A method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis

Abstract

This application relates to the field of epitaxial wafer defect detection technology, specifically to a VCSEL epitaxial wafer surface defect detection method based on optical analysis. The method includes: acquiring a substrate image; acquiring all epitaxial images and temperature data in real time; extracting suspected defect regions from all images; obtaining a growth influence sequence for each suspected defect region in the last epitaxial image based on the overlap and density characteristics of the suspected defect regions in each epitaxial image and the substrate image; obtaining the significance value of small pit particles in each suspected defect region in the last epitaxial image based on the changing trend of the obtained growth influence sequence, its similarity to the changing trend of the temperature data, and the gray-scale mean value of each suspected defect region; and then determining whether each suspected defect region is a small pit particle defect region. This application improves the accuracy of epitaxial wafer surface defect detection by analyzing the characteristics of small pit defects under equipment interference and temperature changes.

Description

Technical Field

[0001] This application relates to the field of epitaxial wafer defect detection technology, specifically to a method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis. Background Technology

[0002] Vertical-cavity surface-emitting lasers (VCSELs) are experiencing rapid market growth due to their excellent beam quality, simple design, and compact size. They are commonly used in 3D cameras and mobile devices for facial recognition and machine vision technologies. Epitaxial wafers are a key material in the manufacture of VCSEL equipment, and their processing quality directly affects the performance of the VCSEL. Furthermore, various defects are easily generated during the processing of epitaxial wafers, making surface defect detection particularly important.

[0003] Optical analysis and inspection methods are widely used because they are non-contact, do not cause physical damage to epitaxial wafers, and can quickly identify various defects on the epitaxial wafer surface. However, in actual inspection processes, insufficient consistency and stability of the illumination intensity of the inspection equipment may make it difficult to accurately identify minute defects. In addition, substrate crystal defects and epitaxial growth parameters may interfere, leading to further defects on the epitaxial wafer surface. Conventional methods do not fully consider the interference of the inspection equipment and the influence of the epitaxial state during defect detection, resulting in inaccurate detection of surface defects on epitaxial wafers. Summary of the Invention

[0004] To address the aforementioned technical problems, this application provides a method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis, thereby resolving existing issues.

[0005] The method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis in this application adopts the following technical solution:

[0006] One embodiment of this application provides a method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis. The method includes the following steps:

[0007] Acquire substrate images and obtain epitaxial images and temperature data in real time throughout the entire epitaxial growth process;

[0008] Extract all suspected defect regions from the substrate image and all epitaxial images; based on the degree of overlap between each suspected defect region in each epitaxial image and the suspected defect region in the substrate image, and the density of suspected defect regions in the neighborhood of each suspected defect region in each epitaxial image, obtain the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image.

[0009] Based on the epitaxial growth influence coefficients of all suspected defect regions in all epitaxial images that are at the same location as the suspected defect regions in the last epitaxial image, the growth influence sequence of each suspected defect region in the last epitaxial image is obtained. Based on the changing trend of the obtained growth influence sequence and the gray-scale mean of each suspected defect region, the first outlier of each suspected defect region in the last epitaxial image is obtained. Combined with the similarity of the changing trend between the growth influence sequence of each suspected defect region and all temperature data, the significance value of small pit particles in each suspected defect region in the last epitaxial image is obtained, and then it is determined whether each suspected defect region in the last epitaxial image is a small pit particle defect region.

[0010] Preferably, the specific process of extracting all suspected defect regions in the substrate image and all epitaxial images is as follows: pixels with gray values ​​greater than a preset segmentation threshold in the substrate image and all epitaxial images are recorded as suspected defect pixels, and then all suspected defect pixels in the substrate image and all epitaxial images are used as inputs to a region growing algorithm to obtain all suspected defect regions in the substrate image and all epitaxial images.

[0011] Preferably, the method for obtaining the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is as follows: statistically analyzing the substrate defect influence coefficient of each suspected defect region in each epitaxial image; statistically analyzing the distribution concentration coefficient of each suspected defect region in each epitaxial image; the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is positively correlated with both the substrate defect influence coefficient and the distribution concentration coefficient.

[0012] Preferably, the substrate defect influence coefficient of each suspected defect region in each epitaxial image refers to the ratio of the total number of suspected defect pixels in the same region of the substrate image to the total number of pixels in each suspected defect region.

[0013] Preferably, the method for obtaining the distribution concentration coefficient of each suspected defect region in each extensional image is as follows: obtaining the center pixel of each suspected defect region in each extensional image; clustering all center pixels in each extensional image to obtain multiple clusters and the local density corresponding to each center pixel; calculating the Euclidean distance between the center pixel of each suspected defect region in each extensional image and the cluster center of its respective cluster; and recording the ratio of the local density of the center pixel of each suspected defect region in each extensional image to the Euclidean distance as the distribution concentration coefficient of each suspected defect region in each extensional image.

[0014] Preferably, the method for obtaining the growth influence sequence of each suspected defect region in the last epitaxial image is as follows: using the synchronous positioning function of the particle tester, the suspected defect regions in all epitaxial images that belong to the same position as each suspected defect region in the last epitaxial image are obtained, and the epitaxial growth influence coefficients of all the suspected defect regions belonging to the same position are arranged in ascending order of time to obtain the growth influence sequence of each suspected defect region in the last epitaxial image.

[0015] Preferably, the method for obtaining the first outlier of each suspected defect region in the last extensional image is as follows: obtaining the statistics of the growth influence sequence of each suspected defect region in the last extensional image; calculating the ratio of the average gray value of all pixels in each suspected defect region in the last extensional image to the maximum gray value in its image; and using the product of the normalized result of the statistics and the ratio as the first outlier of each suspected defect region in the last extensional image.

[0016] Preferably, the method for obtaining the significant value of the pit particles in each suspected defect region in the last epitaxial image is as follows: count the second outlier value of each suspected defect region in the last epitaxial image; and record the ratio of the first outlier value to the second outlier value of each suspected defect region in the last epitaxial image as the significant value of the pit particles in each suspected defect region in the last epitaxial image.

[0017] Preferably, the second outlier of each suspected defect region in the last epitaxial image refers to the DTW distance between the temperature sequence and the growth influence sequence of each suspected defect region in the last epitaxial image; the temperature sequence refers to the sequence composed of all temperature data arranged in chronological order throughout the entire epitaxial growth process.

[0018] Preferably, the specific process of determining whether each suspected defect region in the last epitaxial image is a pit particle defect region is as follows: if the significant value of the pit particles in any suspected defect region in the last epitaxial image is greater than a preset defect threshold, then the suspected defect region is determined to be a pit particle defect region.

[0019] This application has at least the following beneficial effects:

[0020] This application proposes a VCSEL epitaxial wafer surface defect detection method based on optical analysis. By deeply analyzing the degree of influence of substrate material defects on different regions in each epitaxial image and the distribution characteristics of each suspected defect region, an epitaxial growth influence coefficient is constructed, which can reflect the probability that each suspected defect region in each epitaxial image belongs to a small pit defect. Furthermore, considering the changing trend of the epitaxial growth influence coefficient of each position in the epitaxial wafer over time, as well as the brightness characteristics of each position, and combining the changing characteristics of epitaxial growth temperature and defect characteristics, a small pit particle significance value is constructed to evaluate the probability that each suspected defect region belongs to a small pit defect. This achieves the detection of defects on the epitaxial wafer surface. Its beneficial effects are that it can reduce the influence of equipment interference and epitaxial state, more accurately identify small pit defects on the epitaxial wafer surface, and improve the accuracy of epitaxial wafer surface defect detection. Attached Figure Description

[0021] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating the steps of the VCSEL epitaxial wafer surface defect detection method based on optical analysis provided in this application;

[0023] Figure 2 A flowchart illustrating the process of obtaining the significance values ​​of small pit particles in each suspected defect region in the last epitaxial image provided in this application. Detailed Implementation

[0024] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the VCSEL epitaxial wafer surface defect detection method based on optical analysis proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0026] The following description, in conjunction with the accompanying drawings, details the specific scheme of the VCSEL epitaxial wafer surface defect detection method based on optical analysis provided in this application.

[0027] This application provides an embodiment of a VCSEL epitaxial wafer surface defect detection method based on optical analysis. Specifically, the method is described below. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0028] Step 1: Acquire substrate images and obtain epitaxial images and temperature data in real time throughout the entire epitaxial growth process.

[0029] With the development of the semiconductor industry, the linewidth of processes is constantly shrinking and the integration density is constantly increasing. Therefore, identifying and controlling defects during epitaxial wafer fabrication has become particularly important. Particle analyzers detect defects on the surface of epitaxial wafers by collecting the intensity of reflected light. The intensity of the reflected light characterizes the degree of defects present on the epitaxial wafer surface.

[0030] Epitaxial wafers are high-quality single-crystal thin films formed on substrate materials through epitaxial growth technology. Considering that defects in the substrate material itself may cause surface defects in the epitaxial wafer, a particle analyzer is used to acquire substrate images, and surface images of the epitaxial wafer are acquired in real time during the epitaxial growth process, which are subsequently recorded as epitaxial images. Surface defects on the epitaxial wafer are usually formed by the continuous stacking of atoms during epitaxial growth. Epitaxial parameters can also lead to surface defects, among which temperature is an important epitaxial parameter that causes defects. Therefore, the temperature during the epitaxial processing is acquired in real time. The epitaxial images and temperature data are acquired synchronously, with an acquisition time interval of Tmin, where T is a preset time interval, and in this embodiment, T is set to 1.

[0031] Step 2: Extract all suspected defect regions from the substrate image and all epitaxial images; based on the degree of overlap between each suspected defect region in each epitaxial image and the suspected defect region in the substrate image, and the density of suspected defect regions in the neighborhood of each suspected defect region in each epitaxial image, obtain the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image.

[0032] Surface defects on epitaxial wafers can manifest as papillae, small pits, polishing cloth imprints, and sintering features. When using a particle analyzer for inspection, small pit defects are difficult to identify accurately due to their small size. Therefore, this application focuses on detecting small pit defects on the surface of epitaxial wafers. Considering that the formation of surface defects on epitaxial wafers is influenced by various factors, such as defects in the substrate material and the parameter state during epitaxial growth, both of which can lead to the formation of small pits, and that equipment instability during inspection can easily cause misjudgments, the collected data is processed as follows to improve the accuracy of surface defect detection.

[0033] First, there is a certain correspondence between the substrate material and the pits on the epitaxial wafer surface. For example, if a defect exists at a certain location in the substrate material, this defect may grow to the epitaxial layer during the growth process, resulting in a corresponding defect at the same location on the epitaxial wafer surface. Therefore, the location of a defect in the substrate material is more likely to result in a pit defect at the same location on the epitaxial wafer after epitaxial growth. Given this, taking the i-th epitaxial image as an example, the pixels corresponding to surface defects have higher brightness, while the pixels in non-defective areas have lower brightness. Therefore, the grayscale values ​​of all pixels in the i-th epitaxial image are used as input to the Kittler minimum error method to obtain a segmentation threshold, denoted as the preset segmentation threshold. Pixels with grayscale values ​​greater than the preset segmentation threshold are designated as suspected defect pixels. Then, all suspected defect pixels are used as input to a region growing algorithm to obtain all suspected defect regions in the i-th epitaxial image. Using the same steps, replacing the epitaxial image with a substrate image allows the acquisition of all suspected defect pixels and suspected defect regions in the substrate image. Then, using the synchronous positioning function of the particle tester, the differences in defects at the same location on the substrate image and the epitaxial image are compared. Specifically, for a suspected defect area in the epitaxial image, if a suspected defect area also exists in the same area of ​​the substrate image, the higher the degree of overlap between the suspected defect areas, the greater the probability that the location on the epitaxial wafer surface belongs to a small pit particle. Therefore, taking the j-th suspected defect area in the i-th epitaxial image as an example, the total number of suspected defect pixels in the same area of ​​the substrate image is calculated. The ratio of the obtained total number of suspected defect pixels to the total number of pixels in the j-th suspected defect area is denoted as the substrate defect influence coefficient of the j-th suspected defect area in the i-th epitaxial image, and is denoted as [missing information]. The result The larger the value, the greater the influence of substrate defects on the suspected defect region on the epitaxial wafer surface. The Kittler minimum error method and region growth algorithm are both well-known techniques, and their specific processes will not be elaborated further.

[0034] Furthermore, due to the uneven growth rate of the epitaxial growth reaction, the pit particles in the epitaxial wafer exhibit a centrally clustered distribution. Specifically, the distribution of pit particles is denser at the center of the epitaxial wafer, while it is sparser at the edges. Areas with denser distribution typically experience greater thermal stress, making them more prone to pit particle defects. Therefore, for the i-th epitaxial image, this application first uses OpenCV's `moments` function to obtain the center pixel of each suspected defect region. Then, it uses a peak density clustering algorithm to cluster all center pixels in the i-th epitaxial image, obtaining multiple clusters and the local density corresponding to each center pixel. The peak density clustering algorithm is a well-known technique, and its specific process will not be elaborated further. The Euclidean distance between the center pixel of the j-th suspected defect region in the i-th epitaxial image and the cluster center of its respective cluster is calculated. The ratio of the local density of the center pixel of the j-th suspected defect region in the i-th epitaxial image to the Euclidean distance is then used as the distribution concentration coefficient of the j-th suspected defect region in the i-th epitaxial image, denoted as [equation missing]. The local density reflects the density of each central pixel within a local area; the density of the cluster centers of each cluster is relatively high. The smaller the obtained Euclidean distance, the closer the central pixel is to a high-density region, and the more likely the suspected defect area corresponding to that central pixel is a small pit particle defect. The larger the value, the denser the distribution of suspected defect regions in the neighborhood of the j-th suspected defect region in the i-th extrapolation image, and the more likely the suspected defect region is to be a small pit particle defect.

[0035] Thus, the substrate defect influence coefficient and distribution concentration coefficient of each suspected defect region in the i-th epitaxial image are obtained. These two coefficients reflect the probability that the corresponding suspected defect region belongs to a small pit particle defect from the perspectives of substrate material influence and distribution characteristics of the suspected defect region, respectively. Therefore, as a preferred embodiment, the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is obtained based on the degree of overlap between each suspected defect region in each epitaxial image and the suspected defect region in the substrate image, and the density of suspected defect regions in the neighborhood of each suspected defect region in each epitaxial image. This coefficient is used to characterize the probability that each suspected defect region in each epitaxial image contains a small pit particle defect. The method for obtaining the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is as follows: statistically analyze the substrate defect influence coefficient of each suspected defect region in each epitaxial image; statistically analyze the distribution concentration coefficient of each suspected defect region in each epitaxial image; the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is positively correlated with both the substrate defect influence coefficient and the distribution concentration coefficient. It should be noted that the positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases).

[0036] Preferably, in this embodiment, the epitaxial growth influence coefficient of the j-th suspected defect region in the i-th epitaxial image is denoted as... Its specific expression is: In the formula, The epitaxial growth influence coefficient of the j-th suspected defect region in the i-th epitaxial image; Let be the substrate defect influence coefficient of the j-th suspected defect region in the i-th epitaxial image; Let be the distribution concentration coefficient of the j-th suspected defect region in the i-th epitaxial image. The larger the obtained epitaxial growth influence coefficient, the greater the probability that the corresponding suspected defect region contains small pit particle defects.

[0037] Step 3: Based on the epitaxial growth influence coefficients of all suspected defect regions in all epitaxial images that are at the same location as the suspected defect regions in the last epitaxial image, obtain the growth influence sequence of each suspected defect region in the last epitaxial image; based on the changing trend of the obtained growth influence sequence and the gray-scale mean of each suspected defect region, obtain the first outlier of each suspected defect region in the last epitaxial image, and combine the similarity of the changing trend between the growth influence sequence of each suspected defect region and all temperature data to obtain the significance value of small pit particles in each suspected defect region in the last epitaxial image, and then determine whether each suspected defect region in the last epitaxial image is a small pit particle defect region.

[0038] Furthermore, during the epitaxial growth process, as the surface atomic layers are continuously stacked, the number and size of the corresponding small pit particle defects may increase, and the influence and distribution characteristics of the corresponding substrate defects will become more significant. Based on this characteristic, it is helpful to assess whether it belongs to small pit particle defects.

[0039] Therefore, taking the k-th suspected defect region in the last epitaxial image as an example, the synchronous positioning function of the particle tester is used to obtain suspected defect regions in all epitaxial images that belong to the same location as the k-th suspected defect region in the last epitaxial image. The epitaxial growth influence coefficients of all the obtained suspected defect regions are arranged in ascending order of time to obtain the growth influence sequence of the k-th suspected defect region in the last epitaxial image. The Mankendall trend test is then used to obtain the change characteristics of the growth influence sequence. The output of this algorithm is the trend test statistic. The larger the obtained statistic, the more obvious the small pit particle defect characteristics exhibited by the k-th suspected defect region in the last epitaxial image with epitaxial growth. The Mankendall trend test is a well-known technique, and the specific process will not be elaborated here.

[0040] Furthermore, in suspected defect areas with significant pit particle characteristics, the greater the intensity of reflected light collected by the particle tester, the higher the brightness of the corresponding pixel in the epitaxial surface image. Therefore, in the last epitaxial image during the entire epitaxial growth process, the ratio of the average gray value of all pixels in the k-th suspected defect area to the maximum gray value in its corresponding image is taken as the brightness significance value of the k-th suspected defect area. The obtained brightness significance value reflects the significant brightness characteristics of this suspected defect area.

[0041] Furthermore, the product of the normalized result of the statistics of the growth influence sequence of the k-th suspected defect region in the last epitaxial image and the brightness significance value is recorded as the first outlier of the k-th suspected defect region in the last epitaxial image. The obtained first outlier reflects the abnormal characteristics of the epitaxial growth change and brightness state of the suspected defect region. In this embodiment, the sigmoid function is used for normalization.

[0042] Furthermore, epitaxial growth temperature has a significant impact on the formation of pit defects. At high temperatures, atomic migration on the epitaxial wafer surface accelerates, promoting the formation of more pit particles. These pit particles then tend to gradually diffuse and extend. Therefore, the more similar the variation characteristics between the temperature data and the growth influence sequences of each suspected defect region, the more significant the influence of the epitaxial growth temperature on the formation of pit particles in the corresponding suspected defect region. Thus, all temperature data during the entire epitaxial growth process are arranged in chronological order to obtain a temperature sequence. The DTW distance between the temperature sequence and the growth influence sequence of the k-th suspected defect region in the last epitaxial image is calculated and denoted as the second outlier of the k-th suspected defect region in the last epitaxial image. The smaller the second outlier, the more significantly the defect characteristics of the k-th suspected defect region are affected by temperature changes. The calculation of the DTW distance is a well-known technique, and the specific process will not be elaborated further.

[0043] Furthermore, the ratio of the first outlier to the second outlier in the k-th suspected defect region of the last extrapolation image is recorded as the saliency value of the pit particles in the k-th suspected defect region of the last extrapolation image. The larger the saliency value, the more likely the suspected defect region is to belong to the pit particle defect region. The flowchart for obtaining the saliency value of the pit particles in each suspected defect region of the last extrapolation image is as follows. Figure 2 As shown.

[0044] Similarly, the saliency values ​​of small pit particles in all suspected defect areas of the last epitaxial image are obtained, and these values ​​are used as input to the Otsu thresholding algorithm to obtain a segmentation threshold, which is then recorded as the preset defect threshold. The Otsu thresholding algorithm is a well-known technique, and its specific process will not be elaborated further. If the saliency value of small pit particles in any suspected defect area of ​​the last epitaxial image is greater than the preset defect threshold, then the suspected defect area is determined to be a small pit particle defect area. Thus, the defect areas in the epitaxial wafer are obtained, completing the detection of surface defects on the epitaxial wafer. Using the above method to detect surface defects on epitaxial wafers helps improve the accuracy of surface defect detection.

[0045] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0046] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0047] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them; modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some of the technical features, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting surface defects in VCSEL epitaxial wafers based on optical analysis, characterized in that, The method includes the following steps: Acquire substrate images and obtain epitaxial images and temperature data in real time throughout the entire epitaxial growth process; Extract all suspected defect regions from the substrate image and all epitaxial images; based on the degree of overlap between each suspected defect region in each epitaxial image and the suspected defect region in the substrate image, and the density of suspected defect regions in the neighborhood of each suspected defect region in each epitaxial image, obtain the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image. Based on the epitaxial growth influence coefficients of all suspected defect regions in all epitaxial images that are at the same location as the suspected defect regions in the last epitaxial image, the growth influence sequence of each suspected defect region in the last epitaxial image is obtained. Based on the changing trend of the obtained growth influence sequence and the gray-scale mean of each suspected defect region, the first outlier of each suspected defect region in the last epitaxial image is obtained. Combined with the similarity of the changing trend between the growth influence sequence of each suspected defect region and all temperature data, the significance value of small pit particles in each suspected defect region in the last epitaxial image is obtained, and then it is determined whether each suspected defect region in the last epitaxial image is a small pit particle defect region. The method for obtaining the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is as follows: statistically analyze the substrate defect influence coefficient of each suspected defect region in each epitaxial image; statistically analyze the distribution concentration coefficient of each suspected defect region in each epitaxial image; the epitaxial growth influence coefficient of each suspected defect region in each epitaxial image is positively correlated with both the substrate defect influence coefficient and the distribution concentration coefficient. The substrate defect influence coefficient of each suspected defect region in each epitaxial image refers to the ratio of the total number of suspected defect pixels in the same region of the substrate image to the total number of pixels in each suspected defect region. The method for obtaining the distribution concentration coefficient of each suspected defect region in each extensional image is as follows: obtain the center pixel of each suspected defect region in each extensional image; cluster all center pixels in each extensional image to obtain multiple clusters and the local density corresponding to each center pixel; calculate the Euclidean distance between the center pixel of each suspected defect region in each extensional image and the cluster center of its respective cluster; and record the ratio of the local density of the center pixel of each suspected defect region in each extensional image to the Euclidean distance as the distribution concentration coefficient of each suspected defect region in each extensional image. The method for obtaining the growth influence sequence of each suspected defect region in the last epitaxial image is as follows: using the synchronous positioning function of the particle tester, the suspected defect regions in all epitaxial images that belong to the same position as each suspected defect region in the last epitaxial image are obtained. The epitaxial growth influence coefficients of all the suspected defect regions that belong to the same position are arranged in ascending order of time to obtain the growth influence sequence of each suspected defect region in the last epitaxial image. The method for obtaining the first outlier of each suspected defect region in the last extensional image is as follows: obtain the statistics of the growth influence sequence of each suspected defect region in the last extensional image; calculate the ratio of the average gray value of all pixels in each suspected defect region in the last extensional image to the maximum gray value in its image; and take the product of the normalized result of the statistics and the ratio as the first outlier of each suspected defect region in the last extensional image. The method for obtaining the significant value of the small pit particles in each suspected defect region in the last extensional image is as follows: count the second outlier value of each suspected defect region in the last extensional image; and record the ratio of the first outlier value to the second outlier value of each suspected defect region in the last extensional image as the significant value of the small pit particles in each suspected defect region in the last extensional image.

2. The method for detecting surface defects of VCSEL epitaxial wafers based on optical analysis as described in claim 1, characterized in that, The specific process for extracting all suspected defect regions in the substrate image and all epitaxial images is as follows: pixels with gray values ​​greater than a preset segmentation threshold in the substrate image and all epitaxial images are recorded as suspected defect pixels. Then, all suspected defect pixels in the substrate image and all epitaxial images are used as inputs to a region growing algorithm to obtain all suspected defect regions in the substrate image and all epitaxial images.

3. The method for detecting surface defects of VCSEL epitaxial wafers based on optical analysis as described in claim 1, characterized in that, The second outlier in each suspected defect region of the last epitaxial image refers to the DTW distance between the temperature sequence and the growth influence sequence of each suspected defect region in the last epitaxial image; the temperature sequence refers to the sequence composed of all temperature data arranged in chronological order throughout the entire epitaxial growth process.

4. The method for detecting surface defects of VCSEL epitaxial wafers based on optical analysis as described in claim 1, characterized in that, The specific process for determining whether each suspected defect region in the last epitaxial image is a small pit particle defect region is as follows: if the significant value of small pit particles in any suspected defect region in the last epitaxial image is greater than a preset defect threshold, then the suspected defect region is determined to be a small pit particle defect region.

Images