Wafer quality analysis method based on image recognition

Abstract

The application discloses a wafer quality analysis method based on image recognition, and particularly relates to the field of semiconductor manufacturing quality detection, which comprises the following steps: a multispectral and high-resolution fusion camera combination model is used to synchronously collect wafer two-dimensional images under visible light, infrared and ultraviolet light spectrums; surface three-dimensional topographic data are obtained by using a structured light three-dimensional reconstruction technology; an adaptive noise suppression algorithm is applied; noise characteristics are identified and filter parameters are dynamically adjusted through frequency domain analysis; targeted noise reduction is realized; a multiscale feature pyramid and a quantum dot label recognition fusion model are constructed; defect outlines and material heterogeneous regions are extracted layer by layer; microstructure characteristic parameters are extracted in combination with quantum dot fluorescence labeling technology; micro features and three-dimensional topographic data are mapped to a multi-modal space model; wafer three-property parameters are generated; a three-dimensional evaluation system is established; quality grades are comprehensively judged; and an analysis report is outputted, so that the precision and efficiency of wafer quality detection are significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor manufacturing quality detection, and more particularly to a wafer quality analysis method based on image recognition. BACKGROUND

[0002] With the continuous progress of semiconductor manufacturing process, wafer quality detection as a key link to ensure chip performance and reliability is increasingly valued by the industry, and the tiny defects of wafer surface and internal structure, such as lattice dislocation, impurity pollution and surface fluctuation, can significantly affect the subsequent process and the performance of the final product. Therefore, developing a high-precision and high-efficiency wafer quality detection method has become an important research direction to improve the level of semiconductor manufacturing.

[0003] The current technical solution mainly relies on the method of combining traditional optical imaging and machine vision to detect wafer defects. This solution usually uses a single spectrum high-resolution camera to collect wafer surface images, then extracts defect features through image processing algorithms, and supplements with simple three-dimensional measurement means to obtain surface topography data. Some systems also introduce standard template comparison or statistical analysis to determine wafer quality grade, and complete wafer quality analysis.

[0004] However, in actual use, it still has some disadvantages, such as image acquisition relying on a single spectrum band, which is difficult to capture the microscopic characteristics of wafer internal material heterogeneity and lattice defects, resulting in one-sided detection dimension, fixed parameter denoising algorithm, unable to dynamically adapt to different noise types such as high-frequency noise and periodic interference, easy to cause defect feature loss or noise residue, feature extraction is mainly based on a single scale, lacking precise recognition ability of microscopic structures such as nanoscale holes and impurity distribution, and no correlation model between defect features and wafer physical performance and electrical performance. Finally, the quality evaluation only stays at the surface defect judgment, which is difficult to reflect the comprehensive performance of the wafer and the matching degree with the process standard, and cannot meet the fine detection needs of high-end semiconductor manufacturing. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the present application provides a wafer quality analysis method based on image recognition, which solves the problems raised in the background art by the following scheme.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a wafer quality analysis method based on image recognition, comprising S1: constructing a multi-dimensional image acquisition model through a combination model of multi-spectrum and high-resolution fusion camera, collecting wafer two-dimensional images under visible light, infrared and ultraviolet spectrum, and simultaneously starting structure light three-dimensional reconstruction technology to obtain wafer surface three-dimensional topography data, wherein the structure light three-dimensional reconstruction technology uses 405nm laser to project sinusoidal fringes, the frame rate is 30fps, and the single-frame three-dimensional point cloud density is ≥100 points / mm².

[0007] S2: Construct a dynamic adjustment model, apply an adaptive noise suppression algorithm to the collected wafer two-dimensional image, identify noise characteristics through frequency domain analysis, and dynamically adjust filtering parameters based on the identification results to achieve targeted noise reduction, wherein the frequency domain analysis includes determining the noise dominant frequency band by calculating the power spectral density and quantifying the noise energy proportion to judge the interference strength;

[0008] S3: Construct a fusion model of multi-scale feature pyramid and quantum dot marker identification, extract the defect profile and material heterogeneity area of the wafer two-dimensional image through multi-scale feature pyramid layering, and use CdSe / ZnS core-shell structure quantum dots to specifically mark the lattice defects, nanoscale holes and gaps on the wafer surface. The particle size of the quantum dots is 5-8 nm, the excitation wavelength is 488 nm, and the emission wavelength is 655 nm. Combined with fluorescence intensity analysis, wafer microstructure characteristic parameters including lattice defect density, micro stress distribution parameters, and nanoscale hole and gap parameters are extracted.

[0009] S4: Construct a multi-modal space mapping model, map the wafer microstructure characteristic parameters and surface three-dimensional topography data to the correlation mapping function model, and generate wafer three-property parameters representing the wafer physical performance, electrical performance, and process standard matching degree.

[0010] S5: Construct a quality evaluation model, establish a three-dimensional evaluation system according to the wafer three-property parameters, determine the quality grade, and output a wafer quality analysis report.

[0011] Preferably, the combination model has a spectral range of 380-1100 nm, a minimum spectral bandwidth of 10 nm, and a combination mode of self-defined channel combination. When collecting images, different spectral channels are selected for synchronous acquisition according to the preset spectral combination mode.

[0012] Preferably, the fusion model includes a feature pyramid unit and a quantum dot marker identification unit, and realizes real-time data transmission through a data interaction interface with a transmission delay of ≤10 ms. Each layer scale of the feature pyramid unit is associated through a Gaussian difference operator, and each layer is provided with an independent feature filter to automatically filter invalid features with a signal-to-noise ratio <3.

[0013] Preferably, the quantum dots have a marking density of 10 ∧ 6 ∧ -10 ∧ 7 ∧ mm² on the wafer surface and a marking specificity of ≥99.5%. The specific binding groups on the surface of the quantum dots combine with the atoms at the lattice defects to form stable covalent bonds.

[0014] Preferably, the wafer microstructure feature parameters further include an impurity atom distribution parameter, and the micro stress distribution parameter is calculated by measuring the offset of the quantum dot fluorescence emission wavelength and according to the quantitative relationship between the offset and the quantum dot stress sensitivity coefficient.

[0015] Preferably, the physical performance parameter includes wafer surface flatness, the electrical performance parameter includes carrier mobility, resistivity deviation rate, and the process standard matching degree parameter includes photolithography accuracy and thin film thickness deviation rate.

[0016] Preferably, the carrier mobility is calculated based on the lattice defect density and the impurity atom distribution concentration, and the resistivity deviation rate is calculated based on the maximum stress in the micro stress distribution parameter and the volume fraction of the nanoscale hole and gap parameter.

[0017] Preferably, the three-dimensional evaluation system specifically includes a physical performance dimension, an electrical performance dimension and a process standard matching degree dimension, and the quality level is determined according to the comprehensive score of the three dimensions.

[0018] Preferably, the comprehensive score is calculated by weighting the scores of the physical performance dimension, the electrical performance dimension and the process standard matching degree dimension according to the weights of 30%, 40% and 30%, and a quality analysis report including a three-dimensional evaluation radar chart is output.

[0019] Technical effects and advantages of the present application:

[0020] The present application realizes multi-dimensional analysis of wafer quality through multi-dimensional image acquisition and multi-modal data fusion. The multi-dimensional image acquisition model covers visible light, infrared and ultraviolet spectra, and synchronously acquires three-dimensional topographic data. Combined with the hierarchical extraction of defect profiles and material heterogeneity regions and the specific identification of quantum dot markers by a feature pyramid fusion model, macro three-dimensional defects and microstructure features can be captured simultaneously. Compared with traditional single spectrum detection, the integrity and accuracy of feature extraction are greatly improved, and more abundant basic data is provided for subsequent quality evaluation.

[0021] The present application effectively guarantees the reliability of the analysis results through dynamic adjustment of the noise reduction mechanism and refined feature processing. The dynamic adjustment model identifies noise types through frequency domain analysis and adjusts filter parameters accordingly, thereby suppressing noise while preserving defect details to the greatest extent. Independent filters are set at each layer of the feature pyramid to filter invalid features. The quantum dot marker technology combines with the defect area with high specificity, significantly reducing the influence of interference factors on the extraction of microstructure feature parameters, and improving the measurement accuracy of key parameters such as lattice defect density and impurity distribution.

[0022] The application realizes comprehensive quantitative determination of wafer quality through a three-dimensional evaluation system of wafer three-property parameters, and a multi-modal space mapping model maps microscopic features and three-dimensional topography data into physical performance, electrical performance and process standard matching degree parameters; a quality evaluation model obtains an objective quality grade through weighted calculation, compared with a traditional grading method depending only on surface defects, the application can more comprehensively reflect actual performance of the wafer and matching degree with process requirements, provides a more scientific basis for quality control in the semiconductor manufacturing process, and helps to improve the yield and reliability of chip production. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 It is a schematic diagram of the overall structure of the application;

[0024] Figure 2 It is a schematic diagram of the targeted noise reduction structure of the application;

[0025] Figure 3 It is a schematic diagram of the wafer microscopic structure feature parameter mapping structure of the application. DETAILED DESCRIPTION

[0026] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0027] As shown in the image recognition-based wafer quality analysis method shown in Figure 1 , attached Figure 2 and attached Figure 3 , S1: a multi-dimensional image acquisition model is constructed through a combination model of a multi-spectrum and a high-resolution fusion camera, two-dimensional images of the wafer under visible light, infrared and ultraviolet light spectrum are acquired, and a structure light three-dimensional reconstruction technology is started synchronously to acquire three-dimensional topography data of the wafer surface, wherein the structure light three-dimensional reconstruction technology adopts 405nm laser projection of a sinusoidal fringe, the frame rate is 30fps, and the single-frame three-dimensional point cloud density is greater than or equal to 100 points / mm².

[0028] It needs to be specifically pointed out that the multi-dimensional image acquisition model is a combination model of multi-spectrum and high-resolution fusion camera, wherein the spectral detection range covers a wide band interval of 380-1100nm, the visible light region of 380-760nm is used to capture wafer surface defects, the near-infrared region of 760-1100nm is used to penetrate the surface coating to detect internal structure abnormalities, and the ultraviolet band of 200-380nm is used to identify the fluorescent characteristic defects of the material, wherein the minimum spectral bandwidth of the system is 10nm, and the narrow band channel of 10nm is subdivided in the visible light interval to realize fine band division, obtain high-resolution imaging components and clear images of each band, and present micron-level wafer details.

[0029] In terms of channel combination, the model adopts a self-defined channel combination mechanism, is equipped with a preset spectral combination mode and a synchronous triggering module, and the preset spectral combination mode includes three scene-based schemes: for regular surface detection, a visible light three-channel red, green and blue and a near-infrared single-channel combination synchronous acquisition is enabled, wafer surface topography and basic material information are quickly obtained, for high-precision defect screening, an ultraviolet channel plus five narrow-band near-infrared channel combination synchronous acquisition is enabled, and subtle material inhomogeneity is identified through multi-band comparison, and for wafers of special process, user-defined channel parameters are supported, for example, three discrete wavebands of 400nm, 800nm and 1000nm are specified for special detection.

[0030] In terms of acquisition control, the high-resolution fusion camera and the structured light three-dimensional reconstruction system realize precise synchronization in time and space, the structured light three-dimensional reconstruction technology adopts 405nm laser projection of sinusoidal stripes, the frame rate is 30fps, and the single-frame three-dimensional point cloud density is ≥100 points / mm², when the wafer is in a position to be detected, the system first triggers the structured light projector to project an encoded grating pattern to the wafer surface, and simultaneously starts the high-resolution fusion camera to capture the deformed grating image, the three-dimensional topography data of the wafer surface is quickly calculated through the triangulation principle, the three-dimensional topography data includes surface relief and height data, curvature and slope data, and microstructure three-dimensional model data, at the same time, the high-resolution fusion camera synchronously acquires two-dimensional image data of multiple spectral channels in the same field of view range according to the preset spectral combination mode, the two-dimensional image includes surface defect contour image and material heterogeneity region image.

[0031] It needs to be further pointed out that the multi-dimensional image acquisition model synchronously acquires three-dimensional topography data when acquiring two-dimensional images of the wafer surface, wherein the two-dimensional image reflects the characteristics of material chemical component difference and surface pollution, and the three-dimensional topography data represents the geometric parameters of wafer warpage, flatness and defect degree.

[0032] S2: Construct a dynamic adjustment model, apply an adaptive noise suppression algorithm to the collected wafer two-dimensional image, identify noise characteristics through frequency domain analysis, and dynamically adjust filter parameters based on the identification results to achieve targeted noise reduction. The frequency domain analysis includes determining the noise dominant frequency band by calculating the power spectral density and quantifying the noise energy proportion to determine the interference strength.

[0033] It needs to be specifically pointed out that the targeted noise reduction is to perform two-dimensional fast Fourier transform (2D-FFT) on the wafer two-dimensional image and perform frequency spectrum shift, identify noise characteristics by analyzing the amplitude spectrum, determine the noise dominant frequency band by calculating the power spectral density, judge the interference strength by quantifying the noise energy proportion, distinguish the types of Gaussian noise and periodic noise according to the frequency spectrum mode, dynamically adjust the filter parameters based on the identification results, enhance the standard deviation of the adaptive Gaussian filter for high-frequency noise, calibrate the center frequency and bandwidth of the notch filter for periodic interference, optimize the wavelet threshold and decomposition level for mixed noise. After the filtering process is completed in the frequency domain, the inverse Fourier transform is used to convert back to the spatial domain, and a real-time feedback mechanism is introduced, with peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) as indicators, the noise residual is increased by 10% filter strength, and the detail loss is reduced by 10% filter strength, achieving targeted noise suppression and precise preservation of wafer defect details.

[0034] It needs to be further explained that the standard for identifying noise characteristics by analyzing the amplitude spectrum is that high-energy distribution appears in the high-frequency region (the Euclidean distance from any point in the region to the center is greater than one-fourth of the distance between the edges of the entire region), which is determined as high-frequency noise, a specific frequency point appears as an isolated high-energy peak, which is determined as periodic noise, and the energy distribution has no obvious pattern and covers multiple frequency bands, which is determined as mixed noise.

[0035] The standard for determining the noise dominant frequency band by calculating the power spectral density is to square the amplitude spectrum to obtain the power spectral density (PSD), the formula is , slide the window along the frequency axis, calculate the PSD mean value in each window, and when the window mean value is more than 3 times the signal frequency band mean value, mark the frequency range corresponding to the window as the noise dominant frequency band.

[0036] The standard for quantifying the noise energy proportion to determine the interference strength is to calculate the ratio of the total energy in the noise dominant frequency band to the total energy in the full frequency band to obtain the noise energy proportion, and to determine the interference strength by comparing the proportion with the set threshold. The specific analysis method includes:

[0037] Extract the frequency coordinate , the total energy of the full frequency band ;

[0038] Analyze the total energy in the noise dominant frequency band , where This represents the region in the frequency domain corresponding to the noise-dominant frequency band. Represents coordinates in the frequency domain Power spectral density at;

[0039] Analysis of noise energy ratio When r < 10%, it is considered low interference; when 10% ≤ r ≤ 30%, it is considered medium interference; and when r > 30%, it is considered high interference.

[0040] The criteria for distinguishing Gaussian noise, periodic noise, and mixed noise based on spectral patterns are as follows: Gaussian noise: PSD is continuously and uniformly distributed in the high-frequency band, with no obvious peaks, and the proportion of noise energy gradually decreases as the frequency increases; Periodic noise: PSD has sharp peaks at specific frequencies, and the peak energy is more than 5 times that of the surrounding area; Mixed noise: There are both high-frequency uniformly distributed areas and discrete peaks, and the energy proportion of both types of areas exceeds 5%.

[0041] It should be further explained that the standard for dynamically adjusting the filtering parameters based on the recognition results is the standard deviation of the adaptive Gaussian filter for enhancing high-frequency noise, under low interference conditions. Interference High interference The center frequency and bandwidth of the notch filter are calibrated for periodic interference. The calibrated center frequency is the peak frequency coordinate of the periodic noise. (Regarding the bandwidth...) When there is low interference Interference High interference Optimize wavelet threshold and decomposition level for mixed noise, focusing on the number of decomposition levels. When there is low interference Interference High interference For wavelet threshold Specific analysis methods include:

[0042] Analyze the standard deviation of noise ;

[0043] Analysis of wavelet threshold .

[0044] S3: Construct a fusion model of multi-scale feature pyramid and quantum dot labeling recognition. The defect contours and material heterogeneous regions of the wafer's two-dimensional image are extracted layer by layer through the multi-scale feature pyramid. At the same time, CdSe / ZnS core-shell structured quantum dots are used to specifically fluorescently label the lattice defects, nanoscale pores and gaps on the wafer surface. The quantum dots have a particle size of 5-8 nm, an excitation wavelength of 488 nm, and an emission wavelength of 655 nm. Combined with fluorescence intensity analysis, wafer microstructure feature parameters including lattice defect density, micro-stress distribution parameters, and nanoscale pore and gap parameters are extracted.

[0045] It should be particularly pointed out that the multi-scale feature pyramid and the quantum dot marker recognition fusion model includes a feature pyramid unit and a quantum dot marker recognition unit, real-time data transmission is realized through a data interaction interface and the transmission delay is ≤10 ms, the scale of each layer of the feature pyramid unit is realized through a Gaussian difference operator to realize hierarchical correlation and the feature mapping error of adjacent hierarchical layers is ≤2%, an independent feature filter is arranged at each layer to automatically filter invalid features with a signal-to-noise ratio < 3.

[0046] The quantum dot fluorescence labeling technology adopts CdSe / ZnS core-shell structure quantum dots, the particle size of the quantum dots is 5-8 nm, the excitation wavelength is 488 nm, the emission wavelength is 655 nm, the labeling density on the wafer surface is 10 6 -10 7 / mm² and the labeling specificity is ≥99.5%, the specific binding groups on the surface of the quantum dots for the wafer lattice defects form stable covalent bonds with the atoms at the lattice defects to improve the fluorescence labeling combination rate, the wafer microstructure characteristic parameters include the lattice defect density, the micro stress distribution parameter, the nanoscale hole and gap parameter, and the impurity atom distribution parameter.

[0047] It should be further pointed out that a 5-layer multi-scale feature pyramid is constructed, the defect profile and the material heterogeneity region of the two-dimensional image are extracted from the scale level of 1 um x 1 um to 100 um x 100 um, the wafer surface lattice defects are specifically labeled through the quantum dot fluorescence labeling technology, the number of labeled points in the two-dimensional image is counted to convert the defect number per unit area, which is recorded as the lattice defect density, the fluorescence emission wavelength drift of the quantum dots at different positions on the wafer is measured, and the stress distribution parameter is converted and obtained, which is recorded as the micro stress distribution parameter, the nanometer holes and gaps are filled with quantum dots, the geometric parameters of the holes and gaps are measured by using a high-resolution microscope combined with fluorescence intensity, which are recorded as the nanoscale hole and gap parameter, and the spatial distribution of the quantum dot fluorescence intensity of the labeled impurity atoms is obtained to obtain the distribution of the impurity atoms in the wafer microstructure, which is recorded as the impurity atom distribution parameter.

[0048] It should be further pointed out that the method for converting the stress distribution parameter from the fluorescence emission wavelength drift is to calculate the stress value and spatial distribution of the corresponding area of the wafer according to the quantitative relationship between the offset of the fluorescence emission wavelength of the quantum dots and the stress sensitivity coefficient, and the specific analysis method includes:

[0049] Measuring the reference fluorescence emission wavelength of the quantum dots in the stress-free state The actual emission wavelength after being affected by stress , obtaining the wavelength offset , calling the stress sensitivity coefficient of the quantum dots of this type , according to the wavelength offset The ratio of the stress sensitivity coefficient The local stress size is calculated, the above measurement and calculation are repeated for different positions of the wafer, and the stress distribution parameters are generated in combination with the spatial coordinate information.

[0050] S4: Construct a multi-modal space mapping model, map the wafer microstructure feature parameters and the surface three-dimensional topography data to the correlation mapping function model, and generate wafer three-property parameters representing the wafer physical performance, electrical performance and process standard matching degree.

[0051] Need to be specified, the multi-modal space mapping model is a multi-dimensional image data and wafer three-property correlation mapping function model, the wafer three-property is physical performance, electrical performance and process standard matching degree, the wafer microstructure feature parameters and the surface three-dimensional topography data are mapped to different function models, and the wafer three-property parameters are generated, wherein the physical performance parameters include wafer surface flatness, the electrical performance parameters include carrier mobility, resistivity deviation rate, and the process standard matching degree parameters include lithography precision and film thickness deviation rate.

[0052] It needs to be further explained that the wafer surface flatness is specifically analyzed in the following way: in the surface fluctuation and height data, the height value of each sampling point on the wafer surface is ( is the average value of the height, the standard deviation of the surface fluctuation is analyzed , and the wafer surface flatness is further analyzed , wherein is a proportional coefficient determined by the wafer material characteristics and the detection standard.

[0053] The specific analysis method of wafer carrier mobility includes: extracting wafer lattice defect density and impurity atom distribution concentration, respectively denoted as , and analyzing wafer carrier mobility , wherein is the mobility of an ideal defect-free and impurity-free crystal, and a and b are proportional coefficients, the specific values of which are determined by wafer material characteristics and detection standards.

[0054] The specific analysis method of wafer resistivity deviation rate includes: extracting the maximum stress in wafer micro stress distribution parameters as , extracting the volume fraction of nanoscale pores and gaps as , and analyzing the wafer resistivity deviation rate , wherein are the target threshold values of wafer micro stress maximum value and nanoscale pore and gap parameter volume fraction respectively, which are determined by wafer material characteristics and detection standards.

[0055] ​The wafer lithography precision analysis method specifically includes: measuring wafer curvature radius and slope data , calculating defocus amount and projection offset, both of which reflect wafer lithography precision, the defocus amount , wherein is the diameter of the lithography field of view, and the projection offset , wherein is the distance from the lithography lens to the wafer surface.

[0056] The wafer film thickness deviation rate analysis method specifically includes: obtaining actual thickness of each point of the film through microstructure three-dimensional model data in wafer three-dimensional topography, and the ratio of the difference between the actual thickness and the target thickness to the target thickness is the thickness deviation rate.

[0057] S5: Construct a quality evaluation model, establish a three-dimensional evaluation system according to wafer three-property parameters, determine the quality grade, and output a wafer quality analysis report.

[0058] It should be specifically noted that the three-dimensional evaluation system in the quality evaluation model specifically includes a physical performance dimension, an electrical performance dimension, and a process standard matching degree dimension, wafer quality three-dimensional evaluation is performed according to wafer three-property parameters, and the quality grade is determined, and finally a quality analysis report is output, specifically including a three-dimensional evaluation radar chart and a quality grade.

[0059] It should be further noted that, in terms of three-dimensional evaluation system construction, wafer physical performance scores adopt a 100-point deduction system, surface flatness is taken as a reference value with total thickness variation TTV≤5um, and actual value is deducted by 5 points for every 10% exceeding the reference value, 0 points are deducted for exceeding 100% and above, and the total score of physical performance is composed.

[0060] The electrical performance carrier mobility takes the target lower limit (determined by wafer material characteristics and detection standards) as a reference, and is deducted by 10 points for every 5% below the target lower limit, and 0 points are deducted for 50% below, the resistivity deviation rate takes ±5% as a reference, and is deducted by 5 points for every 1% exceeding, and 0 points are deducted for reaching ±25% and above, and the two parameters each account for 50 points, and together constitute the total score of electrical performance.

[0061] The process standard matching degree lithography precision takes ±2nm as a reference, and is deducted by 5 points for every 0.5nm exceeding, and 0 points are deducted for deviation ≥±8nm, the film thickness deviation rate takes ±3% as a reference, and is deducted by 5 points for every 1% exceeding, and 0 points are deducted for reaching ±10% and above, and together constitute the total score of process standard matching degree.

[0062] The quality grade is divided into four grades of excellent, good, qualified and unqualified, the three-dimensional evaluation system comprehensive score is calculated by weighting three dimension scores according to the weight of 30% of physical performance, 40% of electrical performance and 30% of process standard matching degree, the full score is 100 points, the comprehensive score of 90 points and above is excellent, 80-89 points is good, 60-79 points is qualified, and less than 60 points is unqualified.

[0063] The three-dimensional evaluation radar chart takes three dimensions as coordinate axes, marks and connects the comprehensive scores of each dimension, and directly displays the performance of the wafer in each aspect.

[0064] Secondly: the drawings in the disclosed embodiments of the application only involve structures related to the disclosed embodiments, other structures can refer to the usual design, and in the case of no conflict, the same embodiment and different embodiments of the application can be combined with each other;

[0065] Finally: the above only describes the preferred embodiments of the application and is not used to limit the application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application should be included in the protection scope of the application.

Claims

1. A wafer quality analysis method based on image recognition, characterized in that, Includes the following steps: S1: By combining a multispectral and high-resolution camera model, a multi-dimensional image acquisition model is constructed to acquire two-dimensional images of the wafer under visible light, infrared and ultraviolet spectra. Simultaneously, structured light three-dimensional reconstruction technology is launched to obtain three-dimensional morphology data of the wafer surface. The structured light three-dimensional reconstruction technology uses a 405nm laser to project sinusoidal stripes at a frame rate of 30fps, with a single-frame three-dimensional point cloud density ≥100 points / mm². S2: Construct a dynamic adjustment model, apply an adaptive noise suppression algorithm to the acquired two-dimensional wafer images, identify noise features through frequency domain analysis, and dynamically adjust the filtering parameters based on the identification results to achieve targeted noise reduction. The frequency domain analysis includes determining the dominant noise frequency band by calculating the power spectral density and quantifying the noise energy ratio to determine the interference intensity. S3: Construct a fusion model of multi-scale feature pyramid and quantum dot labeling recognition. The defect contours and material heterogeneous regions of the wafer's two-dimensional image are extracted layer by layer through the multi-scale feature pyramid. At the same time, CdSe / ZnS core-shell structured quantum dots are used to specifically fluorescently label the lattice defects, nanoscale pores and gaps on the wafer surface. The quantum dots have a particle size of 5-8nm, an excitation wavelength of 488nm, and an emission wavelength of 655nm. Combined with fluorescence intensity analysis, wafer microstructure feature parameters including lattice defect density, micro-stress distribution parameters, and nanoscale pore and gap parameters are extracted. S4: Construct a multimodal spatial mapping model to map the wafer microstructure feature parameters and surface three-dimensional morphology data to the correlation mapping function model, and generate wafer three-property parameters characterizing the wafer's physical properties, electrical properties and process standard matching degree; S5: Construct a quality assessment model, establish a three-dimensional assessment system based on the three wafer properties parameters, determine the quality level, and output a wafer quality analysis report.

2. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The combined model has a spectral range of 380-1100nm and a minimum spectral bandwidth of 10nm. The combination method is a custom channel combination. When acquiring images, different spectral channels are selected for synchronous acquisition according to the preset spectral combination mode.

3. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The fusion model includes a feature pyramid unit and a quantum dot tag recognition unit, and achieves real-time data transmission with a transmission latency of ≤10ms through a data interaction interface. The feature pyramid unit achieves hierarchical association between its various levels through the difference of Gaussian operator, and each level is equipped with an independent feature filter to automatically filter out invalid features with a signal-to-noise ratio of <3.

4. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The quantum dots are labeled at a density of 10 on the wafer surface. ∧ 6 ∧ -10 ∧ 7 ∧ With a labeling specificity of ≥99.5%, the specific binding groups modified on the quantum dot surface combine with atoms at lattice defects to form stable covalent bonds.

5. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The wafer microstructure characteristic parameters also include impurity atom distribution parameters. The micro-stress distribution parameters are calculated by measuring the offset of the quantum dot fluorescence emission wavelength and based on the quantitative relationship between the offset and the quantum dot stress sensitivity coefficient.

6. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The physical properties include wafer surface flatness; the electrical properties include carrier mobility and resistivity deviation; and the process standard matching degree includes photolithography accuracy and thin film thickness deviation.

7. The wafer quality analysis method based on image recognition according to claim 6, characterized in that: The carrier mobility is calculated based on the lattice defect density and the concentration of impurity atoms, and the resistivity deviation rate is calculated based on the maximum stress value and the volume fraction of nanoscale pores and gaps in the micro-stress distribution parameters.

8. The wafer quality analysis method based on image recognition according to claim 1, characterized in that: The three-dimensional evaluation system specifically includes physical performance, electrical performance, and process standard matching. The quality level is determined based on the comprehensive score of the three dimensions.

9. The wafer quality analysis method based on image recognition according to claim 8, characterized in that: The overall score is calculated by weighting the scores of the physical performance dimension, electrical performance dimension, and process standard matching degree dimension with weights of 30%, 40%, and 30%, respectively, and outputs a quality analysis report including a three-dimensional evaluation radar chart.

Images