VCSEL epitaxial wafer thickness uniformity detection method

Abstract

The invention belongs to the technical field of image processing, and particularly relates to a VCSEL epitaxial wafer thickness uniformity detection method comprising the following steps: obtaining a thickness map and a yield map of a VCSEL epitaxial wafer after space alignment; obtaining the global background thickness of each position in the thickness map; calculating the fluctuation standard deviation of the thickness deviation in the neighborhood of each position; obtaining a neighborhood abnormal score and a neighborhood failure rate, and obtaining a neighborhood reference parameter; calculating a problem attribution risk index according to the product of the neighborhood anomaly score and the neighborhood failure rate of each position, the neighborhood reference parameter and the Pearson's correlation coefficient and the relative deviation and correlation weighting relation; and inferring an abnormal mode type according to the problem attribution risk index, and generating an attribution report. According to the invention, the problem that the correlation attribution capability of the thickness and the yield of the VCSEL epitaxial wafer in the prior art is lacked is solved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method for detecting the thickness uniformity of VCSEL epitaxial wafers. Background Technology

[0002] As a core light source device for cutting-edge technologies such as 3D sensing, optical communication, and lidar, the performance of VCSELs is highly dependent on the quality of epitaxial wafer growth. Among them, the uniformity of epitaxial wafer thickness is a fundamental physical parameter that determines key performance indicators such as chip wavelength consistency, threshold current, and photoelectric conversion efficiency. Therefore, accurate and effective detection of epitaxial wafer thickness uniformity is a prerequisite for ensuring the yield and performance stability of VCSEL chips in large-scale mass production.

[0003] Currently, the commonly used detection method in the industry is to use high-precision optical measurement equipment, such as ellipsometers and white light interferometers, to scan the entire epitaxial wafer after epitaxial growth is completed, and generate a thickness distribution map. Engineers manually interpret the map to find areas of abnormal thickness, and combine it with the sampling test results of the terminal chip to infer process problems. This method can intuitively show the physical distribution of thickness and achieve preliminary monitoring of thickness uniformity to a certain extent.

[0004] However, this method has significant limitations in in-depth analysis and attribution localization. First, manual interpretation of thickness maps relies on the personal experience of engineers. For complex or inconspicuous thickness anomalies, the judgment criteria vary, leading to strong subjectivity and a high risk of missed or incorrect diagnoses. More importantly, this method separates thickness anomalies from yield losses, failing to answer whether a thickness deviation in a particular area is the true cause of the chip yield decline in that area, as yield losses can also be caused by other factors such as substrate defects and process contamination. Existing methods lack an effective algorithm to assess the strength of the causal relationship between thickness anomalies and yield losses, thus failing to free analysts from sifting through massive thickness fluctuations to accurately pinpoint the true cause of the problem, resulting in low efficiency in problem attribution and long process optimization cycles. Summary of the Invention

[0005] To address the technical problem of the lack of attribution ability of existing technologies regarding the correlation between VCSEL epitaxial wafer thickness and yield, this invention provides a method for detecting the thickness uniformity of VCSEL epitaxial wafers, comprising: Obtain the thickness map and yield map of the VCSEL epitaxial wafer after spatial alignment; for each position in the thickness map, calculate the weighted average of the original thickness values ​​in its neighborhood based on the Euclidean distance from each point in the neighborhood to that position to obtain the global background thickness at that position; calculate the difference between the original thickness value at each position in the thickness map and the global background thickness to obtain the thickness deviation; calculate the standard deviation of the thickness deviation fluctuation in the neighborhood of each position; use the ratio of the absolute value of the thickness deviation at each position to the standard deviation of the fluctuation as the neighborhood anomaly score; calculate the yield map of each position... The average yield value of each point within the neighborhood of a given location is used, and the difference between 1 and the average yield value is taken as the neighborhood failure rate. Within the neighborhood of each location, the average of the product of the neighborhood anomaly score and the neighborhood failure rate is calculated to obtain the neighborhood baseline parameter. Based on the relative deviation and correlation weighting relationship between the product of the neighborhood anomaly score and the neighborhood failure rate, the neighborhood baseline parameter, and the Pearson correlation coefficient, a problem attribution risk index is calculated. The anomaly pattern type is inferred based on the problem attribution risk index. An attribution report is generated based on the anomaly pattern type.

[0006] In existing technologies, the standards for manually judging epitaxial wafer thickness anomalies are inconsistent, easily leading to missed or incorrect judgments. Furthermore, it's impossible to establish an effective correlation between thickness and yield, making it difficult to determine whether thickness anomalies truly cause yield loss. Failures caused by other factors such as test malfunctions or impurities may also be attributed to thickness issues, resulting in low attribution efficiency and blind process optimization. This invention solves this problem through a multi-step design: first, accurately aligning thickness and yield graphs to ensure data matching; second, calculating reasonable thickness benchmarks and fluctuation ranges to distinguish between normal and abnormal fluctuations; third, statistically analyzing regional failures and setting regional benchmarks to eliminate accidental interference and inherent fluctuations; and finally, accurately determining the causal relationship between thickness and failures through a risk index, identifying abnormal patterns, matching attributions, and providing suggestions. The entire process eliminates the need for subjective human judgment, accurately pinpointing the true thickness problem, avoiding blind adjustments, shortening process improvement time, significantly improving the accuracy and efficiency of thickness detection and problem attribution, reducing the generation of defective products, and addressing the core pain point of existing technologies' lack of ability to correlate thickness and yield.

[0007] Preferably, obtaining the thickness map and yield map of the VCSEL epitaxial wafer after spatial alignment includes: For a single VCSEL epitaxial wafer, the thickness distribution data after its growth is completed is obtained and recorded as a thickness map; at the same time, the yield distribution data after the same epitaxial wafer has gone through the chip manufacturing and testing process is obtained and recorded as a yield map; an image registration algorithm based on feature points is used to accurately align the thickness map and the yield map in space.

[0008] Preferably, the global background thickness satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map Global background thickness; Indicates position in thickness map The set of all neighboring points is denoted as the neighborhood set, and the size of the neighborhood is a radius of 3. λ A circular area; Representing the neighborhood One point of responsibility The original thickness value; In the thickness map, the neighboring points With position The neighborhood distance; This represents the smooth bandwidth parameter.

[0009] This invention calculates the normal thickness baseline at each location using a specific method, clearly distinguishing between overall thickness variations and local anomalies in the epitaxial wafer. It avoids interference from natural thickness variations, such as those from the edge to the center, preventing the misinterpretation of normal overall changes as local anomalies. Simultaneously, the defined neighborhood range reduces the number of points requiring calculation while maintaining accuracy, accelerating computation and minimizing time spent on complex calculations. This provides an accurate normal baseline for subsequent identification of local thickness anomalies, improving the reliability of anomaly detection.

[0010] Preferably, the standard deviation of the fluctuation satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map The standard deviation of the fluctuation; Indicates the midpoint of the thickness map The thickness deviation of point j in the neighborhood set; Indicates position in thickness map The neighborhood set; Midpoint of the thickness map The number of points in the neighborhood set, >1; Position in the thickness map The neighborhood average deviation.

[0011] This invention calculates the fluctuation of thickness deviation in each region to understand the normal fluctuation range of different regions. For example, some regions naturally have small fluctuations, while others have slightly larger fluctuations. It avoids mistaking normal deviations in regions with large fluctuations for abnormalities, and also avoids overlooking minor anomalies in regions with small fluctuations. This prevents misjudging normal conditions as abnormalities or missing genuine anomalies, making the assessment of thickness deviation more consistent with the actual conditions of different regions, and providing a reasonable basis for subsequent measurement of the degree of anomaly.

[0012] Preferably, the neighborhood anomaly score includes: Calculate the midpoint of the thickness map The absolute difference between the original thickness value and its global background thickness, and the midpoint of the thickness map. The absolute difference between the original thickness value and its global background thickness and the midpoint of the thickness map. Compare the standard deviations of the fluctuations within the same neighborhood, and add a very small positive number to the denominator of the ratio. The results are used as neighborhood anomaly scores.

[0013] Preferably, the neighborhood failure rate satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map Neighborhood failure rate; This represents the number of points in the neighborhood set. Represent the yield profile and locate the midpoint of the thickness profile. points in the neighborhood set Yield value; Indicates position in thickness map The neighborhood set.

[0014] This invention calculates the overall failure status of a region, avoiding being misled by the accidental failure of a single point. For example, if a certain point is occasionally unqualified, it will not lead to the conclusion that the entire region is problematic. By statistically analyzing the overall situation within the region, it can reflect the true performance level of the region, help establish the relationship between thickness and regional failure, and prevent the judgment from being affected by the special circumstances of a single point, making the evaluation of regional performance more objective.

[0015] Preferably, obtaining neighborhood reference parameters includes: Point in the thickness map Within the neighborhood set, calculate the average of the product of the neighborhood failure rate and the neighborhood anomaly score of all neighborhood points, and use this as the neighborhood baseline parameter.

[0016] Preferably, the problem attribution risk index satisfies the following expression: ; In the formula, This represents the attribution risk index of point k in the thickness map. , Indicates the midpoint of the thickness map Neighborhood anomaly score, neighborhood failure rate; The Pearson correlation coefficient represents the data sequence of neighborhood anomaly scores of all neighboring points in the neighborhood set of point k in the thickness map, arranged by position, and the data sequence of neighborhood failure rates of all neighboring points arranged by the same position. Represents the neighborhood reference parameters; Represents the maximum value function; Represents an exponential function with the natural constant as its base; It represents a very small positive number, and guarantees that the denominator is not 0.

[0017] This invention calculates a risk index using this method, accurately distinguishing whether a thickness issue truly caused the failure. It avoids attributing failures caused by test malfunctions, impurities, or other factors to thickness problems, eliminating irrelevant interference. Simultaneously, it amplifies the risk differences between different regions, facilitating the rapid identification of key problem areas and preventing errors in calculations due to extreme cases. This allows engineers to quickly pinpoint problems truly caused by thickness, rather than blindly searching through vast amounts of data, improving the accuracy and efficiency of problem localization.

[0018] Preferably, the type of abnormal pattern is inferred based on the problem attribution risk index, including: The problem attribution risk index of all points in the thickness map is rendered into a pseudo-color map to form a problem attribution map; the highlighted areas in the problem attribution map are segmented to extract independent abnormal regions; morphological features are extracted from each independent abnormal region to form an abnormal feature parameter set; the abnormal feature parameter set is input into a pre-trained pattern recognition model to identify the abnormal pattern type of the abnormal region.

[0019] Preferably, generating an attribution report based on the anomaly pattern type includes: The system matches the identified abnormal patterns with a pre-established process knowledge base to infer potential process causes leading to yield loss and generates an attribution report. Based on the attribution report, the system automatically generates targeted process optimization suggestions.

[0020] The beneficial effects of this invention are as follows: VCSEL products, as core components in cutting-edge fields such as 3D sensing, optical communication, and LiDAR, directly impact the implementation and development of downstream technologies due to their quality and stability. Traditional inspection methods relying on manual experience are not only inefficient and prone to errors, but also easily delay process improvement opportunities, resulting in significant production waste. This invention, through a systematic and intelligent thickness detection and problem attribution scheme, changes the traditional extensive inspection mode. It can efficiently and accurately identify thickness problems in production, precisely pinpoint the underlying process causes, and provide specific improvement suggestions, making production adjustments more targeted. This not only stabilizes the production quality of VCSEL epitaxial wafers, increases the proportion of qualified products, and reduces material and time costs, but also helps companies improve production management capabilities and reduce market risks caused by quality issues. From an industry perspective, the application of this invention can promote the transformation of VCSEL production from experience-driven to data-driven, improve the overall industry's production efficiency and product competitiveness, provide reliable device support for technological breakthroughs in related cutting-edge fields, and contribute to the high-quality development of the entire industry chain. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a method for detecting the thickness uniformity of VCSEL epitaxial wafers according to the present invention. Detailed Implementation

[0022] This invention discloses a method for detecting the thickness uniformity of VCSEL epitaxial wafers, referring to... Figure 1 This includes steps S1-S4: S1: Obtain the thickness map and yield map of the VCSEL epitaxial wafer after spatial alignment; for each position in the thickness map, perform a weighted average of the original thickness values ​​in its neighborhood based on the Euclidean distance from each point in the neighborhood to that position to obtain the global background thickness at that position.

[0023] It's important to note that VCSEL epitaxial wafer production is a complex process involving multiple technological stages, from epitaxial growth to chip manufacturing and testing. To monitor and evaluate the production process and product quality from different perspectives, it's necessary to obtain thickness and yield maps for the VCSEL epitaxial wafers. A thickness map is the thickness distribution data obtained after the growth of a single VCSEL epitaxial wafer using specific measurement methods. It reflects the thickness information at every location on the epitaxial wafer, graphically presenting the spatial distribution of thickness, much like taking a panoramic photo of the epitaxial wafer's thickness. A yield map, on the other hand, is the yield distribution data obtained after the same epitaxial wafer has undergone chip manufacturing and testing. It clearly shows the pass / fail status and performance distribution of the chip at different locations, identifying which areas have high yields and which areas may have problems. Passing areas are marked with 1, and failing areas with 0, reflecting the performance results of different areas on the epitaxial wafer at the time of testing. In the context of VCSEL epitaxial wafer thickness and yield correlation analysis, when performing feature point-based image registration on thickness and yield maps, in order to ensure that the data correspond one-to-one, the statistical analysis is effective, and the interference of invalid regions is eliminated, it is necessary to ensure that the epitaxial wafer regions involved in the two are completely consistent.

[0024] Specifically, for a single VCSEL epitaxial wafer, the thickness distribution data after its growth is completed is obtained and recorded as a thickness map; at the same time, the yield distribution data after the same epitaxial wafer has completed the chip manufacturing and testing process is obtained and recorded as a yield map; an image registration algorithm based on feature points is used to accurately align the thickness map and the yield map in space.

[0025] It should be noted that, to accurately identify local thickness anomalies affecting yield in thickness maps, it is first necessary to define the normal thickness baseline, i.e., the gradually changing global background thickness on the epitaxial wafer. If the original thickness value is used directly for analysis, it is easily affected by the overall thickness trend of the wafer, such as the natural gradient from the edge to the center, making it impossible to distinguish between the overall trend and local anomalies, resulting in a loss of accuracy in subsequent anomaly identification. Therefore, this invention requires calculating the global background thickness at point k.

[0026] Preferably, the original thickness value at each location in the thickness map and the Euclidean distance from each location's neighboring points to that location are obtained, denoted as the neighborhood distance; the smoothing bandwidth parameter is automatically determined based on empirical rules of physical scale according to the wafer's physical dimensions and data resolution; the global background thickness is calculated based on each neighborhood distance and the smoothing bandwidth parameter in the thickness map, including: The global background thickness satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map Global background thickness; Indicates position in thickness map The set of all neighboring points is denoted as the neighborhood set, and the size of the neighborhood is a radius of 3. λ A circular area; Representing the neighborhood One point of responsibility The original thickness value; In the thickness map, the neighboring points With position The neighborhood distance; This represents the smooth bandwidth parameter.

[0027] In the formula, This represents a Gaussian kernel function used as a weighting factor, whose value varies with the neighborhood points. To the center position distance It increases and then decreases, and the rate of decrease is determined by the smoothing bandwidth parameter; Indicates the position in the thickness map The thickness value of each point in the neighborhood Assign appropriate weights; Indicates the position in the thickness map The sum of the original thickness values ​​of all neighboring points in the neighborhood set multiplied by their corresponding weights reflects the weighted thickness accumulation. Indicates position in thickness map The sum of the weights of all points within the neighborhood set is used to normalize the weighted sum of the numerator, making the final result... Within a reasonable numerical range, it can accurately reflect the position in the thickness map. And the background thickness of the neighborhood; Indicates the position in the thickness map The original thickness values ​​of all points in the neighborhood set are weighted according to the rule that the closer the neighborhood point is to position k, the greater the weight, and the farther the distance, the smaller the weight. Then, the weighted sum is divided by the total weight to obtain the global background thickness at position k, which reflects the background thickness of the position and the surrounding area.

[0028] It should be noted that setting the neighborhood size to a circular region with a radius of 3λ is based on the rapid decay characteristic of the Gaussian kernel function, achieving an optimal balance between computational efficiency and numerical accuracy. Specifically, when the distance between a neighboring point and the center point reaches 3λ, the thickness value of points further away contributes very little to the final calculation result of the background thickness and can be ignored. Therefore, by setting this, the theoretically infinite calculation range can be reduced to a finite local region without sacrificing computational accuracy, thereby greatly improving the algorithm's running efficiency.

[0029] Thus, the global background thickness at any point in the thickness map has been obtained.

[0030] S2: Calculate the difference between the original thickness value at each position in the thickness map and the global background thickness to obtain the thickness deviation; calculate the standard deviation of the thickness deviation fluctuation in the neighborhood of each position; take the ratio of the absolute value of the thickness deviation at each position to the standard deviation of the fluctuation as the neighborhood anomaly score.

[0031] It should be noted that after obtaining the global background thickness at any point in the thickness map, the degree of deviation of each point from the background can be directly calculated. However, the absolute magnitude of this deviation value is not sufficient to indicate its abnormality. A deviation value may be significant in a flat area, but may be normal in a region with drastic fluctuations. Therefore, it is necessary to first measure the fluctuation level of the region in which it is located.

[0032] Specifically, based on the difference between the global background thickness and the original thickness value of all points in the thickness map, the thickness deviation of each point is calculated and denoted as the thickness deviation; the number of pixels in the neighborhood set of all points in the thickness map, and the average thickness deviation of the pixels in the neighborhood set are counted and denoted as the neighborhood average deviation; based on the thickness deviation, the number of pixels, and the neighborhood average deviation, the standard deviation of the fluctuation between the thickness deviation and the neighborhood average deviation of all points in the thickness map is calculated and denoted as the fluctuation standard deviation, which includes: The standard deviation of fluctuation satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map The standard deviation of the fluctuation; Indicates the midpoint of the thickness map The thickness deviation of point j in the neighborhood set; Indicates position in thickness map The neighborhood set; Midpoint of the thickness map The number of points in the neighborhood set, >1; Position in the thickness map The neighborhood average deviation.

[0033] In the formula, Indicates the midpoint of the thickness map The square of the difference between each thickness deviation within its neighborhood and the average deviation of its neighborhood is used to evaluate the midpoint of the thickness map. The degree to which the thickness deviation deviates from the average level of the neighborhood is denoted as the degree of deviation; Indicates the midpoint of the thickness map The sum of the deviations of all points within the neighborhood set is denoted as the total deviation. This indicates that the midpoint of the thickness map has been calculated. The variance of the degree to which the thickness deviation within a neighborhood deviates from the average level of the neighborhood is denoted as the neighborhood variance. This represents taking the square root of the neighborhood variance to obtain the neighborhood standard deviation, which is used to evaluate... The degree of fluctuation in the degree of deviation within the neighborhood.

[0034] It's important to note that in VCSEL epitaxial wafer thickness anomaly detection, simply using the thickness deviation—the difference between the original thickness and the global background thickness—is insufficient for accurate anomaly identification. Different regions of the epitaxial wafer exhibit varying thickness fluctuation characteristics. For instance, the central region shows relatively smooth thickness fluctuations, where even minor deviations could indicate anomalies; conversely, edge regions, influenced by the growth process, exhibit larger fluctuations, where larger deviations might be within the normal range. Without considering the regional fluctuation levels, it's easy to misjudge normal as abnormal or miss anomalies as normal. Furthermore, when calculating the standard deviation of fluctuation, some regions may exhibit extremely small fluctuations, with the standard deviation approaching zero. Directly using these as the denominator would render the calculation ineffective. Therefore, a minimum constant needs to be introduced, and the neighborhood anomaly score must be standardized using the ratio of deviation to fluctuation level. This ensures a uniform standard for measuring the degree of anomalies across all regions of the wafer, guaranteeing accurate identification of thickness anomaly locations that have a real impact on subsequent steps.

[0035] Preferably, the midpoint of the thickness map is obtained. The original thickness value and global background thickness, as well as the standard deviation of fluctuation; set a minimum positive number. ; Calculate the midpoint of the thickness map The neighborhood anomaly score, denoted as the neighborhood anomaly score, includes: The neighborhood anomaly score satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map Neighborhood anomaly scores; Indicates the midpoint of the thickness map The original thickness value; Indicates the midpoint of the thickness map Global background thickness; Indicates the midpoint of the thickness map The standard deviation of the fluctuation; It represents a very small positive number, and guarantees that the denominator is not 0.

[0036] In the formula, Indicates the midpoint of the thickness map The absolute difference between the original thickness value and its global background thickness reflects the degree to which the thickness at that point deviates from the neighborhood background. This indicates the midpoint of the thickness map. The standard deviation of the fluctuation plus a very small positive number , used as the denominator for normalization; This indicates the midpoint of the thickness map. The magnitude of the deviation between the original thickness value and the global background thickness, and the midpoint of the thickness map. By comparing the overall fluctuation level of this deviation within its neighborhood, this normalization process achieves standardized analysis of anomalies. The larger the value, the higher the value of the midpoint of the thickness map. The thickness deviation relative to its neighborhood background is more significant against its neighborhood background, so the midpoint of the thickness map The larger the neighborhood anomaly score, the better.

[0037] S3: Calculate the average yield value of each point in the neighborhood of each location, and use the difference between 1 and the average yield value as the neighborhood failure rate; within the neighborhood of each location, calculate the average of the product of the neighborhood anomaly score and the neighborhood failure rate to obtain the neighborhood baseline parameter; based on the relative deviation and correlation weighting relationship between the product of the neighborhood anomaly score and the neighborhood failure rate, the neighborhood baseline parameter, and the Pearson correlation coefficient, calculate the problem attribution risk index.

[0038] It's important to note that the local anomaly score calculated solely from the thickness map only reflects the degree of thickness deviation at that physical location. To determine whether this thickness anomaly truly affects the yield value of the corresponding point in the neighborhood, it's necessary to examine the failure status of the corresponding region in conjunction with the yield map. This is because the yield value of a single region may be affected by random factors, such as test noise or minor local defects, leading to occasional failures. However, by calculating the neighborhood failure rate and statistically analyzing the collective failure of all points within the k-neighborhood of point k in the thickness map, the randomness of individual point failures can be effectively smoothed out. This accurately reflects whether a performance problem of concentrated failure across all points in the neighborhood has occurred in that region, and further analyzes whether there are significant performance failure issues in that region.

[0039] Specifically, the yield map is obtained and the midpoint of the thickness map is located. The yield value of each point within the neighborhood set, and the number of points in the neighborhood set, are used to calculate the points in the thickness map. The neighborhood failure rate, denoted as , includes: The neighborhood failure rate satisfies the following expression: ; In the formula, Indicates the midpoint of the thickness map Neighborhood failure rate; This represents the number of points in the neighborhood set. Represent the yield profile and locate the midpoint of the thickness profile. points in the neighborhood set Yield value; Indicates position in thickness map The neighborhood set.

[0040] In the formula, Indicates to The sum of the yield values ​​of all points within the neighborhood is the number of qualified points in the neighborhood. Midpoint of the thickness map The proportion of qualified chips in the neighborhood area, i.e., the neighborhood average yield; This means subtracting the neighborhood average yield from 1 to obtain the neighborhood average yield. This indicates that the thickness is determined by the midpoint of the statistical thickness map. The collective failure of all points in the neighborhood effectively smooths out the randomness of individual point failures.

[0041] It should be noted that different regions of VCSEL epitaxial wafers are affected by processes such as airflow and temperature, resulting in significant inherent fluctuations. Traditional global benchmarks are prone to misjudging normal fluctuations as problems. This invention uses the average value of the product of the failure rate of the neighborhood of point k and the anomaly score of the neighborhood as the neighborhood benchmark parameter to eliminate fluctuation interference within the neighborhood.

[0042] Preferably, the point in the thickness map Within the neighborhood set, calculate the average of the product of the neighborhood failure rate and the neighborhood anomaly score of all neighborhood points, and use this as the neighborhood baseline parameter.

[0043] It should be noted that VCSEL epitaxial wafer yield loss may be caused by a variety of factors, such as probe testing failure, substrate impurities, and electrode fabrication defects. Not all of these factors are related to thickness. If we do not screen for scenarios where thickness anomalies are positively correlated with yield loss, yield loss that is not due to thickness will be mistakenly attributed to thickness issues. At the same time, although there is a joint deviation between thickness anomalies and yield loss in some areas, the positive correlation between the two is extremely weak. This indicates that yield loss is more likely caused by other factors. If we do not apply positive correlation weighting to this degree of deviation, we will exaggerate the thickness-related risk in that area.

[0044] Preferably, the midpoint of the thickness map is calculated based on the neighborhood baseline parameters, fluctuation standard deviation, neighborhood failure rate, and neighborhood anomaly score. The problem attribution risk index, denoted as the problem attribution risk index, includes: The problem attribution risk index satisfies the following expression: ; In the formula, This represents the attribution risk index of point k in the thickness map. , Indicates the midpoint of the thickness map Neighborhood anomaly score, neighborhood failure rate; The Pearson correlation coefficient represents the data sequence of neighborhood anomaly scores of all neighboring points in the neighborhood set of point k in the thickness map, arranged by position, and the data sequence of neighborhood failure rates of all neighboring points arranged by the same position. Represents the neighborhood reference parameters; Represents the maximum value function; Represents an exponential function with the natural constant as its base; It represents a very small positive number, and guarantees that the denominator is not 0.

[0045] In the formula, This means that only the positive correlation portion of the Pearson correlation coefficient is retained, i.e. When >0, take Otherwise, take 0 to ensure that subsequent calculations only focus on the positive correlation between thickness anomalies and yield loss, and exclude interference from the two being unrelated or negatively related. Midpoint of the thickness map The combined influence strength of neighborhood anomaly score and neighborhood failure rate is such that the larger the neighborhood anomaly score and the larger the neighborhood failure rate, the larger the product and the stronger the correlation between the two problems. Indicates the midpoint of the thickness map The degree of deviation of the product of the neighborhood anomaly score and the neighborhood failure rate relative to the neighborhood baseline parameter; This indicates that the degree of deviation is weighted. In the middle, only the positive correlation is strong. >0, and deviates significantly. The larger the value, the greater the response of the problem attribution risk index. If there is no positive correlation, even Even if it is large, it will not lead to a large response in the problem attribution risk index, thus avoiding misjudgment; In the middle, when The larger, The closer it is to 0, the closer it is to 1; the exponential function amplifies the changes in risk-related assessment data non-linearly. In the middle, the smaller the denominator, The larger the value, the stronger the positive correlation and the greater the deviation from the normal baseline. The larger the value, the higher the risk of attributing the problem to the cause, and vice versa. By identifying a genuine causal relationship between thickness anomalies and yield failures, and outputting a problem attribution risk index, we can determine whether point k in the thickness map truly caused the yield decrease due to thickness issues.

[0046] S4: Infer the type of abnormal pattern based on the problem attribution risk index; generate an attribution report based on the type of abnormal pattern.

[0047] Specifically, the problem attribution risk index of all points in the thickness map is rendered into a pseudo-color map to form a problem attribution map; the highlighted areas in the problem attribution map are processed by image segmentation to extract independent abnormal regions; morphological features are extracted for each independent abnormal region to form an abnormal feature parameter set; the abnormal feature parameter set is input into a pre-trained pattern recognition model to identify the abnormal pattern type of the abnormal region.

[0048] It should be noted that morphological features are extracted for each independent anomalous region, specifically by calculating its area, perimeter, aspect ratio, roundness, or principal axis direction and other geometric parameters; the set of anomalous feature parameters is input into a pre-trained pattern recognition model, specifically a support vector machine; the types of anomalous patterns include ring-shaped anomalous patterns, stripe-shaped anomalous patterns, center point anomalous patterns, or edge collapse.

[0049] Preferably, the identified abnormal pattern types are matched with a pre-established process knowledge base to infer potential process causes leading to yield loss and generate an attribution report; the system automatically generates targeted process optimization suggestions based on the attribution report.

[0050] It should be noted that the process knowledge base stores different anomaly mode types and VCSEL epitaxial growth process parameters, such as MOCVD reaction temperature, gas flow rate, and wafer rotation speed. Based on the attribution report, the system automatically generates targeted process optimization suggestions, such as adjusting the temperature field distribution of the MOCVD equipment, optimizing gas flow uniformity, or calibrating the wafer rotation mechanism, to eliminate or mitigate thickness anomalies that lead to yield losses.

[0051] This completes the uniformity test of the VCSEL epitaxial wafer thickness.

[0052] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for detecting the thickness uniformity of VCSEL epitaxial wafers, characterized in that, include: Obtain the thickness map and yield map of the VCSEL epitaxial wafer after spatial alignment; for each position in the thickness map, perform a weighted average of the original thickness values ​​in its neighborhood based on the Euclidean distance from each point in the neighborhood to that position to obtain the global background thickness at that position. The difference between the original thickness value at each location in the thickness map and the global background thickness is calculated to obtain the thickness deviation; the standard deviation of the fluctuation of the thickness deviation in the neighborhood of each location is calculated; the ratio of the absolute value of the thickness deviation at each location to the standard deviation of the fluctuation is used as the neighborhood anomaly score. Calculate the average yield value of each point in the neighborhood of each location, and use the difference between 1 and the average yield value as the neighborhood failure rate; within the neighborhood of each location, calculate the average of the product of the neighborhood anomaly score and the neighborhood failure rate to obtain the neighborhood baseline parameter; based on the relative deviation and correlation weighting relationship between the product of the neighborhood anomaly score and the neighborhood failure rate, the neighborhood baseline parameter, and the Pearson correlation coefficient, calculate the problem attribution risk index. Inferring the type of abnormal pattern based on the problem attribution risk index; Generate attribution reports based on the anomaly pattern type.

2. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The acquisition of the thickness map and yield map of the VCSEL epitaxial wafer after spatial alignment includes: For a single VCSEL epitaxial wafer, the thickness distribution data after its growth is completed is obtained and recorded as a thickness map; at the same time, the yield distribution data after the same epitaxial wafer has gone through the chip manufacturing and testing process is obtained and recorded as a yield map; an image registration algorithm based on feature points is used to accurately align the thickness map and the yield map in space.

3. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The global background thickness satisfies the following expression: ； In the formula, Indicates the midpoint of the thickness map Global background thickness; Indicates position in thickness map The set of all neighboring points is denoted as the neighborhood set, and the size of the neighborhood is a radius of 3. λ A circular area; Representing the neighborhood One point of responsibility The original thickness value; In the thickness map, the neighboring points With position The neighborhood distance; This represents the smooth bandwidth parameter.

4. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The standard deviation of the fluctuation satisfies the following expression: ； In the formula, Indicates the midpoint of the thickness map The standard deviation of the fluctuation; Indicates the midpoint of the thickness map The thickness deviation of point j in the neighborhood set; Indicates position in thickness map The neighborhood set; Midpoint of the thickness map The number of points in the neighborhood set, >1; Position in the thickness map The neighborhood average deviation.

5. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The neighborhood anomaly score includes: Calculate the midpoint of the thickness map The absolute difference between the original thickness value and its global background thickness, and the midpoint of the thickness map. The absolute difference between the original thickness value and its global background thickness and the midpoint of the thickness map. Compare the standard deviations of the fluctuations within the same neighborhood, and add a very small positive number to the denominator of the ratio. The results are used as neighborhood anomaly scores.

6. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The neighborhood failure rate satisfies the following expression: ； In the formula, Indicates the midpoint of the thickness map Neighborhood failure rate; This represents the number of points in the neighborhood set. Represent the yield profile and locate the midpoint of the thickness profile. points in the neighborhood set Yield value; Indicates position in thickness map The neighborhood set.

7. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The process of obtaining neighborhood reference parameters includes: Point in the thickness map Within the neighborhood set, calculate the average of the product of the neighborhood failure rate and the neighborhood anomaly score of all neighborhood points, and use this as the neighborhood baseline parameter.

8. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The problem attribution risk index satisfies the following expression: ； In the formula, This represents the attribution risk index of point k in the thickness map. , Indicates the midpoint of the thickness map Neighborhood anomaly score and neighborhood failure rate; The Pearson correlation coefficient represents the data sequence of neighborhood anomaly scores of all neighboring points in the neighborhood set of point k in the thickness map, organized by position, and the data sequence of neighborhood failure rates of all neighboring points organized by the same position. Represents the neighborhood reference parameters; Represents the maximum value function; Represents an exponential function with the natural constant as the base; It represents a very small positive number, and guarantees that the denominator is not 0.

9. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The method of inferring the type of abnormal pattern based on the problem attribution risk index includes: The problem attribution risk index of all points in the thickness map is rendered into a pseudo-color map to form a problem attribution map; the highlighted areas in the problem attribution map are segmented to extract independent abnormal regions; morphological features are extracted from each independent abnormal region to form an abnormal feature parameter set; the abnormal feature parameter set is input into a pre-trained pattern recognition model to identify the abnormal pattern type of the abnormal region.

10. The method for detecting the thickness uniformity of a VCSEL epitaxial wafer according to claim 1, characterized in that, The generation of attribution reports based on anomaly pattern types includes: The system matches the identified abnormal patterns with a pre-established process knowledge base to infer potential process causes leading to yield loss and generates an attribution report. Based on the attribution report, the system automatically generates targeted process optimization suggestions.

Images