Wafer surface contamination defect detection method and device

Abstract

The application relates to the technical field of wafer detection, and particularly provides a wafer surface pollution defect detection method and device. Dark field shooting is performed on a target wafer to obtain a surface dark field scattering image of the wafer. Each pixel in the dark field image is traversed, an SNR estimation graph is calculated through setting of a neighborhood and SNR estimation, global threshold segmentation is performed on the SNR estimation graph through a double threshold principle to obtain a low threshold binary graph and a high threshold binary graph. Connected regions in the low threshold binary graph are traversed. If at least one pixel in the connected region is marked as a defect in the high threshold binary graph, all pixels in the connected region are marked as real defects; otherwise, the pixels are marked as false defects. Under the conditions of large background noise dynamic range, small number of defect pixels and weak defect signals, the application can stably detect defects and defect outlines.

Description

Technical Field

[0001] This invention belongs to the field of wafer inspection technology, and particularly relates to a method and apparatus for detecting surface contamination defects on wafers. Background Technology

[0002] Wafer surface contamination defect detection is a crucial part of the semiconductor manufacturing process. Its main purpose is to detect contamination on the wafer surface to ensure the quality of chip manufacturing and the performance of the final product. Contamination defects include particle contamination, scratches, cracks, etc., which can have a significant impact on chip manufacturing and the performance of the final product.

[0003] Dark-field inspection is a method for detecting defects on wafer surfaces. It works by using a laser beam to illuminate the wafer surface from above and to the side, then a detector collects the scattered light and converts it into an image signal. Because reflected light from the wafer surface is avoided, only the scattered light from defects reaches the detector, making defective areas appear brighter and defect-free areas appear darker. Due to its high sensitivity to scattered light, dark-field inspection is highly effective for detecting minute surface defects.

[0004] A major challenge in current dark-field inspection processes is the varying cleanliness standards and yield requirements of wafer surfaces, resulting in a wide range of defect and contaminant sizes to be detected. This places high demands on the sensitivity limits of inspection systems and algorithms. Furthermore, the diversity of wafer fabrication processes further increases the difficulty of inspection. While dark-field imaging technology can obtain images of wafer scattering intensity, the background signal (haze) varies considerably. This large dynamic range often masks subtle defects and contaminant signals on the wafer surface, significantly interfering with accurate detection.

[0005] In industrial production environments, detection algorithms not only need to possess high stability but also need to balance efficiency and yield standards. This undoubtedly places more stringent demands on the design and optimization of algorithms. Therefore, there is an urgent need for a wafer surface contamination defect detection method that can effectively solve the problems of complex background signals and weak defect signals in dark field detection. Summary of the Invention

[0006] In view of this, the present invention aims to provide a wafer surface contamination defect detection method that uses the neighborhood SNR (signal-to-noise ratio) estimation principle to accurately estimate the background signal intensity around the current signal, effectively segment and distinguish the background signal and the defect signal, and then uses a set of high and low thresholds for dual threshold segmentation to improve detection stability and enhance the ability to detect minute defects and contaminants on the wafer surface.

[0007] To achieve the above objectives, the technical solution created by this invention is implemented as follows:

[0008] The present invention provides a method for detecting contamination defects on the wafer surface, comprising: for the SNR estimation map of the wafer dark field image, setting at least two thresholds to segment it, marking pixels with gray values ​​greater than the thresholds as defects, and marking the remaining pixels as normal, thereby forming multiple binary images corresponding to different thresholds;

[0009] Combine any two thresholds;

[0010] In each threshold combination, traverse the connected regions in the low-threshold binary image. If there is at least one pixel in the connected region that is marked as a defect in the high-threshold binary image, then mark all pixels in the connected region as real defects; otherwise, mark them as pseudo defects.

[0011] Preferably, dark field imaging is performed on the target wafer to obtain a dark field image of the wafer surface.

[0012] Preferably, the method for obtaining the SNR estimation map is as follows:

[0013] Iterate through each pixel in the dark field image and select an n×n neighborhood containing that pixel to calculate the SNR estimate for each pixel.

[0014] Preferably, the method for obtaining the SNR estimation map is as follows:

[0015] Iterate through each pixel in the dark field image and select an n×m neighborhood containing that pixel to calculate the SNR estimate for each pixel.

[0016] Preferably, the formula for calculating the SNR estimate is:

[0017] ;

[0018] in, This represents the estimated SNR value of a pixel. Represents the grayscale value of a pixel in a dark-field image. This represents the mean background noise in the pixel neighborhood. This represents the amplitude of background noise in the pixel neighborhood. This represents the maximum background noise value in the pixel neighborhood. This represents the minimum background noise value in the pixel neighborhood.

[0019] Preferably, the average background noise of the pixel neighborhood The value can be: the median gray value of all pixels in the pixel neighborhood, or the average gray value of all pixels in the pixel neighborhood, or calculated using a quadratic std(standard deviation) mean estimation algorithm. .

[0020] Preferably, the quadratic std mean estimation algorithm is used to calculate... The steps are as follows:

[0021] Calculate the average gray value and standard deviation of all pixels in the neighborhood of a pixel;

[0022] Remove pixels smaller than (mean - Theta × standard deviation) and pixels larger than (mean + Theta × standard deviation), where Theta is a set constant;

[0023] Calculate the average grayscale value of the remaining pixels and use this average grayscale value as the mean background noise value of the pixel neighborhood. .

[0024] Preferably, the background noise amplitude of the pixel neighborhood The value can be: (median gray value of the pixel neighborhood - minimum gray value of the pixel neighborhood) × 2, or calculated using the 6std algorithm. .

[0025] Preferably, the 6std algorithm is used to calculate. The steps are as follows:

[0026] Calculate the standard deviation of gray levels of all pixels in the neighborhood of a pixel;

[0027] =6 × grayscale standard deviation.

[0028] Preferably, increasing the size of the neighborhood window improves the accuracy of the SNR estimate.

[0029] Preferably, the method for enlarging or shrinking the neighborhood window size is as follows:

[0030] The aspect ratio of the neighborhood window is fixed and it is expanded or reduced proportionally.

[0031] Alternatively, the aspect ratio of the neighboring window may be expanded or reduced non-proportionally.

[0032] Preferably, when two thresholds are set to segment the SNR estimation map, the pixels marked as real defects in the low-threshold binary map are the defects on the wafer surface.

[0033] Preferably, when setting more than two thresholds to segment the SNR estimation map, each threshold combination is marked to obtain a binary map that completes the marking of real defects and pseudo defects; all threshold combinations are subjected to pixel OR operation to obtain the final binary map, and the pixels marked as real defects in the final binary map are the defects on the wafer surface.

[0034] Preferably, in each threshold combination, the high threshold is greater than or equal to 3, and the low threshold is less than or equal to 2.

[0035] A wafer surface contamination defect detection device includes a wafer surface contamination defect detection module, which uses a wafer surface contamination defect detection method to detect contamination defects.

[0036] An electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor.

[0037] The memory stores instructions that can be executed by at least one processor. When the instructions are executed by at least one processor, the at least one processor performs a wafer surface contamination defect detection method.

[0038] A computer-readable storage medium storing computer instructions, which, when executed by a processor, cause the processor to perform a method for detecting contamination defects on a wafer surface.

[0039] Compared with the prior art, the present invention can achieve the following beneficial effects:

[0040] This invention employs the neighborhood SNR estimation principle and optimizes the SNR estimation formula to accurately estimate the background signal intensity in the neighborhood of each pixel. This effectively segments and distinguishes between background signals and defect signals. Furthermore, the dual-threshold calibration method improves the ability to identify extremely weak defect intensity signals in dark-field images. Even under conditions of small defect size and low image signal-to-noise ratio, it can accurately and stably detect minute defects and contaminants on the wafer surface. In addition, the method of this invention has good robustness. In complex environments with a large dynamic range of background signals, it can still stably detect defect signals with small pixel size and weak signal-to-noise ratio, and stably detect defect contours, effectively overcoming the influence of background noise on the detection results.

[0041] The method of this invention has certain advantages in defect detection efficiency and meets the comprehensive requirements of industrial production for the stability, efficiency and yield of detection algorithms.

[0042] Furthermore, the parameter values ​​in the SNR estimation formula proposed in this invention can be adjusted to varying degrees according to actual application conditions, for example... Regarding the value of , the gray average value or the second std mean estimation result can be used to replace the gray median value, making it more suitable for situations where the dynamic range of the original background is small and the algorithm efficiency requirement is high. The value can also be obtained using the 6std algorithm, making it more suitable for situations where the dynamic range of the original image background is small and the algorithm efficiency is required to be high. Furthermore, the size of the neighborhood window can be set according to the calculation efficiency, defect contour type, etc., and can be reduced by means of proportional or non-proportional methods to improve the detection accuracy while minimizing the reduction in detection efficiency.

[0043] Based on the dual-threshold segmentation design, it can be modified into multi-threshold logic with different segments according to the defect contour type and light intensity distribution, further improving the defect detection capability in complex scenarios. Attached Figure Description

[0044] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0045] Figure 1 This is a flowchart of a wafer surface contamination defect detection method provided according to an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram illustrating the principle of SNR estimation calculation according to an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of a strip-shaped defect according to an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of defects on a wafer dark field image provided according to an embodiment of the present invention;

[0049] Figure 5 This is an SNR estimation map of a wafer dark field image provided according to an embodiment of the present invention;

[0050] Figure 6 It is a mask diagram provided according to an embodiment of the present invention;

[0051] Figure 7 This is a maker diagram provided according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and do not constitute a limitation thereof. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, some operations related to the invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0053] It should be noted that, unless otherwise specified, the embodiments and features described in this invention can be combined to form various implementations. Furthermore, the order of the steps or actions in the method description can be changed or adjusted in a manner readily apparent to those skilled in the art. Therefore, the various orders in the specification and drawings are merely for the clear description of a particular embodiment and do not imply a mandatory order, unless otherwise stated that a particular order must be followed.

[0054] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0055] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0056] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0057] Please see Figure 1 In one embodiment of the present invention, a method for detecting wafer surface contamination defects is provided. Based on dual-threshold neighborhood SNR estimation, defect signals and background signals are segmented from the wafer dark-field scattering image. This effectively solves the problems of complex background signals and weak defect signals in dark-field detection, meeting the high-precision and high-efficiency requirements of semiconductor manufacturing for wafer surface contamination defect detection. The specific detection process is as follows:

[0058] First, dark-field imaging is performed on the target wafer to obtain a dark-field scattering image of its surface, providing raw data for subsequent contamination defect detection. Dark-field detection is an existing technology, and the specific detection process is not within the scope of this invention. The main principle of dark-field imaging is: the wafer is illuminated by a side-above laser, causing scattered light to be generated at defective areas, while almost no scattered light is generated in defect-free areas. This forms an image on the detector where the defects are brighter and the background is darker, highlighting the contamination defects on the wafer surface and laying the foundation for subsequent signal processing and defect identification.

[0059] By traversing every pixel in the dark field image and performing dual-threshold neighborhood SNR estimation for each pixel, defect signals can be distinguished from background noise. When calculating the pixel SNR estimate, a neighborhood needs to be selected first. The neighborhood refers to a window of size n×n pixels or n×m pixels centered on the current pixel when calculating the SNR estimate for each pixel. The grayscale values ​​of the pixels within the neighborhood window are used to calculate the SNR estimate for the current pixel. The size of the neighborhood window is determined based on actual detection needs and defect characteristics. A larger window can cover more background information and improve the accuracy of defect detection, but it increases the computational load; while a smaller window is faster to calculate, but may be more susceptible to local noise. Therefore, the neighborhood can be adjusted based on experience and actual needs. There are two ways to enlarge or reduce the neighborhood window size: one is to fix the aspect ratio of the neighborhood window and enlarge or reduce it proportionally; the other is to not fix the aspect ratio of the neighborhood window and enlarge or reduce it non-proportionally.

[0060] In a preferred embodiment, an n×n neighborhood corresponds to a square neighborhood, while an n×m neighborhood corresponds to a rectangular neighborhood. The specific window shape and size of the neighborhood can be selected based on the shape of the defect. Rectangular neighborhoods are suitable for narrow or elongated defects such as slip or scratches. For example, setting a rectangular neighborhood to be longer horizontally than vertically improves the SNR accuracy for vertical defects. Although increasing the neighborhood size allows for a larger number of estimated pixels and more accurate SNR estimation, it also significantly increases the computational load, resulting in substantial time consumption. Therefore, selecting a specific direction for size expansion based on the defect shape to improve SNR estimation accuracy yields better results, balancing detection capability and detection time.

[0061] by Figure 3Taking the example shown, assume that the black line in the figure represents a vertical defect (such as a narrow defect or a long strip defect, like a slip or scratch), the red and blue boxes are two optional rectangular neighborhoods, and the intersection of the red and blue boxes is an optional square neighborhood. Since the blue and red boxes are larger than the square boxes and contain more pixels, the SNR estimation is less affected by extreme pixels, and the estimation is more stable. Therefore, when dealing with defects with a certain shape tendency, the rectangular neighborhood can be preferred. However, since the red box contains more defect pixels than the blue box, its estimated maximum background grayscale value will be larger, which in turn will increase the denominator of the following SNR estimation formula (1), resulting in a smaller SNR estimation value. The SNR estimation is significantly affected. Therefore, the defect SNR signal obtained by using the blue box neighborhood is better than the defect SNR signal obtained by using the red box neighborhood. Therefore, expansion in a specific direction can be selected to improve the defect detection capability.

[0062] After the neighborhood is selected, the SNR estimate for each pixel is calculated using the following formula (1):

[0063] (1);

[0064] in, This represents the estimated SNR value of the current pixel. Its value mainly reflects the intensity of the current pixel signal relative to the background noise, that is, the magnitude of the current pixel gray value relative to the gray values ​​of its neighbors. This represents the signal strength of the current pixel in the dark-field image, that is, the gray value of the current pixel in the dark-field image. This represents the mean background noise in the pixel neighborhood. This represents the amplitude of background noise in the pixel neighborhood. This represents the maximum background noise value in the pixel neighborhood. This represents the minimum background noise value in the pixel neighborhood. and Generally, it cannot be obtained directly, so it cannot be obtained through... Specific values ​​and The background noise amplitude is obtained by subtracting the specific values. Therefore, algorithm design is needed to... Perform an overall estimate, rather than actually obtaining the calculation. Specific values ​​and The specific value.

[0065] As a preferred embodiment, the mean background noise of the pixel neighborhood Generally, the median value is taken as the grayscale value of all pixels in the pixel's neighborhood. The median has a certain degree of noise resistance and representativeness of background noise. Therefore, when the background dynamic range of a dark-field image is large and the accuracy of contamination and defect detection is required to be high, The value is taken as the median grayscale value of all pixels in the neighborhood. However, in dark-field images where the dynamic range is small and the primary focus is on defect detection efficiency, Alternatively, the value can be the average grayscale value of all pixels in the pixel's neighborhood, or calculated using a quadratic std mean estimation algorithm. Among them, the quadratic std mean estimation algorithm is used to calculate... The specific process is as follows:

[0066] First, calculate the average grayscale value and standard deviation of all pixels in the neighborhood of a given pixel. Then, remove all pixels in the neighborhood whose grayscale value is less than (average value - Theta × standard deviation) and all pixels whose grayscale value is greater than (average value + Theta × standard deviation). Theta is a manually set constant, typically ranging from 1 to 3 based on experience. After removing outlier pixels with large deviations, calculate the average value of the remaining pixels in the neighborhood and assign this average value to the appropriate value. .

[0067] As a preferred embodiment, the background noise amplitude of the pixel neighborhood Generally, the value is taken as twice the difference between the median gray value and the minimum gray value in the neighborhood, that is:

[0068] = (median gray value of the pixel neighborhood - minimum gray value of the pixel neighborhood) × 2.

[0069] When the dynamic range of dark-field images is small and the primary focus is on defect detection efficiency,... The value of can also be replaced by the calculation result of the 6std noise estimation algorithm, and the 6std noise international algorithm can be used to calculate . The specific process is as follows:

[0070] Calculate the standard deviation of gray levels for all pixels in the pixel's neighborhood, and then copy 6 times the standard deviation of gray levels to... ,Right now:

[0071] =6 × grayscale standard deviation.

[0072] Please see Figure 4 This refers to the dark-field image obtained by dark-field photography, where the boxed area is the region selected based on the defect shape, and the punctuated pixels are the actual defect pixels (in practice, it is initially impossible to determine which pixels are the actual defect pixels). By combining the signal strength of the pixel with the mean and amplitude of the background noise in the neighborhood through the selected neighborhood and the SNR estimation formula (1), the signal-to-noise ratio of the current pixel is quantitatively evaluated, thereby highlighting the potential defect signal that is not easily identified, and calculating to obtain the signal-to-noise ratio of the current pixel. Figure 5The SNR estimation plot of the wafer dark field image is shown.

[0073] For the SNR estimation map of the wafer dark field image, this embodiment of the invention uses high and low thresholds for segmentation to achieve defect labeling. Specifically, firstly, based on actual inspection requirements and empirical data, a suitable high threshold and a suitable low threshold are selected to segment the SNR estimation map, where the high threshold T1 is set to T1≥3 and the low threshold T2 is set to T2≤2. The high threshold is used to extract more obvious and stronger defect signals, while the low threshold is used to capture weaker defect signals while minimizing background noise interference. The image obtained through high threshold segmentation is shown below. Figure 6 The binary image shown is denoted as the maker image (defect marker image). The image obtained through low-threshold segmentation is shown below. Figure 7 The binary image shown is denoted as a mask image (defect contour mask image).

[0074] The specific global threshold segmentation process is as follows:

[0075] The SNR estimation map is segmented using a high threshold. Pixels with gray values ​​greater than the high threshold are marked as true (i.e., defects), and the remaining pixels are marked as false (i.e., normal, non-defective pixels), generating a marker binary map. This binary map mainly contains relatively significant defective pixels, which can serve as reference markers for subsequent defect identification.

[0076] The SNR estimation map is segmented using a low threshold. Pixels with gray values ​​greater than the low threshold are marked as true (i.e., defect contour mask), and the remaining pixels are marked as false (i.e., normal), generating a binary mask image. This binary image covers more potential defect areas, including some weaker defect signals and possible pseudo-defect areas, providing more comprehensive information for subsequent connected component analysis. The combination of high and low threshold segmentation and cross-validation effectively improves the detection capability of minute defects.

[0077] Traverse all connected regions in the mask binary graph, and determine whether the connected regions are real or false defects based on the markings in the maker graph. Finally, generate a detected defect binary graph. The specific identification and labeling process for true and false defects is as follows:

[0078] First, in the binary mask image, find and extract all connected regions. Here, a connected region is a region consisting of adjacent pixels that all have a value of true after initial calibration with a low threshold. Each connected region may correspond to a potential defect.

[0079] For each extracted connected region, check if at least one pixel within that region is marked as true in the maker map. That is, the pixel is marked as true by both the high and low thresholds. If the connected region overlaps with the defect markers in the maker map, it indicates that the connected region is likely to contain a real defect. If no pixel in the connected region is marked as true in the maker map, the connected region is likely to be a false defect caused by background noise or other interference factors.

[0080] Based on the above criteria, pixels within connected regions containing true pixels in the maker image are marked as real defects, while connected regions where all pixels are false in the maker image are marked as pseudo-defects. Then, all pixels marked as pseudo-defects in the mask image are reset to false, pseudo-defects are deleted, and pixels containing real defects are marked as true, resulting in a binary image of the detected defects. This binary image clearly shows the location and extent of contamination defects on the wafer surface.

[0081] The above embodiment only uses two thresholds to segment the SNR estimation map. However, in complex cases, based on the defect contour type and light intensity distribution, it can be modified to segment the SNR estimation map using multiple thresholds with different segments. Multi-threshold segmentation still follows the dual-threshold segmentation logic, that is, combining and cascading multiple thresholds in pairs to form multiple dual-threshold combinations. Each dual-threshold combination is processed using the dual-threshold labeling method described above. This will ultimately detect multiple defect binary maps corresponding to different threshold combinations. These multiple defect binary maps are then fused to further improve defect detection capability. To maximize the detection capability of subtle defects and contamination, the fusion method for multiple defect binary maps generally involves performing a pixel OR operation on all defect binary maps, that is, taking the union of all pixels marked as true in all defect binary maps.

[0082] Based on the aforementioned wafer surface contamination defect detection method, this embodiment of the invention also constructs a wafer surface contamination defect detection device. This device includes at least a wafer surface contamination defect detection module. This module uses the aforementioned wafer surface contamination defect detection method to perform a comprehensive scan and analysis of the wafer surface, accurately identifying various minute contamination particles and defects. Furthermore, this device possesses efficient data processing capabilities, enabling it to quickly classify and record the detected contamination defect information and output the detection results.

[0083] An electronic device is designed to provide computational support for a wafer surface contamination defect detection method. This electronic device is equipped with at least one high-performance processor for computation and data processing, capable of executing the aforementioned wafer surface contamination defect detection method and data analysis tasks. The electronic device also includes a memory closely connected to the processor, storing a set of instructions executable by the processor. When the processor executes these instructions, it can accurately execute each step of the wafer surface contamination defect detection method, from data acquisition and image processing to defect identification and classification, ensuring the efficiency and accuracy of the entire detection process.

[0084] A computer-readable storage medium internally stores computer instructions specifically for invoking the detection of contamination defects on wafer surfaces. When these computer instructions are loaded and executed by a processor, the processor drives the entire detection system to work according to a preset process and algorithm based on the wafer surface contamination defect detection method, thereby achieving automated and intelligent detection of contamination defects on wafer surfaces.

[0085] In summary, the above description is merely a preferred embodiment of this specification and is not intended to limit the scope of protection of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

[0086] The systems, apparatuses, modules, or units described in one or more of the above embodiments may be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, a computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0087] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0088] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0089] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

Claims

1. A method for detecting a contamination defect on a wafer surface, comprising the steps of: The SNR estimation map of the wafer dark field image is segmented by setting at least two thresholds, and pixels with a gray value greater than the threshold are marked as defects, and the remaining pixels are marked as normal, thereby forming a plurality of binary maps corresponding to different thresholds. ​ The method for obtaining the SNR estimation map is as follows: Each pixel on the dark field image is traversed, and an n*n neighborhood or an n*m neighborhood containing the pixel is selected, and the SNR estimation value of each pixel is calculated; the calculation formula of the SNR estimation value is as follows: Any two thresholds are combined; ； wherein, represents the SNR estimate of a pixel, represents the grey value of a pixel in the dark field image, represents the mean of the background noise of the pixel neighborhood; represents the amplitude of the background noise of the pixel neighborhood, represents the maximum of the background noise of the pixel neighborhood, represents the minimum of the background noise of the pixel neighborhood; In each threshold combination, the connected regions in the low threshold binary map are traversed, and if there is at least one pixel in the connected region that is marked as a defect in the high threshold binary map, all pixels in the connected region are marked as real defects, otherwise, they are marked as false defects; When two thresholds are set to segment the SNR estimation map, the pixels marked as real defects in the low threshold binary map are the defects on the wafer surface; When more than two thresholds are set to segment the SNR estimation map, each threshold combination marks a binary map with real defect and false defect labels; all threshold combinations are pixel ORed to obtain a final binary map, and the pixels marked as real defects in the final binary map are the defects on the wafer surface. The target wafer is dark field photographed to obtain a dark field image of the wafer surface.

2. The wafer surface contamination defect detection method according to claim 1, wherein The average value and the standard deviation of the gray values of all pixels in the pixel neighborhood are calculated; 3. The wafer surface contamination defect detection method of claim 1, wherein Background noise mean value of the pixel neighborhood Value: median of the gray values of all pixels in the pixel neighborhood, or mean of the gray values of all pixels in the pixel neighborhood, or computed by a quadratic std mean value estimation algorithm .

4. The wafer surface contamination defect detection method according to claim 3, wherein The step of calculating the quadratic std mean estimation algorithm is: The step is: Pixels less than (average value-Theta*standard deviation) and greater than (average value+Theta*standard deviation) are removed, where Theta is a constant; The standard deviation of the gray values of all pixels in the pixel neighborhood is calculated; calculating a gray level average of the remaining pixels and using the gray level average as a background noise mean value for the pixel neighborhood .

5. The wafer surface contamination defect detection method according to claim 3, wherein background noise amplitude of the pixel neighborhood = (median gray value of the pixel neighborhood - minimum gray value of the pixel neighborhood) x 2, or computed by the 6std algorithm .

6. The wafer surface contamination defect detection method according to claim 5, wherein The steps of calculating the 6std algorithm are: The steps of calculating the 6std algorithm are: The accuracy of the SNR estimation value is improved by increasing the size of the neighborhood window. = 6 x gray scale standard deviation.

7. The wafer surface contamination defect detection method of claim 1, wherein The expansion or reduction of the neighborhood window size is as follows:

8. The wafer surface contamination defect detection method according to claim 7, wherein The aspect ratio of the neighborhood window is fixed to expand or reduce proportionally; Or, the aspect ratio of the neighborhood window is not fixed to expand or reduce non-proportionally. In each threshold combination, the high threshold value is greater than or equal to 3, and the low threshold value is less than or equal to 2.

9. The wafer surface contamination defect detection method of claim 1, wherein A wafer surface contamination defect detection module is included, and the wafer surface contamination defect detection module uses the wafer surface contamination defect detection method of any one of claims 1 to 9 for contamination defect detection.

10. A wafer surface contamination defect detection device, characterized in that, The SNR estimation map of the wafer dark field image is segmented by setting at least two thresholds, and pixels with a gray value greater than the threshold are marked as defects, and the remaining pixels are marked as normal, thereby forming a plurality of binary maps corresponding to different thresholds.

11. An electronic device, comprising: The method for obtaining the SNR estimation map is as follows: Each pixel on the dark field image is traversed, and an n*n neighborhood or an n*m neighborhood containing the pixel is selected, and the SNR estimation value of each pixel is calculated; the calculation formula of the SNR estimation value is as follows: Any two thresholds are combined; In each threshold combination, the connected regions in the low threshold binary map are traversed, and if there is at least one pixel in the connected region that is marked as a defect in the high threshold binary map, all pixels in the connected region are marked as real defects, otherwise, they are marked as false defects; When two thresholds are set to segment the SNR estimation map, the pixels marked as real defects in the low threshold binary map are the defects on the wafer surface; 12. A computer readable storage medium storing computer instructions, wherein, When more than two thresholds are set to segment the SNR estimation map, each threshold combination marks a binary map with real defect and false defect labels; all threshold combinations are pixel ORed to obtain a final binary map, and the pixels marked as real defects in the final binary map are the defects on the wafer surface. The target wafer is dark field photographed to obtain a dark field image of the wafer surface. The average value and the standard deviation of the gray values of all pixels in the pixel neighborhood are calculated; Pixels less than (average value-Theta*standard deviation) and greater than (average value+Theta*standard deviation) are removed, where Theta is a constant; The standard deviation of the gray values of all pixels in the pixel neighborhood is calculated; The accuracy of the SNR estimation value is improved by increasing the size of the neighborhood window. The expansion or reduction of the neighborhood window size is as follows: The aspect ratio of the neighborhood window is fixed to expand or reduce proportionally; Or, the aspect ratio of the neighborhood window is not fixed to expand or reduce non-proportionally. In each threshold combination, the high threshold value is greater than or equal to 3, and the low threshold value is less than or equal to 2. A wafer surface contamination defect detection module is included, and the wafer surface contamination defect detection module uses the wafer surface contamination defect detection method of any one of claims 1 to 9 for contamination defect detection. The SNR estimation map of the wafer dark field image is segmented by setting at least two thresholds, and pixels with a gray value greater than the threshold are marked as defects, and the remaining pixels are marked as normal, thereby forming a plurality of binary maps corresponding to different thresholds. The method for obtaining the SNR estimation map is as follows: Each pixel on the dark field image is traversed, and an n*n neighborhood or an n*m neighborhood containing the pixel is selected, and the SNR estimation value of each pixel is calculated; the calculation formula of the SNR estimation value is as follows: Any two thresholds are combined; In each threshold combination, the connected regions in the low threshold binary map are traversed, and if there is at least one pixel in the connected region that is marked as a defect in the high threshold binary map, all pixels in the connected region are marked as real defects, otherwise, they are marked as false defects; When two thresholds are set to segment the SNR estimation map, the pixels marked as real defects in the low threshold binary map are the defects on the wafer surface; When more than two thresholds are set to segment the SNR estimation map, each threshold combination marks a binary map with real defect and false defect labels; all threshold combinations are pixel ORed to obtain a final binary map, and the pixels marked as real defects in the final binary map are the defects on the wafer surface. The target wafer is dark field photographed to obtain a dark field image of the wafer surface. The average value and the standard deviation of the gray values of all pixels in the pixel neighborhood are calculated; Pixels less than (average value-Theta*standard deviation) and greater than (average value+Theta*standard deviation) are removed, where Theta is a constant; The standard deviation of the gray values of all pixels in the pixel neighborhood is calculated; The accuracy of the SNR estimation value is improved by increasing the size of the neighborhood window. The expansion or reduction of the neighborhood window size is as follows: The aspect ratio of the neighborhood window is fixed to expand or reduce proportionally; Or, the aspect ratio of the neighborhood window is not fixed to expand or reduce non-proportionally. In each threshold combination, the high threshold value is greater than or equal to 3, and the low threshold value is less than or equal to 2. A wafer surface contamination defect detection module is included, and the wafer surface contamination defect detection module uses the wafer surface contamination defect detection method of any one of claims 1 to 9 for contamination defect detection.

Images