UV light source precise control method and related equipment for wafer defect detection

Abstract

This application relates to the field of light source control technology and discloses a method for precisely controlling a UV light source for wafer defect detection and related equipment. The method includes: determining initial detection parameters for wafer defect detection; obtaining a wafer surface image of a target wafer based on the initial detection parameters; identifying defect information in the wafer surface image; evaluating the detection effect value of the initial detection parameters based on the defect information; and, if the detection effect value is lower than a preset detection threshold, generating a UV light source control strategy based on the defect information. Through the implementation of this application scheme, a closed-loop feedback mechanism is established by obtaining initial detection parameters and evaluating the detection effect value, promptly identifying situations where the detection effect is not ideal, and adjusting the UV light source parameters accordingly, ensuring that defect detection is always maintained in an optimal state and improving the accuracy of wafer defect detection.

Description

Technical Field

[0001] The present application relates to the field of light source control technology, and in particular to a method for precisely controlling a UV light source for wafer defect detection and related equipment. Background Art

[0002] With the rapid development of integrated circuit technology and the continuous reduction of process nodes, defect detection during wafer manufacturing has become increasingly important and challenging. As a key step in semiconductor manufacturing, wafer defect detection directly impacts product yield and quality. Traditional wafer defect detection relies primarily on manual visual inspection and automated optical inspection equipment with fixed parameters. These methods often suffer from missed detections and false detections when encountering minute defects, making them difficult to meet the precision requirements of modern semiconductor manufacturing.

[0003] As a crucial optical component in wafer defect inspection, the performance and parameter settings of UV (ultraviolet) light sources have a crucial impact on detection results. However, the industry currently relies on empirical, fixed parameter settings, lacking a dynamic, precise control mechanism for UV light sources. This static inspection method is unable to adapt to the varying characteristics of different wafer types and struggles to address new defect signatures introduced by process changes, limiting detection accuracy and reliability. Summary of the Invention

[0004] The present application provides a method for precisely controlling UV light sources for wafer defect detection and related equipment, which can at least solve the problem of low detection accuracy caused by using UV light sources with fixed parameters to detect different types of wafers in related technologies.

[0005] In a first aspect, the present application provides a method for precisely controlling a UV light source for wafer defect detection, the method comprising:

[0006] Determine initial inspection parameters for wafer defect detection;

[0007] Acquire a wafer surface image of the target wafer according to the initial detection parameters;

[0008] identifying defect information of the wafer surface image;

[0009] Evaluate the detection effect value of the initial detection parameter according to the defect information;

[0010] If the detection effect value is lower than a preset detection threshold, a UV light source control strategy is generated according to the defect information.

[0011] Optionally, in a first implementation of the first aspect of the present application,

[0012] The step of identifying defect information of the wafer surface image includes:

[0013] Preprocessing the wafer surface image;

[0014] Extracting defect features from pre-processed wafer surface images;

[0015] Identifying a corresponding defect type according to the defect characteristics;

[0016] Defect information is marked on the wafer surface image according to the defect type.

[0017] Optionally, in a second implementation of the first aspect of the present application, the step of evaluating the detection effect value of the initial detection parameter according to the defect information includes:

[0018] generating a defect statistics report based on the defect information;

[0019] Determine the detection rate, false detection rate, and missed detection rate of the wafer defect detection according to the defect statistics report;

[0020] The detection effect value is determined by converting the false detection rate and the missed detection rate into a penalty factor and performing exponential penalty on the detection rate.

[0021] Optionally, in a third implementation of the first aspect of the present application, the step of generating a UV light source control strategy according to the defect information includes:

[0022] extracting characteristic parameters of the defect information;

[0023] Determining light source requirement data corresponding to the defect characteristics according to the characteristic parameters;

[0024] A UV light source control strategy corresponding to the light source demand data is generated through an adaptive optimization algorithm.

[0025] Optionally, in a fourth implementation of the first aspect of the present application, the method further includes:

[0026] Detect the actual output parameters of the UV light source through the light sensor;

[0027] Determining a first deviation value between the actual output parameter and the initial detection parameter;

[0028] The UV light source control strategy is determined according to the deviation type and deviation degree corresponding to the first deviation value.

[0029] Optionally, in a fifth implementation of the first aspect of the present application, the method further includes:

[0030] Detecting environmental parameters and material parameters of the target wafer by a multi-dimensional sensor;

[0031] Determining a reference value for wafer defect detection according to the environmental parameters and the material parameters;

[0032] determining a second deviation value between the reference value and the initial detection parameter;

[0033] The UV light source control strategy is determined according to the second deviation value.

[0034] Optionally, in a sixth implementation of the first aspect of the present application, the method further includes:

[0035] Determining the dielectric response characteristics of the target wafer under UV light irradiation based on Jonscher's general power law model;

[0036] Analyzing the correlation between the dielectric response characteristics and the defect type based on the Dissado-Hill model;

[0037] Establishing a mapping model between wafer defect characteristics and UV light source parameters based on the correlation relationship;

[0038] The UV light source control strategy is optimized and iterated based on the mapping model until the detection effect value reaches a preset detection threshold.

[0039] A second aspect of the present application provides a UV light source precision control device for wafer defect detection, the UV light source precision control device for wafer defect detection comprising:

[0040] A determination module, used to determine initial detection parameters for wafer defect detection;

[0041] an acquisition module, configured to acquire a wafer surface image of a target wafer according to the initial detection parameters;

[0042] an identification module, configured to identify defect information of the wafer surface image;

[0043] An evaluation module, configured to evaluate a detection effect value of the initial detection parameter based on the defect information;

[0044] A generation module is used to generate a UV light source control strategy according to the defect information if the detection effect value is lower than a preset detection threshold.

[0045] A third aspect of an embodiment of the present application provides an electronic device, including a memory and a processor, wherein the processor is used to execute a computer program stored on the memory. When the processor executes the computer program, it implements each step of the method for precise control of UV light sources for wafer defect detection provided in the first aspect of the embodiment of the present application.

[0046] The fourth aspect of the embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, each step of the method for precisely controlling the UV light source for wafer defect detection provided in the first aspect of the embodiment of the present application is implemented.

[0047] In summary, according to the method and related equipment for precisely controlling a UV light source for wafer defect detection provided by this application, initial detection parameters for wafer defect detection are determined; a wafer surface image of a target wafer is acquired based on the initial detection parameters; defect information in the wafer surface image is identified; and the detection effect value of the initial detection parameters is evaluated based on the defect information. If the detection effect value is lower than a preset detection threshold, a UV light source control strategy is generated based on the defect information. Through the implementation of this application, a closed-loop feedback mechanism is established by acquiring initial detection parameters and evaluating the detection effect value, enabling timely detection of unsatisfactory detection results and adjusting the UV light source parameters accordingly, ensuring that defect detection is always maintained in an optimal state and improving the accuracy of wafer defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic flow chart of a method for precisely controlling a UV light source for wafer defect detection provided in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of dielectric response test data of a non-defective area in wafer defect detection is provided for an embodiment of the present application;

[0050] Figure 3 A schematic diagram of dielectric response test data of surface contamination areas in wafer defect detection is provided for an embodiment of the present application;

[0051] Figure 4 A schematic diagram of a program module of a UV light source precision control device for wafer defect detection provided in an embodiment of the present application;

[0052] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the scope of protection of this application.

[0054] In order to solve the problem of low detection accuracy caused by using UV light sources with fixed parameters to detect different types of wafers in related technologies, the embodiment of the present application provides a method for accurately controlling UV light sources for wafer defect detection, such as Figure 1 The flowchart of the UV light source precise control method for wafer defect detection provided in this embodiment includes the following steps:

[0055] Step 110: Determine initial detection parameters for wafer defect detection.

[0056] Specifically, in this embodiment, determining the initial detection parameters for wafer defect detection is the first step in the entire detection process. These initial detection parameters include the UV light source's wavelength, intensity, exposure time, and spot size. These parameters directly impact the sensitivity and accuracy of the detection system. Analyzing historical data and previous test results can help determine the most effective parameter settings. Calibration experiments on multiple samples can then be performed to determine the detection performance under different parameter settings. Wafers made of different materials may respond differently to the UV light source, so parameters need to be adjusted based on the material's characteristics. For example, for silicon wafers, the light source intensity is set to 500mW / cm²; for gallium arsenide wafers, the light source intensity is set to 300mW / cm². Equipment specifications and capabilities also influence parameter selection, such as the maximum light source intensity and adjustable wavelength range. Therefore, when selecting initial detection parameters, factors such as historical data, wafer material, and equipment capabilities should be fully considered to ensure the effectiveness of the initial detection parameters ultimately determined at the beginning of wafer defect detection.

[0057] Step 120 : Acquire a wafer surface image of the target wafer according to the initial detection parameters.

[0058] Specifically, in this embodiment, the wavelength, intensity, exposure time, and spot size of the UV light source are set according to the initial detection parameters. During image acquisition, the UV light source is adjusted according to the detection parameters to ensure uniform illumination of the wafer surface. The camera is calibrated to ensure image clarity and accuracy, and multiple images of the wafer surface are captured to ensure that every area is fully covered. After image acquisition, the image is preprocessed to improve the visibility of defective areas. An edge detection algorithm is used to highlight possible defect edges in the image, and a preprocessed image is generated.

[0059] Step 130: Identify defect information of the wafer surface image.

[0060] Specifically, in this embodiment, the shape corresponding to the defect in the wafer surface image is determined by the defect edge identified by the edge detection algorithm, and an SVM classifier is used to identify different types of defects. For example, the SVM classifier is trained to identify scratches and point defects, and the defects are divided into types such as scratches, point defects, and particles according to the shape and size of the defects. For example, the length of the scratch is greater than 100μm, and the diameter of the point defect is less than 50μm. Then, an image annotation tool (such as LabelImg) is used to annotate the detected defects on the image, and the annotation information includes the location, type, and size of the defect. The detected defect information is stored in a defect database, and the defect information is managed using a data management system. When defining unknown defects later, the defect characteristics of the unknown defects can be used to query the defect database.

[0061] In an optional implementation of this embodiment, the step of identifying defect information of a wafer surface image includes: preprocessing the wafer surface image; extracting defect features from the preprocessed wafer surface image; identifying corresponding defect types based on the defect features; and marking defect information on the wafer surface image based on the defect type.

[0062] Specifically, in this embodiment, during the preprocessing phase, the acquired wafer surface image undergoes a series of processing steps to improve image quality and highlight defect features. These include using Gaussian filtering to remove image noise, enhancing image contrast through histogram equalization, and employing an edge detection algorithm to highlight edge information within the image. For example, for a wafer image containing a scratch defect, a Gaussian filter with a parameter of σ = 1.0 can be used for noise reduction, followed by histogram equalization to enhance the contrast between the scratch area and the background. After preprocessing, defect features are extracted from the optimized image, including shape, texture, and color features. By calculating shape features such as the area, perimeter, and circularity of the defect area, combined with texture features extracted using the gray-level co-occurrence matrix (GLCM) and color distribution features reflected by the color histogram, a comprehensive description of the defect's characteristic information can be achieved. For example, for a point defect, its shape regularity can be determined by calculating its circularity (4π·area / perimeter²), while its texture features can be analyzed using the GLCM to determine its surface characteristics. With complete feature information, defect type identification can be performed, such as using a support vector machine (SVM) or deep learning network for classification. By training with a large amount of annotated data, the classifier can learn the characteristic patterns of different defect types and accurately identify new defect samples. For example, an SVM classifier with an RBF kernel function can be trained on 3,000 defect images of different types in a training set, and then evaluated with 1,000 validation images, ultimately achieving a classification accuracy of over 90%. Finally, the identified defect information is annotated on the wafer surface image. This includes automatic annotation of the defect's location, type, and size, combined with manual verification to ensure annotation accuracy. This annotation information can be stored in JSON format, containing the defect's coordinate location, type, and size. For example, for a scratch defect at coordinates (100, 200), a box can be drawn on the image and its information stored in the format {"position": [100, 200], "type": "scratch", "size": 50}. This annotation information is ultimately stored in a database for subsequent query and analysis.

[0063] Step 140: Evaluate the detection effect value of the initial detection parameters according to the defect information.

[0064] Specifically, in this embodiment, after identifying and classifying the defect information of the wafer surface image, it is necessary to evaluate the detection effect value of the initial detection parameters, and the evaluation annotations include but are not limited to the detection rate, false detection rate and missed detection rate, wherein the number of detected defects, the number of falsely reported defects, the number of missed detection defects and the total number of actual defects are counted, and the detection rate is determined according to the ratio of the number of detected defects to the actual number of defects, the false detection rate is determined according to the ratio of the number of falsely reported defects to the actual number of no defects, and the missed detection rate is determined according to the ratio of the number of missed detection defects to the actual number of defects. According to the calculated indicators, the detection effect value of the initial detection parameters is comprehensively evaluated according to the detection rate, false detection rate and missed detection rate.

[0065] In an optional implementation of this embodiment, the step of evaluating the detection effect value of the initial detection parameters based on the defect information includes: generating a defect statistical report based on the defect information; determining the detection rate, false detection rate and missed detection rate of the wafer defect detection based on the defect statistical report; and performing exponential penalty on the detection rate by converting the false detection rate and missed detection rate into a penalty factor to determine the detection effect value.

[0066] Specifically, in this embodiment, in the wafer defect detection system, it is first necessary to generate a detailed statistical report based on the acquired defect information. This statistical report contains key information such as the number, type, and distribution location of the defects. By systematically organizing and analyzing this data, a complete defect statistical result can be obtained. For example, in the inspection of a batch of 100 wafers, the system detected 90 real defects, 5 false detection defects, and missed 10 actual defects. These raw data are organized into a structured statistical report, including information such as the number distribution of each type of defect, the spatial distribution characteristics of the defects, and the severity of the defects. Based on the generated defect statistical report, three key indicators need to be calculated next: detection rate, false detection rate, and missed detection rate. The detection rate represents the ratio of the number of defects correctly detected to the total number of defects that actually exist, the false detection rate reflects the ratio of the number of defects incorrectly identified as defects to the total number of areas that are actually defect-free, and the missed detection rate represents the ratio of the number of actual defects that cannot be detected to the total number of defects that actually exist. In the above example, the detection rate is 90 / 100 = 0.9, indicating that 90% of actual defects are correctly detected; the false positive rate is 5 / 200 = 0.025, indicating that 5 out of 200 defect-free areas are incorrectly identified as defects; and the missed detection rate is 10 / 100 = 0.1, indicating that 10% of actual defects are not detected. To obtain more representative detection effect values, it is necessary to convert the false positive rate (FPR) and missed detection rate (FNR) into a penalty factor and use an exponential function to increase the penalty effect. The penalty factor is calculated as follows:

[0067] ,

[0068] When the false detection rate and missed detection rate are small, the penalty factor changes relatively little; when the false detection rate and missed detection rate increase, the penalty factor increases rapidly. The final detection effect value is calculated by the following formula:

[0069] ,

[0070] Where DR is the detection rate and PF is the penalty factor. Substituting the above values into the formula, we can get: OS= This detection performance value comprehensively considers the impact of the detection rate, false positive rate, and missed detection rate, and can more comprehensively reflect the performance of the detection system. The closer the detection performance value is to 1, the better the detection system performance. In practical applications, when the detection performance value falls below a preset threshold, such as 0.95, the detection system parameters need to be adjusted and optimized to improve detection performance. This approach enables quantitative evaluation and continuous optimization of detection system performance. In actual wafer production lines, this evaluation method can provide clear guidance for detection system tuning. For example, when the detection performance value of 0.79 is found to be lower than the preset threshold of 0.95, analysis of the penalty factor reveals that the main problem lies in the high missed detection rate (0.1). This indicates that the main direction of system optimization is to improve detection sensitivity to reduce missed detections. At the same time, a low false detection rate (0.025) indicates good system specificity, a characteristic that should be maintained during the optimization process. This quantitative evaluation method can make the optimization process of the detection system more targeted and efficient.

[0071] Step 150: If the detection effect value is lower than the preset detection threshold, a UV light source control strategy is generated according to the defect information.

[0072] Specifically, in this embodiment, if the detection effect value is lower than the preset detection threshold, it is necessary to generate a UV light source control strategy based on the defect information to optimize the detection effect. The control strategy includes adjusting parameters such as the wavelength, intensity, exposure time, and spot size of the light source. The preset detection threshold is to set the threshold of the detection effect value based on historical data and experimental results. The calculated detection effect value is compared with the preset threshold to determine whether it is necessary to generate a control strategy. The reasons for the poor detection effect are analyzed based on the defect information, such as the deviation of the defect position from the center of the spot, insufficient light source intensity, etc. The control strategy is generated based on the analysis results to adjust parameters such as the wavelength, intensity, exposure time, and spot size of the light source. The effectiveness of the control strategy is verified through simulation experiments to ensure that the adjusted parameters can improve the detection effect.

[0073] In an optional implementation of this embodiment, the step of generating a UV light source control strategy based on defect information includes: extracting characteristic parameters of the defect information; determining light source demand data corresponding to the defect characteristics based on the characteristic parameters; and generating a UV light source control strategy corresponding to the light source demand data through an adaptive optimization algorithm.

[0074] Specifically, in this embodiment, in a wafer defect detection system, extracting characteristic parameters from acquired defect information is a key step in achieving precise light source control. First, a comprehensive analysis of the defect information is required to extract key parameters, including the defect's shape, size, and position characteristics. Specifically, shape characteristics include the defect's contour curvature and edge sharpness; size characteristics include the defect's area, perimeter, and maximum diameter; and position characteristics include the defect's spatial distribution and density on the wafer. For example, for a point defect with an area of 200 square microns, a contour curvature of 0.85, and an edge sharpness of 0.92, these characteristic parameters will directly influence the subsequent formulation of the light source control strategy. Based on the extracted characteristic parameters, the corresponding light source requirement data must be determined. This process requires establishing a mapping relationship between the characteristic parameters and the light source parameters, translating the defect's physical characteristics into specific light source requirements. Light source requirement data primarily includes wavelength requirements, intensity requirements, and illumination angle requirements. For example, for the point defect described above, its characteristic parameters can determine that a UV light source with a wavelength of 365 nm, an intensity requirement of 500 mW / cm², and an optimal illumination angle of 75 degrees is required. Establishing this mapping relationship requires considering multiple factors, including the difficulty of defect detection, the optical properties of the material, and the hardware limitations of the inspection system. After obtaining the light source requirement data, an adaptive optimization algorithm is used to generate the final UV light source control strategy. This algorithm employs a particle swarm optimization (PSO) algorithm combined with an adaptive weight adjustment mechanism to achieve global optimization of light source parameters. This algorithm uses light source parameters such as wavelength, intensity, and illumination angle as optimization variables and performs iterative optimization with detection performance as the objective function. During the optimization process, the algorithm dynamically adjusts parameter weights based on the results of each iteration, enabling faster convergence to the optimal solution. For example, in one optimization, after 50 iterations, the optimal light source control strategy was obtained: the wavelength was adjusted to 362nm, the light intensity was set to 520mW / cm², and the illumination angle was fine-tuned to 73 degrees. This set of parameters is significantly optimized compared to the initial settings and better meets defect detection requirements. The entire process forms a closed-loop optimization system, from extracting characteristic parameters to determining light source requirement data to generating the final control strategy. This approach achieves adaptive light source control based on defect characteristics, significantly improving the accuracy and efficiency of defect detection. In actual wafer production lines, this adaptive control method can effectively deal with defects of different types and characteristics, ensuring that the detection system always maintains the best detection state.

[0075] In an optional implementation of this embodiment, the method also includes: detecting the actual output parameters of the UV light source through a light-sensing sensor; determining a first deviation value between the actual output parameters and the initial detection parameters; and determining the UV light source control strategy based on the deviation type and deviation degree corresponding to the first deviation value.

[0076] Specifically, in this embodiment, in a wafer inspection system, because UV light sources may age over time, a light sensor is required to monitor the light source's actual output parameters in real time. Using high-precision photoelectric conversion elements, the light sensor accurately captures key parameters of the UV light source, such as wavelength, intensity, and spot uniformity. After acquiring the actual output parameters, the system compares them with pre-set initial detection parameters to calculate a first deviation value. This deviation value incorporates information from multiple dimensions, primarily regarding wavelength deviation, intensity deviation, and uniformity deviation. For example, the initial set parameters for a UV light source are: wavelength 365nm, intensity 500mW / cm², and uniformity deviation less than 3%. The actual output parameters detected by the sensor are: wavelength 368nm, intensity 460mW / cm², and uniformity deviation 4.5%. Calculation yields a wavelength deviation of +3nm, an intensity deviation of -40mW / cm², and a uniformity deviation exceeding the standard by 1.5%. Based on this calculated first deviation value, the system further analyzes the type and extent of the deviation to formulate appropriate control strategies. Deviations are primarily categorized into three types: wavelength drift, intensity attenuation, and uniformity degradation. Different control strategies are required for each type of deviation. In the example above, significant intensity attenuation (8%) is typically due to light source aging; wavelength drift (+0.8%) is relatively minor and likely caused by temperature fluctuations; and uniformity degradation (50% above the specified value) warrants significant attention. Based on these deviation analyses, the system generates corresponding UV light source control strategies. Intensity attenuation can be compensated by increasing the input power, raising the light source's drive current to 110% of its original value. Wavelength drift can be compensated by fine-tuning the light source's temperature control system, reducing the operating temperature by 2°C. Uniformity issues require adjustments to the collimator and diffuser positions in the optical system. These control strategies are automatically executed by the system, and the results are monitored in real time. In practical applications, this deviation analysis-based control method can effectively address issues caused by UV light source aging. For example, after six months of use, this method detected a 12% decrease in light intensity from a UV light source on a production line. The system automatically increased the light source's drive current to restore the output intensity to its initial level, ensuring the stability of the detection system. The system also records these deviation data and control history, establishing a light source aging curve to provide a basis for predictive maintenance. If the deviation value of a certain parameter continues to increase and exceeds the controllable range, the system will promptly issue a warning to replace the light source, thereby avoiding a significant decline in inspection quality. This real-time monitoring and automatic control method can not only promptly detect and correct UV light source output deviations, but also predict the light source life cycle, providing a scientific basis for equipment maintenance, and ultimately ensuring the long-term stable operation of the wafer inspection system.

[0077] It should be noted that to protect the reliability of the light source while compensating for light attenuation, the temperature and light output of the light source are gradually adjusted and monitored when increasing the current to ensure that the current remains within a safe range and to avoid a sudden, large increase in current. Furthermore, a current cap is set in the system to ensure that this safe range is not exceeded even when light intensity compensation is required. Furthermore, the operating temperature of the LED is monitored in real time and linked to current regulation. When the temperature exceeds a preset safety threshold, the current is automatically reduced or additional cooling measures are implemented.

[0078] In an optional implementation of this embodiment, the method also includes: detecting environmental parameters and material parameters of the target wafer through a multi-dimensional sensor; determining a baseline value for wafer defect detection based on the environmental parameters and material parameters; determining a second deviation value between the baseline value and the initial detection parameter; and determining a UV light source control strategy based on the second deviation value.

[0079] Specifically, in this embodiment, in the wafer defect detection system, a multi-dimensional sensor network uses temperature sensors, humidity sensors, air pressure sensors, and material property sensors deployed in the detection environment to collect environmental parameters and wafer material parameters in real time. These sensors can accurately capture environmental factors such as temperature changes (accuracy of ±0.1°C), relative humidity (accuracy of ±1%), and atmospheric pressure (accuracy of ±0.1kPa) within the detection environment. They can also obtain material properties such as the wafer's refractive index, reflectivity, and dielectric constant. For example, in a certain detection environment, the temperature sensor detects an ambient temperature of 23.5°C, the humidity sensor indicates a relative humidity of 45%, the air pressure sensor indicates an atmospheric pressure of 101.3kPa, and the material property sensor measures the wafer's refractive index as 3.42 and reflectivity as 0.35. Based on the acquired environmental and material parameters, the system calculates a benchmark value for wafer defect detection using an established physical model. This benchmark value includes parameters in multiple dimensions, such as the light source wavelength benchmark, intensity benchmark, and illumination angle benchmark. The calculation process takes into account the impact of environmental factors on the propagation characteristics of light, as well as the influence of material parameters on the interaction between light and the wafer surface. Specifically, temperature changes will affect the wavelength stability of the light source, humidity changes will affect the scattering characteristics of light, and the optical properties of the material directly determine the interaction effect between light and the wafer surface. , strength benchmark and illumination angle reference The calculation formulas are:

[0080] ,

[0081] ,

[0082] ,

[0083] in, is the initial wavelength, is the influence coefficient of temperature on wavelength, is the reference temperature, is the initial strength, is the influence coefficient of humidity on strength, is the influence coefficient of air pressure on strength, For reference humidity, is the reference air pressure, is the initial irradiation angle, is the refractive index of the wafer material.

[0084] In the above example, the calculated baseline values are: wavelength 367nm, intensity 485mW / cm², and irradiation angle 72 degrees. Comparing these calculated baseline values with the system's pre-set initial detection parameters yields a second deviation value. This deviation reflects deviations in detection parameters due to environmental and material factors. For example, if the initial detection parameters are set as: wavelength 365nm, intensity 500mW / cm², and irradiation angle 75 degrees, the calculation yields a wavelength deviation of +2nm, an intensity deviation of -15mW / cm², and an angle deviation of -3 degrees. These deviations clearly indicate the extent to which the UV light source needs to be adjusted. Based on the calculated second deviation value, the system generates a corresponding UV light source control strategy. The development of this control strategy requires comprehensive consideration of the changing trends of environmental and material parameters, as well as the extent to which these changes affect detection results. For the above example, the system generates the following control strategy: adjust the light source wavelength to 367nm to account for temperature fluctuations; reduce the light intensity to 485mW / cm² to match the material's reflectivity; and adjust the irradiation angle to 72 degrees for optimal detection results. In practical applications, this approach, based on multi-dimensional sensing and real-time control, can effectively address the impact of environmental and material changes. For example, on a semiconductor production line, when the ambient temperature rises from 20°C to 25°C, the system automatically adjusts the UV light source parameters to maintain stable detection results. When wafers with different doping concentrations are replaced, the system automatically optimizes detection parameters based on the changes in material parameters to ensure detection accuracy. The system also records the correlation between environmental and material parameters and detection results, establishing a parameter impact model to provide data support for subsequent optimization of detection parameters.

[0085] In an optional implementation of this embodiment, the method also includes: determining the dielectric response characteristics of the target wafer under UV light source irradiation based on the Jonscher general power law model; analyzing the correlation between the dielectric response characteristics and the defect type based on the Dissado-Hill model; establishing a mapping model between wafer defect characteristics and UV light source parameters based on the correlation relationship; and optimizing and iterating the UV light source control strategy based on the mapping model until the detection effect value reaches a preset detection threshold.

[0086] Specifically, in this embodiment, during the wafer defect detection process, it is first necessary to build a dielectric response characteristic analysis system based on the Jonscher general power law model. The system uses a UV light source to irradiate the wafer surface to stimulate the dielectric response of the material, and at the same time uses a high-precision capacitive sensor array to collect dielectric response data on the wafer surface. For silicon wafers, when the UV light source frequency varies in the range of 200-400nm, the dielectric response of the material follows the Jonscher general power law model. By measuring the real and imaginary parts of the dielectric constant at different frequencies, the complete dielectric loss characteristic curve of the wafer material can be obtained. The Jonscher general power law model is expressed by the following formula:

[0087] ,

[0088] in, is the dielectric loss, is the low-frequency dielectric constant, are high-frequency dielectric constants, which reflect the dielectric properties of the material in static electric field and rapidly changing electric field respectively. is the excitation frequency of the UV light source, which determines the frequency range of the system response. is the relaxation time constant, the time scale of the dielectric relaxation process, n is the power exponent, which describes the rate at which the dielectric loss changes with frequency, and j is the imaginary unit. For different types of defects, the value of n may be different, thus providing a basis for defect identification. For example, during the inspection process, if it is found that the dielectric loss in a certain frequency range is significantly higher than that in the defect-free area, and the dielectric loss obtained by fitting is significantly higher than that in the defect-free area, then the dielectric loss in a certain frequency range may be significantly higher than that in the defect-free area, and the dielectric loss in a certain frequency range may be ... A large value may indicate that there are large structural defects in the area, such as cracks or deep scratches. The dielectric response characteristics refer to the electrical polarization characteristics of a material under the action of an external electric field, including the change pattern of parameters such as dielectric constant and dielectric loss with the frequency of the external electric field. In wafer inspection, when a UV light source irradiates the wafer surface, the photon energy excites the material to produce electron-hole pairs, causing the local charge distribution to change, thereby causing changes in dielectric properties. First, standard dielectric response data is obtained in a defect-free area. When the UV light source is irradiated at a specific frequency, the dielectric constant and dielectric loss of the wafer surface are measured to obtain a baseline dielectric spectrum. This baseline dielectric spectrum reflects the normal dielectric response characteristics under a complete lattice structure. Secondly, dielectric response data is measured in areas with defects. Different types of defects will change the local electronic structure and energy level distribution, thereby affecting the dielectric response. For example, for surface contamination, the introduction of new surface states will change the interface charge distribution, resulting in a change in the dielectric constant. The measured data is then substituted into the Dissado-Hill model equation:

[0089] ,

[0090] in, is the complex permittivity, describing the overall dielectric response of the material, is the dielectric relaxation strength, reflecting the degree of polarization, is the characteristic relaxation time, which characterizes the time scale of the polarization process, a is the short-order parameter, which describes the correlation within the local structural unit, and b is the long-order parameter, which describes the correlation between different structural units. By comparing the dielectric response differences between defect-free areas and areas with different types of defects, the corresponding relationship between defect characteristics and dielectric response parameters is established. For example, when there are micro scratches on the wafer surface, its dielectric loss curve will have a characteristic peak in a specific frequency range, while particle contamination may cause the curve to change abnormally at other frequency points. Figure 2 、 Figure 3Shown are the dielectric response test data for defect-free areas and surface contamination areas during wafer defect detection. After obtaining sufficient experimental data, a machine learning algorithm was used to establish a mapping relationship between wafer defect characteristics and UV light source parameters. Specifically, characteristics such as defect type, size, and depth were used as input variables, and parameters such as the UV light source's wavelength, intensity, and illumination angle were used as output variables. A neural network training algorithm was used to obtain an optimal parameter mapping model. For micro-scratch detection, for example, when a scratch depth of 100 nm was detected, the model automatically adjusted the UV light source wavelength to approximately 280 nm and set the illumination angle within a range of 15-20 degrees to achieve optimal detection results. Based on the established mapping model, an iterative optimization algorithm was designed to dynamically adjust the UV light source parameters. The algorithm first performs defect detection based on the initial detection parameters and evaluates the detection performance by calculating metrics such as the detection rate, false detection rate, and missed detection rate. If the detection performance value falls below a preset threshold, the UV light source is adjusted based on the optimal parameters predicted by the mapping model. In actual applications, assuming the preset detection threshold is 95%, when the effectiveness value of a particular test is only 90%, the optimization algorithm will predict a more optimal UV light source parameter combination based on the current test results and historical data, and verify the effect in the next round of testing, repeating this process until the preset threshold is reached. Throughout this process, the real-time monitoring system continuously collects changes in environmental and material parameters and inputs these factors into the model as correction terms to ensure stable detection results under different operating conditions. This closed-loop control mechanism enables the detection system to adapt to different types of wafer materials and diverse defect characteristics, greatly improving the accuracy and reliability of detection.

[0091] It should be noted that this application utilizes UV LEDs for UV light source detection. However, considering that UV LEDs typically have a specific wavelength range and are limited by material and process limitations in manufacturing, achieving a wide range of continuously adjustable wavelengths in a single device is difficult. In practical applications, a multi-wavelength UV LED array or a combination of multiple LEDs with different wavelengths can be used to achieve multi-band detection. During the actual detection process, a control system can sequentially switch between LEDs of different wavelengths, irradiating and measuring the wafer surface multiple times. After each irradiation, a high-precision capacitive sensor array collects dielectric response data at the corresponding wavelength. For example, a 280nm UV LED is first used for irradiation to obtain a set of dielectric response data; then, a 310nm UV LED is switched to obtain another set of data. In this way, after measuring at multiple wavelengths, a complete set of multi-band dielectric response data can be obtained. This data is then input into an analysis system based on the Jonscher general power law model and the Dissado-Hill model. By analyzing the dielectric response characteristics at different wavelengths, the characteristic responses of different defect types can be further analyzed. For example, if a certain defect exhibits a significant dielectric loss peak at a wavelength of 280nm but has no obvious characteristics at a wavelength of 310nm, the defect type can be identified more accurately by comparing the difference between the two sets of data.

[0092] According to the proposed method for precisely controlling a UV light source for wafer defect detection, initial detection parameters for wafer defect detection are determined; a surface image of the target wafer is acquired based on the initial detection parameters; defect information in the wafer surface image is identified; and the detection effect value of the initial detection parameters is evaluated based on the defect information. If the detection effect value is lower than a preset detection threshold, a UV light source control strategy is generated based on the defect information. Through the implementation of this proposed method, a closed-loop feedback mechanism is established by acquiring initial detection parameters and evaluating the detection effect value, promptly identifying unsatisfactory detection results and adjusting the UV light source parameters accordingly, ensuring that defect detection is always maintained in an optimal state and improving the accuracy of wafer defect detection.

[0093] Figure 4 The embodiment of the present application provides a UV light source precise control device for wafer defect detection, which can be used to implement the UV light source precise control method for wafer defect detection in the aforementioned embodiment. Figure 4 As shown, the UV light source precision control device for wafer defect detection mainly includes:

[0094] A determination module 10 is used to determine initial detection parameters for wafer defect detection;

[0095] An acquisition module 20 is configured to acquire a wafer surface image of a target wafer according to initial detection parameters;

[0096] Identification module 30, used to identify defect information of wafer surface image;

[0097] An evaluation module 40 is used to evaluate the detection effect value of the initial detection parameters based on the defect information;

[0098] The generating module 50 is configured to generate a UV light source control strategy based on the defect information if the detection effect value is lower than a preset detection threshold.

[0099] In an optional embodiment of the present application, the recognition module is specifically used to: preprocess the wafer surface image; extract defect features from the preprocessed wafer surface image; identify the corresponding defect type based on the defect features; and mark defect information on the wafer surface image based on the defect type.

[0100] In an optional embodiment of the present application, the evaluation module is specifically used to: generate a defect statistical report based on defect information; determine the detection rate, false detection rate and missed detection rate of wafer defect detection based on the defect statistical report; and perform exponential penalty on the detection rate by converting the false detection rate and missed detection rate into a penalty factor to determine the detection effect value.

[0101] In an optional embodiment of the present application, the generation module is specifically used to: extract characteristic parameters of defect information; determine light source demand data corresponding to the defect characteristics based on the characteristic parameters; and generate a UV light source control strategy corresponding to the light source demand data through an adaptive optimization algorithm.

[0102] In an optional embodiment of the present application, the UV light source precision control device for wafer defect detection further includes a detection module. The detection module is configured to detect actual output parameters of the UV light source using a light sensor. The determination module is further configured to determine a first deviation between the actual output parameters and initial detection parameters; and determine a UV light source control strategy based on the deviation type and degree corresponding to the first deviation value.

[0103] In an optional embodiment of the present application, the detection module is further configured to detect environmental parameters and material parameters of the target wafer using a multi-dimensional sensor. The determination module is further configured to determine a baseline value for wafer defect detection based on the environmental parameters and material parameters; determine a second deviation value between the baseline value and the initial detection parameter; and determine a UV light source control strategy based on the second deviation value.

[0104] In an optional embodiment of the present application, the UV light source precision control device for wafer defect detection further includes: an analysis module and an adjustment module. The determination module is further configured to: determine the dielectric response characteristics of the target wafer under UV light source irradiation based on the Jonscher general power law model. The analysis module is configured to: analyze the correlation between the dielectric response characteristics and the defect type based on the Dissado-Hill model. The adjustment module is configured to: establish a mapping model between wafer defect characteristics and UV light source parameters based on the correlation relationship; and iteratively optimize the UV light source control strategy based on the mapping model until the detection effect value reaches a preset detection threshold.

[0105] According to the proposed solution, a UV light source precision control device for wafer defect detection determines initial detection parameters for wafer defect detection; obtains a wafer surface image of the target wafer based on the initial detection parameters; identifies defect information in the wafer surface image; and evaluates the detection effect value of the initial detection parameters based on the defect information. If the detection effect value is lower than a preset detection threshold, a UV light source control strategy is generated based on the defect information. Through the implementation of the proposed solution, a closed-loop feedback mechanism is established by obtaining initial detection parameters and evaluating the detection effect value, promptly identifying unsatisfactory detection results and adjusting the UV light source parameters accordingly, ensuring that defect detection is always maintained in an optimal state and improving the accuracy of wafer defect detection.

[0106] According to the application plan Figure 5 An electronic device provided in an embodiment of the present application can be used to implement the UV light source precise control method for wafer defect detection in the aforementioned embodiment, mainly comprising:

[0107] Memory 501, processor 502, and computer program 503 stored in memory 501 and executable on processor 502. Memory 501 and processor 502 are connected via communication. When processor 502 executes computer program 503, the method for precisely controlling a UV light source for wafer defect detection described in the aforementioned embodiment is implemented. The number of processors may be one or more.

[0108] The memory 501 can be a high-speed random access memory (RAM) memory or a non-volatile memory such as a disk drive. The memory 501 is used to store executable program code. The processor 502 is coupled to the memory 501 .

[0109] Furthermore, the embodiment of the present application also provides a computer-readable storage medium, which can be provided in the electronic device in the above embodiments. The computer-readable storage medium can be the above Figure 5Memory in the illustrated embodiment.

[0110] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method for precisely controlling a UV light source for wafer defect detection described in the aforementioned embodiments. Furthermore, the computer-readable storage medium may be a USB flash drive, a removable hard drive, a read-only memory (ROM), RAM, a magnetic disk, or an optical disk, among other media capable of storing program code.

[0111] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0112] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0113] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for precisely controlling a UV light source for wafer defect detection, characterized in that: include: Determine initial inspection parameters for wafer defect detection; Acquire a wafer surface image of the target wafer according to the initial detection parameters; identifying defect information of the wafer surface image; Evaluate the detection effect value of the initial detection parameter according to the defect information; The step of evaluating the detection effect value of the initial detection parameter according to the defect information includes: generating a defect statistics report based on the defect information; Determine the detection rate, false detection rate, and missed detection rate of the wafer defect detection according to the defect statistics report; The detection effect value is determined by converting the false detection rate and the missed detection rate into a penalty factor and performing an exponential penalty on the detection rate; If the detection effect value is lower than the preset threshold, the UV light source is adjusted according to the optimal parameters predicted by the mapping model until the detection effect value reaches the preset detection threshold; The dielectric response characteristics of the target wafer under UV light irradiation are determined based on the Jonscher general power law model; wherein the Jonscher general power law model is expressed by the following formula: ， in, is the dielectric loss, is the low-frequency dielectric constant, are high-frequency dielectric constants, which reflect the dielectric properties of the material in static electric field and rapidly changing electric field respectively. is the excitation frequency of the UV light source, which determines the frequency range of the system response. is the relaxation time constant, the time scale of the dielectric relaxation process, n is the power exponent, which describes the rate at which the dielectric loss changes with frequency, and j is the imaginary unit; Dielectric response data is measured in the defective area and substituted into the Dissado-Hill model. The relationship between the dielectric response characteristics and the defect type is analyzed based on the Dissado-Hill model. The Dissado-Hill model is expressed by the following formula: ， in, is the complex permittivity, describing the overall dielectric response of the material, is the dielectric relaxation strength, reflecting the degree of polarization, is the characteristic relaxation time, which characterizes the time scale of the polarization process; a is the short-order parameter, which describes the correlation within the local structural unit; b is the long-order parameter, which describes the correlation between different structural units. By comparing the dielectric response differences between defect-free areas and areas with different types of defects, the correlation between defect characteristics and dielectric response parameters is established; A mapping model between wafer defect characteristics and UV light source parameters is established based on the association relationship; the defect type, size, and depth are used as input variables, and the wavelength, intensity, and irradiation angle of the UV light source are used as output variables, and the optimal parameter mapping model is obtained through neural network training.

2. The UV light source precise control method for wafer defect detection according to claim 1, characterized in that: The step of identifying defect information of the wafer surface image includes: Preprocessing the wafer surface image; Extracting defect features from pre-processed wafer surface images; Identifying a corresponding defect type according to the defect characteristics; Defect information is marked on the wafer surface image according to the defect type.

3. A UV light source precision control device for wafer defect detection, characterized in that: The UV light source precise control device for wafer defect detection includes: A determination module, used to determine initial detection parameters for wafer defect detection; an acquisition module, configured to acquire a wafer surface image of a target wafer according to the initial detection parameters; an identification module, configured to identify defect information of the wafer surface image; An evaluation module, configured to evaluate a detection effect value of the initial detection parameter based on the defect information; The evaluation module is further configured to generate a defect statistical report based on the defect information; determine a detection rate, a false detection rate, and a missed detection rate of the wafer defect detection based on the defect statistical report; perform an exponential penalty on the detection rate by converting the false detection rate and the missed detection rate into a penalty factor to determine a detection effect value; and if the detection effect value is lower than a preset threshold, adjust the UV light source according to the optimal parameters predicted by the mapping model until the detection effect value reaches a preset detection threshold; The determination module is further configured to determine the dielectric response characteristics of the target wafer under UV light irradiation based on a Jonscher general power law model; wherein the Jonscher general power law model is represented by the following formula: ， in, is the dielectric loss, is the low-frequency dielectric constant, are high-frequency dielectric constants, which reflect the dielectric properties of the material in static electric field and rapidly changing electric field respectively. is the excitation frequency of the UV light source, which determines the frequency range of the system response. is the relaxation time constant, the time scale of the dielectric relaxation process, n is the power exponent, which describes the rate at which the dielectric loss changes with frequency, and j is the imaginary unit; The analysis module is used to measure dielectric response data in the defective area and substitute the dielectric response data into the Dissado-Hill model; based on the Dissado-Hill model, the correlation between the dielectric response characteristics and the defect type is analyzed; the Dissado-Hill model is expressed by the following formula: ， in, is the complex permittivity, describing the overall dielectric response of the material, is the dielectric relaxation strength, reflecting the degree of polarization, is the characteristic relaxation time, which characterizes the time scale of the polarization process; a is the short-order parameter, which describes the correlation within the local structural unit; b is the long-order parameter, which describes the correlation between different structural units. By comparing the dielectric response differences between defect-free areas and areas with different types of defects, the correlation between defect characteristics and dielectric response parameters is established; The adjustment module is used to establish a mapping model between wafer defect characteristics and UV light source parameters based on the association relationship; the defect type, size, and depth are used as input variables, and the wavelength, intensity, and irradiation angle of the UV light source are used as output variables, and the optimal parameter mapping model is obtained through neural network training.

4. An electronic device, characterized in that: Comprising a memory and a processor, wherein: The processor is configured to execute a computer program stored in the memory; When the processor executes the computer program, the steps of the method for precisely controlling a UV light source for wafer defect detection according to any one of claims 1 to 2 are implemented.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for precisely controlling a UV light source for wafer defect detection according to any one of claims 1 to 2 are implemented.

Images