Method and apparatus for detecting EBI online monitoring system capabilities

Abstract

The invention relates to a method for detecting the capability of an EBI online monitoring system and a computer readable storage medium. The method comprises the following steps: acquiring wafer defect data detected by the EBI online monitoring system; processing the wafer defect data to obtain prediction data corresponding to the defect data; and comparing the prediction data with a preset threshold value of a machine table of the semiconductor manufacturing system, and if the prediction data is smaller than the threshold value, indicating that the EBI online monitoring system does not completely acquire the defect data of the wafer. According to the scheme, the performance of the semiconductor manufacturing tool or the chip can be compared more sensitively, and the chip design can be optimized before the marginal DOI leak detection problem is detected online. According to the scheme of the invention, a marginal EBI result can be marked, and negative scores can be avoided when the overall online estimation yield performance is evaluated.

Description

Technical Field

[0001] The present invention relates to the field of defect detection in semiconductor manufacturing systems, and in particular to a method and device for detecting the capability of an EBI (E-Beam Inspection) online monitoring system, as well as a computer-readable medium and a 3D NAND manufacturing system. Background Art

[0002] Electron beam (E-Beam) VC (Voltage Contrast) defect detection is the detection of the difference in electronic signals between good contacts (CT) and hard open CT in 3D NAND products. In the EBI (E-Beam Inspection) monitoring method, the electrical connection status of the contacts is detected by the voltage intensity contrast (VC) mode, which is good at finding open contacts. EBI pulls out / pushes electrons (conductive / insulating) in the contact area, accumulates voltage when charging, and the voltage change further produces a contrast in voltage intensity. EBI is used to detect the surface potential difference of the defective contact hole CT (Contract) to identify bad CT. In this process, the defect signal is quantified by the grayscale (GLV) distribution of the defective CT image.

[0003] Currently, due to the GLV shift of the wafer caused by the change of wafer / manufacturing tool conditions, the online electron beam will intermittently miss the marginal DOI (Defect of Interest), which will reduce the stability of the detected chip (die). However, neither the companies (vendors) that produce and sell semiconductor chips nor the local sites have effective detection methods to monitor / control whether the performance of the chip has changed. How to evaluate whether the EBI online monitoring system can detect all DOIs is a black box, which is very risky for the purity and quality verification of online electron beam indicators.

[0004] Therefore, the existing technology requires a solution that can effectively detect, evaluate and quantify the capabilities and performance of EBI online monitoring systems.

[0005] The foregoing content is only intended to help understand the background of the embodiments of the present invention, and is not intended to mean that the embodiments of the present invention fall within the scope of the relevant prior art known to those skilled in the art. Summary of the invention

[0006] The present invention relates to a method and a computer-readable medium for detecting the capability of an EBI online monitoring system. The solution of the present invention can solve the problem in the prior art that it is difficult to quantify and evaluate the defect detection performance of an EBI online monitoring system.

[0007] A first aspect of the present invention provides a method for detecting the capability of an EBI online monitoring system, the method comprising: acquiring wafer defect data detected by the EBI online monitoring system; processing the wafer defect data to obtain prediction data corresponding to the defect data; comparing the prediction data with a threshold preset for a machine in a semiconductor manufacturing system, and if the prediction data is less than the threshold, determining that the EBI online monitoring system has not fully acquired the defect data of the wafer.

[0008] A second aspect of the present invention provides a non-transitory computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, implement the above-mentioned method for detecting the capability of an EBI online monitoring system.

[0009] A third aspect of the present invention provides a 3D NAND manufacturing system, which uses the above-mentioned method for detecting the capability of an EBI online monitoring system, or includes the device for detecting the capability of an EBI online monitoring system of the present invention.

[0010] The fourth aspect of the present invention provides a device for detecting the capability of an EBI online monitoring system, the device comprising: an acquisition device for acquiring wafer defect data detected by the EBI online monitoring system; a processing device for processing the wafer defect data to obtain predicted data corresponding to the defect data; the detection device for comparing the predicted data with a threshold preset for a machine in a semiconductor manufacturing system, and if the predicted data is less than the threshold, determining that the EBI online monitoring system has not fully acquired the defect data of the wafer.

[0011] The solution for detecting the capability of an EBI online monitoring system of the present invention can more sensitively compare the performance of manufacturing tools or chips (die), which helps to optimize chip design before online detection encounters marginal DOI missed detection problems. The solution of the present invention can also mark marginal EBI results, which can avoid negative scores when evaluating the overall online estimated yield performance.

[0012] The advantages of the embodiments of the present invention are not limited to the above advantages. Through the following description, those skilled in the art can clearly understand other advantages not described herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A method for detecting the capability of an EBI online monitoring system according to an embodiment of the present invention is shown.

[0014] FIG. 2( a ) and FIG. 2( b ) are diagrams showing the calculation principle of defect intensity DI according to an embodiment of the present invention.

[0015] 3(a), 3(b), 3(c) and 3(d) are schematic diagrams showing whether there are any problems in the detection results of the HEBI online monitoring system according to an embodiment of the present invention.

[0016] Figures 4(a) and 4(b) show partial peak distribution and unimodal distribution of defect intensity and number, respectively. Figures 4(c) and 4(d) show schematic diagrams of predicting defect intensity after processing Figures 4(a) and 4(b) by KDE and MLS, respectively.

[0017] Figure 5 The diagram shows a distribution diagram of defect intensity when a HEBI online monitoring system according to an embodiment of the present invention detects multiple wafers.

[0018] (1) to (6) of FIG. 6 are schematic diagrams showing defect result predictions in six wafers according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] Detailed descriptions of known technologies related to the embodiments disclosed in this specification may be omitted to avoid confusing the subject matter of the embodiments disclosed in this specification. In addition, the accompanying drawings are only for facilitating the understanding of the embodiments disclosed in this specification, but do not limit the technical spirit disclosed herein, and it should be understood that these embodiments include all modifications, equivalents and substitutes within the spirit and scope of the present invention.

[0020] As used herein, terms including ordinal numbers such as "first" and "second" may be used to describe various components without limiting these components. These terms are only used to distinguish one component from another.

[0021] It will be understood that when a component is referred to as being "connected to" or "coupled to" another component, it can be directly connected to or coupled to the other component, or intermediate components may exist. Conversely, when a component is referred to as being "directly connected to" or "directly coupled to" another component, there are no intermediate components.

[0022] As used herein, singular forms shall include plural forms as well unless the context clearly indicates otherwise.

[0023] It should be further understood that the terms "including" or "having", when used in this specification, specify the existence of the stated features, numbers, steps, operations, components, elements, or a combination thereof, but they do not exclude the existence or addition of one or more other features, numbers, steps, operations, components, elements, or a combination thereof.

[0024] Hereinafter, the embodiments disclosed in this specification are described with reference to the accompanying drawings, in which the same reference numerals are given to refer to the same or similar components and redundant descriptions thereof are omitted.

[0025] The present invention proposes a scheme for predicting and quantifying the detection capability of HEBI (Hermes Microvision Inc. E-Beam Inspection EBI) online monitoring system. This is very helpful for predicting and quantifying invisible defects beyond the detection capability of HEBI, especially in the HVM (High Volume Manufacturing) stage as a batch output product.

[0026] HEBI (Hermes Microvision Inc. E-Beam Inspection) monitoring solution can detect the contrast of voltage intensity on the wafer surface by pulling out / pushing electrons (conductive / insulating) through the contact wafer area. In the display interface of HEBI detector, the contrast of voltage intensity of a single through hole (pillar) in the wafer in a dark (dark) state can be displayed.

[0027] According to one or more embodiments of the present invention, the HMI (Hermes Microvision Inc.) EBI detection method is adopted in the present invention, and other EBI detection methods may also be adopted in implementation. The image of the chip surface is displayed by detecting the contrast of the voltage intensity on the chip surface, so as to judge the location of the defect according to the brightness or darkness of the image.

[0028] According to one or more embodiments of the present invention, an online detection scheme in a semiconductor manufacturing system can obtain defects of interest (DOI) in a wafer according to a HEBI detection method. The defect of interest DOI is positively linearly correlated with a difference ΔGLV between a gray level (GLV) of a defect image and a reference gray level (Reference GLV), and a curve of the number of DOIs in a determined chip (die) and ΔGLV (or its threshold) should present a complete Gaussian distribution. If only the dark (dark) voltage intensity contrast (Dark VC) of through holes (such as through holes in 3D NAND) in the HEBI detection scheme is considered, any distribution category that does not conform to the complete Gaussian distribution may represent an abnormality in the HEBI detection scheme, that is, some DOIs are not detected by the HEBI online monitoring system, or there is a problem with the capability of the HEBI online monitoring system, which will have a great impact on the subsequent HVM of the semiconductor manufacturing system.

[0029] According to one or more embodiments of the present invention, the present invention proposes a scheme for detecting the capability of the EBI online monitoring system, using kernel density estimation (KDE) and least squares method (OLS) to achieve accurate prediction. KDE constructs a continuous probability density by linearly adding discrete points to obtain a data sample distribution of DOI information smoothing. The truncated data distribution is obtained by calculating the extreme points of the continuous distribution, and then the curve is fitted using OLS to obtain the data distribution with the minimum error compared with the real data. The data with the minimum sampling is used to estimate the real defect distribution and monitor the performance of the HEBI online detection system.

[0030] Figure 1 A method for detecting the capability of an EBI online monitoring system according to an embodiment of the present invention is shown.

[0031] like Figure 1 As shown, in block S101, the wafer defect data is obtained, that is, the data of the wafer DOI defect (or other defects) detected by the HEBI online monitoring system is obtained. In block S102, the defect data is processed, that is, the wafer defect data is processed to obtain the prediction data corresponding to the defect data; in block S103, the capability of the HEBI online monitoring system is determined, that is, the prediction data is compared with the threshold preset by the machine of the semiconductor manufacturing system. If the prediction data is less than the threshold, it is determined that the EBI online monitoring system has not completely obtained the defect data of the wafer.

[0032] Specifically, in block S101, when the machine of the semiconductor manufacturing system is running, the HEBI online monitoring system is sampled to scan the wafer to obtain defect data information. The wafer defect data includes: the image grayscale level GLV corresponding to each defect in the wafer, the defect intensity DI calculated by the GLV, and the number of defects in the wafer. Then, the DGLV (Dark GLV), BGLV (Bright GLV), signal intensity DI, and number of defects corresponding to each defect obtained are output to the database. Among them, BGLV and RGLV are essentially the image grayscale levels corresponding to the defects, except that DGLV indicates that the dark representation is better on the interface of the HEBI online monitoring system, while BGLV indicates that the bright representation is better.

[0033] According to one or more embodiments of the present invention, DI (Defect intensity) represents the intensity of a defect, and DI is positively correlated with the size of ΔGLV, wherein ΔGLV is the difference between the GLV of each defect and the reference GLV.

[0034] FIG. 2( a ) and FIG. 2( b ) are diagrams showing the calculation principle of defect intensity DI according to an embodiment of the present invention.

[0035] FIG2(a) shows the principle of difference operation. Two images A and B represent two images of a person. The intensity value is obtained by performing a difference operation. For example, the middle image represents the result of the difference operation between image A and image B. It can be seen from the middle image that the difference between the two images can be clearly obtained by the difference operation.

[0036] FIG2(b) shows a schematic diagram of performing difference calculations with the first reference image (Ref 1) and the second reference image (Ref 2) and the defect image. As shown in FIG2(b), the three images above represent the defect image, the first reference image, and the second reference image, respectively, wherein the first reference image is an image of a detection position located on the left side of the wafer pattern adjacent to the current defect, and the second reference image is an image of a detection unit located on the right side of the wafer pattern adjacent to the current defect. The two images in the middle are: the black image on the left represents the first difference image obtained by subtracting the first reference image from the defect image, and the black image on the right represents the second difference image obtained by subtracting the second reference image from the defect image. The black image at the bottom calculates the same part in the first difference image and the second difference image, and finally considers the same part to be the location where the last defect occurs.

[0037] The formula for the difference operation is: The DI of each defect is calculated by the difference operation, wherein the difference operation is:

[0038]

[0039] Where DI represents the intensity of the current defect, I2(i, j) represents the GLV value of the current defect image, I1(i, j) represents the GLV value of the first reference image, I3(i, j) represents the GLV value of the second reference image, σ 12 and σ 23 A custom constant.

[0040] According to one or more embodiments of the present invention, ΔGLV is a grayscale image value represented by grayscale numbers from 0 to 255, where 0 represents the darkest and 255 represents the brightest. Therefore, DI in the present invention is also a dimensionless number.

[0041] According to one or more embodiments of the present invention, in block 102, wafer defect (e.g., DOI) data is processed. Specifically, ΔGLV is a default parameter for obtaining the minimum intensity limit of the defect DOI of interest on the wafer. Therefore, how to accurately predict the starting point of the DOI threshold Gaussian distribution is a key parameter, and the DOI threshold is linearly related to ΔGLV. In the present invention, KDE and OLS are used to achieve accurate prediction. KDE constructs a continuous probability density by linearly adding discrete points to obtain a smooth data sample distribution. The truncated data distribution is obtained by calculating the extreme points of the continuous distribution (kernel density = 0), and then curve fitting is performed using OLS to obtain a data distribution with the smallest error compared to the real data, estimate the real defect distribution and monitor the capabilities of the HEBI online monitoring system.

[0042] FIG3 shows a schematic diagram of whether there are problems with the detection results of the HEBI online monitoring system according to an embodiment of the present invention. FIG3(a) shows a schematic diagram of wafer defects with problems in the detection results of the HEBI online monitoring system, and FIG3(b) shows a schematic diagram of wafer defects with no problems in the detection results of the HEBI online monitoring system. It can be seen from FIG3(a) and FIG3(b) that the grayscale level of the correction image (path image) of the defect of the through hole DVC (Dark VC) in FIG3(a) is less than the grayscale level in FIG3(b).

[0043] Figures 3(c) and 3(d) show the distribution of DI and DOI defect numbers under the defects of Figures 3(a) and 3(b), respectively. The horizontal axis of the bar graphs in Figures 3(c) and 3(d) represents the defect intensity, which refers to the specific value of DI. The DI on the left is small and the DI on the right is large. The larger the value, the stronger the defect intensity is, and the easier it is to be caught by the machine of the semiconductor manufacturing system. The vertical axis refers to the number of DOI defects in a 0.02DI step year (usually, a wafer has more than several thousand defects). The threshold refers to the threshold set by the machine. When DI<threshold, the defect will be deleted by the machine; when DI>threshold, the machine will report that the defect has been caught. The peak value in Figures 3(c) and 3(d) refers to the defect intensity DI size corresponding to the largest number of defects when falling into a certain compensation range under the same DOI defect intensity DI step value condition.

[0044] According to one or more embodiments of the present invention, it can be seen from FIG. 3(d) that the defect distribution of the wafer in the ideal case is similar to the second bar graph, that is, the defect distribution presents an approximate complete Gaussian distribution. The complete Gaussian distribution indicates that regardless of the strength of the defect signal, it has been captured by the detection machine, wherein the defect signal on the left is weak and the defect signal on the right is strong. Relatively speaking, if the threshold is set too high, all defects with weak signals will be filtered out. This situation indicates that defects will be missed in the machine detection results (i.e., the detection results of the HEBI online monitoring system), which is unacceptable to the semiconductor manufacturing system. However, in the actual semiconductor system defect detection, there is another situation, that is, the threshold setting is unified. Due to the stability problem of the machine, the quality of the defect correction image (patch image) deteriorates. The quality deterioration will directly affect the GLV distribution corresponding to the image, which will cause the ΔGLV of the difference operation to become smaller, resulting in a smaller defect intensity DI. The overall DI signal becomes weaker, which will also cause the weaker part to be missed. Therefore, the threshold is set in advance according to the actual situation of the semiconductor manufacturing system and balances the above two situations.

[0045] According to one or more embodiments of the present invention, in block 102, processing the wafer defect data includes: establishing a two-dimensional distribution of DI and the number of defects, performing a kernel density operation on the two-dimensional distribution to obtain a predicted smooth continuous curve of the two-dimensional distribution, and then fitting the continuous smooth curve by the least squares method to obtain a Gaussian distribution curve of the two-dimensional distribution, and solving twice the standard deviation based on the Gaussian distribution curve to obtain a predicted value of DI.

[0046] Figures 4(a) and 4(b) show the partial peak distribution and single peak distribution of defect intensity and quantity, respectively. All data in Figures 4(a) and 4(b) are defect data, which are the overall distribution of all defects in a wafer. The purpose of the two-dimensional distribution in Figures 4(a) and 4(b) is to predict the corresponding Gaussian distribution through the actual data distribution, and calculate the value of the horizontal axis defect intensity DI corresponding to the corresponding 2 times standard deviation (2 times Sigma) point based on the Gaussian distribution. Then compare the predicted DI value with the threshold (Threshold) pre-set by the semiconductor manufacturing system machine. If the DI corresponding to 2 times the standard deviation (2 times Sigma) is greater than the preset threshold (Threshold), it means that there is no problem with the defect results of this batch of wafers detected by the HEBI online monitoring system (that is, the complete Gaussian distribution is captured); on the contrary, if DI is less than the preset threshold (Threshold), it means that there is a problem with the results of this batch of wafers detected by the HEBI online monitoring system, that is, only part of the Gaussian distribution is captured, which means that the HEBI online monitoring system has missed some defects and has not fully obtained all the defects on the wafer.

[0047] In Figure 4(a) and Figure 4(b), the horizontal axis represents the defect intensity DI, and the vertical axis represents the number of defects in a 0.02DI step year, which can be understood as the density of defects.

[0048] FIG4(c) and FIG4(d) are schematic diagrams showing the predicted defect intensity after FIG4(a) and FIG4(b) are processed by KDE and MLS, respectively.

[0049] As shown in Figures 4(c) and 4(d), the predicted DI value is obtained by processing the two-dimensional distribution of DI and the number of defects in Figures 4(a) and 4(b), that is, the kernel density operation is performed on the two-dimensional distribution to obtain the predicted smooth continuous curve of the two-dimensional distribution, and then the continuous smooth curve is fitted by the least squares method to obtain the Gaussian distribution curve of the two-dimensional distribution, and the predicted value of DI is obtained by solving twice the standard deviation (2 times Sigma) based on the Gaussian distribution curve. The standard deviation (Sigma) represents the standard deviation of the Gaussian distribution, 2 times Sigma represents 2 times the standard deviation of the offset center, which can contain 95% of the data; 3 times Sigma represents 3 times the standard deviation of the offset center, which can contain 99.7% of the data. It can be seen from Figures 4(c) and 4(d) that 2 times the standard deviation is closer to the desired expected value. Therefore, when evaluating the HEBI online monitoring system, 2 times Sigma is used for comparison with the threshold set by the machine.

[0050] In Figure 4(c) and Figure 4(d), the horizontal axis represents the defect intensity DI, and the vertical axis represents the kernel probability density. The value of the final integral sum of the horizontal axis multiplied by the vertical axis is 1, which means the probability distribution of falling into different DI steps. The discrete real data is completed to a smooth curve by predicting the probability. The KDE line is the kernel probability density curve calculated based on the real data, and the OLS line is the standard Gaussian distribution curve predicted based on the kernel probability density curve. Figure 4(c) and Figure 4(d) show the position where the kernel probability density (kernel density) = 0, which means the point where the first-order derivative of the kernel probability density curve is 0, which represents the minimum value on the predicted curve. In the actual prediction calculation, such extreme value points are excluded from the interference of the measurement, so that the predicted Gaussian distribution will be more accurate. By comparing the DI values ​​predicted by Figure 4(c) and Figure 4(d) with the preset threshold of the machine, it can be judged whether the HEBI online monitoring system has completely captured the defects of the wafer.

[0051] Figure 5 The diagram shows a distribution diagram of defect intensity when a HEBI online monitoring system according to an embodiment of the present invention detects multiple wafers.

[0052] Depend on Figure 5 As shown, the horizontal axis represents time, and the vertical axis is the defect intensity DI value corresponding to 2 times the standard deviation (Sigma) calculated based on the kernel density and least squares method. The dotted line in the figure represents the threshold set by the machine, and a point in the figure represents the defect intensity DI value corresponding to 2 times Sigma predicted for a wafer. If the predicted DI value is higher than the threshold, it means that the complete Gaussian distribution is captured (that is, on the dotted line). If it is lower than the threshold represented by the dotted line, it means that the HEBI online monitoring system has captured part of the Gaussian distribution, which also indicates that the data of this wafer detected by the HEBI online monitoring system is problematic.

[0053] FIG. 6 is a schematic diagram showing defect result prediction in six wafers according to an embodiment of the present invention.

[0054] As shown in (1) to (6) of FIG. 6, Figure 5 The defect prediction results of the six randomly selected circled wafers in the figure. In Figure 6 (1) to (6), each figure shows the actual defect distribution of a wafer, where the horizontal axis is the defect intensity DI and the vertical axis refers to the number of defects within the corresponding DI step range. The dotted line represents Figure 5 The corresponding vertical axis value in is the predicted 2 times standard deviation (Sigma) value of DI. Because the result distribution of each wafer is different, and a 2 times standard deviation (Sigma) prediction operation is performed each time based on the actual data distribution, the value will be different.

[0055] In Figures 6 (1) to (6), the starting value of the horizontal axis on the left side of each figure is 0.36, which is the threshold (Threshold) set by the semiconductor manufacturing system machine. The peak value (Peak) represents the defect intensity DI size corresponding to the largest number of defects within a certain compensation range under the same defect intensity DI step value condition. In the prediction result, the peak value must be on the right side of the red dotted line (i.e., the value of 2 times the standard deviation (Sigma) of the predicted DI). The farther the red dotted line is from the starting value on the left side (i.e., the preset threshold), the more the data tends to be normally distributed, and the healthier the batch of wafer results predicted by the HEBI online monitoring system. On the contrary, if the red dotted line is closer to the starting value on the left side (i.e., the preset threshold), it means that there is a greater possibility that only part of the Gaussian distribution is present, indicating that the batch of wafer results predicted by the HEBI online monitoring system may have problems (for example, the results of the wafer defects predicted by Figures 6 (1) and 6 (2) are problematic). The ideal result is to provide a reference for machine performance for the semiconductor manufacturing system through the scheme of the present invention, and to point out the direction for continuously optimizing machine performance. By adjusting the machine parameters, the red dotted line (i.e., the value of 2 times the standard deviation (Sigma) of the predicted DI) is moved away from the left starting value (i.e., the preset threshold).

[0056] The following are further examples of the present invention:

[0057] Example 1. A method for detecting the capability of an EBI online monitoring system, the method comprising: acquiring wafer defect data detected by the EBI online monitoring system; processing the wafer defect data to obtain prediction data corresponding to the defect data; comparing the prediction data with a threshold preset for a machine in a semiconductor manufacturing system, and if the prediction data is less than the threshold, determining that the EBI online monitoring system has not fully acquired the wafer defect data.

[0058] Example 2. The method according to Example 1, wherein the wafer defect data includes: an image grayscale level GLV corresponding to each defect in the wafer, a defect intensity DI calculated by the GLV, and the number of defects in the wafer.

[0059] Example 3. The method according to Example 2, wherein the DI is positively correlated with the magnitude of ΔGLV, wherein ΔGLV is the difference between the GLV of each defect and the reference GLV.

[0060] Example 4. The method of Example 3, wherein the DI of each defect is calculated by a difference operation, wherein the difference operation is:

[0061]

[0062] Where DI represents the intensity of the current defect, I2(i, j) represents the GLV value of the current defect image, I1(i, j) represents the GLV value of the first reference image, I3(i, j) represents the GLV value of the second reference image, σ 12 and σ 23 is a user-defined constant, wherein the first reference image is an image of a detection unit located on the left side of the current defect in the wafer pattern, and the second reference image is an image of a detection unit located on the right side of the current defect in the wafer pattern.

[0063] Example 5. A method according to Example 1, wherein processing the wafer defect data includes: establishing a two-dimensional distribution of DI and the number of defects, performing a kernel density operation on the two-dimensional distribution to obtain a predicted smooth continuous curve of the two-dimensional distribution, and then fitting the continuous smooth curve by the least squares method to obtain a Gaussian distribution curve of the two-dimensional distribution, and solving twice the standard deviation based on the Gaussian distribution curve to obtain a predicted value of DI.

[0064] Example 6. The method according to Example 5, wherein comparing the predicted data with a threshold preset for a machine of the semiconductor manufacturing system comprises: comparing the predicted value of the DI with a threshold preset for a machine of the semiconductor manufacturing system.

[0065] Example 7. The method according to Example 6, wherein, when the predicted value of the DI is greater than the preset threshold, it indicates that the EBI online monitoring system obtains the complete Gaussian distribution of defects and can fully obtain the defect data of the wafer; when the predicted value of the DI is less than or equal to the preset threshold, it indicates that the EBI online monitoring system obtains the partial Gaussian distribution of defects and cannot fully obtain the defect data of the wafer.

[0066] Example 8. The method of Example 1, wherein the defect of the wafer is a defect of interest (DOI).

[0067] Example 9. The method according to Example 8, wherein the defect of interest DOI is a single through-hole defect in a wafer, and the single through-hole is displayed as a dark color in the interface of the EBI online monitoring system.

[0068] Example 10. The method according to Example 3, wherein the ΔGLV is a grayscale image value represented by a digital grayscale of 0-255, where 0 represents the darkest and 255 represents the brightest.

[0069] Example 11. A method according to Example 5, wherein when obtaining the predicted value of DI, the predicted value of DI corresponding to a kernel probability density of zero is excluded.

[0070] Example 12. A non-transitory computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, implements the method for detecting the capability of an EBI online monitoring system according to any one of Examples 1-11.

[0071] Example 13. A device for detecting the capability of an EBI online monitoring system, the device comprising: an acquisition device for acquiring wafer defect data detected by the EBI online monitoring system; a processing device for processing the wafer defect data to obtain prediction data corresponding to the defect data; and the detection device for comparing the prediction data with a threshold preset for a machine in a semiconductor manufacturing system, and if the prediction data is less than the threshold, determining that the EBI online monitoring system has not fully acquired the wafer defect data.

[0072] Example 14. The device according to Example 13, wherein the wafer defect data includes: an image gray level GLV corresponding to each defect in the wafer, a defect intensity DI calculated by the GLV, and the number of defects in the wafer.

[0073] Example 15. The apparatus of Example 13, wherein the DI is positively correlated with the magnitude of ΔGLV, wherein ΔGLV is the difference between the GLV of each defect and a reference GLV.

[0074] Example 16. The apparatus of Example 13, wherein the DI for each defect is calculated by a difference operation, wherein the difference operation is:

[0075]

[0076] Where DI represents the intensity of the current defect, I2(i, j) represents the GLV value of the current defect image, I1(i, j) represents the GLV value of the first reference image, I3(i, j) represents the GLV value of the second reference image, σ 12 and σ 23 is a user-defined constant, wherein the first reference image is an image of a detection unit located on the left side of the current defect in the wafer pattern, and the second reference image is an image of a detection unit located on the right side of the current defect in the wafer pattern.

[0077] Example 17. An apparatus according to Example 13, wherein processing the wafer defect data includes: establishing a two-dimensional distribution of DI and the number of defects, performing a kernel density operation on the two-dimensional distribution to obtain a predicted smooth continuous curve of the two-dimensional distribution, then fitting the continuous smooth curve by the least squares method to obtain a Gaussian distribution curve of the two-dimensional distribution, and solving twice the standard deviation based on the Gaussian distribution curve to obtain a predicted value of DI.

[0078] Example 18. The device according to Example 17, wherein comparing the predicted data with a threshold preset for a tool of the semiconductor manufacturing system comprises: comparing the predicted value of the DI with a threshold preset for a tool of the semiconductor manufacturing system.

[0079] Example 19. The device according to Example 18, wherein, when the predicted value of the DI is greater than the preset threshold, it indicates that the EBI online monitoring system obtains the complete Gaussian distribution of defects and can fully obtain the defect data of the wafer; when the predicted value of the DI is less than or equal to the preset threshold, it indicates that the EBI online monitoring system obtains a partial Gaussian distribution of defects and cannot fully obtain the defect data of the wafer.

[0080] Example 20. The apparatus of Example 13, wherein the defect of the wafer is a defect of interest (DOI).

[0081] Example 21. The apparatus according to Example 20, wherein the defect of interest DOI is a single through-hole defect in a wafer, and the single through-hole is displayed as a dark color in the interface of the EBI online monitoring system.

[0082] Example 22. In the device according to Example 16, the ΔGLV is a grayscale image value represented by a digital grayscale of 0-255, where 0 represents the darkest and 255 represents the brightest.

[0083] Example 23. The device according to Example 18, wherein when obtaining the predicted value of DI, the predicted value of DI corresponding to a kernel probability density of zero is excluded.

[0084] Example 24. A 3D NAND manufacturing system, wherein the manufacturing system uses the above-mentioned method for detecting the capability of an EBI online monitoring system, or includes the device for detecting the capability of an EBI online monitoring system of the present invention.

[0085] While specific embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various modifications and variations may be made thereto without departing from the technical spirit of the present invention as provided by the claims hereinafter appended.

Claims

1. A method for detecting the capability of an EBI online monitoring system, the method comprising: Acquiring wafer defect data detected by the EBI online monitoring system; Processing the wafer defect data to obtain prediction data corresponding to the defect data; The predicted data is compared with a threshold preset for a machine in a semiconductor manufacturing system. If the predicted data is less than the threshold, it is determined that the EBI online monitoring system has not completely acquired defect data of the wafer.

2. The method according to claim 1, wherein: The wafer defect data includes: an image gray level GLV corresponding to each defect in the wafer, a defect intensity DI calculated by the GLV, and the number of defects in the wafer.

3. The method according to claim 2, wherein: The DI is positively correlated with the magnitude of ΔGLV, wherein ΔGLV is the difference between the GLV of each defect and the reference GLV.

4. The method according to claim 3, wherein: The DI for each defect is calculated by a difference operation: Where DI represents the intensity of the current defect, I2(i, j) represents the GLV value of the current defect image, I1(i, j) represents the GLV value of the first reference image, I3(i, j) represents the GLV value of the second reference image, σ 12 and σ 23 is a user-defined constant, wherein the first reference image is an image of a detection unit located on the left side of the current defect in the wafer pattern, and the second reference image is an image of a detection unit located on the right side of the current defect in the wafer pattern.

5. The method according to claim 1, wherein: Processing the wafer defect data includes: A two-dimensional distribution of DI and the number of defects is established, a kernel density operation is performed on the two-dimensional distribution to obtain a predicted smooth continuous curve of the two-dimensional distribution, and then the continuous smooth curve is fitted by the least squares method to obtain a Gaussian distribution curve of the two-dimensional distribution, and two times the standard deviation is solved based on the Gaussian distribution curve to obtain a predicted value of DI.

6. The method according to claim 5, wherein: Comparing the predicted data with a threshold value preset by a tool of the semiconductor manufacturing system includes: The predicted value of the DI is compared with a threshold value preset in a tool of a semiconductor manufacturing system.

7. The method according to claim 6, wherein: When the predicted value of the DI is greater than the preset threshold, it indicates that the EBI online monitoring system obtains a complete Gaussian distribution of defects and is able to fully obtain defect data of the wafer; When the predicted value of the DI is less than or equal to the preset threshold, it means that the EBI online monitoring system obtains a partial Gaussian distribution of defects and cannot completely obtain defect data of the wafer.

8. The method according to claim 1, wherein: The defect of the wafer is a defect of interest DOI. 9 . The method according to claim 8 , wherein the defect of interest DOI is a single through hole defect in a wafer, and the single through hole is displayed as a dark color in the interface of the EBI online monitoring system.

10. The method according to claim 3, wherein: The ΔGLV is a grayscale image value represented by a grayscale number of 0-255, where 0 represents the darkest and 255 represents the brightest.

11. The method according to claim 5, wherein: When obtaining the predicted value of DI, the predicted value of DI corresponding to the kernel probability density of zero is excluded.

12. A non-transitory computer-readable storage medium having instructions stored thereon, which, when executed by one or more processors, implement the method for detecting the capability of an EBI online monitoring system according to any one of claims 1-11.

13. A device for detecting the capability of an EBI online monitoring system, the device comprising: An acquisition device, used to acquire wafer defect data detected by the EBI online monitoring system; A processing device, used for processing the wafer defect data to obtain prediction data corresponding to the defect data; The detection device is used to compare the predicted data with a threshold preset by a machine in a semiconductor manufacturing system. If the predicted data is less than the threshold, it is determined that the EBI online monitoring system has not completely acquired defect data of the wafer.

14. A 3D NAND manufacturing system, wherein the 3D NAND manufacturing system uses the method for detecting the capability of an EBI online monitoring system according to any one of claims 1 to 11, or includes the non-transitory computer-readable storage medium according to claim 12, or includes the device for detecting the capability of an EBI online monitoring system according to claim 13.

Images