Wafer processing defect testing method and apparatus

Abstract

The application relates to the technical field of defect detection, and provides a wafer processing defect testing method and device. Through region division according to wafer design, the wafer is subjected to hierarchical light intensity irradiation and color light irradiation, image collection, construction of a light intensity channel set and a color channel set, selection of a standard color channel and determination of region division position mapping to the two channel sets; interest region division of the light intensity channel set is carried out based on the design, and defect identification is carried out on the interest region according to the division result and the mapping result to obtain a first defect set; color channel set defect identification is carried out, and a second defect set is determined according to the mapping result, and a defect identification result is generated according to the two defect sets. The technical problem that the wafer defect identification method is single in the prior art, wafer defect identification omission is prone to occur, and the wafer defective product rate is increased is solved, the wafer defect identification accuracy and the defect identification efficiency are improved, and the technical effect that the wafer production defective product rate is reduced is achieved.

Description

Technical Field

[0001] This application relates to the field of defect detection technology, and in particular to a defect testing method and apparatus for wafer processing. Background Technology

[0002] As the core material for semiconductor device fabrication, the surface quality of wafers directly affects the performance and reliability of devices. Therefore, wafer defect identification is crucial. Currently, wafer defect identification mainly relies on image processing technology and manual inspection, but these methods all have certain limitations and shortcomings.

[0003] Differences in the shape and size of defects on the wafer surface, as well as their sensitivity to colored light, can affect the accuracy of detection. Moreover, with the continuous increase in wafer size and the continuous advancement of manufacturing processes, defects on wafers are becoming more and more complex and diverse, which brings greater challenges to defect identification.

[0004] In summary, existing technologies suffer from limitations in wafer defect identification methods, which are prone to omissions and lead to increased wafer defect rates. Summary of the Invention

[0005] Therefore, it is necessary to provide a wafer processing defect testing method and apparatus that can improve the accuracy and efficiency of wafer defect identification and reduce the defect rate in wafer production, in order to address the above-mentioned technical problems.

[0006] A defect testing method for wafer fabrication includes: reading wafer design information and performing wafer region division, setting separation features, wherein the region division result includes a first region and a second region, the first region being an important region and the second region being a regular region; irradiating the wafer under test with hierarchical light intensity and acquiring images of the irradiation results to construct a light intensity channel set; irradiating the wafer under test with colored light and acquiring images of the irradiation results to construct a color channel set; selecting a standard color channel in the color channel set and performing feature matching of the standard color channel based on the separation features to determine the regional positions of the first region and the second region, and mapping them to the light intensity channel set and the color channel set; dividing the light intensity channel set into regions of interest for each light intensity channel based on the design information, and locating the regions of interest according to the division results and mapping results; identifying defects in the regions of interest and determining a first defect set according to the mapping results; identifying defects in the color channel set and determining a second defect set according to the mapping results; and generating a defect identification result for the wafer under test based on the first defect set and the second defect set.

[0007] A defect testing device for wafer fabrication, the device comprising: a wafer region segmentation module for reading wafer design information and performing wafer region segmentation, setting separation features, wherein the region segmentation result includes a first region and a second region, the first region being an important region and the second region being a regular region; a light intensity channel construction module for irradiating the wafer under test with hierarchical light intensity, acquiring images of the irradiation results, and constructing a light intensity channel set; a color channel construction module for irradiating the wafer under test with colored light, acquiring images of the irradiation results, and constructing a color channel set; and a feature matching execution module for selecting a standard color channel in the color channel set and performing feature matching based on the separation features. The system employs a feature matching module to determine the location of the first and second regions and maps them to the light intensity channel set and the color channel set. A region of interest (ROI) localization module is used to divide the light intensity channel set into ROIs for each light intensity channel based on the design information, and to locate the ROIs according to the division and mapping results. A defect identification execution module is used to identify defects in the ROIs and determine a first defect set based on the mapping results. A defect set determination module is used to identify defects in the color channel set and determine a second defect set based on the mapping results. A defect identification execution module is used to generate defect identification results for the wafer under test based on the first and second defect sets.

[0008] The aforementioned wafer processing defect testing method and apparatus solves the technical problem in the prior art where the wafer defect identification method is singular, easily leading to omissions in wafer defect identification and resulting in an increase in wafer defect rate. It achieves the technical effect of improving the accuracy and efficiency of wafer defect identification and reducing the wafer production defect rate.

[0009] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a defect testing method for wafer fabrication in one embodiment.

[0011] Figure 2 This is a schematic diagram of the process for determining a second defect set in a defect testing method for wafer fabrication in one embodiment;

[0012] Figure 3 This is a schematic diagram of a wafer processing control and early warning process in a defect testing method for wafer processing, as described in one embodiment.

[0013] Figure 4 This is a structural block diagram of a defect testing device for wafer processing in one embodiment;

[0014] Figure labeling: 1. Wafer region division module; 2. Light intensity channel construction module; 3. Color channel construction module; 4. Feature matching execution module; 5. Region of interest localization module; 6. Defect recognition execution module; 7. Defect set determination module; 8. Defect recognition execution module. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0016] like Figure 1 As shown, this application provides a defect testing method for wafer fabrication, the method comprising:

[0017] S100: Read the wafer design information and perform wafer region division, set the separation features, wherein the region division result includes a first region and a second region, the first region is an important region, and the second region is a regular region;

[0018] Specifically, it should be understood that during wafer production, multiple batches of the same wafers are often produced based on wafer design information. The design information of the wafer under test is read. The design information is the layout design of the functional areas on the wafer surface designed by the wafer designer based on the target application field of the wafer or the needs of the planned chip production.

[0019] Based on the design information, the key circuit area, test area, input / output port and pin area are located. The above three functional areas are regarded as important areas of the wafer surface. The other functional areas of the wafer surface excluding the above three functional areas are regarded as regular areas. This embodiment does not limit the setting of important area composition and regular area composition, and can be adjusted according to the concentration of electronic devices on the wafer and the importance of electronic devices.

[0020] Based on the preset important region composition and the normal region composition, the wafer is divided into regions by combining the design information, thereby dividing the wafer surface into important regions and non-important regions, and obtaining a region division result including a first region and a second region, wherein the first region is an important region and the second region is a normal region.

[0021] The segmentation feature is a positioning feature used to assist in dividing the surface of the wafer under test into regions according to the important and regular region components determined above. For example, the segmentation feature is a reference point or positioning point in wafer design, including but not limited to rectangular reference points, circular reference points, and cross reference points. The wafer angle and the dividing line position for important and regular regions are located based on the segmentation feature, enabling angle adjustment and mapping of important and regular regions when obtaining a wafer entity or photograph.

[0022] S200: Irradiates the wafer under test with hierarchical light intensity, acquires images of the irradiation results, and constructs a set of light intensity channels;

[0023] S300: Irradiate the wafer under test with colored light, and acquire images of the irradiation results to construct a color channel set;

[0024] Specifically, in this embodiment, the wafer to be tested is a wafer product obtained by random sampling during the production process of multiple batches of wafers with the same design information, which is to be inspected for defects.

[0025] It should be understood that to achieve the different functions and characteristics of various functional areas of a wafer, wafers are typically composed of multiple materials. The differences in refractive index and transmittance of these materials result in variations in the light transmittance of different parts of the wafer surface. Furthermore, the different structural designs of different functional areas of the wafer, including differences in shape, size, density, and other parameters, also lead to variations in the light transmittance of different areas of the wafer surface. Consequently, due to the differences in refractive index and light transmittance across different areas of the wafer surface, various types of defects generated during the wafer manufacturing process exhibit differences in clarity and completeness when displayed under different colors of light.

[0026] Therefore, in this embodiment, the functional area layout of the wafer under test is obtained according to the design information, and different levels of light intensity are set for each functional area. Based on the different levels of light intensity, the wafer under test is irradiated through, and images of the wafer under test are acquired under different levels of light intensity.

[0027] A set of light intensity channels is constructed based on images of the wafer under test acquired under different levels of light intensity irradiation. Each light intensity channel in the set corresponds to an image of the wafer under test under a certain level of light intensity irradiation.

[0028] While setting the hierarchical light intensity for penetration illumination and image acquisition, this embodiment controls the light intensity to a standard light intensity (for example, controlling the light intensity to the lowest, average, or intermediate level of the hierarchical light intensity), changes the illumination color, illuminates the wafer under test with different colors of light, and acquires images of the illumination results to construct a color channel set. Each color channel in the color channel set corresponds to an image of the wafer under test illuminated by one color of light.

[0029] For example, if there are K types of light intensity levels, and the surface of the wafer under test is illuminated and images are acquired based on the K types of light intensity levels, then the light intensity channel set consists of K light intensity channels. If the wafer under test is illuminated with M colors of light, then the color channel set consists of M color channels.

[0030] This embodiment uses layered light intensity irradiation and different color light irradiation on the wafer under test to provide a valid reference for subsequent cross-verification and to determine the defects on the surface of the wafer under test.

[0031] S400: Select a standard color channel in the color channel set, and perform feature matching of the standard color channel based on the segmentation feature to determine the regional positions of the first region and the second region, and map them to the light intensity channel set and the color channel set;

[0032] Specifically, in this embodiment, the standard color channel is the color channel obtained by illuminating the wafer under test with RGB light of 255, 225, 255 and acquiring an image, and the image of the wafer under test for which the standard color channel is extracted is selected from the set of color channels.

[0033] It should be noted that in this embodiment, the wafer under test is subjected to hierarchical light intensity irradiation and image acquisition. During the color light irradiation and image acquisition process, the position of the wafer under test on the stage remains fixed. Therefore, under different colors of light and different light intensities, the relative positions of the defect features in the acquired images are fixed, and thus the positions of the wafer partition features in each image of the wafer under test are relatively consistent.

[0034] Therefore, in this embodiment, after obtaining the wafer image to be tested corresponding to the standard color channel, feature matching is performed on the surface image of the wafer to be tested in the standard color channel based on the segmentation features. The feature matching is to adjust the angle of the surface image of the wafer to be tested in the standard color channel based on the reference point or positioning point of the wafer design in the segmentation features, so that the placement angle of the wafer in each surface image of the wafer to be tested is consistent, so as to avoid the same defect being identified as a defect at different positions on the wafer surface when performing defect feature identification based on multiple images to be tested. Simultaneously, after adjusting the angle of the wafer image under test, based on the relative positional relationship between the reference point and the region division lines in the partitioning feature, the regional positions of the first region and the second region in the wafer surface image under test in the standard color channel are determined. Based on the obtained first region position and second region position, the image of the wafer surface image under test is divided to obtain a first region position image set and a second region position image set. Multiple local images of the wafer surface under test in each region position image set correspond to the same region position on the wafer surface under test. The first region position image set and the second region position image set are mapped to the light intensity channel set and the color channel set to achieve synchronous division of the first region and the second region of all wafer images under test in the light intensity channel set and the color channel set.

[0035] S500: Based on the design information, the interest region of each light intensity channel in the light intensity channel set is divided, and the interest region is located according to the division result and mapping result;

[0036] Specifically, it should be understood that to achieve different functions and characteristics in different functional areas of a wafer, wafers are typically composed of multiple materials. The differences in refractive index and transmittance of these materials result in variations in the light transmittance of different parts of the wafer surface. Consequently, the clarity and completeness of various types of defects generated during the wafer manufacturing process vary under different levels of light intensity. The region of interest is the local area of ​​the wafer under test that can accurately reflect wafer defects when illuminated by any level of light intensity.

[0037] This embodiment obtains the material composition and electronic component integration density of each functional area of ​​the wafer under test based on the design information, obtains the light intensity required for clear display of defects under different materials and electronic component integration densities, thereby obtaining the mapping relationship between functional areas and layer light intensities, and then aggregates several functional areas based on layer light intensities to obtain one or more functional areas corresponding to each layer light intensity.

[0038] Each level of light intensity corresponds to one or more functional regions, which represent the region of interest (ROI) division for each light intensity channel in the light intensity channel set. In this embodiment, based on the segmentation of the wafer image under test within the light intensity channel set according to the first and second regions, a secondary segmentation is performed based on the ROI segmentation results. After this secondary segmentation, the local image of the wafer under test retained by each light intensity channel in the light intensity channel set is the ROI image. The ROI image may contain the boundary line between the first and second regions, and the presence of defects on the local wafer surface can be clearly identified based on the ROI image.

[0039] S600: Perform defect identification in the region of interest, and determine a first defect set based on the mapping result;

[0040] Specifically, in this embodiment, a defect recognition model is pre-built based on a BP neural network. The input data of the defect recognition model is wafer images, and the output result is the wafer defect recognition result. The training method of the defect recognition model is as follows: defect images of multiple wafers of the same model in history are collected according to the design information and used as a sample defect image set. The defect types in the sample defect image set are identified and labeled to obtain a sample defect recognition result set. The sample defect image set and the sample defect recognition result set are used as model training data and divided into a training set, a test set, and a validation set in an 8:1:1 ratio. The defect recognition model is trained based on the training set and the test set, and the accuracy of the defect recognition model output is verified based on the validation set. When the defect recognition model's defect recognition accuracy is higher than 95%, the model is considered to have been successfully trained.

[0041] Based on the region of interest, a local image of the wafer to be tested is extracted and input into the defect recognition model to perform wafer defect recognition and obtain the first defect set. The first defect set includes the defect type recognition results of the region of interest corresponding to each level of light intensity channel in the light intensity channel set.

[0042] S700: Perform defect identification on the color channel set and determine a second defect set based on the mapping result;

[0043] In one embodiment, such as Figure 2 As shown, the method steps provided in this application further include:

[0044] S710: Construct a defect dataset for wafers based on big data, wherein the defect dataset carries a region association identifier for the first region and the second region;

[0045] S720: Perform color light sensitivity identification on the defect dataset and determine the color light sensitivity correlation coefficient, wherein the sensitivity correlation coefficient includes a positive correlation coefficient;

[0046] S730: Perform defect identification verification on the color channel set using the sensitive correlation coefficient, and determine the second defect set based on the identification verification result.

[0047] Specifically, in this embodiment, the big data refers to the model and specification information of the wafer under test obtained based on the design information, and the information of historically produced wafers of the same model and specification extracted from historical wafer production data, including but not limited to images of various surface defects on the wafer. Multiple historical wafer surface images are obtained based on the big data, and various surface defects are identified in these images. The first and second regions are mapped to the multiple historical wafer images using the distinguishing features to obtain the frequency of occurrence of various surface defects in the first and second regions. The percentage of the frequency of occurrence of various surface defects in the first and second regions relative to the total frequency of surface defects on the wafer is used as an identifier for each surface defect. This identifier is a regional association identifier between the first and second regions. The defect dataset is constructed based on the multiple surface defects with regional association identifiers.

[0048] Multiple historical wafer entities are obtained, and the multiple historical wafer entities are illuminated with the color light. The integrity of the display of defects under different color lights is quantified to obtain the sensitivity correlation coefficient.

[0049] The specific method for numerically determining the sensitivity correlation coefficient of defects is as follows: Based on the various surface defects, a first surface defect is randomly selected. Multiple images of the first surface defect under various colored lights are obtained based on the colored light illumination from step S200. The maximum width and maximum length values ​​of the multiple first surface defect images are collected, and multiple sets of length-width values ​​that have a mapping relationship with the multiple first surface defect images are obtained. The multiple sets of length-width values ​​are serialized. The colored light corresponding to the first surface defect image with the maximum length and width data is marked as 100. The percentage deviation between the second-ranked length and width data and the maximum length and width data is calculated and multiplied by 100 to obtain the sensitivity correlation coefficient of the colored light corresponding to the second-ranked length and width data. This process is repeated to obtain the sensitivity correlation coefficient of the first surface defect's display completeness under various colored lights. The same method is used to identify the sensitivity correlation coefficients of various surface defects in the defect dataset under various colored lights. It should be understood that, theoretically, no defect is completely undisplayed under a certain light; therefore, the sensitivity correlation coefficients only include positive correlation coefficients.

[0050] Multiple images of the wafer surface to be tested corresponding to the color channel set are input into the defect recognition model constructed in step S600 to perform wafer defect recognition, obtaining the defect recognition result (defect type identification markers of multiple images of the wafer surface to be tested) of the color channel set. The defect recognition result of the color channel set is verified using the sensitive correlation coefficient, and the second defect set is determined based on the verification result. This embodiment will be described in detail in the following description as the optimal embodiment for verifying the defect recognition result of the color channel set using the sensitive correlation coefficient and determining the second defect set based on the verification result.

[0051] This embodiment achieves the technical effect of identifying and locating defects based on their sensitivity to different colors of light by irradiating a wafer under test with different colors of light and identifying defects. This improves the accuracy of defect identification and enhances the reliability and reference value of defect data when optimizing wafer manufacturing processes or detecting defective wafers based on defect identification results.

[0052] S800: Generate the defect identification result of the wafer under test based on the first defect set and the second defect set.

[0053] In one embodiment, such as Figure 3 As shown, the method steps provided in this application further include:

[0054] S810: Record the defect identification results and generate processing feedback information;

[0055] S820: Control and early warning of wafer processing based on the processing feedback information.

[0056] Specifically, in this embodiment, the defect identification result of the wafer under test is generated by combining the first defect set and the second defect set. The defect identification result accurately reflects the types of surface defects of the wafer under test. The defect identification result is recorded, and the processing feedback information is generated and sent to the wafer production line manager. The wafer production line manager determines the processing steps corresponding to the types of surface defects of the wafer under test in the processing feedback information to optimize the wafer production process, or suspends the production of wafers in the same batch under test for production equipment maintenance management, thereby achieving the technical effect of improving the yield of wafer production based on the design information.

[0057] In one embodiment, the method steps provided in this application further include:

[0058] S731: When a defect is identified in any channel of the color channel set, the corresponding sensitivity correlation coefficient is matched according to the defect, and the positive correlation order of the color channels corresponding to the sensitivity correlation coefficient is sorted.

[0059] S732: Extract the first defect identification feature of the corresponding color channel with respect to the defect location, wherein the first defect identification feature includes feature type and feature value;

[0060] S733: When the first defect identification features include non-identical features, an anomaly is marked at the current position, and the second defect set is determined based on the anomaly marking result.

[0061] In one embodiment, the method steps provided in this application further include:

[0062] S733-1: Call the color channel corresponding to the sensitive correlation coefficient of the non-identical features, and extract the second defect identification feature of the color channel;

[0063] S733-2: Determine the number of positive correlation coefficients that satisfy a predetermined threshold based on the sensitive correlation coefficients of the non-identical features, and set a comparison pass threshold based on the corresponding number;

[0064] S733-3: When the number of recognitions of the second defect recognition feature in the color channel exceeds the comparison pass threshold, the corresponding area is determined to be a combined defect area, and the second defect set is determined according to the first defect recognition feature and the second defect recognition feature.

[0065] In one embodiment, the method steps provided in this application further include:

[0066] S732-1: Based on the positive correlation order, classify the coefficient levels of the corresponding positive correlation coefficients and determine the level differences;

[0067] S732-2: If the difference between the positive correlation coefficient level of the first order and the positive correlation coefficient level of the second order exceeds two levels, and the first defect identification features do not include dissimilar features, then the second defect set is obtained based on the feature value corresponding to the first order.

[0068] In one embodiment, the method steps provided in this application further include:

[0069] S732-2-1: Set the sequential aggregation interval, and perform level aggregation of the level division results according to the sequential aggregation interval, and filter to obtain the target aggregation result;

[0070] S732-2-2: Perform correlation analysis between the hierarchical classification results based on the target aggregation results and the feature values;

[0071] S732-2-3: Determine the second defect set based on the results of the correlation analysis.

[0072] This embodiment is a refinement of step S700, and it also verifies the defect identification results of the color channel set through the sensitive correlation coefficient, and determines the optimal embodiment of the second defect set based on the identification verification results.

[0073] Specifically, in this embodiment, the surface image of the wafer under test corresponding to each color channel in the color channel set is input one by one into the defect recognition model constructed in step S600 for wafer defect recognition. When a defect recognition result is obtained after inputting any color channel into the model, it indicates that a defect has been identified in the wafer under test from that color channel, meaning that a defect must exist on the surface of the wafer under test. When a defect is identified, the operation of the defect recognition model is stopped. After determining that a defect exists on the surface of the wafer under test, this embodiment further obtains precise dimensional data of the defect.

[0074] Specifically, in step S600, this embodiment obtains the sensitivity correlation coefficient identifiers of the sensitivity of various surface defects in the defect dataset to illumination by various colors of light. Therefore, this embodiment matches the corresponding surface defect type in the defect dataset and the correlation sensitivity coefficient corresponding to the current color channel. Based on the mapping relationship between the sensitivity correlation coefficient and the color light, and the sensitivity correlation coefficient (positive correlation coefficient) of a surface defect type corresponding to the defect on the surface of the wafer under test under different colors of light, the color channels in the color channel set are sorted in a positive correlation order to serialize the color channel set.

[0075] Based on the identification of the surface defect locations of the wafer under test, the images of the defect locations in all color channels of the color channel set are mapped to them, thereby extracting images of the defect locations in all color channels and obtaining a color channel image set.

[0076] Defect identification is performed based on the set of color channel images. Specifically, the set of color channel images is input one by one into the defect identification model constructed in step S600 to identify wafer defects, and the defect type feature identification results corresponding to multiple color channel images in the set of color channel images are obtained. It should be understood that since the display integrity and clarity of defects are different under different color lights, there may be a case where a defect is displayed under one color light but not under another color light. Therefore, there may be a case where the defect type feature identification result of a certain color channel image includes the aforementioned defect types as well as newly added defect types.

[0077] Based on the defect type feature identification results, the feature values ​​(defect size parameters) of the defect type feature identification results are obtained using existing defect size parameter measurement methods. The defect type feature identification results of the obtained color channel image set are collectively referred to as the feature types, and the defect size parameters of the defect type feature identification results are collectively referred to as feature values. The feature types and the feature values ​​constitute the first defect identification feature.

[0078] In this embodiment, a threshold for classifying positive correlation coefficient levels is preset. For example, the positive correlation coefficient from 0 to 100 is divided into 10 levels, and the threshold for the second level of positive correlation coefficient is [10, 20), and so on, to obtain the classification of positive correlation coefficient levels.

[0079] Based on the positive correlation order, the coefficient levels of the corresponding positive correlation coefficients are divided to obtain the level division result of each color channel image in the color channel image set, and the level difference between two adjacent color channel images is calculated based on the level division result.

[0080] A sequential aggregation interval is set, which is a threshold value for retaining one item from two adjacent color channel images whose level differences are within a certain range and whose positive correlation order is used. The level classification results are aggregated according to the sequential aggregation interval, and a target aggregation result is obtained by filtering. The target aggregation result is a set of color channel images with a reduced number of images obtained by deleting the smaller item from adjacent color channel images whose level differences satisfy the sequential aggregation interval. The sorting of multiple color channel images in the target aggregation result is based on the positive correlation order sorting with carry-over.

[0081] The positive correlation coefficient levels of the first and second orders are the ranking results of two adjacent positive correlation coefficients in the target aggregation result. Based on the first defect identification feature extraction, defect type feature identification results are obtained for the color channel image sets corresponding to the first and second orders.

[0082] If the difference between the positive correlation coefficient level of the first order and the positive correlation coefficient level of the second order exceeds two levels, determine whether the defect type feature identification results of the color channel image sets corresponding to the first order and the second order are consistent, that is, the first defect identification features do not include different features.

[0083] If the difference between the positive correlation coefficient level of the first order and the positive correlation coefficient level of the second order exceeds two levels, and the first defect identification features do not include different features, it indicates that the size of the same type of defect (the same defect) contained in the two adjacent color channel image sets corresponding to the first and second orders is significantly different, and the defect feature value of the color channel image corresponding to the second order is not reliable.

[0084] Therefore, in this embodiment, the second defect set is obtained based on the feature values ​​corresponding to the first sequence. The second defect set is the defect feature location and defect feature value of the wafer surface to be tested.

[0085] When the first defect identification feature includes dissimilar features, it indicates that the defect location contains a defect that only appears under certain colors of light, and the defect location is a mixture of two defects. Therefore, this embodiment identifies the locations containing dissimilar features in the defect location as anomalies to obtain the anomaly identification result.

[0086] The method for obtaining the first defect identification feature is adopted. By calling the color channel corresponding to the sensitive correlation coefficient of the non-identical feature, the color channel is a number of color channels of interest to the defects of the non-identical feature. The second defect identification feature of the currently obtained number of color channels is extracted. The logic for obtaining the second defect identification feature is the same as the logic for obtaining the first defect identification feature.

[0087] The sensitivity correlation coefficients of the defect types corresponding to the non-identical features in the defect dataset under multiple colors of light are obtained. Based on the obtained sensitivity correlation coefficients of multiple colors of light, positive correlation coefficients are extracted. Multiple positive correlation coefficients are compared with a predetermined threshold (the value of a positive correlation coefficient set based on requirements) to obtain the number of samples that meet the predetermined threshold. A comparison pass threshold is set according to the corresponding number. The comparison pass threshold is the standard for recognizing the existence of the non-identical features at the defect location.

[0088] Specifically, when the number of recognized second defect identification features in a color channel exceeds the comparison pass threshold (i.e., when all color channels meeting the pass threshold requirement can identify the second defect identification features with different characteristics), the corresponding region is determined to be a combined defect region with multiple defect types. The feature types and feature values ​​of the different characteristics in the combined defect region are then added to the second defect identification features, updating the second defect identification features. The second defect set is then determined based on the first and second defect identification features.

[0089] This embodiment uses a set of light intensity channels with different levels of light intensity and a set of color channels with different colors to perform multi-dimensional defect detection and verification on the wafer under test. This achieves efficient and high-accuracy wafer defect identification results, thereby reducing the defect rate in wafer production and providing highly reliable reference data for wafer production quality control.

[0090] In one embodiment, such as Figure 4 As shown, a defect testing device for wafer fabrication is provided, comprising: a wafer region segmentation module 1, a light intensity channel construction module 2, a color channel construction module 3, a feature matching execution module 4, a region of interest localization module 5, a defect recognition execution module 6, a defect set determination module 7, and a defect recognition execution module 8, wherein:

[0091] The wafer region partitioning module 1 is used to read the wafer design information, perform wafer region partitioning, and set partitioning features. The region partitioning result includes a first region and a second region, where the first region is an important region and the second region is a regular region.

[0092] The light intensity channel construction module 2 is used to irradiate the wafer under test with layered light intensity, and to acquire images of the irradiation results to construct a set of light intensity channels.

[0093] Color channel construction module 3 is used to illuminate the wafer under test with colored light, and to acquire images of the illumination results to construct a color channel set;

[0094] Feature matching execution module 4 is used to select a standard color channel in the color channel set, and perform feature matching of the standard color channel based on the segmentation feature to determine the regional positions of the first region and the second region, and map them to the light intensity channel set and the color channel set;

[0095] The region of interest (ROI) localization module 5 is used to divide the ROI of each light intensity channel in the light intensity channel set based on the design information, and to locate the ROI based on the division results and mapping results.

[0096] The defect identification execution module 6 is used to identify defects in the region of interest and determine a first defect set based on the mapping result;

[0097] The defect set determination module 7 is used to identify defects in the color channel set and determine a second defect set based on the mapping result;

[0098] The defect identification execution module 8 is used to generate the defect identification result of the wafer under test based on the first defect set and the second defect set.

[0099] In one embodiment, the apparatus further includes:

[0100] A defect set construction unit is used to construct a defect dataset for wafers based on big data, wherein the defect dataset carries a region association identifier for the first region and the second region;

[0101] A sensitive labeling execution unit is used to perform sensitive labeling of the color light on the defect dataset and determine the sensitive correlation coefficient of the color light, wherein the sensitive correlation coefficient includes a positive correlation coefficient;

[0102] The defect identification verification unit is used to verify the defect identification results of the color channel set through the sensitive correlation coefficient, and to determine the second defect set based on the identification verification results.

[0103] In one embodiment, the apparatus further includes:

[0104] The defect association sorting unit is used to sort the color channels in the color channel set according to the positive correlation order of the color channels when a defect is identified in any channel of the color channel set.

[0105] The feature extraction unit is used to extract a first defect identification feature of the corresponding color channel with respect to the defect location, wherein the first defect identification feature includes feature type and feature value;

[0106] An anomaly identification execution unit is used to identify an anomaly at the current position when the first defect identification features include non-identical features, and to determine the second defect set based on the anomaly identification result.

[0107] In one embodiment, the apparatus further includes:

[0108] The defect feature extraction unit is used to call the color channel corresponding to the sensitive correlation coefficient of the non-identical features and extract the second defect identification feature of the color channel.

[0109] The threshold setting unit is used to determine the number of positive correlation coefficients that meet a predetermined threshold based on the sensitive correlation coefficients of the non-identical features, and to set a comparison pass threshold based on the corresponding number.

[0110] The defect set determination unit is used to determine the corresponding region as a combined defect region when the number of recognitions of the second defect recognition feature in the color channel exceeds the comparison pass threshold, and to determine the second defect set based on the first defect recognition feature and the second defect recognition feature.

[0111] In one embodiment, the apparatus further includes:

[0112] The grade difference determination unit is used to classify the coefficient grades of the corresponding positive correlation coefficients according to the positive correlation order and determine the grade difference.

[0113] The defect set acquisition unit is used to obtain the second defect set based on the feature value corresponding to the first order if the difference between the positive correlation coefficient level of the first order and the positive correlation coefficient level of the second order exceeds two levels, and the first defect identification features do not include dissimilar features.

[0114] In one embodiment, the apparatus further includes:

[0115] The hierarchical aggregation execution unit is used to set the sequential aggregation interval, perform hierarchical aggregation of the hierarchical division results according to the sequential aggregation interval, and filter to obtain the target aggregation result;

[0116] A correlation analysis execution unit is used to perform correlation analysis between the hierarchical classification results of the target aggregation results and the feature values;

[0117] The defect set determination unit is used to determine the second defect set based on the results of relevant analysis.

[0118] In one embodiment, the apparatus further includes:

[0119] A processing feedback generation unit is used to record defects based on the defect identification results and generate processing feedback information.

[0120] The processing control early warning unit is used to control and warn of wafer processing based on the processing feedback information.

[0121] For a specific embodiment of a defect testing device for wafer fabrication, please refer to the embodiment of a defect testing method for wafer fabrication described above, which will not be repeated here. Each module in the aforementioned defect testing device for wafer fabrication can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0122] In summary, any of the methods or steps described above can be stored as computer instructions or programs in various types of computer memory, and the computer instructions or programs can be recognized by various types of computer processors to implement any of the above methods or steps.

[0123] Based on the above specific embodiments of the present invention, any improvements and modifications made to the present invention by those skilled in the art without departing from the principle of the present invention shall fall within the patent protection scope of the present invention.

Claims

1. A defect testing method for wafer fabrication, characterized in that, The method includes: Read the wafer design information and perform wafer region division, set the segmentation features, wherein the region division result includes a first region and a second region, the first region is an important region and the second region is a regular region; The wafer under test is irradiated with hierarchical light intensity, and the irradiation results are image acquired to construct a set of light intensity channels; The wafer under test is irradiated with colored light, and the irradiation result is image acquired to construct a color channel set; A standard color channel in the color channel set is selected, and feature matching of the standard color channel is performed based on the segmentation feature to determine the regional positions of the first region and the second region, and then mapped to the light intensity channel set and the color channel set. Based on the design information, the interest region of each light intensity channel in the light intensity channel set is divided, and the interest region is located according to the division result and mapping result. Perform defect identification in the region of interest, and determine a first defect set based on the mapping result; Perform defect identification on the color channel set, and determine a second defect set based on the mapping result; The defect identification results of the wafer under test are generated based on the first defect set and the second defect set.

2. The method as described in claim 1, characterized in that, The method further includes: A defect dataset for wafers is constructed based on big data, wherein the defect dataset carries a regional association identifier between the first region and the second region; The defect dataset is subjected to color light sensitivity labeling, and the color light sensitivity correlation coefficient is determined, wherein the sensitivity correlation coefficient includes positive correlation coefficient; The defect identification results of the color channel set are verified by the sensitive correlation coefficient, and the second defect set is determined based on the identification verification results.

3. The method as described in claim 2, characterized in that, The method further includes: When a defect is identified in any channel of the color channel set, the corresponding sensitivity correlation coefficient is matched according to the defect, and the color channels corresponding to the sensitivity correlation coefficient are sorted in positive correlation order. Extract a first defect identification feature for the corresponding color channel with respect to the defect location, wherein the first defect identification feature includes feature type and feature value; When the first defect identification features include non-identical features, the current position is marked as abnormal, and the second defect set is determined based on the abnormality marking result.

4. The method as described in claim 3, characterized in that, The method further includes: Call the color channel corresponding to the sensitive correlation coefficient of the non-identical features, and extract the second defect identification feature of the color channel; The number of positive correlation coefficients that satisfy a predetermined threshold is determined based on the sensitive correlation coefficients of the non-identical features, and a comparison pass threshold is set according to the corresponding number. When the number of the second defect identification features identified in the color channel exceeds the comparison pass threshold, the corresponding area is determined to be a combined defect area, and the second defect set is determined based on the first defect identification feature and the second defect identification feature.

5. The method as described in claim 4, characterized in that, The method further includes: Based on the positive correlation order, the coefficient levels of the corresponding positive correlation coefficients are divided, and the level differences are determined. If the difference between the positive correlation coefficient level of the first order and the positive correlation coefficient level of the second order exceeds two levels, and the first defect identification features do not include dissimilar features, then the second defect set is obtained based on the feature value corresponding to the first order.

6. The method as described in claim 5, characterized in that, The method further includes: Set sequential aggregation intervals, aggregate the level division results according to the sequential aggregation intervals, and filter to obtain the target aggregation results; A correlation analysis is performed between the hierarchical classification results of the target aggregation results and the feature values; The second defect set was determined based on the results of the relevant analysis.

7. The method as described in claim 1, characterized in that, The method further includes: The defect identification results are recorded, and processing feedback information is generated. Control and early warning of wafer processing are performed based on the processing feedback information.

8. A defect testing device for wafer processing, characterized in that, The device includes: The wafer region partitioning module is used to read the wafer design information, perform wafer region partitioning, and set partitioning features. The region partitioning result includes a first region and a second region, where the first region is an important region and the second region is a regular region. The light intensity channel construction module is used to irradiate the wafer under test with layered light intensity, acquire images of the irradiation results, and construct a set of light intensity channels. The color channel construction module is used to illuminate the wafer under test with colored light, acquire images of the illumination results, and construct a color channel set. The feature matching execution module is used to select a standard color channel in the color channel set, and perform feature matching of the standard color channel based on the segmentation feature to determine the regional positions of the first region and the second region, and map them to the light intensity channel set and the color channel set; The region of interest (ROI) localization module is used to divide the ROI of each light intensity channel in the light intensity channel set based on the design information, and to locate the ROI based on the division results and mapping results. The defect recognition execution module is used to perform defect recognition in the region of interest and determine a first defect set based on the mapping result; The defect set determination module is used to identify defects in the color channel set and determine a second defect set based on the mapping result; The defect identification execution module is used to generate the defect identification result of the wafer under test based on the first defect set and the second defect set.

Images