge is greater and more uniform, and the offset is smaller, that is, F i The smaller it is, the less interference the gloss of the circuit edge is subject to and the less likely etching glue residue is.
[0057] So far, the gloss disturbance index F of each circuit edge can be obtained through the above method.
[0058] Further, all the gloss disturbance indexes of the circuit edges are taken as input, and the K-means clustering algorithm is used. The number of clusters is set to 2, and the measured distance is the absolute value of the gloss disturbance index difference of each circuit edge, and the output is two clusters. The K-means clustering algorithm is a well-known technology, and its process is not repeated in this embodiment. The mean of the gloss disturbance index of the two clusters is calculated, and the cluster with the largest gloss disturbance index mean is taken as the circuit disturbed cluster, and the circuit edge corresponding to the gloss disturbance index contained therein is taken as the etching glue interference edge; the cluster with the smallest gloss disturbance index mean is taken as the circuit clean cluster, and the circuit edge corresponding to the gloss disturbance index contained therein is taken as the circuit stripe clean edge. The larger the proportion of the etching glue interference edge, the poorer the cleaning effect of the wafer after etching.
[0059] Calculate image A n Wafer cleaning effect evaluation parameters (β):
[0060] ;
[0061] In the formula, β n For image A n Wafer cleaning effect evaluation parameter, γ n ,ε n They are image A n The total number of clean edges of circuit stripes and interference edges of etching glue. The greater the number of clean edges of circuit stripes in the wafer and the fewer interference edges of etching glue, the better the wafer cleaning effect.
[0062] At this point, the above method can be used to calculate the evaluation parameters of the wafer cleaning effect in each image, which can be used to accurately determine the residual condition of the etching glue film on the wafer surface in the subsequent process.
[0063] Furthermore, due to the particularity of the etching glue, it has a strong adhesion ability on the wafer surface. During the cleaning process, there may be a situation where only a thin film of etching glue remains on the wafer surface. Since the thickness of the film is small, the gloss of the wafer surface is less affected. It is difficult to accurately distinguish the residual situation of the etching glue film only by the cleaning effect evaluation parameters. In the process of wafer image acquisition, the direction of the light source is usually fixed, and the metallic gloss on the wafer surface will have a certain gloss deviation direction, that is, the metallic gloss has a certain change direction, and the gloss will gradually increase from the end away from the light source to the end close to the light source. When etching glue remains on the wafer surface, although the etching glue has a small impact on the gloss, due to its different refraction effect on light, the position where the etching glue remains will have a deflection of the light path and uneven distribution.
[0064] Take image A n For example, the center point of the light source is projected into image A, and the location of the light source projection point is recorded as the light source point. n Connect all the pixels in the Lth pixel with the light source point. Take the Lth pixel as an example. After connecting it with the light source point, take the 5 pixels on the line with the closest Euclidean distance to the pixel, and then take the 5 pixels on the extended line of the line to jointly construct the gloss transition disturbance judgment sequence of the Lth pixel, and record it as X L , and obtain the gloss transition disturbance judgment sequence of the remaining pixels by the same method. The gloss transition disturbance judgment sequence of each pixel is used as the input of the ApEn algorithm, the window length is set to 3, and the output is the ApEn value of the gloss transition disturbance judgment sequence of each pixel. Then, a differential sequence of the gloss transition disturbance judgment sequence of each pixel is used as input, and the Pettitt mutation point detection algorithm is used to output the mutation point of a differential sequence of the gloss transition disturbance judgment sequence of each pixel. The ApEn algorithm and the Pettitt mutation point detection algorithm are well-known technologies, and the process is not repeated in this embodiment. Then, taking each pixel as the center, a 5×5 gloss distribution judgment window is constructed, and the gloss mean in each judgment window is calculated. Construct the etching glue residue confidence level for each pixel ( ):
[0065] ;
[0066] ;
[0067] ;
[0068] Where V L is the gloss variation coefficient of the Lth pixel, ApEn(X L ) is the sequence X Lis the ApEn value, For sequence X L The number of mutation points in a differential sequence of a section, sum(X L ) is the calculation sequence X L The total number of pixels, θ L (l),θ L (l-1) is the sequence X L The glossiness of the lth and l-1th pixel points in μ. L is the local gloss consistency index of the Lth pixel, exp( ) is an exponential function with a natural constant as the base, is the total number of pixels in the gloss distribution judgment window, b L is the gray value of the Lth pixel, b ρ(ω) is the gray value of the ω-th pixel in the gloss distribution judgment window, that is, the glossiness. σ is the total number of pixels in the image, , are the grayscale means of the gloss distribution judgment windows of the Lth and τth pixels respectively. is the etching glue residue confidence of the Lth pixel.
[0069] When the ApEn value of the gloss transition disturbance judgment sequence of the Lth pixel point is smaller and the mutation in a differential sequence is less, The larger the value of is, the greater the difference in the change of each pixel in the gloss transition disturbance judgment sequence is, that is, The larger the value is, the gloss gradient of the Lth pixel point is consistent with the gloss variation law of the wafer under the light source for image acquisition. When the difference between the gloss distribution of the Lth pixel point and the gloss of the other pixels in the gloss distribution judgment window is smaller, and the difference between the gloss mean of the window and the gloss mean of the other windows in the image is smaller, that is, μ L The larger the gloss change of the pixel point is, the more consistent the gloss change of the pixel point is with that of other positions. The smaller it is, the less likely it is that there is etching glue residue at the Lth pixel.
[0070] So far, the image A can be calculated by the above method n The confidence level of the etching glue residue at all pixels in the image. n The detected wafer cleaning effect evaluation parameters are used to construct image A n Etching cleaning residual coefficient (η):
[0071] ;
[0072] ;
[0073] In the formula, For image A n The etching glue residual correction coefficient in image A is n The total number of pixels in is the residual confidence of the etching glue at the Lth pixel. n For image A n The etching cleaning residue coefficient, β n For image A n The wafer cleaning effect evaluation parameters in the figure are as follows: Norm() is the normalization process. n The greater the possibility of etching glue residue in the image, the worse the cleaning effect. n The greater the possibility of residual etching glue on the surface of the wafer being tested, the more necessary it is to further clean the wafer.
[0074] So far, we have obtained image A n The extraction diagram of the etching cleaning residual coefficient is as follows: Figure 2 shown.
[0075] Step S003 , detecting the wafer cleaning residues according to the constructed etching cleaning residue coefficient.
[0076] Furthermore, this embodiment uses the wafer etching cleaning residue coefficient to determine whether there is any etching glue residue in the cleaned wafer, so as to judge the cleaning effect of the wafer more carefully and accurately. Specifically, this embodiment will set a residue judgment threshold, and the wafer corresponding to the image whose etching cleaning residue coefficient is greater than the residue judgment threshold will be judged as having cleaning residue; otherwise, the etcher surface has a good cleaning effect and no residue exists. In the case where there is cleaning residue on the wafer surface, a second cleaning process can be performed to ensure the cleaning quality of the wafer surface.
[0077] At this point, according to the above process of this embodiment, accurate detection of cleaning residues on the wafer during the cleaning process of the etcher can be achieved.
[0078] It should be noted that the above sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above is a description of a specific embodiment of this specification. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0079] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.
[0080] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifications to the technical solutions recorded in the aforementioned embodiments, or equivalent replacement of some of the technical features therein, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A method for detecting wafer residues in a cleaning etching machine, characterized in that: The method comprises the following steps: Collect wafer surface images and extract the edges of each circuit using edge detection algorithms; The etching cleaning residual coefficient is constructed based on the situation of each circuit edge and the difference between each pixel point. The construction process of the etching cleaning residual coefficient includes: Determine the gloss disturbance index of each circuit edge based on the grayscale information of each circuit edge and the difference relationship between the edge and the adjacent circuit edge; determine the wafer cleaning effect evaluation parameter based on the gloss disturbance index of all circuit edges; Determine the gloss variation coefficient of each pixel based on the grayscale change of the pixel in the direction of the line connecting each pixel and the projection point of the light source, wherein the light source is the projection point of the center point of the light source for image acquisition in the image; determine the local gloss consistency index of each pixel based on the grayscale difference and distribution in the local neighborhood of each pixel; determine the etching glue residue confidence of each pixel based on the gloss variation coefficient and the local gloss consistency index, and determine the etching glue residue correction coefficient based on the etching glue residue confidence of all pixels; Determine the etching cleaning residue coefficient based on the wafer cleaning effect evaluation parameter and the etching glue residue correction coefficient; when the etching cleaning residue coefficient is greater than a preset threshold, residues exist on the wafer surface, and the detection of wafer residues is completed; The gloss disturbance index is determined as follows: The grayscale values of all pixels on each circuit edge constitute the circuit glossiness sequence of each circuit edge; all detected circuit edges are divided into two categories: horizontal edges and vertical edges. For horizontal edges, the grayscale values of all pixels on the line connecting the two pixels with the largest and smallest horizontal coordinates constitute a direction connection sequence; for vertical edges, the grayscale values of all pixels on the line connecting the two pixels with the largest and smallest vertical coordinates constitute a direction connection sequence; A perpendicular line is drawn through each pixel point on the circuit edge, and the angle between the tangent line of each pixel point on the circuit edge and the corresponding perpendicular line is counted; the Euclidean distance between the pixel points at the corresponding position in the circuit gloss sequence of each circuit edge and the circuit edge closest to it is calculated as the adjacent offset distance of each pixel point on each circuit edge, and then the parallel variation coefficient of each circuit edge is calculated, and the parallel variation coefficient B of the i-th circuit edge is i The corresponding calculation formula is: ; Where P is the total number of pixels at the edge of the i-th circuit, λ i,a ,λ g,a are the angles between the ith circuit edge and the ath pixel in the nearest circuit edge g, respectively. i,a is the adjacent offset distance of the ath pixel point in the i-th circuit edge, is the mean of the adjacent offset distances of the i-th circuit edge; The Euclidean distance between the circuit glossiness sequence of each circuit edge and the pixel point at the corresponding position in the direction connection sequence is used as the offset distance of each pixel point on each circuit edge, and the directional offset of each circuit edge is determined based on the distribution of the offset distance of each circuit edge and the cosine similarity between the circuit glossiness sequence and the direction connection sequence; The gloss disturbance index of each circuit edge is determined based on the parallel variation coefficient and the directional offset.
2. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The process of determining the direction offset is as follows: Calculate the cosine similarity between the circuit glossiness sequence and the directional connection sequence of the circuit edge, obtain the mean offset distance of all pixel points on the circuit edge, calculate the cumulative result of the absolute value of the difference between the offset distance of all pixel points on the circuit edge and the mean offset distance, use the negative number of the ratio of the cosine similarity to the cumulative result as the exponent of an exponential function with a natural constant as the base, and use the calculation result of the exponential function as the directional offset of the circuit edge.
3. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The gloss disturbance index is the product of the parallel variation coefficient and the directional offset.
4. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The method for determining the wafer cleaning effect evaluation parameters is as follows: Cluster the gloss disturbance index of all circuit edges, and take the circuit edge corresponding to the gloss disturbance index in the cluster with the largest mean value of gloss disturbance index as the etching glue interference edge, and the circuit edge corresponding to the gloss disturbance index in the cluster with the smallest mean value as the circuit stripe clean edge; Calculate the sum of the number of interference edges of the etching glue and the preset value; The number of clean edges of the circuit stripes is divided by the sum result to obtain a wafer cleaning effect evaluation parameter.
5. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The method for determining the gloss gradient coefficient is as follows: Obtaining a line connecting each pixel point and the projection point of the light source, forming a gloss disturbance judgment sequence of each pixel point with a plurality of points on the line with the closest Euclidean distance to each pixel point, and calculating the cumulative sum of the absolute values of the gray value differences between any two adjacent pixel points in the sequence; The ApEn algorithm is used to calculate the ApEn value of the sequence, and the result of adding the reciprocal of the sum of the ApEn value and the preset value to the accumulated sum is used as the gloss transition regularity coefficient of each pixel point.
6. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The local gloss consistency index is determined by: With each pixel as the center, a gloss distribution judgment window with a preset side length is constructed, and the local gloss consistency index of the Lth pixel is recorded as μ L , the corresponding calculation formula is: ; In the formula, exp( ) is an exponential function with a natural constant as the base, is the total number of pixels in the gloss distribution judgment window, b L is the gray value of the Lth pixel, b ρ(ω) is the gray value of the ω-th pixel in the gloss distribution judgment window, σ is the total number of pixels in the image, are the grayscale means of the gloss distribution judgment windows of the Lth and τth pixels respectively.
7. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The etching glue residue confidence is the reciprocal of the sum of the gloss gradient law coefficient and the local gloss consistency index.
8. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The etching glue residue correction coefficient is the average of the etching glue residue confidence levels of all pixels in the wafer surface image.
9. A wafer residue detection method for cleaning an etching machine as claimed in claim 1, characterized in that: The etching cleaning residue coefficient is a normalized value of the product of the inverse of the wafer cleaning effect evaluation parameter and the etching glue residue correction coefficient.
