High-resolution semiconductor surface and interior flaw detection method and device based on multiphoton nonlinear fluorescence effect

Abstract

The invention discloses a high-resolution semiconductor surface and interior flaw detection method and device based on a multi-photon nonlinear fluorescence effect, and relates to nonlinear optics and optical imaging semiconductor detection. The short-wave infrared light penetrates through a semiconductor to be detected, a surface or an internal defect to trigger scattering, the transmittance / reflectivity and the power are changed, the thin film containing the high-order nonlinear fluorescent nanoparticles is excited, the luminance generates magnitude order change, and the defect detection contrast ratio and the imaging resolution ratio are improved. Full-automatic control is achieved through an FPGA and LabVIEW, sub-images are automatically collected and spliced, and gray scale abnormal areas are analyzed to judge defect types and distribution. The optical imaging device comprises an exciting light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control analysis module, realizes short-wave infrared light excitation and visible light-near infrared detection, and has the advantages of intelligence, high resolution, low cost and large-area full-automatic flaw detection. The method is suitable for large-scale semiconductor efficient flaw detection, gets rid of dependence on an infrared imaging camera, and reduces system cost.

Description

Technical Field

[0001] The present invention relates to the field of nonlinear optics and optical imaging semiconductor detection, and in particular to a low-cost, high-resolution semiconductor surface and internal flaw detection method and device based on multi-photon nonlinear fluorescence effect. Background Art

[0002] With the development of modern electronic technology and chip industry, semiconductors are important basic materials, and their quality directly affects device performance and production yield. However, during the manufacturing and processing of semiconductor materials, tiny defects such as bubbles, scratches, and cracks are prone to occur on their surface and inside, which may cause device failure or performance degradation. Therefore, it is very important to detect semiconductor surface and internal defects efficiently and accurately.

[0003] Traditional semiconductor flaw detection methods rely on microscopic optics or laser interferometry technology, but there are problems such as high equipment cost, slow detection speed, and small detection field of view. Due to the good transmittance of semiconductor materials to infrared light, short-wave infrared imaging technology has become a very promising semiconductor quality detection method in recent years. However, this technology relies on high-cost short-wave infrared cameras and is difficult to promote on a large scale.

[0004] Nonlinear optical technology has received extensive attention in recent years. Nonlinear optical materials represented by lanthanide-doped upconversion fluorescent materials can generate visible light emission through infrared excitation and exhibit excellent optical nonlinear characteristics, providing new solutions for flaw detection technology. The present invention proposes a large-area semiconductor surface and internal flaw detection method and device based on short-wave infrared excitation and visible light-near infrared detection. By introducing multi-photon nonlinear effects and intelligent control systems, low-cost, high-resolution, large-viewing-field, and efficient semiconductor surface and internal defect detection can be achieved. Summary of the invention

[0005] The purpose of the present invention is to address the deficiencies of the prior art and provide a high-resolution semiconductor surface and internal flaw detection method and device based on multi-photon nonlinear fluorescence effect that integrates short-wave infrared excitation and visible light-near infrared detection. Through the deep integration of intelligent technology, high-resolution, low-cost, and large-area fully automatic semiconductor surface and internal flaw detection can be achieved, changing the traditional flaw detection method and providing technical support for semiconductor production quality control.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect, comprising: an excitation light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control and analysis module; under the excitation of short-wave infrared light, based on the overall coordination of the intelligent control module, the remaining modules are driven to complete visible light-near infrared detection, intelligent, high-resolution, low-cost, large-area fully automatic semiconductor flaw detection technology.

[0008] The excitation light generation module includes a continuous short-wave infrared laser, a pulsed short-wave infrared laser and a short-wave infrared LED light source, as well as a filter group and a reflector placed in sequence along the direction of the light beam emitted by the excitation light source; the excitation light source generates a steady-state short-wave infrared light beam output; the filter group is used to purify the light beam; the reflector is used to adjust the light angle to maximize the space utilization of the detection device; the continuous short-wave infrared laser is used for high-stability detection, the pulsed short-wave infrared laser is used for transient defect analysis, and the short-wave infrared LED light source is used for low-cost rapid screening;

[0009] The sample scanning module is divided into a transmission sample scanning module or a reflection sample scanning module; wherein:

[0010] The transmission sample scanning module includes a three-dimensional electric translation stage and a semiconductor sample; the three-dimensional electric translation stage is used for spatial positioning of the semiconductor sample, moves in the XYZ direction according to the set resolution, and gradually scans the entire sample area; the semiconductor sample is fixed on the three-dimensional electric translation stage and can be penetrated by short-wave infrared light, and the transmitted light is used for subsequent excitation of nonlinear fluorescent film sample imaging, so as to realize the detection of defects such as bubbles, microcracks, crystal defects, etc. on the surface and inside of the semiconductor;

[0011] The reflective sample scanning module includes a beam splitter, a three-dimensional electric translation stage and a semiconductor sample; the beam splitter is used to extract imaging reflected light for subsequent excitation of nonlinear fluorescent film sample imaging; the three-dimensional electric translation stage is used for spatial positioning of the semiconductor sample, moves in the XYZ direction according to the set resolution, and gradually scans the entire sample area; the semiconductor sample is fixed on the three-dimensional electric translation stage and can reflect short-wave infrared light; these reflected lights contain information on semiconductor surface and internal defects, such as scratches, pits, particle contamination, and edge collapse and cracks, so as to realize the detection of semiconductor surface and internal defects;

[0012] The nonlinear fluorescent film imaging module includes a beam correction lens group, a nonlinear fluorescent film, a filter group, an imaging lens and a camera; the beam correction lens group is used to match the imaging lens parameter design, correct optical aberrations, and improve the beam utilization rate; the nonlinear fluorescent film is evenly mixed with luminescent materials with nonlinear fluorescent effects to generate emission fluorescence with orders of magnitude changes to improve imaging contrast and resolution; the filter group is used to screen imaging bands to match different application scenarios; the imaging lens is used to adjust the imaging aperture and field of view to ensure that the image is completely covered on the sensor to avoid image edge distortion or shearing; the camera is used to record imaging pictures.

[0013] The intelligent control and analysis module includes a computer and an FPGA chip; the computer has built-in LabVIEW software to complete parameter setting, real-time data transmission, visual display, image stitching and defect analysis, and coordinate the various modules of the drive system; the FPGA chip is connected to the excitation light source and the imaging camera to change the excitation and imaging modes, and communicates with the computer through a USB or Ethernet interface to send pulses to trigger the camera, and cooperates with the three-dimensional electric translation stage to achieve cyclic scanning to obtain large-area, high-resolution stitched images.

[0014] Furthermore, the filter set of the excitation light generating module is assembled by a rotatable bracket for adjusting and matching different excitation light wavelengths.

[0015] Furthermore, the filter set of the data acquisition and analysis module can be assembled by a rotatable bracket to adjust and match different collected fluorescence wavelengths.

[0016] Furthermore, the operating wavelength of the excitation light source of the excitation light generating module is mainly located in the range of 1100 to 3000 nm where the semiconductor's light absorption rate is significantly reduced, and the operating wavelength also matches the absorption region of the nonlinear fluorescent material.

[0017] Furthermore, the excitation light source has a built-in light intensity automatic adjustment device, which can dynamically adjust the light intensity to meet the detection requirements of semiconductors with different thicknesses.

[0018] Furthermore, the semiconductor sample has defects caused by the manufacturing process, such as surface defects (scratches, pits, particle contamination, etc.), internal defects (bubbles, microcracks, crystal defects, etc.), and edge defects (collapse, cracks, etc.). These defects will cause the scattering effect of the incident short-wave infrared light, thereby reducing its transmittance / reflectivity. The modulated excitation light will excite the nonlinear fluorescent film to achieve imaging of the defects.

[0019] Furthermore, the fluorescent film of the nonlinear fluorescent film imaging module uses a transparent polymer material as a film substrate, has good transparency and mechanical properties, and can be uniformly mixed with a fluorescent material with a nonlinear effect to achieve a response to short-wave infrared light. The materials used include but are not limited to polystyrene (PS), polydimethylsiloxane (PDMS), polyvinyl alcohol (PVA), polyacrylic acid (PAA), polymethyl methacrylate (PMMA), polyparaxylene (PET), siloxane, and acrylic resin polymers.

[0020] Furthermore, the fluorescent material has good optical nonlinear response and can be efficiently excited by short-wave infrared light. Such fluorescent materials include common organic dyes, quantum dots and lanthanide-doped upconversion fluorescent materials. The excitation process of upconversion fluorescent materials includes continuous excited state absorption upconversion process (fluorescence nonlinearity is generally less than 5), energy transfer upconversion process (fluorescence nonlinearity is generally less than 10) and photon avalanche upconversion process (fluorescence nonlinearity is generally greater than 10).

[0021] Furthermore, the lanthanide elements include but are not limited to neodymium (Nd) that can be excited by 1350-1400 nm and 1750-1850 nm, erbium (Er) that can be excited by 1450-1650 nm, holmium (Ho) that can be excited by 1500-2000 nm, thulium (Tm) that can be excited by 1200-1500 nm and 1650-1950 nm, and dysprosium (Dy) that can be excited by 1200-1300 nm. In addition, by designing the core-shell structure and modulation of the excitation mode, lanthanide elements such as praseodymium (Pr), promethium (Pm), samarium (Sm), europium (Eu), and terbium (Tb) that match short-wave infrared excitation are also suitable for the selection range of the above-mentioned fluorescent materials.

[0022] Furthermore, the lanthanide-doped fluorescent material can be doped with a single lanthanide luminescent element or co-doped with multiple lanthanide elements. The nonlinear fluorescence energy transfer effect of multiple lanthanide doping will amplify the final fluorescence nonlinearity, which can further improve the semiconductor surface and internal flaw detection performance.

[0023] Furthermore, the FPGA chip can be replaced by a microcontroller and a single-chip microcomputer control system to realize the conversion and control of the trigger signal; the LabVIEW can be replaced by Python, Matlab, LabWindows / CVI development environment to realize UI interface design and coordinate the various modules of the control system.

[0024] The present invention also provides a high-resolution semiconductor surface and internal flaw detection method based on multi-photon nonlinear fluorescence effect, comprising the following steps:

[0025] S1: Equipment initialization: Ensure that the excitation light source module outputs stable short-wave infrared light, adjust the beam output power and collimation, purify through filters, ensure that the beam intensity and radiation angle are adapted to the characteristics of semiconductor samples and nonlinear fluorescent films; ensure that the three-dimensional electric translation stage is connected to the computer through a suitable communication protocol, and the software initialization settings are successful; fix the semiconductor sample to be tested and the fluorescent film uniformly mixed with nonlinear effect fluorescent materials for subsequent detection;

[0026] It should be noted that in step S1, the excitation light generation module adjusts the output power to penetrate the semiconductor to achieve a specific light intensity, ensuring that the luminescent response of the fluorescent material during the excitation process is within the nonlinear region, thereby effectively improving the image contrast and defect detection accuracy.

[0027] It should be noted that in step S1, the thickness of the fluorescent film is uniform and the fluorescent material is evenly distributed. The surface ligands of the fluorescent material are modified (such as coating a hydrophilic or lipophilic surface layer) to ensure the dispersion of the fluorescent material and the compatibility with the membrane matrix solution. The fluorescent material is added to the membrane matrix solution, and the solution is fully stirred using a magnetic stirrer or a high-speed vortex oscillator to ensure that the fluorescent material is evenly dispersed, and the time and power of the ultrasonic treatment are appropriately increased to increase the dispersibility.

[0028] It should be noted that in step S1, the fluorescent film is prepared by coating a film matrix solution containing fluorescent materials on a glass substrate using methods such as spin coating, scraper coating, and flow coating, and a fluorescent film with uniform thickness and uniform distribution of fluorescent materials is obtained by natural molding or heat curing.

[0029] It should be noted that in step S1, the fluorescent materials include quantum dots, organic dyes, traditional upconversion fluorescent nanoparticles, and photon avalanche effect upconversion nanoparticles; the nonlinearity of quantum dots and organic dyes is generally less than 3, the nonlinearity of traditional upconversion fluorescent nanoparticles is generally less than 6, and the nonlinearity of photon avalanche effect rare earth doped upconversion nanoparticles is generally greater than 10.

[0030] S2: Thin film excitation and imaging detection: After being scattered by defects, the transmission / reflection power of short-wave infrared light penetrating the semiconductor sample changes. Specifically, in the transmission flaw detection mode, the excitation light that passes through the semiconductor is modulated by the corrective lens group and irradiated to the fluorescent film, generating a visible light-near infrared signal with an emission intensity that varies in magnitude; in the reflection flaw detection mode, the beam splitter separates the reflected excitation light, which is modulated by the corrective lens group and irradiated to the fluorescent film, also generating a visible light-near infrared signal with an emission intensity that varies in magnitude. The FPGA chip triggers the camera to capture images and records the first frame of image data and related position information; the LabVIEW program controls the three-dimensional electric translation stage through instructions to accurately move the semiconductor sample to the preset starting position to ensure that every area on its surface is excited and the image is captured. At each displacement stop point, the excitation-collection process is repeated until the entire sample surface is scanned. After each scan, a frame of image will be captured for subsequent stitching and analysis;

[0031] It should be noted that in step S2, the FPGA chip realizes the generation, triggering gating and processing functions of the pulse signal; the LabVIEW realizes parameter setting, real-time data transmission and visual display; the detection process realizes the fully automatic operation from three-dimensional electric translation stage positioning to excitation light output, camera image acquisition and data analysis through the LabVIEW control system and automation program, thereby improving the efficiency, accuracy and automation of the entire flaw detection process.

[0032] It should be noted that in step S2, the three-dimensional electric translation stage is used for spatial positioning of the sample. With the help of high-precision pulses generated inside the FPGA chip, the XYZ axis stepping is controlled by setting the stepping interval and frequency, thereby realizing the stepping signal control of the three-dimensional electric translation stage and the timing management of image acquisition, and realizing rapid detection of the entire sample area.

[0033] S3: Image stitching: After the 3D electric translation stage completes the image scanning process, the FPGA chip stores multiple sub-images with specific position information. Based on the unique position information and precise image stitching algorithm, LabVIEW stitches multiple sub-images in sequence into a complete semiconductor flaw detection image. During the stitching process, any image overlap areas caused by displacement errors will be identified and processed to ensure the continuity and seamless transition of the stitched image, and ultimately generate a large-area flaw detection image with high resolution and precise stitching;

[0034] It should be noted that in step S3, the image stitching process uses a stitching algorithm based on the coordinate data of the three-dimensional electric translation stage to spatially locate the image captured in each frame and stitch them into a complete flaw detection image. During the stitching process, the stitching error caused by inaccurate displacement is automatically eliminated to improve the image synthesis accuracy.

[0035] It should be noted that, in step S3, the stitching algorithm ensures accurate alignment between images by matching feature points in multiple sub-images, thereby generating a seamless large-area image.

[0036] It should be noted that in step S3, the stitching algorithm includes a series of sub-algorithms to perform stitching error correction, including automatic calibration and alignment adjustment, to ensure error-free and high-quality stitching results.

[0037] S4: Defect analysis: Grayscale distribution analysis is performed based on the stitched large-area images to detect areas in the image where the grayscale value changes suddenly or deviates from the normal range; bubbles, scratches, cracks and other defects on the surface and inside of the semiconductor are automatically identified through image processing methods; machine vision algorithms classify and count the detected abnormal areas, and calculate the type, distribution and size of the defects. This information will serve as the basis for further analysis and evaluation of semiconductor quality. The system can quantify the results according to the set parameters, such as the minimum size of the defect, the number of defects allowed, etc., automatically generate reports, and provide intelligent quality assessment;

[0038] It should be noted that in step S4, the defect analysis step automatically detects brightness discontinuities or mutation areas in the image through an analysis algorithm based on the grayscale distribution of the image, and classifies and automatically marks the defect types through a machine learning algorithm. The system is continuously trained and optimized through an artificial intelligence model to improve its recognition ability and speed for new types of defects, and maintain a high recognition rate under dynamically changing process conditions.

[0039] It should be noted that in step S4, the defect analysis process will provide a defect detection report in real time, and automatically classify and count different types of defects on the surface or inside of the semiconductor to facilitate further analysis and quality assessment.

[0040] Compared with the prior art, the technical effects and advantages of the present invention are:

[0041] 1. Significantly improve the accuracy of flaw detection: By introducing multi-photon nonlinear effects and high-order nonlinear fluorescent materials, the present invention significantly improves the detection sensitivity and contrast of defects. By using short-wave infrared excitation to achieve fluorescence detection in the visible light range, even tiny surface and internal defects such as bubbles, scratches and cracks can be clearly distinguished, thereby improving the accuracy and reliability of detection to sub-ten-micron level;

[0042] 2. Transmission realizes three-dimensional flaw detection function: Short-wave infrared light source can effectively penetrate semiconductor materials to realize deep identification of defects on the surface and inside of semiconductors. Through the excitation and fluorescence detection of short-wave infrared light, it can not only detect defects on the surface of semiconductors, but also identify potential cracks, bubbles and other defects inside semiconductors, realizing three-dimensional flaw detection function. This deep detection capability provides more dimensional information for semiconductor detection, which helps to improve the accuracy and comprehensiveness of defect detection;

[0043] 3. Realize low-cost detection: The present invention uses short-wave infrared excitation technology and nonlinear fluorescent materials to replace the traditional high-cost short-wave infrared camera, and only needs to be equipped with an ordinary silicon-based CCD or COMS camera to complete defect detection. By reducing hardware requirements, the equipment cost is greatly reduced, providing economic advantages for the large-scale application of technology;

[0044] 4. Broaden the detection field of view: Combining the multi-photon effect with intelligent scanning and stitching technology, the present invention realizes seamless detection of large-area semiconductor surface and internal defects. By precisely controlling the displacement platform and image stitching algorithm, the device can effectively solve the efficiency problem caused by the small detection field of view in traditional flaw detection methods and meet the needs of modern industry for large-area detection;

[0045] 5. High-contrast detection imaging: The present invention utilizes the sensitive response of highly nonlinear fluorescent materials to excitation power, which can not only detect traditional defects such as scratches, but also capture complex micro defects such as hidden cracks or sub-surface defects with high sensitivity. The nonlinear enhancement of luminous intensity is achieved under lower power excitation. By greatly improving the range of luminous brightness variation, the contrast between semiconductor surface and internal defects and background signals is increased several times, making the defects appear more intuitive and clear;

[0046] 6. Simple operation and strong scalability: The device is modular and intelligent in design. It can be quickly deployed in a standard industrial environment by integrating high-efficiency light sources, displacement platforms, nonlinear fluorescent materials and control systems. The system has a simple structure and is easy to maintain. It also has the ability to flexibly adapt to a variety of semiconductor specifications. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Schematic diagram of the optical path of the transmission-type flaw detection mode device in Examples 1, 4 and 5 of the present invention.

[0048] Figure 2 Schematic diagram of the optical path of the device based on the reflection type flaw detection mode in Example 2 of the present invention.

[0049] Figure 3 The following is a design flow and logic diagram of the software in the intelligent control module in embodiments 1, 2, 4 and 5 of the present invention.

[0050] Figure 4Schematic diagram of the fluorescent film described in Examples 1, 2, 4 and 5 of the present invention.

[0051] Figure 5 This is a typical up-conversion fluorescent nanoparticle NaErF according to Example 3 of the present invention. 4 Schematic diagram of the energy transfer process of luminescence energy levels in the system.

[0052] Figure 6 The NaErF synthesized in Examples 3, 4 and 5 of the present invention 4 Electron microscope image of the system. Where a is NaErF 4 Transmission electron micrograph of b; b is NaErF 4 @NaYF 4 Transmission electron microscopy image.

[0053] Figure 7 The NaErF synthesized in Example 3 of the present invention 4 @NaYF 4 The visible light spectrum and 655 nm emission peak power curve and fitting results. Where a is the visible light spectrum; b is the 655 nm emission peak power curve and fitting results.

[0054] Figure 8 Schematic diagram of the silicon wafer to be tested in Example 4 of the present invention. Wherein, a is the size layout of the USAF1951 resolution plate artificially etched on the silicon wafer; b is the actual picture of the USAF1951 resolution plate artificially etched on the silicon wafer.

[0055] Fig. 9 These are the silicon wafer flaw detection imaging images in Example 4 of the present invention. Among them, a is the camera imaging image without fluorescent film as the excitation target; b is the imaging image when the short-wave infrared laser penetrates the silicon wafer and excites the fluorescent film to improve the imaging contrast; c is the large-area flaw detection image after the short-wave infrared laser penetrates the silicon wafer to take several sub-images and stitch them together.

[0056] Fig.10 This is a chip flaw detection imaging diagram of an integrated circuit in Example 5 of the present invention, wherein the squares are marked with some defects in the processing process.

[0057] Figure 1 and 2 The marks are: 1. short-wave infrared laser; 2. first filter; 3. reflector; 4. three-dimensional electric translation stage; 5. semiconductor sample; 6. beam correction lens group; 7. nonlinear fluorescent film; 8. second filter; 9. imaging lens; 10. camera; 11. computer end; 12. FPGA chip; 13. beam splitter. DETAILED DESCRIPTION

[0058] The present invention will be further described below. It should be noted that the following examples are based on the technical solution and provide detailed implementation methods and specific operating processes, but the protection scope of the present invention is not limited to the following examples.

[0059] Example 1

[0060] like Figure 1 As shown, an embodiment of the present invention describes a low-cost, high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect. This embodiment uses transmission mode imaging, including: an excitation light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control and analysis module; under the excitation of short-wave infrared light, based on the overall coordination of the intelligent control module, the remaining modules are driven to complete visible light-near infrared detection, intelligent, high-resolution, low-cost, large-area fully automatic semiconductor flaw detection technology.

[0061] The excitation light generation module includes a laser 1, and a first filter 2 and a reflector 3 placed in sequence along the direction of the laser beam emitted by the laser. The laser 1 generates a steady-state short-wave infrared laser beam output; the first filter 2 is used to purify the laser; and the reflector 3 is used to adjust the angle of the light to maximize the space utilization of the detection device. The laser 1 can be a short-wave infrared laser.

[0062] The sample scanning module includes a three-dimensional electric translation stage 4 and a semiconductor sample 5. The three-dimensional electric translation stage 4 is used for spatial positioning of the semiconductor sample 5, and moves in the XYZ direction according to the set resolution to gradually scan the entire sample area; the semiconductor sample 5 is fixed on the three-dimensional electric translation stage 4, and can be penetrated by the short-wave infrared laser, and the transmitted light is used for subsequent excitation of the nonlinear fluorescent film 7 for imaging.

[0063] The nonlinear fluorescent film imaging module includes a beam correction lens group 6, a nonlinear fluorescent film 7, a second filter 8, an imaging lens 9 and a camera 10 coaxially placed along a semiconductor sample 5; the beam correction lens group 6 is used to match the imaging lens parameter design, correct optical aberrations, and improve beam utilization; the nonlinear fluorescent film 7 is uniformly mixed with fluorescent materials with nonlinear effects to produce emission fluorescence with orders of magnitude changes to improve imaging contrast and resolution; the second filter 8 is used to screen imaging bands to match different application scenarios; the imaging lens 9 is used to adjust the imaging aperture and field of view to ensure that the image is completely covered on the sensor to avoid image edge distortion or shearing; the camera 10 is used to record imaging pictures.

[0064] The intelligent control and analysis module includes a computer terminal 11 and an FPGA chip 12; the computer terminal 11 has built-in LabVIEW software to complete parameter setting, real-time data transmission, visual display, image stitching and defect analysis, and coordinate the various modules of the drive system; the FPGA chip 12 is connected to the laser 1 and the camera 10 to change the excitation and imaging mode, and communicates with the computer terminal 11 through the USB interface to send a pulse to trigger the camera 10, and cooperates with the three-dimensional electric translation stage 4 to achieve cyclic scanning to obtain a large-area, high-resolution stitched image;

[0065] The logic block diagram of the above device written based on FPGA-LabVIEW can be found in Figure 3 . A typical workflow is: start the LabVIEW program, specify the starting point, end point and step length of the 3D scanning area through the operation interface, establish the coordinate range, and the software calculates the scanning time to evaluate whether the running time is reasonable; the calibration status of the electric translation stage is initialized, and the program self-checks the accuracy of the communication between the high-speed camera and the FPGA to ensure the synchronization of the trigger signal; the program self-checks the storage path to ensure that the shooting data can be saved to the specified position in real time, initializes the stitching parameters, and performs pixel offset calibration / coordinate conversion matrix initialization. The functional area is divided into position control module, image acquisition module, data storage module and image stitching module. LabVIEW updates the scanning path diagram in real time, and the pulse signal generated by FPGA controls the high-speed camera to capture the image. FPGA records the position (encoder value) and timestamp at the same time, binds it to the image data, and feeds it back to LabVIEW; the program uses the XYZ position information to calculate the pixel offset, stitches the sub-image into a large image, and establishes a real-time preview window to ensure that there is no serious image omission in the current stitching. Move each step length one by one and repeat the shooting until all positions are scanned.

[0066] The embodiment of the present invention provides a low-cost, high-resolution semiconductor surface and internal flaw detection method based on multi-photon nonlinear fluorescence effect, comprising the following steps:

[0067] S1: Equipment initialization: Ensure that the laser 1 outputs stable short-wave infrared light, adjust the beam output power and collimation, purify it through the first filter 2, ensure that the beam intensity and radiation angle are adapted to the characteristics of the semiconductor sample 5 and the nonlinear fluorescent film 7; ensure that the three-dimensional electric translation stage 4 is connected to the computer terminal 11 normally, and the software initialization settings are successful; fix the semiconductor sample 5 to be tested and the nonlinear fluorescent film 7 uniformly mixed with nonlinear effect fluorescent materials for subsequent detection;

[0068] It should be noted that in step S1, the laser 1 adjusts the output power to penetrate the semiconductor sample 5 to achieve a specific light intensity, ensuring that the luminous response of the nonlinear fluorescent film 7 during the excitation process is within the nonlinear region, thereby effectively improving the image contrast and defect detection accuracy.

[0069] It should be noted that in step S1, the nonlinear fluorescent film 7 has a uniform thickness and the fluorescent material is evenly distributed: the surface ligands of the modified fluorescent material are coated with a hydrophilic surface layer to ensure the dispersion of the fluorescent material and the compatibility with the membrane matrix solution. The fluorescent material is added to the membrane matrix solution, and the solution is fully stirred using a magnetic stirrer or a high-speed vortex oscillator to ensure that the fluorescent material is evenly dispersed, and the time and power of the ultrasonic treatment are appropriately increased to increase the dispersibility.

[0070] It should be noted that in step S1, the nonlinear fluorescent film 7 is prepared by a scraping method, where a film matrix solution containing fluorescent materials is coated on a glass substrate, and a fluorescent film with uniform thickness and uniform distribution of fluorescent materials is obtained by a natural forming method.

[0071] It should be noted that, in step S1, the fluorescent material in the nonlinear fluorescent film 7 is conventional up-conversion fluorescent nanoparticles, and its nonlinearity is generally less than 6.

[0072] S2: Thin film excitation and imaging detection: After being scattered by defects, the transmission power of short-wave infrared light penetrating the semiconductor sample 5 changes. Specifically, in the transmission flaw detection mode, the laser that penetrates the semiconductor sample 5 is modulated by the correction lens group 6 and irradiated onto the nonlinear fluorescent film 7, generating a visible light-near infrared signal with an emission intensity that varies in magnitude; The FPGA chip 12 triggers the camera 10 to capture images and records the first frame of image data and related position information; The LabVIEW program controls the three-dimensional electric translation stage 4 through instructions to accurately move the semiconductor sample 5 to the preset starting position to ensure that every area on its surface is excited and the image is captured. At each displacement stop point, the excitation-collection process is repeated until the entire sample surface is scanned. After each scan, a frame of image will be captured for subsequent stitching and analysis;

[0073] It should be noted that in step S2, the FPGA chip 12 realizes the generation, triggering gating and processing functions of the pulse signal; the LabVIEW realizes parameter setting, real-time data transmission and visual display; the detection process realizes the fully automatic operation from the positioning of the three-dimensional electric translation stage 4 to the output of the excitation light, the camera image acquisition and the data analysis through the LabVIEW control system and the automation program, thereby improving the efficiency, accuracy and automation of the entire flaw detection process.

[0074] It should be noted that in step S2, the three-dimensional electric translation stage 4 is used for spatial positioning of the sample. With the help of the high-precision pulses generated inside the FPGA chip 12, the XYZ axis stepping is controlled by setting the stepping interval and frequency, so as to realize the stepping signal control of the three-dimensional electric translation stage 4 and the timing management of fluorescence signal acquisition, thereby realizing rapid detection of the entire sample area.

[0075] S3: Image stitching: After the three-dimensional electric translation stage 4 completes the image scanning process, the FPGA chip 12 stores multiple sub-images with specific position information. Based on the unique position information and precise image stitching algorithm, LabVIEW stitches multiple sub-images in sequence into a complete semiconductor flaw detection image. During the stitching process, any image overlap areas caused by displacement errors will be identified and processed to ensure the continuity and seamless transition of the stitched image, and finally generate a large-area flaw detection image with high resolution and precise stitching;

[0076] It should be noted that in step S3, the image stitching process uses a stitching algorithm based on the coordinate data of the three-dimensional electric translation stage 4 to spatially locate the image captured in each frame and stitch them into a complete flaw detection image. During the stitching process, the stitching error caused by inaccurate displacement is automatically eliminated to improve the image synthesis accuracy.

[0077] It should be noted that, in step S3, the stitching algorithm ensures accurate alignment between images by matching feature points in multiple sub-images, thereby generating a seamless large-area image.

[0078] It should be noted that in step S3, the stitching algorithm includes a series of sub-algorithms to perform stitching error correction, including automatic calibration and alignment adjustment, to ensure error-free and high-quality stitching results.

[0079] S4: Defect analysis: Grayscale distribution analysis is performed based on the stitched large-area images to detect areas in the image where the grayscale value changes suddenly or deviates from the normal range; bubbles, scratches, cracks and other defects on the surface and inside of the semiconductor are automatically identified through image processing methods; machine vision algorithms (such as CNN, SVM) classify and count the detected abnormal areas, and calculate the type, distribution and size of the defects. This information will serve as the basis for further analysis and evaluation of semiconductor quality. The system can quantify the results according to the set parameters, such as the minimum size of the defect, the number of defects allowed, etc., automatically generate reports, and provide intelligent quality assessment;

[0080] It should be noted that in step S4, the defect analysis step automatically detects brightness discontinuities or mutation areas in the image through an analysis algorithm based on the grayscale distribution of the image, and classifies and automatically marks the defect types through a machine learning algorithm. The system is continuously trained and optimized through an artificial intelligence model to improve its recognition ability and speed for new types of defects, and maintain a high recognition rate under dynamically changing process conditions.

[0081] It should be noted that in step S4, the defect analysis process will provide a defect detection report in real time, and automatically classify and count different types of defects on the surface or inside of the semiconductor to facilitate further analysis and quality assessment.

[0082] Example 2

[0083] like Figure 2 As shown, an embodiment of the present invention describes a low-cost, high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect. This embodiment uses reflective mode imaging, including: an excitation light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control and analysis module; under the excitation of short-wave infrared light, based on the overall coordination of the intelligent control module, the remaining modules are driven to complete visible light-near infrared detection, intelligent, high-resolution, low-cost, large-area fully automatic semiconductor flaw detection technology.

[0084] The excitation light generation module includes a laser 1, and a first filter 2 and a reflector 3 placed in sequence along the direction of the laser beam emitted by the laser. The laser 1 generates a steady-state short-wave infrared laser beam output; the first filter 2 is used to purify the laser; and the reflector 3 is used to adjust the angle of the light to maximize the space utilization of the detection device. The laser 1 can be a short-wave infrared laser.

[0085] The sample scanning module includes a beam splitter 13, a three-dimensional electric translation stage 4 and a semiconductor sample 5; the beam splitter 13 is used to extract imaging reflected light for subsequent excitation of the nonlinear fluorescent film 7 for imaging; the three-dimensional electric translation stage 4 is used for spatial positioning of the semiconductor sample 5, moves in the XYZ direction according to the set resolution, and gradually scans the entire sample area; the semiconductor sample 5 is fixed on the three-dimensional electric translation stage 4 and can reflect short-wave infrared light.

[0086] The nonlinear fluorescent film imaging module includes a beam correction lens group 6, a nonlinear fluorescent film 7, a second filter 8, an imaging lens 9 and a camera 10; the beam correction lens group 6 is used to match the imaging lens parameter design, correct optical aberrations, and improve beam utilization; the nonlinear fluorescent film 7 evenly mixes fluorescent materials with nonlinear effects to produce emission fluorescence with orders of magnitude changes to improve imaging contrast and resolution; the second filter 8 is used to screen imaging bands to match different application scenarios; the imaging lens 9 is used to adjust the imaging aperture and field of view to ensure that the image is completely covered on the sensor to avoid image edge distortion or shearing; the camera 10 is used to record imaging pictures.

[0087] The intelligent control and analysis module includes a computer terminal 11 and an FPGA chip 12; the computer terminal 11 has built-in LabVIEW software to complete parameter setting, real-time data transmission, visual display, image stitching and defect analysis, and coordinate the various modules of the drive system; the FPGA chip 12 is connected to the laser 1 and the camera 10 to change the excitation and imaging mode, and communicates with the computer terminal 11 through the USB interface to send a pulse to trigger the camera 10, and cooperates with the three-dimensional electric translation stage 4 to achieve cyclic scanning to obtain large-area, high-resolution stitching images.

[0088] The logic block diagram of the above device written based on FPGA-LabVIEW can be found in Figure 3 . A typical workflow is: start the LabVIEW program, specify the starting point, end point and step length of the 3D scanning area through the operation interface, establish the coordinate range, and the software calculates the scanning time to evaluate whether the running time is reasonable; the calibration status of the electric translation stage is initialized, and the program self-checks the accuracy of the communication between the high-speed camera and the FPGA to ensure the synchronization of the trigger signal; the program self-checks the storage path to ensure that the shooting data can be saved to the specified position in real time, initializes the stitching parameters, and performs pixel offset calibration / coordinate conversion matrix initialization. The functional area is divided into position control module, image acquisition module, data storage module and image stitching module. LabVIEW updates the scanning path diagram in real time, and the pulse signal generated by FPGA controls the high-speed camera to capture the image. FPGA records the position (encoder value) and timestamp at the same time, binds it to the image data, and feeds it back to LabVIEW; the program uses the XYZ position information to calculate the pixel offset, stitches the sub-image into a large image, and establishes a real-time preview window to ensure that there is no serious image omission in the current stitching. Move each step length one by one and repeat the shooting until all positions are scanned.

[0089] The embodiment of the present invention provides a low-cost, high-resolution semiconductor surface and internal flaw detection method based on multi-photon nonlinear fluorescence effect, comprising the following steps:

[0090] S1: Equipment initialization: Ensure that the laser 1 outputs stable short-wave infrared light, adjust the beam output power and collimation, purify it through the first filter 2, ensure that the beam intensity and radiation angle are adapted to the characteristics of the semiconductor sample 5 and the nonlinear fluorescent film 7; ensure that the three-dimensional electric translation stage 4 is connected to the computer terminal 11 normally, and the software initialization settings are successful; fix the semiconductor sample 5 to be tested and the nonlinear fluorescent film 7 uniformly mixed with nonlinear effect fluorescent materials for subsequent detection;

[0091] It should be noted that in step S1, the laser 1 adjusts the output power so that the short-wave infrared light is reflected by the semiconductor sample 5 to reach a specific light intensity, ensuring that the luminous response of the nonlinear fluorescent film 7 during the excitation process is within the nonlinear region, thereby effectively improving the image contrast and defect detection accuracy.

[0092] It should be noted that in step S1, the nonlinear fluorescent film 7 has a uniform thickness and the fluorescent material is evenly distributed: the surface ligands of the modified fluorescent material are coated with a hydrophilic surface layer to ensure the dispersion of the fluorescent material and the compatibility with the membrane matrix solution. The fluorescent material is added to the membrane matrix solution, and the solution is fully stirred using a magnetic stirrer or a high-speed vortex oscillator to ensure that the fluorescent material is evenly dispersed, and the time and power of the ultrasonic treatment are appropriately increased to increase the dispersibility.

[0093] It should be noted that in step S1, the nonlinear fluorescent film 7 is prepared by a scraping method, where a film matrix solution containing fluorescent materials is coated on a glass substrate, and a fluorescent film with uniform thickness and uniform distribution of fluorescent materials is obtained by a natural forming method.

[0094] It should be noted that, in step S1, the fluorescent material in the nonlinear fluorescent film 7 is conventional up-conversion fluorescent nanoparticles, and its nonlinearity is generally less than 6.

[0095] S2: Thin film excitation and imaging detection: After being scattered by defects, the transmission power of short-wave infrared light penetrating the semiconductor sample 5 changes. Specifically, in the reflective flaw detection mode, the beam splitter 13 separates the reflected excitation light, which is modulated by the beam correction lens group 6 and irradiated to the nonlinear fluorescent film 7, generating a visible light-near infrared signal with an emission intensity that varies in magnitude; the FPGA chip 12 triggers the camera 10 to capture images and records the first frame of image data and related position information; the LabVIEW program controls the three-dimensional electric translation stage 4 through instructions to accurately move the semiconductor sample 5 to the preset starting position to ensure that every area on its surface is excited and the image is captured. At each displacement stop point, the excitation-collection process is repeated until the entire sample surface is scanned. After each scan, a frame of image will be captured for subsequent stitching and analysis;

[0096] It should be noted that in step S2, the FPGA chip 12 realizes the generation, triggering gating and processing functions of the pulse signal; the LabVIEW realizes parameter setting, real-time data transmission and visual display; the detection process realizes the fully automatic operation from the positioning of the three-dimensional electric translation stage 4 to the output of the excitation light, the camera image acquisition and the data analysis through the LabVIEW control system and the automation program, thereby improving the efficiency, accuracy and automation of the entire flaw detection process.

[0097] It should be noted that in step S2, the three-dimensional electric translation stage 4 is used for spatial positioning of the sample. With the help of the high-precision pulses generated inside the FPGA chip 12, the XYZ axis stepping is controlled by setting the stepping interval and frequency, so as to realize the stepping signal control of the three-dimensional electric translation stage 4 and the timing management of fluorescence signal acquisition, thereby realizing rapid detection of the entire sample area.

[0098] S3: Image stitching: After the three-dimensional electric translation stage 4 completes the image scanning process, the FPGA chip 12 stores multiple sub-images with specific position information. Based on the unique position information and precise image stitching algorithm, LabVIEW stitches multiple sub-images in sequence into a complete semiconductor flaw detection image. During the stitching process, any image overlap areas caused by displacement errors will be identified and processed to ensure the continuity and seamless transition of the stitched image, and finally generate a large-area flaw detection image with high resolution and precise stitching;

[0099] It should be noted that in step S3, the image stitching process uses a stitching algorithm based on the coordinate data of the three-dimensional electric translation stage 4 to spatially locate the image captured in each frame and stitch them into a complete flaw detection image. During the stitching process, the stitching error caused by inaccurate displacement is automatically eliminated to improve the image synthesis accuracy.

[0100] It should be noted that, in step S3, the stitching algorithm ensures accurate alignment between images by matching feature points in multiple sub-images, thereby generating a seamless large-area image.

[0101] It should be noted that in step S3, the stitching algorithm includes a series of sub-algorithms to perform stitching error correction, including automatic calibration and alignment adjustment, to ensure error-free and high-quality stitching results.

[0102] S4: Defect analysis: Grayscale distribution analysis is performed based on the stitched large-area images to detect areas in the image where the grayscale value changes suddenly or deviates from the normal range; bubbles, scratches, cracks and other defects on the surface and inside of the semiconductor are automatically identified through image processing methods; machine vision algorithms classify and count the detected abnormal areas, and calculate the type, distribution and size of the defects. This information will serve as the basis for further analysis and evaluation of semiconductor quality. The system can quantify the results according to the set parameters, such as the minimum size of the defect, the number of defects allowed, etc., automatically generate reports, and provide intelligent quality assessment;

[0103] It should be noted that in step S4, the defect analysis step automatically detects brightness discontinuities or mutation areas in the image through an analysis algorithm based on the grayscale distribution of the image, and classifies and automatically marks the defect types through a machine learning algorithm. The system is continuously trained and optimized through an artificial intelligence model to improve its recognition ability and speed for new types of defects, and maintain a high recognition rate under dynamically changing process conditions.

[0104] It should be noted that in step S4, the defect analysis process will provide a defect detection report in real time, and automatically classify and count different types of defects on the surface or inside of the semiconductor to facilitate further analysis and quality assessment.

[0105] Example 3

[0106] In the embodiment of the present invention, NaErF is prepared by coprecipitation method. 4 Upconversion fluorescent nanoparticles, and by coating NaYF 4 The shell further optimizes its luminescence properties.

[0107] NeA 4 Matrix utilization Er 3+ The multiphoton excitation characteristics of ions can generate a variety of visible light emission signals under 1550 nm excitation, including two-photon emission at 980 nm (near infrared) and 800 nm (near infrared) and three-photon emission at 655 nm (red light), 544 nm (green light) and 530 nm (green light). In order to reduce the luminescence quenching phenomenon, thick-shell NaErF 4 @NaYF 4 The luminescence efficiency of the sample was significantly improved by suppressing surface non-radiative relaxation.

[0108] Figure 5 The up-conversion fluorescent nanoparticles NaErF in the embodiment of the present invention are 4 Schematic diagram of the energy transfer process of the luminescence energy level in the system. Under 1550 nm laser excitation, Er 3+ The ions are initially excited by absorbing photons through ground state transitions. The energy is then transferred through lattice vibration multi-phonon processes or cross relaxation to accumulate the number of particles in other excited state energy levels, thereby generating 980 nm (near infrared) and 800 nm (near infrared) and three-photon emissions of 655 nm (red), 544 nm (green) and 530 nm (green). Thick-shelled NaYF 4 The layer effectively isolates surface defects and non-radiative channels, making Er 3+ The ions maintain efficient multiphoton luminescence.

[0109] Figure 6 The NaErF synthesized in the embodiment of the present invention 4 Electron microscope image of the system. Figure 6 Figure a in the figure is NaErF 4 The transmission electron microscopy image shows that the particle size of the nanoparticles is 19.1±1.2 nm, with good size uniformity; Figure 6 Figure b in the figure is NaErF 4 @NaYF 4 Transmission electron microscopy images show that after thick shell coating, the particle size increases to 38.6±1.5nm, indicating that the shell thickness is about 10 nm and the particles still maintain a regular spherical structure.

[0110] Figure 7The NaErF synthesized in the embodiment of the present invention 4 @NaYF 4 Visible light spectrum and 655 nm emission peak power curve and fitting results. Figure 7 a in the figure shows that the thick shell significantly suppresses the non-radiative relaxation channel, which significantly improves the luminescence intensity and efficiency; Figure 7 The results in b verify the optical nonlinear behavior of the sample, especially its excellent performance in 655 nm red light emission. The three-photon fitting nonlinearity reaches 2.8, close to the theoretical level of 3, showing good three-photon response capability.

[0111] Example 4

[0112] Polyvinyl alcohol (PVA) was used as the membrane matrix solution. NaErF 4 @NaYF 4 Nonlinear fluorescent nanoparticles are dispersed in a 25 wt% PVA solution and coated into a uniform film with a thickness of 20 μm and high optical transparency. Under the excitation of a 1550 nm short-wave infrared laser, the film exhibits significant three-photon red emission (655 nm) and achieves an order of magnitude change in brightness. Figure 4 It should be noted that similar effects can be achieved by using PS, PDMS, PAA, PMMA, etc. as membrane matrix solutions. Similar to PVA, their role is to provide a flexible transparent film as a substrate to disperse fluorescent nanoparticles with high-order nonlinearity.

[0113] The silicon wafer with the USAF1951 resolution plate etched manually was used as the test object. The size distribution and position of the etched USAF1951 resolution plate were recorded as follows: Figure 8 The actual picture of etching USAF1951 resolution plate is shown in a. Figure 8 As shown in b, the physical morphology of regular line pairs can be seen.

[0114] The etched silicon wafer is placed under the fluorescent film to form the structure to be tested. The transmission flaw detection mode device is used for demonstration. The silicon wafer allows the 1550nm laser light source to penetrate the silicon wafer and reach the fluorescent film. The fluorescent film is excited by the short-wave infrared laser, and the filter extracts the three-photon 655nm red light signal of the film. The significant contrast of the luminous intensity improves the contrast and imaging resolution of the detection. Fig. 9 a in the figure is the camera imaging image without fluorescent film as the excitation target; Fig. 9The image b in the figure shows that the short-wave infrared laser penetrates the silicon wafer and excites the fluorescent film to improve the imaging contrast. The resolution reaches 128 lp / mm, which can achieve an imaging resolution of ~4μm line width, which is 5 times higher than the 20μm resolution of traditional infrared imaging cameras. In addition, thanks to the high-order nonlinearity of upconversion nanoparticles, the imaging contrast reaches 5, which is better than traditional infrared camera imaging. Fig. 9 The image c in the figure is a large-area flaw detection image obtained by stitching together several sub-images taken by a short-wave infrared laser penetrating a silicon wafer, and the imaging field of view reaches 10mm×10mm.

[0115] Example 5

[0116] Polyvinyl alcohol (PVA) was used as the membrane matrix solution, and a certain amount of NaErF 4 @NaYF 4 Nonlinear fluorescent nanoparticles are uniformly mixed to form a fluorescent film with high optical transparency. Under the excitation of 1550 nm short-wave infrared laser, the film exhibits significant three-photon red light emission (655 nm) and achieves an order of magnitude change in light brightness. Figure 4 shown.

[0117] The chip etched with micro-integrated circuits is used as the test object. The chip to be tested is placed under the fluorescent film to form a test structure. The transmission flaw detection mode device is used for demonstration. The chip allows the 1550 nm laser light source to penetrate the chip and reach the fluorescent film. The fluorescent film is excited by the short-wave infrared laser, and the filter extracts the three-photon 655 nm red light signal of the film. The significant contrast of the luminous intensity improves the contrast and imaging resolution of the detection. Fig.10 This is an imaging diagram of chip flaw detection of an integrated circuit, in which the boxes indicate some defects in the processing process.

[0118] The present invention constructs an optical imaging device composed of an excitation light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control and analysis module, realizing short-wave infrared light excitation, visible light-near infrared detection, intelligent, high-resolution, low-cost, large-area fully automatic semiconductor flaw detection technology. The present invention is suitable for efficient flaw detection of large-scale semiconductors and has broad application prospects in the rapid location of semiconductor surface and internal defects. The method gets rid of the dependence on infrared imaging cameras, greatly reduces system costs, and provides a new way for online defect detection in semiconductor manufacturing processes.

[0119] For those skilled in the art, various corresponding changes and modifications can be made according to the above technical solutions and concepts, and all of these changes and modifications should be included in the protection scope of the claims of the present invention.

Claims

1. A high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect, characterized in that It includes an excitation light generation module, a sample scanning module, a nonlinear fluorescent film imaging module and an intelligent control and analysis module; the internal flaw detection device is divided into a transmission flaw detection mode and a reflection flaw detection mode; under the excitation of short-wave infrared light, based on the overall coordination of the intelligent control module, it drives the other modules to complete visible light-near infrared detection, intelligent, high-resolution, large-area fully automatic semiconductor flaw detection technology; The excitation light generation modules all include a continuous short-wave infrared laser, a pulsed short-wave infrared laser and a short-wave infrared LED light source, as well as a filter set and a reflector placed in sequence along the direction of the light beam emitted by the excitation light source; the excitation light source generates a steady-state short-wave infrared light beam output; The filter set is used to purify the light beam; the reflector is used to adjust the light angle to maximize the space utilization of the detection device; In the transmission flaw detection mode, the sample scanning module includes a three-dimensional electric translation stage and a semiconductor sample; the three-dimensional electric translation stage is used for spatial positioning of the semiconductor sample, moves in the XYZ direction according to the set resolution, and gradually scans the entire sample area; the semiconductor sample is fixed on the three-dimensional electric translation stage and can be penetrated by short-wave infrared light, and the transmitted light is used for subsequent excitation of nonlinear fluorescent film sample imaging; In the reflection flaw detection mode, the sample scanning module includes a beam splitter, a three-dimensional electric translation stage and a semiconductor sample; the beam splitter is used to extract imaging reflection light for subsequent excitation of nonlinear fluorescent film sample imaging; the three-dimensional electric translation stage is used for spatial positioning of the semiconductor sample, moves in the XYZ direction according to the set resolution, and gradually scans the entire sample area; the semiconductor sample is fixed on the three-dimensional electric translation stage and can reflect short-wave infrared light; The nonlinear fluorescent film imaging module includes a beam correction lens group, a nonlinear fluorescent film, a filter group, an imaging lens and a camera; The beam correction lens group is used to match the imaging lens parameter design, correct optical aberrations, and improve beam utilization; The nonlinear fluorescent film is uniformly mixed with luminescent materials having nonlinear fluorescent effects to generate emission fluorescence with orders of magnitude changes to improve imaging contrast and resolution; in the transmission flaw detection mode, the nonlinear fluorescent film is arranged above the sample to receive the transmitted light excitation; in the reflection flaw detection mode, the nonlinear fluorescent film is arranged behind the beam splitter to receive the reflected light excitation; The filter set is used to filter the imaging bands to match different application scenarios; the imaging lens is used to adjust the imaging aperture and field of view to ensure that the image is fully covered on the sensor and avoid image edge distortion or clipping; the camera is used to record the imaging pictures; The intelligent control and analysis module includes a computer and an FPGA chip; the computer has built-in LabVIEW software to complete parameter setting, real-time data transmission, visual display, image stitching and defect analysis, and coordinate the various modules of the drive system; the FPGA chip is connected to the excitation light source and the imaging camera to change the excitation and imaging modes, and communicates with the computer through a USB or Ethernet interface to send pulses to trigger the camera, and cooperates with the three-dimensional electric translation stage to achieve cyclic scanning to obtain large-area, high-resolution stitched images.

2. A high-resolution semiconductor surface and internal flaw detection device based on multiphoton nonlinear fluorescence effect as claimed in claim 1, characterized in that The filter set of the excitation light generating module is assembled by a rotatable bracket, which is used to adjust and match different excitation light wavelengths; the filter set of the nonlinear fluorescent film imaging module can be assembled by a rotatable bracket, which is used to adjust and match different collected fluorescence wavelengths; the working wavelength of the excitation light source of the excitation light generating module is mainly located in the range of 1100 to 3000 nm where the semiconductor's absorption rate of light is significantly reduced, and the working wavelength also matches the absorption region of the nonlinear fluorescent material; the excitation light source has a built-in automatic light intensity adjustment device, which can dynamically adjust the light intensity to adapt to the detection requirements of semiconductors of different thicknesses.

3. A high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect as claimed in claim 1, characterized in that The semiconductor sample has defects caused by the manufacturing process, including surface defects, internal defects, and edge defects. These defects cause a scattering effect on the incident short-wave infrared light, thereby reducing its transmittance / reflectivity. The modulated excitation light will excite the nonlinear fluorescent film to achieve imaging of the defects.

4. A high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect as claimed in claim 1, characterized in that The fluorescent film of the nonlinear fluorescent film imaging module uses a transparent polymer material as a film substrate, has good transparency and mechanical properties, and can be uniformly mixed with a fluorescent material with a nonlinear effect to achieve a response to short-wave infrared light; the materials used include polystyrene, polydimethylsiloxane, polyvinyl alcohol, polyacrylic acid, polymethyl methacrylate, polyparaxylene, siloxane or acrylic resin polymer; The fluorescent material has good optical nonlinear response and can be efficiently excited by short-wave infrared light; the fluorescent material includes common organic dyes, quantum dots and lanthanide element-doped up-conversion fluorescent materials; The excitation process of upconversion fluorescent materials includes continuous excited state absorption upconversion process, energy transfer upconversion process and photon avalanche upconversion process; The lanthanide elements include: neodymium that can be excited by 1350-1400 nm and 1750-1850 nm, erbium that can be excited by 1450-1650 nm, holmium that can be excited by 1500-2000 nm, thulium that can be excited by 1200-1500 nm and 1650-1950 nm, and dysprosium that can be excited by 1200-1300 nm; in addition, by designing the core-shell structure and modulation of the excitation mode, lanthanide elements such as praseodymium, promethium, samarium, europium, and terbium that match short-wave infrared excitation are also suitable for the selection range of the above-mentioned fluorescent materials; The lanthanide-doped fluorescent material is doped with a single lanthanide luminescent element, or co-doped with multiple lanthanide elements; the nonlinear fluorescence energy transfer effect of multiple lanthanide doping will amplify the final fluorescence nonlinearity, further improving the semiconductor surface and internal flaw detection performance.

5. A high-resolution semiconductor surface and internal flaw detection device based on multi-photon nonlinear fluorescence effect as claimed in claim 1, characterized in that The FPGA chip is replaced by a microcontroller and a single-chip microcomputer control system to realize the conversion and control of the trigger signal; the LabVIEW is replaced by Python, Matlab, and LabWindows / CVI development environment to realize the UI interface design and coordinate the various modules of the control system.

6. A high-resolution semiconductor surface and internal flaw detection method based on multi-photon nonlinear fluorescence effect, characterized in that The following steps are involved: S1: Equipment initialization: Ensure that the excitation light source module outputs stable short-wave infrared light, adjust the beam output power and collimation, purify through filters, ensure that the beam intensity and radiation angle are adapted to the characteristics of semiconductor samples and nonlinear fluorescent films; ensure that the three-dimensional electric translation stage is connected to the computer through a suitable communication protocol, and the software initialization settings are successful; fix the semiconductor sample to be tested and the fluorescent film uniformly mixed with nonlinear effect fluorescent materials for subsequent detection; S2: Thin film excitation and imaging detection: After being scattered by defects, the transmission / reflection power of short-wave infrared light penetrating the semiconductor sample changes; specifically, in the transmission flaw detection mode, the excitation light that passes through the semiconductor is modulated by the correction lens group and irradiated to the fluorescent film, generating a visible light-near infrared signal with an emission intensity that varies in magnitude; in the reflection flaw detection mode, the beam splitter separates the reflected excitation light, which is modulated by the correction lens group and irradiated to the fluorescent film, also generating a visible light-near infrared signal with an emission intensity that varies in magnitude; the FPGA chip triggers the camera to capture images and records the first frame of image data and related position information; The LabVIEW program controls the 3D electric translation stage through instructions to accurately move the semiconductor sample to the preset starting position to ensure that every area on its surface is excited and the image is collected; at each displacement stop point, the excitation-collection process is repeated until the entire sample surface is scanned; after each scan, a frame of image is collected for subsequent stitching and analysis; S3: Image stitching: After the 3D electric translation stage completes the image scanning process, the FPGA chip stores multiple sub-images with specific position information; based on the unique position information and precise image stitching algorithm, LabVIEW stitches multiple sub-images in sequence into a complete semiconductor flaw detection image; during the stitching process, any image overlap areas caused by displacement errors will be identified and processed to ensure the continuity and seamless transition of the stitched image, and ultimately generate a large-area flaw detection image with high resolution and precise stitching; S4: Defect analysis: Grayscale distribution analysis is performed based on the stitched large-area images to detect areas in the image where the grayscale value changes suddenly or deviates from the normal range; bubbles, scratches, and cracks on the surface and inside of the semiconductor are automatically identified through image processing methods; machine vision algorithms classify and count the detected abnormal areas, and calculate the type, distribution, and size of the defects; this information will serve as the basis for further analysis and evaluation of semiconductor quality; the system quantifies the results according to the set parameters, such as the minimum size of the defect and the allowable number of defects, automatically generates a report, and provides intelligent quality assessment.

7. A high-resolution semiconductor surface and internal flaw detection method based on multiphoton nonlinear fluorescence effect as claimed in claim 6, characterized in that In step S1, the excitation light generation module adjusts the output power to achieve a specific light intensity through the semiconductor, ensuring that the luminescent response of the fluorescent material during the excitation process is within the nonlinear region, thereby effectively improving the image contrast and defect detection accuracy; In step S1, the fluorescent film has a uniform thickness and the fluorescent material is evenly distributed; Modify the surface ligands of the fluorescent material to ensure the dispersion of the fluorescent material and its compatibility with the membrane matrix solution; Add the fluorescent material to the membrane matrix solution, use magnetic stirring or high-speed vortex oscillator to fully stir the solution to ensure that the fluorescent material is evenly dispersed, and appropriately increase the time and power of ultrasonic treatment to increase dispersibility; In step S1, the fluorescent film is prepared by spin coating, scraping coating, or flow coating to coat a film matrix solution containing fluorescent materials on a glass substrate; a fluorescent film with uniform thickness and uniform distribution of fluorescent materials is obtained by natural molding or heating curing; In step S1, the fluorescent materials include quantum dots, organic dyes, traditional upconversion fluorescent nanoparticles, and photon avalanche effect upconversion nanoparticles; the nonlinearity of quantum dots and organic dyes is generally less than 3, the nonlinearity of traditional upconversion fluorescent nanoparticles is generally less than 6, and the nonlinearity of photon avalanche effect rare earth doped upconversion nanoparticles is generally greater than 10.

8. A high-resolution semiconductor surface and internal flaw detection method based on multiphoton nonlinear fluorescence effect as claimed in claim 6, characterized in that In step S2, the FPGA chip realizes the generation, triggering gating and processing functions of pulse signals; the LabVIEW realizes parameter setting, real-time data transmission and visual display; the detection process realizes full-automatic operation from three-dimensional electric translation stage positioning to excitation light output, camera image acquisition and data analysis through the LabVIEW control system and automation program, thereby improving the efficiency, accuracy and automation of the entire flaw detection process; In step S2, the three-dimensional electric translation stage is used for spatial positioning of the sample. With the help of high-precision pulses generated inside the FPGA, the XYZ axis stepping is controlled by setting the stepping interval and frequency, thereby realizing the stepping signal control of the three-dimensional electric translation stage and the timing management of image acquisition, and realizing rapid detection of the entire sample area.

9. A high-resolution semiconductor surface and internal flaw detection method based on multiphoton nonlinear fluorescence effect as claimed in claim 6, characterized in that In step S3, the image stitching process spatially locates each frame of the acquired image through a stitching algorithm based on the coordinate data of the three-dimensional electric translation stage, and stitches them into a complete flaw detection image. During the stitching process, the stitching error caused by inaccurate displacement is automatically eliminated to improve the image synthesis accuracy; In step S3, the image stitching algorithm ensures accurate alignment between images by matching feature points in multiple sub-images, thereby generating a seamless large-area image; In step S3, the image stitching algorithm includes a series of sub-algorithms to perform stitching error correction, including automatic calibration and alignment adjustment, to ensure error-free and high-quality stitching results.

10. A high-resolution semiconductor surface and internal flaw detection method based on multiphoton nonlinear fluorescence effect as claimed in claim 6, characterized in that In step S4, the defect analysis step automatically detects brightness discontinuities or mutation areas in the image through an analysis algorithm based on the grayscale distribution of the image, and classifies and automatically marks the defect types through a machine learning algorithm, wherein the system is continuously trained and optimized through an artificial intelligence model to improve its recognition capability and recognition speed for new types of defects, and maintain a high recognition rate under dynamically changing process conditions; In step S4, the defect analysis process will provide a defect detection report in real time, and automatically classify and count different types of defects on the surface or inside of the semiconductor to facilitate further analysis and quality assessment.

Images