Wafer detecting and sorting system
By conducting multi-dimensional evaluation of the polishing surface, thickness and profile of the wafer, combined with defect density, and designing a comprehensive quality evaluation system, the problem of insufficient detection accuracy and efficiency in the existing technology is solved, and more accurate wafer quality is achieved Evaluation and sorting.
Patent Information
- Application Number
- CN202410967390.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art fails to effectively consider the surface roughness and overall flatness of thickness detection in wafer sorting, resulting in insufficient detection accuracy and efficiency, and thickness detection relies on partial area data to lead to deviations.
By obtaining the depth image of the polished surface of the wafer for segmentation, the flatness index is calculated; obtaining the thickness data to calculate the thickness uniformity index; obtaining the contour image and the standard to calculate the deviation index; combining with the defect density index, the design of the comprehensive quality evaluation system is used for classification and sorting.
It improves sorting accuracy and efficiency, avoids single indicator errors, provides multi-dimensional quality evaluation, enhances production decision support, and improves product quality and production efficiency.
Smart Images

Figure CN118926128A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of wafer classification, and in particular to a wafer detection and sorting system. Background Art
[0002] Wafer manufacturing is the core of the semiconductor industry and is of great significance to my country's technological independence, economic development and national security. Mastering wafer manufacturing and sorting technology can reduce dependence on foreign supply chains, promote independent innovation, and improve overall technological levels. At the same time, the high added value and technological content of the semiconductor industry will help drive the upgrading of related industrial chains, promote economic development, and create high-skilled employment opportunities.
[0003] The Chinese patent with application number CN202111176439.6 proposes a classification and packaging method for multi-grade chips on wafers. The method tests the wafers, grades the chips and generates a distribution map, and engraves the corresponding QR code on the substrate; then the chips of different grades are produced using one work order, and then all chips are patched; during this period, the grade of the chip corresponding to each product on the substrate is recorded to generate a substrate map.
[0004] A Chinese patent with application number CN202110419968.8 proposes a cloud computing-based intelligent wafer detection and classification method, device and system. The method obtains the capacitance-voltage characteristic parameters of the wafer and calculates the deviation resistivity of the wafer, obtains multiple detection thicknesses corresponding to multiple positions of the wafer to determine the average thickness of the wafer, obtains an image of the wafer to determine whether the wafer has defects, and finally controls the sorting device to classify the wafer through the above three characteristics.
[0005] The Chinese patent application number CN202110223090.0 proposes a method for improving the yield rate of semiconductor chips using a machine learning classifier. By analyzing the preliminary data of the wafer bare die before packaging, the finished product test results of the wafer bare die after packaging are predicted, and the quality of the wafer bare die is classified according to the predicted results. Finally, the die is packaged according to the quality classification. This method can significantly improve the accuracy of the quality classification of the wafer bare die before packaging, thereby improving the yield rate of the semiconductor chip after packaging.
[0006] Although the current method can solve the problems of insufficient technical level, low degree of automation, and low detection accuracy faced by wafer sorting to a certain extent, the surface roughness of the wafer is an important parameter of wafer quality. The above method only considers defect detection and does not consider the overall polishing surface flatness. The thickness detection method only relies on partial area data, which may lead to deviations. The detection accuracy and sorting efficiency in the wafer sorting process still need to be improved.
[0007] To this end, the present invention proposes a wafer inspection and sorting system. Summary of the invention
[0008] The object of the present invention is to provide a wafer inspection and sorting system, which obtains a wafer flatness index by acquiring a polished surface depth image of a wafer and performing segmentation, performing hypothesis tests on all pixel values in the segmented image; acquiring thickness data of the wafer and performing segmentation, calculating a thickness uniformity coefficient for all data in a segmented data group, and then calculating a wafer thickness uniformity index; acquiring a wafer contour image, and overlapping it with a standard wafer at a specified position shape, and calculating a deviation index; acquiring wafer image data and extracting features, detecting defects, and calculating a defect density index; designing a comprehensive quality evaluation system for the wafer according to the above four indexes, obtaining a comprehensive quality score for the wafer, and classifying and sorting the wafers according to a quality classification standard, thereby improving sorting accuracy and efficiency.
[0009] To achieve the above object, the present invention provides the following technical solutions:
[0010] A wafer inspection and sorting system, comprising:
[0011] The wafer flatness calculation module is used to obtain a polished surface depth image of the wafer, segment the polished surface depth image to obtain a segmented image; segment and position encode the segmented image point by point to obtain pixel depth data, perform hypothesis testing on the pixel depth data, and determine whether the segmented image meets the flatness standard; determine whether all the segmented images meet the flatness standard, and calculate the wafer flatness index according to the compliance ratio;
[0012] Further, calculating the wafer flatness index includes:
[0013] Acquire the polished surface depth image of the wafer, segment the polished surface depth image using a sliding window, with a window size of h×w and a step size of s, to obtain N segmented images; perform point-by-point segmentation and position encoding on the segmented image to obtain pixel depth data, and extract the foreground area depth value in the pixel depth data and background area depth value The differences in mean depth values were analyzed using a two-sample t test;
[0014] Assumptions:
[0015] H0: There is no significant difference in the average depth values of the foreground area and the background area; H1: There is a significant difference in the average depth values of the foreground area and the background area;
[0016] Calculate the average depth value of the foreground area and standard deviation s f; Calculate the average depth value of the background area and standard deviation s b ;
[0017] Calculate the t statistic using the formula:
[0018]
[0019] Among them, f n represents the number of samples in the foreground area, b n represents the number of samples in the background area; according to the degree of freedom, the t distribution table is searched, and the degree of freedom calculation formula is:
[0020]
[0021] Wherein, v represents the degree of freedom; determine the corresponding p value, if the p value is less than the preset significance level α, reject H0, it is considered that there is a significant difference between the average depth values of the foreground area and the background area, and the segmented image does not meet the standard; if the p value is greater than the preset significance level α, accept H0, it is considered that there is no significant difference between the average depth values of the foreground area and the background area, and the segmented image meets the standard;
[0022] Perform hypothesis testing on all the segmented images to obtain the number M of segmented images that meet the standard, and calculate the wafer flatness index using the formula:
[0023]
[0024] Wherein, r1 represents the wafer flatness index.
[0025] A wafer thickness uniformity calculation module, used to obtain the thickness data of the wafer, segment the thickness data to obtain segmented data, calculate the average thickness and standard deviation of the segmented data, calculate the uniformity coefficient according to the average thickness and the standard deviation, judge whether the segmented data meets the thickness uniformity standard according to the uniformity coefficient, judge whether all the segmented data groups meet the thickness uniformity standard, and calculate the wafer thickness uniformity index according to the compliance ratio;
[0026] Further, calculating the wafer thickness uniformity index includes:
[0027] Obtain the thickness data of the wafer, and segment the thickness data to obtain N segmentation data groups {h i}, where i = 1, 2, ..., X; Calculate the mean of the segmented data set, the formula is:
[0028]
[0029] in, represents the mean of the segmented data set, and X represents the number of samples in the segmented data set; the standard deviation of the segmented data set is calculated as follows:
[0030]
[0031] Wherein, σ represents the standard deviation of the segmented data set; the uniformity coefficient of the segmented data set is calculated as follows:
[0032]
[0033] Wherein, c represents the uniformity coefficient; if the uniformity coefficient is less than the preset threshold value C, the thickness uniformity of the segmented data group meets the standard; the uniformity coefficient is calculated for all the segmented data groups and judged to obtain the number P of the segmented data groups that meet the standard; the wafer thickness uniformity index is calculated, and the formula is:
[0034]
[0035] Wherein, r2 represents the wafer thickness uniformity index.
[0036] Furthermore, the thickness data is uniformly collected on the wafer, and data segmentation is performed on the thickness data according to the position area of the segmented image.
[0037] A wafer edge deviation calculation module is used to obtain image data of the wafer, process the image data, and obtain a wafer contour shape; overlap the wafer contour shape with a standard wafer shape at a specified position, obtain an error set, and calculate a wafer edge deviation index;
[0038] Further, calculating the wafer edge deviation index includes:
[0039] The image data of the wafer is acquired, and the image data is processed to obtain the wafer contour shape; the standard wafer shape is composed of line segments and major arcs; the wafer contour shape and the standard wafer shape are partially overlapped with the line segment, and the midpoints of the line segments of the wafer contour shape and the standard wafer shape are overlapped at the same position; T points are uniformly selected on the major arc of the standard wafer shape to obtain a point set {t i}, where i = 1, 2, ..., T; starting from the center of the superior arc of the standard wafer shape, the direction is t i The ray and the wafer outline shape have intersection points, and the point set {t i ′}, where i = 1, 2, ..., T; calculate the point set {t i} and the point set {t iThe Euclidean distance between the corresponding points of ′} is as follows:
[0040] l i =||t i -t i ′||;
[0041] Among them, l i Represents the error value; the error set A={l1,l2,…,l T}; If the maximum value max(A) in the error set is greater than the preset error threshold L, the wafer contour shape does not meet the standard; If the maximum value max(A) in the error set is less than or equal to the preset error threshold L, the mean of the error set is calculated, and the formula is:
[0042]
[0043] Among them, l i represents the error value, Represents the error mean, T represents the number of data; calculate the mean square error of the error set, the formula is:
[0044]
[0045] Wherein, MSE represents mean square error, l represents the error mean; the wafer edge deviation index is calculated to determine whether the wafer contour shape meets the standard, and the formula is:
[0046]
[0047] Wherein, r3 represents the wafer edge deviation index, ranging from 0 to 100, and α represents , β represents the weight parameter of MSE, and k1 represents the adjustment parameter, which is used to control the influence of the mean square error on the wafer edge deviation index.
[0048] A wafer defect density detection module is used to obtain image data of the wafer, extract features from the image data, detect defect size and defect position, calculate defect density, and calculate a wafer defect density index based on the defect density;
[0049] Further, calculating the wafer defect density index includes:
[0050] Acquire the image data of the wafer, denoise the image data to obtain denoised image data; use the Canny edge detection operator to extract image edges from the denoised image data to obtain the number of defective areas; calculate the circumscribed rectangle that can completely surround the defective area to obtain the area of the defective area;
[0051] Further, if the number of defective regions is greater than a preset threshold value E, the quality of the wafer does not meet the standard; if the area of the defective region is greater than a preset threshold value S, the quality of the wafer does not meet the standard; if the number of defective regions is less than or equal to the preset threshold value E and the area of the defective region is less than or equal to the preset threshold value S, the defect density ρ is calculated, and the defect density index of the wafer is calculated;
[0052] When the number of defective regions is less than a preset threshold value E and the area of the defective region is less than a preset threshold value S, the defect density ρ is calculated; the wafer defect density index is designed according to the defect density to determine whether the defects of the wafer meet the quality standards, and the formula is:
[0053]
[0054] Among them, r4 represents the wafer defect density index, ranging from 0 to 100, ρ represents the defect density, and k2 represents an adjustment parameter used to control the influence of the defect density on the wafer defect density index.
[0055] A wafer quality comprehensive evaluation module, used to obtain the wafer flatness index, the thickness uniformity index, the edge deviation index and the defect density index, calculate the comprehensive quality evaluation score of the wafer, and classify the wafer according to a preset quality classification standard;
[0056] Furthermore, the calculation of the comprehensive evaluation score of wafer quality includes:
[0057] The wafer flatness index, the wafer thickness uniformity index, the wafer edge deviation index and the wafer defect density index are obtained, and a comprehensive quality evaluation system is designed, and the formula is:
[0058] r=ω1r1+ω2r2+ω3r3+ω4r4;
[0059] Among them, r represents the comprehensive quality evaluation score of the wafer, ω1 represents the weight coefficient of r1, ω2 represents the weight coefficient of r2, ω3 represents the weight coefficient of r3, and ω4 represents the weight coefficient of r4.
[0060] Furthermore, the weight coefficients satisfy the formula:
[0061] ω1+ω2+ω3+ω4=1;
[0062] Used to adjust the parameter size of different weight coefficients;
[0063] Obtain a comprehensive quality evaluation score of the wafer, classify the wafer according to a preset quality classification standard, and obtain a first classification result.
[0064] Furthermore, the preset quality classification standards include:
[0065] When the comprehensive quality evaluation score of the wafer is less than a preset threshold value Q, the quality of the wafer is unqualified; when the comprehensive quality evaluation score of the wafer is greater than or equal to the preset threshold value Q, the wafer is divided into five grades of S, A, B, C and D according to the comprehensive quality evaluation score.
[0066] The wafer sorting module is used to receive the first classification result and perform sorting.
[0067] Furthermore, information of the unqualified wafers is recorded and the wafers are reprocessed to obtain processed wafers; and comprehensive quality evaluation and sorting are performed on the processed wafers.
[0068] Compared with the prior art, the present invention has the following beneficial effects:
[0069] 1. The sliding window method is used to segment the image, which can quickly process large-area images without losing image details. Each window can be processed independently and in parallel, which significantly improves the processing speed. The statistical method is used to perform hypothesis testing on the pixel depth data, which can effectively determine the flatness of the segmented image, distinguish normal areas from abnormal areas, and reduce the possibility of misjudgment and missed judgment. By performing point-by-point segmentation and hypothesis testing on all segmented images, the entire polished surface is comprehensively inspected, avoiding the deviation caused by relying only on data from some areas.
[0070] 2. A novel method for judging whether the wafer contour shape meets the standards has been designed. This method has extremely high adaptability and can flexibly cope with the contour detection of wafers of different shapes and sizes. In addition to judging whether the wafer meets the standards, this method can also provide detailed contour deviation data and shape error distribution, etc. By analyzing these data, it is possible to understand where the shape deviation exists on the wafer, so as to make targeted process adjustments. Based on the error data, a wafer edge deviation index has been designed to achieve efficient and accurate detection of the wafer contour shape.
[0071] 3. By introducing the wafer flatness index, wafer thickness uniformity index, wafer edge deviation index and wafer defect density index, a comprehensive wafer quality evaluation system is constructed. The multi-dimensional evaluation method avoids the errors that may be caused by a single indicator, and more comprehensively and accurately evaluates the overall quality of the wafer, providing rich data support for process improvement and production decisions, which helps to improve wafer production efficiency and product quality; according to the preset quality classification standards, through the automated classification and sorting system, wafers can be quickly identified and classified, improving sorting efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] Figure 1 A schematic diagram of the structure of a standard single crystal silicon wafer provided in an embodiment of the present invention;
[0073] Figure 2 A schematic diagram of the structure of the outline shape of a single crystal silicon wafer to be inspected provided in an embodiment of the present invention;
[0074] Figure 3 A schematic diagram of the overlapping structure of a standard single crystal silicon wafer and a single crystal silicon wafer to be tested provided in an embodiment of the present invention;
[0075] Figure 4 A schematic diagram of the intersection of the ray and the outline of the single crystal silicon wafer to be tested and the intersection of the ray and the outline of the standard single crystal silicon wafer;
[0076] Figure 5 A schematic diagram of the structure of a wafer inspection and sorting system provided in an embodiment of the present invention.
[0077] In the figure: 1. The midline segment of the standard single crystal silicon wafer shape; 2. The major arc of the standard single crystal silicon wafer shape; 3. The ray from the center of the circle to the point on the major arc in the standard single crystal silicon wafer shape; 4. The center of the circle corresponding to the major arc in the intersection of the outline of the single crystal silicon wafer to be tested; 5. The midpoint of the midline segment of the standard single crystal silicon wafer shape; 6. The intersection of the ray and the outline of the single crystal silicon wafer to be tested; 7. Uniformly selected points on the major arc in the standard single crystal silicon wafer shape. DETAILED DESCRIPTION
[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0079] Example 1
[0080] This embodiment applies a wafer inspection and sorting system proposed by the present invention to the sorting process of 6-inch single crystal silicon wafers, specifically including:
[0081] The wafer flatness calculation module is used to obtain the polished surface depth image of a 6-inch single-crystal silicon wafer, segment the polished surface depth image to obtain a segmented image; segment and position encode the segmented image point by point to obtain pixel depth data, perform hypothesis testing on the pixel depth data to determine whether the segmented image meets the flatness requirements; determine whether all the segmented images meet the flatness requirements, and obtain a wafer flatness index according to the compliance ratio;
[0082] Furthermore, the wafer flatness calculation module includes:
[0083] Obtain the polished surface depth image of the 6-inch single crystal silicon wafer, segment the polished surface depth image using a sliding window method, with a window size of h×w and a step size of s, to obtain N segmented images; perform point-by-point segmentation and position encoding on the segmented image to obtain pixel depth data, and extract the foreground area depth value in the pixel depth data and background area depth value Two-sample t-test was used to analyze the differences in mean depth values;
[0084] Assumptions:
[0085] H0: There is no significant difference in the average depth values between the foreground area and the background area; H1: There is a significant difference in the average depth values between the foreground area and the background area;
[0086] Calculate the average depth value of the foreground area and standard deviation s f ; Calculate the average depth value of the background area and standard deviation s b ;
[0087] Calculate the t statistic using the formula:
[0088]
[0089] Among them, f n represents the number of samples in the foreground area, b n Represents the number of samples in the background area; find the t distribution table according to the degree of freedom, and the degree of freedom calculation formula is:
[0090]
[0091] Wherein, v represents the degree of freedom; determine the corresponding p value, if the p value is less than the preset significance level α, reject H0, it is considered that there is a significant difference in the average depth value of the foreground area and the background area, and the segmented image does not meet the standard; if the p value is greater than the preset significance level α, accept H0, it is considered that there is no significant difference in the depth value of the foreground area and the background area, and the segmented image meets the standard;
[0092] Perform hypothesis testing on all the segmented images to obtain the number M of segmented images that meet the standard, and calculate the wafer flatness index using the formula:
[0093]
[0094] Among them, r1 represents the wafer flatness index. Through calculation, the wafer flatness index of a 6-inch single-crystal silicon wafer is 97. A statistical method is used to perform hypothesis testing on pixel depth data to effectively judge flatness and reduce the possibility of misjudgment and missed judgment; through point-by-point segmentation and hypothesis testing, the entire polished surface is fully tested to avoid deviations caused by relying only on data from a partial area.
[0095] A wafer thickness uniformity calculation module is used to obtain the thickness data of the 6-inch single crystal silicon wafer, segment the thickness data to obtain segmented data groups, calculate the average thickness and standard deviation of the segmented data groups, calculate the uniformity coefficient according to the average thickness and the standard deviation, judge whether the segmented data groups meet the thickness uniformity requirements according to the uniformity coefficient, judge whether all the segmented data groups meet the thickness uniformity requirements, and obtain a wafer thickness uniformity index according to the compliance ratio;
[0096] Furthermore, the thickness uniformity calculation module includes:
[0097] Obtain the thickness data of the 6-inch single crystal silicon wafer, and segment the thickness data to obtain N segmentation data groups {h i}, where i = 1, 2, ..., X; Calculate the mean of the segmented data set, the formula is:
[0098]
[0099] in, represents the mean of the segmented data set, and X represents the number of samples in the segmented data set; the standard deviation of the segmented data set is calculated as follows:
[0100]
[0101] Wherein, σ represents the standard deviation of the segmented data set; the uniformity coefficient of the segmented data set is calculated as follows:
[0102]
[0103] Wherein, c represents the uniformity coefficient; if c is less than a preset threshold value C, the thickness uniformity of the segmented data group meets the standard;
[0104] Calculate the uniformity coefficients of all the segmented data groups and determine to obtain P segmented data groups that meet the standards; calculate the wafer thickness uniformity index, the formula is:
[0105]
[0106] Wherein, r2 represents the wafer thickness uniformity index.
[0107] Table 1 is a set of 6-inch monocrystalline silicon wafer thickness data, calculate the average thickness Standard deviation σ=0.35 (two significant figures have been retained), uniformity coefficient (Two significant figures have been retained); The results show that the mean of this set of single crystal silicon wafer thickness data is 625.0, the standard deviation is 0.35, and the uniformity coefficient is 0.06%, showing very high uniformity and consistency.
[0108] Table 1. Thickness data of 6-inch single crystal silicon wafer
[0109] serial number Thickness h(um) serial number Thickness h(um) 1 624.3 9 625.3 2 625.7 10 625.5 3 624.8 11 624.7 4 625.2 12 625.1 5 625.0 13 624.8 6 624.6 14 625.0 7 625.1 15 624.6 8 624.9
[0110] The final wafer thickness uniformity index of the 6-inch single crystal silicon wafer is 97. Using the uniformity coefficient, the uniformity problem of the thickness segmentation data group can be accurately identified and evaluated; by evaluating all segmentation data groups, it is possible to comprehensively determine whether the overall thickness of the 6-inch single crystal silicon wafer is uniform.
[0111] Furthermore, the thickness data of the 6-inch single crystal silicon wafer is uniformly collected on the wafer, and the thickness data is segmented in the area where the segmented image is located according to the segmentation method of the depth image to ensure the consistency of the segmented area, that is, each segmented area corresponds to the same physical area in the thickness data, ensuring that the regional division of thickness measurement is more scientific and reasonable.
[0112] A wafer edge deviation calculation module is used to obtain image data of the 6-inch single crystal silicon wafer, process the image data, and obtain a wafer contour shape; overlap the wafer contour shape with a standard wafer shape at a specified position, calculate the degree of deviation between the wafer contour shape and the standard wafer shape, and obtain a wafer edge deviation index;
[0113] Furthermore, the wafer edge deviation calculation module includes:
[0114] The image data of a 6-inch single crystal silicon wafer is obtained, and the image data is processed to obtain the outline shape of the wafer. The structure of the outline shape of the single crystal silicon wafer to be detected is as follows: Figure 2 As shown; the standard wafer shape is composed of line segment 1 and major arc 2, and the structure of the standard single crystal silicon wafer shape is as follows Figure 1 As shown; the wafer outline shape and the standard wafer shape are partially overlapped by line segment 1, and the midpoint 5 of the line segment of the wafer outline shape and the standard wafer shape overlaps at the same position, and the overlap structure of the standard single crystal silicon wafer and the single crystal silicon wafer to be tested is as shown Figure 3 As shown; T points are uniformly selected on the superior arc of the standard wafer shape to obtain the point set {t i}, where i = 1, 2, ..., T; starting from the center 4 of the superior arc of the standard wafer shape, the direction is t i The ray 3 and the wafer outline have an intersection 6, the ray 3 and the outline of the single crystal silicon wafer to be tested have an intersection 6, and the ray and the outline of the standard single crystal silicon wafer have an intersection 7. Figure 4 As shown. We get the point set {t i ′}, where i = 1, 2, ..., T; calculate the point set {t i} and the point set {t i The Euclidean distance between the corresponding points of ′} is as follows:
[0115] l i =||t i -t i ′||;
[0116] Among them, l i Represents the error value; the error set A={l1,l2,…,l T}; If the maximum value max(A) in the error set is greater than the preset error threshold L, the wafer contour shape does not meet the standard; If the maximum value max(A) in the error set is less than or equal to the preset error threshold L, the mean of the error set is calculated, and the formula is:
[0117]
[0118] Among them, l i represents the error value, l represents the error mean, and T represents the number of data; the mean square error of the error set is calculated as follows:
[0119]
[0120] Among them, MSE represents mean square error, represents the error mean; the wafer edge deviation index is designed to determine whether the wafer contour shape meets the standard, and the formula is:
[0121]
[0122] Wherein, r3 represents the wafer edge deviation index, ranging from 0 to 100, and α represents , β represents the weight parameter of MSE, and k1 represents the adjustment parameter, which is used to control the influence of the mean square error on the wafer edge deviation index.
[0123] Table 2 is a set of measurement error data, and the average error is calculated. (Two significant figures have been retained), the mean square error MSE=868.18 (two significant figures have been retained), and the wafer edge deviation index of this set of error data is calculated to be 97 based on the wafer edge deviation function.
[0124] Table 2. Error set
[0125] serial number Error(um) serial number Error(um) 1 50.1 9 70.1 2 75.3 10 120.3 3 60.2 11 85.2 4 90.0 12 40.0 5 45.8 13 95.6 6 130.5 14 35.7 7 80.7 15 110.9 8 55.4
[0126] The wafer edge deviation index can not only determine whether the contour shape of the 6-inch single crystal silicon wafer meets the standard, but also provide detailed contour deviation data and shape error distribution, etc., which is of great reference value for process improvement and quality control.
[0127] A wafer defect density detection module is used to obtain image data of the 6-inch single crystal silicon wafer, extract features from the image data, detect defect size and defect position, and calculate defect density to obtain a wafer defect density index;
[0128] Furthermore, the wafer defect detection module includes:
[0129] Acquire image data of the 6-inch single crystal silicon wafer, denoise the image data to obtain denoised image data; use the Canny edge detection operator to extract image edges from the denoised image data to obtain the number of defective areas; calculate a circumscribed rectangle that can completely surround the defective area to obtain the area of the defective area;
[0130] Furthermore, if the number of defective regions is greater than a preset threshold value E, the quality of the wafer does not meet the standard; if the area of the defective region is greater than a preset threshold value S, the quality of the wafer does not meet the standard; if the number of defective regions is less than or equal to the preset threshold value E and the area of the defective region is less than or equal to the preset threshold value S, the defect density ρ is calculated, and the defect density index of the wafer is calculated. By pre-detecting the number and area of the defective regions, the quality of the wafer can be quickly judged to see whether it meets the quality standards, thereby achieving early screening and reducing the subsequent processing costs of defective products.
[0131] Furthermore, when the number of defective regions is less than or equal to a preset threshold value E and the area of the defective region is less than or equal to a preset threshold value S, the defect density ρ is calculated; and a wafer defect density index function is designed according to the defect density to measure whether the defects of the wafer meet the quality standards, and the formula is:
[0132]
[0133] Among them, r4 represents the wafer defect density index, ranging from 0 to 100, ρ represents the defect density, and k2 represents the adjustment parameter, which controls the degree of influence of the defect density on the score. The number of defect areas and the area of defect areas of 6-inch monocrystalline silicon wafers are both less than the preset threshold. After calculation, its defect density is 0.07, and the wafer defect density index is 96. By extracting features from image data, multi-dimensional defect analysis including defect size, shape, location, etc. can be performed to fully understand the defect characteristics; by calculating the defect density index, the defect degree of the wafer is quantitatively evaluated, and the overall quality of the wafer is quantified in the form of a defect density index.
[0134] A wafer quality comprehensive evaluation module, used to obtain the flatness index, thickness uniformity index, edge deviation index and defect density index of the 6-inch single crystal silicon wafer, calculate the comprehensive quality evaluation score of the wafer according to the comprehensive quality evaluation system of the wafer, and classify the wafer according to a preset quality classification standard;
[0135] Furthermore, the wafer quality comprehensive evaluation module includes:
[0136] The data of the wafer flatness index, the wafer thickness uniformity index, the wafer edge deviation index and the wafer defect density index of the 6-inch single crystal silicon wafer are obtained, and a comprehensive quality evaluation system of the wafer is designed, and the formula is:
[0137] r=ω1r1+ω2r2+ω3r3+ω4r4;
[0138] Among them, r represents the comprehensive quality evaluation score of the wafer, ω1 represents the weight coefficient of r1, ω2 represents the weight coefficient of r2, ω3 represents the weight coefficient of r3, and ω4 represents the weight coefficient of r4. After calculation, the final comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is 97. By introducing a comprehensive evaluation system of wafer flatness index, wafer thickness uniformity index, wafer edge deviation index and wafer defect density index, the overall quality of the wafer is evaluated more comprehensively and accurately, avoiding the errors that may be caused by a single indicator.
[0139] Furthermore, the weight coefficients satisfy the formula:
[0140] ω1+ω2+ω3+ω4=1;
[0141] Used to dynamically adjust the parameter sizes of different weight coefficients.
[0142] Obtain a comprehensive quality evaluation score for the 6-inch single crystal silicon wafer, and classify the wafer according to a preset quality classification standard.
[0143] Further, the quality classification criteria include:
[0144] When the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is between 95 and 100, the wafer is rated as S-grade; when the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is between 90 and 95, the wafer is rated as A-grade; when the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is between 85 and 90, the wafer is rated as B-grade; when the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is between 80 and 85, the wafer is rated as C-grade; when the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is between 75 and 80, the wafer is rated as D-grade; when the comprehensive quality evaluation score of the 6-inch monocrystalline silicon wafer is below 75, the wafer quality is unqualified. The 6-inch monocrystalline silicon wafer with a comprehensive quality evaluation score of 97 is rated as S-grade. Through the fine quality scoring range, wafers can be classified more finely, which helps enterprises to distinguish and manage wafers of different quality grades and maximize the use of resources.
[0145] The wafer sorting module is used to receive the classification results of the 6-inch single crystal silicon wafer quality comprehensive evaluation module and sort the wafers according to the classification results.
[0146] Furthermore, the information of the unqualified 6-inch single crystal silicon wafers is recorded and reprocessed to obtain processed wafers; the processed wafers are subjected to comprehensive quality evaluation and sorting to form a closed-loop feedback system to continuously improve and enhance the production quality of wafers.
[0147] The system contains multiple modules, which respectively detect the flatness, thickness uniformity, edge deviation and defect density of wafers. Each indicator is obtained through special calculation and detection methods. The comprehensive evaluation of multiple indicators can more accurately reflect the overall quality of the wafer; the wafer sorting module automatically sorts wafers according to the classification results of the comprehensive quality evaluation module, reducing manual intervention and improving sorting efficiency and accuracy; the preset quality classification standards make the sorting process more standardized and consistent, ensuring that each wafer is classified according to the same standards, improving production efficiency and product quality, reducing production costs, and enhancing the company's market competitiveness.
[0148] Example 2
[0149] In the modern semiconductor manufacturing process, the quality of wafers is directly related to the performance and reliability of the final product. In order to ensure the high-quality production of wafers and improve the overall production efficiency, a semiconductor company introduced a wafer inspection and sorting system proposed by the present invention. The structure is as follows: Figure 5The system consists of multiple modules, each of which performs different quality inspection tasks, and ultimately achieves accurate classification and sorting of wafers through comprehensive evaluation of various indicators. The following is a detailed description of the system and its specific benefits in practical applications.
[0150] The wafer flatness calculation module obtains the depth image of the polished surface of the wafer and segments it to obtain multiple segmented images. Then, each segmented image is segmented and position-encoded point by point to form pixel depth data. By performing hypothesis tests on these data groups, it can be determined whether each segmented image meets the flatness requirements. Finally, the flatness index of the wafer is calculated based on the compliance ratio of all segmented images. Through detection, the system found that 99% of the segmented images met the flatness requirements, and the calculated flatness index was 99.
[0151] The wafer thickness uniformity calculation module obtains the thickness data of the wafer and divides the data into multiple segmented data groups. The average thickness and standard deviation are calculated for each data group, and then the uniformity coefficient is calculated based on these two parameters. The thickness uniformity of each data group can be judged by the uniformity coefficient, and the thickness uniformity index of the wafer can be calculated based on the compliance ratio. The system detected an average thickness of 625.0 microns, a standard deviation of 0.35 microns, a corresponding uniformity coefficient of approximately 0.06%, and 98% of the segmented data groups met the thickness uniformity requirements, and the final thickness uniformity index was 98.
[0152] The wafer edge deviation calculation module obtains the image data of the wafer and processes it to obtain the outline shape of the wafer. By overlapping the actual outline with the standard outline at the specified position, the degree of deviation between the two is calculated to obtain the wafer edge deviation index. The system measures that the edge deviation of the wafer is within the range of ±200 microns, and the deviation index is calculated to be 95.
[0153] The wafer defect density detection module extracts features from the wafer image data, detects the size and location of the defects, and calculates the defect density to obtain the defect density index of the wafer. The system detects that the defect density per square centimeter is 0.02, and the defect density index is calculated to be 99.
[0154] The wafer quality comprehensive evaluation module obtains the output data of the above modules, calculates the comprehensive quality evaluation score of the wafer through the comprehensive quality evaluation system, and classifies the wafer according to the preset quality classification standards. The comprehensive evaluation score of a wafer is 98 points, which is classified as an S-level product.
[0155] Finally, the wafer sorting module receives the classification results of the comprehensive quality evaluation module and automatically sorts them according to the classification results. Unqualified wafers will be recorded and reprocessed to ensure that the final product meets high quality standards. Of the 200 wafers detected by the system, 190 meet the A-level standard and 10 need to be reprocessed.
[0156] By introducing this wafer inspection and sorting system, a semiconductor company has achieved significant practical benefits:
[0157] The automated inspection and sorting system has greatly reduced manual operation time and increased overall production efficiency by 30%. The time required for manual inspection and sorting was originally 20 hours, but now it has been shortened to 14 hours. Through comprehensive quality inspection and precise sorting, it is ensured that each wafer meets strict quality standards. The qualified rate in the production process has increased from 95.45% to 97.02%, which has significantly improved the quality of the final product. Due to the reduction of scrap rate and resource conservation, the production cost has been reduced by 5.09%. The information recording and analysis of unqualified wafers helps to trace quality problems, optimize production processes, and reduce future quality problems. Through traceability analysis every month, several production links have been improved, further improving product quality.
[0158] In summary, the wafer inspection and sorting system has significantly improved production efficiency and product quality through precise inspection, evaluation and sorting, bringing considerable economic benefits and market competitive advantages to semiconductor companies.
[0159] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A wafer inspection and sorting system, characterized in that: include: A wafer flatness calculation module is used to obtain a polished surface depth image of the wafer, and segment the polished surface depth image to obtain a segmented image; The segmented image is segmented point by point and position-encoded to obtain pixel depth data, a hypothesis test is performed on the pixel depth data to determine whether all the segmented images meet the flatness standard, and a wafer flatness index is calculated according to the compliance ratio; A wafer thickness uniformity calculation module, used to obtain the thickness data of the wafer and calculate the wafer thickness uniformity index; A wafer edge deviation calculation module is used to obtain image data of the wafer, process the image data, and obtain a wafer contour shape; overlap the wafer contour shape with a standard wafer shape at a specified position, obtain an error set, and calculate a wafer edge deviation index; A wafer defect density detection module is used to obtain image data of the wafer, extract features from the image data, detect defect size and position, calculate defect density, and calculate a wafer defect density index based on the defect density; A wafer quality comprehensive evaluation module, used to obtain the wafer flatness index, the wafer thickness uniformity index, the wafer edge deviation index and the wafer defect density index, calculate the comprehensive quality evaluation score, and classify the wafer according to a preset quality classification standard to obtain a first classification result; A wafer sorting module is used for sorting according to the first classification result.
2. A wafer inspection and sorting system according to claim 1, characterized in that: Calculation of wafer flatness index includes: Acquire the polished surface depth image of the wafer, segment the polished surface depth image using a sliding window, with a window size of h×w and a step size of s, to obtain N segmented images; perform point-by-point segmentation and position encoding on the segmented image to obtain the pixel point depth data, and extract the foreground area depth value in the pixel point depth data and background area depth value The differences in mean depth values were analyzed using a two-sample t test; Assumptions: H0: There is no significant difference in the average depth values of the foreground area and the background area; H1: There is a significant difference in the average depth values of the foreground area and the background area; Calculate the average depth value of the foreground area and standard deviation s f ; Calculate the average depth value of the background area and standard deviation s b ; Calculate the t statistic using the formula: Among them, f n represents the number of samples in the foreground area, b n represents the number of samples in the background area; according to the degree of freedom, the t distribution table is searched, and the degree of freedom calculation formula is: Wherein, v represents the degree of freedom; determine the corresponding p value, if the p value is less than the preset significance level α, reject H0, it is considered that there is a significant difference between the average depth values of the foreground area and the background area, and the segmented image does not meet the standard; if the p value is greater than the preset significance level α, accept H0, it is considered that there is no significant difference between the average depth values of the foreground area and the background area, and the segmented image meets the standard; Perform hypothesis testing on all the segmented images to obtain the number M of segmented images that meet the standard, and calculate the wafer flatness index using the formula: Wherein, r1 represents the wafer flatness index.
3. A wafer inspection and sorting system according to claim 1, characterized in that: Calculation of wafer thickness uniformity index includes: Obtain the thickness data of the wafer, and segment the thickness data to obtain N segmentation data groups {h i }, where i = 1, 2, ..., X; Calculate the mean of the segmented data set using the formula: in, represents the mean of the segmented data set, and X represents the number of samples in the segmented data set; the standard deviation of the segmented data set is calculated as follows: Wherein, σ represents the standard deviation of the segmented data set; the uniformity coefficient of the segmented data set is calculated as follows: Wherein, c represents the uniformity coefficient; if the uniformity coefficient is less than the preset threshold value C, the thickness uniformity of the segmented data group meets the standard; the uniformity coefficient is calculated for all the segmented data groups and judged to obtain the number P of the segmented data groups that meet the standard; the wafer thickness uniformity index is calculated, and the formula is: Wherein, r2 represents the wafer thickness uniformity index.
4. The wafer inspection and sorting system according to claim 1, characterized in that: Calculation of wafer edge deviation index includes: The image data of the wafer is acquired, and the image data is processed to obtain the wafer contour shape; the standard wafer shape is composed of line segments and major arcs; the wafer contour shape and the standard wafer shape are partially overlapped with the line segment, and the midpoints of the line segments of the wafer contour shape and the standard wafer shape are overlapped at the same position; T points are uniformly selected on the major arc of the standard wafer shape to obtain a point set {t i }, where i = 1, 2, ..., T; starting from the center of the superior arc of the standard wafer shape, the direction is t i The ray and the wafer outline shape have intersection points, and the point set {t i ′}, where i = 1, 2, ..., T; calculate the point set {t i } and the point set {t i The Euclidean distance between the corresponding points of ′} is as follows: l i =||t i -t i ′||; Among them, l i Represents the error value; the error set A={l1,l2,…,l T }; If the maximum value max(A) in the error set is greater than the preset error threshold L, the wafer contour shape does not meet the standard; If the maximum value max(A) in the error set is less than or equal to the preset error threshold L, the mean of the error set is calculated, and the formula is: Among them, l i represents the error value, Represents the error mean, T represents the number of data; calculate the mean square error of the error set, the formula is: Wherein, MSE represents the mean square error, l represents the error mean; the wafer edge deviation index is calculated to determine whether the wafer contour shape meets the standard, and the formula is: Among them, r3 represents the wafer edge deviation index, ranging from 0 to 100, α represents the weight parameter of l, β represents the weight parameter of MSE, and k1 represents the adjustment parameter used to control the influence of the mean square error on the wafer edge deviation index.
5. The wafer inspection and sorting system according to claim 1, characterized in that: Calculation of wafer defect density index includes: Acquire the image data of the wafer, denoise the image data to obtain denoised image data; use the Canny edge detection operator to extract image edges from the denoised image data to obtain the number of defective areas; calculate the circumscribed rectangle that can completely surround the defective area to obtain the area of the defective area; Calculate the defect density ρ; design the wafer defect density index according to the defect density to determine whether the defects of the wafer meet the standards. The formula is: Among them, r4 represents the wafer defect density index, ranging from 0 to 100, ρ represents the defect density, and k2 represents an adjustment parameter used to control the influence of the defect density on the wafer defect density index.
6. The wafer inspection and sorting system according to claim 1, characterized in that: Calculation of comprehensive wafer quality evaluation score includes: The wafer flatness index, the wafer thickness uniformity index, the wafer edge deviation index and the wafer defect density index are obtained, and a comprehensive quality evaluation system is designed, and the formula is: r=ω1r1+ω2r2+ω3r3+ω4r4; Wherein, r represents the comprehensive quality evaluation score, ranging from 0 to 100, ω1 represents the weight coefficient of r1, ω2 represents the weight coefficient of r2, ω3 represents the weight coefficient of r3, and ω4 represents the weight coefficient of r4; the weight coefficients satisfy the formula: ω1+ω2+ω3+ω4=1; Among them, ω1, ω2, ω3 and ω4 are used to adjust the parameter sizes of different weight coefficients; The comprehensive quality evaluation score is obtained, and the wafers are classified according to the preset quality classification standard.
7. The wafer inspection and sorting system according to claim 1, characterized in that: The preset quality classification standards include: If the comprehensive quality evaluation score of the wafer is less than the preset threshold Q, the quality of the wafer is unqualified; if the comprehensive quality evaluation score of the wafer is greater than or equal to the preset threshold Q, the wafer is divided into five grades: S, A, B, C and D according to the comprehensive quality evaluation score.
8. The wafer inspection and sorting system according to claim 1, characterized in that: include: The thickness data is uniformly collected on the wafer, and data segmentation is performed on the thickness data according to the position area of the segmented image.
9. The wafer inspection and sorting system according to claim 1, characterized in that: include: Recording information of the unqualified wafers and reprocessing them to obtain processed wafers; The processed wafers are subjected to comprehensive quality evaluation and sorting.
10. The wafer inspection and sorting system according to claim 5, characterized in that: include: If the number of defective areas is greater than the preset threshold E, the quality of the wafer does not meet the standard; if the area of the defective area is greater than the preset threshold S, the quality of the wafer does not meet the standard; if the number of defective areas is less than or equal to the preset threshold E and the area of the defective area is less than or equal to the preset threshold S, calculate the defect density ρ and calculate the wafer defect density index.
Citation Information
Cited By
Silicon wafer microdefect testing method and system
CN120927662A
