Wafer detection device and wafer optical detection method based on RCWA improved algorithm

Abstract

The invention provides a wafer detection device and a wafer optical detection method based on an RCWA improved algorithm, and belongs to the field of wafer quantity detection.The wafer detection device comprises an illumination system, a wafer moving table, a microscopic imaging system, an image processor and an imaging simulation system; the wafer optical detection method comprises the steps of wafer image acquisition, image preprocessing and acquisition of a to-be-detected image, light source modeling, material three-dimensional structure and optical parameter definition, RCWA calculation and near-field electric field distribution acquisition, far-field imaging calculation and reference image generation, image pair processing, difference image acquisition, threshold algorithm and convolution kernel processing, and the like. Acquiring defect position and signal intensity information, post-processing and generating a detection report. According to the RCWA improved algorithm based on the CPU coordinated allocation GPU computing cluster operation matrix index calculation, the overall wafer detection efficiency is improved, and the RCWA improved algorithm can be popularized and applied in the fields related to layer structure electromagnetic field calculation, such as wafer detection and photomask detection.

Description

Technical Field

[0001] This invention belongs to the field of wafer optical inspection with improved field computing speed, and specifically relates to a wafer inspection device and a wafer optical inspection method based on an improved RCWA algorithm. Background Technology

[0002] Semiconductor defect detection requires magnifying and imaging the image of deep subwavelength microstructures into a detector to detect physical defects such as surface defects and scratches on wafers. Due to the proximity effect of deep subwavelength microstructures, and because the defects are smaller than the diffraction limit, the detected abnormal signals cannot be easily correlated with the reference signals.

[0003] Therefore, it is necessary to combine simulation calculations with comparative analysis of acquired wafer images to determine defect information. Rigorous Coupled-Wave Analysis (RCWA) is mainly used for numerical simulation calculations of periodic dielectric light scattering, playing a crucial role in scenarios such as optical defect detection, critical dimension measurement, and lithography process simulation optimization in semiconductor manufacturing. The basic principle of RCWA is to expand the periodic dielectric constant distribution and electromagnetic field into spatial Fourier series, establish coupled-wave equations using Maxwell's equations, calculate the characteristic states of the electric and magnetic fields, then calculate the scattering matrix of a single layer, and finally use Redheffer star product operations to obtain the global scattering matrix. The most time-consuming part of the traditional RCWA method is solving large-scale eigenvalue problems. The eigenvalue solving step accounts for more than half of the entire RCWA simulation computation time, becoming a bottleneck restricting simulation efficiency. This also limits the efficiency of quantity detection or lithography processing in specific application scenarios. The parallel computing architecture of modern GPUs provides revolutionary acceleration capabilities for large-scale matrix operations in RCWA. In contrast, the serial architecture of CPUs is often limited by memory bandwidth bottlenecks when handling such problems. Therefore, in practical engineering applications, the direct calculation of eigenvalues ​​and eigenmatrices using GPUs should be avoided as much as possible.

[0004] In summary, a new computational simulation scheme needs to be designed for optical inspection of wafer defects in order to improve the overall inspection efficiency. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, the purpose of this invention is to provide a wafer inspection device and a wafer optical inspection method based on the improved RCWA algorithm, which can solve the above-mentioned problems.

[0006] A wafer inspection device based on an improved RCWA algorithm includes an illumination system, a wafer stage, a microscopic imaging system, an image processor, and an imaging simulation system. The microscopic imaging system is positioned towards the wafer stage to acquire raw wafer images. The imaging simulation system calculates the near-field electric field distribution on the wafer surface based on the improved RCWA algorithm and calculates remote imaging using a simulation model of the imaging system to produce a reference image. The image processor is used for preprocessing the raw wafer images to obtain the image to be inspected, and compares the image to be inspected with the reference image to obtain and output defect detection information.

[0007] Furthermore, the imaging simulation system's simulation model includes a light source model and a wafer 3D structure model. The light source model simulates illumination light of different types and parameters. The wafer 3D structure model includes a near-field optical field sub-model, a far-field optical field sub-model, and an imaging surface sub-model. The near-field optical field sub-model is used to solve the near-field electromagnetic scattering characteristics of the wafer surface, the far-field optical field sub-model is used for far-field calculation of the far-field imaging process of the optical system, and the imaging surface sub-model is used to calculate the simulation results of the simulated optical image and intermediate process simulation information.

[0008] Furthermore, the hardware of the imaging simulation system adopts a distributed cluster architecture GPU computing cluster module. The distributed cluster architecture consists of a master control node and multiple computing nodes working together, and using the RCWA improved algorithm to perform parallel computing to obtain simulated optical images of single-layer or multi-layer structures on the wafer surface.

[0009] Furthermore, the improved RCWA algorithm includes: S1, calculating the matrix exponent E of the transfer matrix T using the fast matrix exponent approximation method; S2, calculating the transfer matrix T after stacking adjacent vacuum dielectric layers; S3, calculating the scattering matrix S corresponding to the transfer matrix T using the layer scattering matrix transformation formula; S4, traversing all layers of the multilayer structure and calculating the scattering matrix of each layer. S5. Using Redheffer star product operation to obtain the global scattering matrix, j=1,2,…,L; .

[0010] The present invention also provides a wafer optical inspection method based on the aforementioned wafer inspection device. The wafer optical inspection method includes: wafer image acquisition, image preprocessing and obtaining the image to be inspected, light source modeling, material three-dimensional structure and optical parameter definition, RCWA calculation and near-field electric field distribution acquisition, far-field imaging calculation and generation of reference image, image pair processing, difference map acquisition, threshold algorithm and convolution kernel processing, defect location and signal intensity information acquisition, post-processing and inspection report generation.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows: The apparatus and method provided in this application, based on the improved RCWA algorithm for matrix exponential calculation of the GPU computing cluster module with CPU coordination and allocation, realize the speed improvement of multi-layer structure field calculation, improve the overall detection efficiency, and facilitate its application in fields involving layer structure electromagnetic field calculation such as wafer inspection and photomask inspection. Attached Figure Description

[0012] Figure 1 and Figure 2 Schematic diagrams of different embodiments of the wafer inspection device; Figure 3 This is a schematic diagram of optical inspection of a periodic structure using a wafer inspection device. Figure 4 A schematic diagram of a multilayer structure on a patterned wafer surface; Figure 5 This is a schematic diagram of the system principle of a wafer inspection device; Figure 6 This is a schematic diagram of a GPU computing cluster module; Figure 7 This is a schematic diagram of the electromagnetic field during the imaging process on the wafer surface. Figure 8 Schematic diagrams of different types of light sources; Figure 9 This is a schematic diagram of the wafer optical inspection method.

[0013] 10. Lighting system; 20. Wafer motion stage; 30. Microscopic imaging system; 31. Wafer; 40. Image Processor; 50. Imaging simulation system. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] A wafer inspection device based on an improved RCWA algorithm, see [link / reference]. Figures 1-8 The wafer inspection device includes an illumination system 10, a wafer motion stage 20, a microscopic imaging system 30, an image processor 40, and an imaging simulation system 50.

[0016] The illumination system 10 provides illumination to the wafer stage 20. The microscopic imaging system 30 is positioned towards the wafer stage 20 to acquire images of the target area on the stage, obtaining raw wafer images. The wafer stage 20 drives the wafer space on the wafer carrier to move according to a set path. The imaging simulation system 50 calculates the near-field electric field distribution on the wafer surface based on an improved RCWA algorithm and calculates remote imaging using the imaging system simulation model, thereby producing a reference image. The image processor 40 preprocesses the raw wafer images to obtain the image to be inspected, compares the image to be inspected with the reference image, and obtains and outputs defect detection information.

[0017] Specifically, the lighting system 10 is used to provide multiple types of switchable illumination light to the wafer motion stage 20, see [link to documentation]. Figure 8 The lighting sources include circular illumination, annular illumination, dipole illumination, and quadar illumination.

[0018] The imaging simulation system 50 includes a light source model and a wafer 3D structure model. The light source model simulates illumination light of different types and parameters. The wafer 3D structure model includes a near-field light field sub-model, a far-field light field sub-model, and an imaging surface sub-model. The near-field light field sub-model is used to solve the near-field electromagnetic scattering characteristics of the wafer surface. The far-field light field sub-model is used for far-field calculation of the far-field imaging process of the optical system. The imaging surface sub-model is used to calculate the simulation results of the simulated optical image and intermediate process simulation information.

[0019] See Figure 7 This diagram illustrates the electromagnetic field process of imaging a multi-layered target (e.g., a wafer) surface, primarily comprising illumination, scattering, collection, refocusing, and image plane. The optical components at each stage (such as condenser lenses, collecting lenses, objective lenses, and focusing lenses) must work precisely together, using optical principles like refraction and reflection to control the light path, ultimately projecting the object's optical characteristics (such as structure, texture, and color) onto the image plane.

[0020] The illumination segment includes the light field from the light source through the condenser pupil of the illumination lens group to the surface of the target.

[0021] The scattering segment, or near-field segment, mainly refers to the electromagnetic field corresponding to the light field on the target surface, and its electromagnetic scattering matrix needs to be calculated.

[0022] The collection section consists of light reflected from the target surface (mainly from the end of the scattering section) passing through the entrance pupil to the objective lens.

[0023] The refocusing section, from the objective lens through the exit pupil to the image plane, is where the imaging light field converges.

[0024] The image plane is where the far-field light field is finally imaged at the end of the collection segment. Examples of image planes include CCD cameras, CMOS cameras, and TDI (Transmission Display Interface).

[0025] Near-field calculations, specifically the calculation of the near-field electromagnetic scattering matrix of multilayer structures based on an improved RCWA algorithm. Specifically, this mainly involves scattering; the RCWA module uses an improved RCWA algorithm based on matrix exponentiation to solve for the near-field electromagnetic scattering matrix of multilayer structures.

[0026] The light source parameters of the light source model include wavelength, shape, and polarization. The wafer parameters of the wafer 3D structure model include thickness, shape, refractive index of each layer, and extinction coefficient.

[0027] The imaging simulation system 50 uses a distributed cluster architecture GPU computing cluster module as its hardware. See [link / reference]. Figure 6 The distributed cluster architecture consists of a master node and multiple computing nodes working together, and uses the RCWA improved algorithm to perform parallel computing to obtain simulated optical images of single-layer or multi-layer structures on the wafer surface.

[0028] An improved RCWA algorithm based on matrix exponent calculation calculates the received information, solves for near-field electromagnetic scattering characteristics, simulates far-field imaging process, and outputs simulated optical images; these calculations run on a GPU computing cluster module.

[0029] The master node and compute nodes of the GPU computing cluster module are interconnected via IB network cards. The master node coordinates task allocation based on the CPU. The compute nodes transmit data with each GPU in the node through the PCIe high-speed bus. The GPU array realizes multi-card collaborative computing based on the high-speed chip interconnect protocol. The computing results are finally transmitted from the corresponding compute nodes to the master node for aggregation and output.

[0030] The improved RCWA algorithm includes the following steps.

[0031] S1. Calculate the matrix exponent E of the transfer matrix T using the fast matrix exponent approximation method. In the formula, This represents the layer thickness of the target structure to be simulated. Let P represent the wavelength of the incident light, and let Q be the electromagnetic field transformation matrix.

[0032] S2. Calculate the transfer matrix T after stacking vacuum dielectric layers on adjacent sides.

[0033] .

[0034] In the formula, It represents the characteristic state of the electric field in free space, such as the vacuum layer. It refers to the characteristic magnetic field state of free space such as the vacuum layer. Represents a unit matrix.

[0035] here It can be represented as: Where x, y, and z represent the spatial coordinates of the magnetic field on the wafer surface. It is not difficult to verify the relation. .

[0036] S3. Using the layer scattering matrix transformation formula, calculate the scattering matrix S corresponding to the transmission matrix T.

[0037] S4. Traverse all layers of the multilayer structure and calculate the scattering matrix of each layer. , j=1,2,…,L.

[0038] S5. Obtain the global scattering matrix by using Redheffer star product operation. .

[0039] The fast matrix exponent approximation method in S1 includes: approximating the matrix exponent by summing an infinite series as follows: The definition of matrix R is... d represents the layer thickness of the optical device, and λ represents the wavelength of the incident light; first, calculate the reduced matrix exponent. The original matrix exponent is then recovered by repeated squaring operations, i.e. Where the integer m takes the following values: That is, the choice of integer m is based on the 1-norm of matrices P and Q.

[0040] GPU computing employs matrix block operations to calculate twice the total harmonics of the previous version, thereby improving storage space utilization.

[0041] For step S3, the transfer matrix T and the scattering matrix S are used together. The transfer matrix T is responsible for handling single-layer propagation, and the scattering matrix S is used for inter-layer coupling, so as to balance computational efficiency and numerical robustness.

[0042] The imaging system simulation model calculates the far-field imaging process, including Collection simulation and Refocus simulation. The Collection simulation characterizes the propagation from the near field to the entrance pupil, using Fraunhofer diffraction. In the formula, x and y are the coordinates on the objective lens, and f and g are the spatial frequencies of the light source's pupil. This represents the spatial frequency at the imaging pupil.

[0043] Refocus calculations simulate the propagation of the objective lens from the exit pupil to the imaging plane, based on the pupil function and lens system aberrations: The corresponding exit pupil electric field is: .

[0044] Finally, the imaging system simulation model was used to perform optical simulations based on the Abbe method to obtain imaging simulation results. .

[0045] .

[0046] In the formula, Fourier transform of the cross-correlation function of the light source The frequency of the light source's pupil. The Fourier transform of the transfer function of the imaging system For the imaging pupil frequency, for Conjugate, for Conjugate The Fourier transform of the transfer function in the frequency domain of the target structure.

[0047] In summary, the detection scheme integrates an imaging simulation system based on an improved RCWA algorithm in image processing. Defect localization and evaluation require accurate simulation calculation of the detection signals corresponding to different defect types. RCWA is used to efficiently calculate the near-field signals corresponding to different defects in the wafer, and the far-field detection signals are obtained through near-field-far-field transformation. The scheme transforms the computational focus of RCWA into a matrix exponential calculation problem, fully leveraging the advantages of GPU parallel computing combined with the coordinated allocation of CPU resources, significantly improving computational efficiency.

[0048] Wafer optical inspection methods A wafer optical inspection method based on the aforementioned wafer inspection device, see [link to relevant documentation]. Figure 9 The process flow of the wafer optical inspection method is as follows.

[0049] Wafer image acquisition: The wafer inspection device acquires wafer images to obtain raw image data. Specifically, using high-resolution imaging equipment—in this case, the aforementioned wafer inspection device—images are acquired from the wafer to obtain raw image data. These images will serve as the basis for subsequent processing and analysis.

[0050] Image preprocessing is performed to obtain the image to be inspected. Preprocessing of the wafer image includes noise reduction and background elimination to obtain the image to be inspected for subsequent comparative analysis and defect identification. Specifically, the acquired wafer image is preprocessed, including but not limited to noise reduction and background elimination. The purpose of preprocessing is to remove noise and interference information from the image and improve image quality. After preprocessing, a clear and high-quality image to be inspected is obtained. These images will be used for subsequent comparative analysis and defect identification.

[0051] Light source modeling involves constructing a model of the light source with parameters including wavelength, shape, and polarization. See also... Figure 9 Meanwhile, in another path, the light source is modeled, including parameters such as wavelength, shape, and polarization.

[0052] The material's three-dimensional structure and optical parameters are defined, including the three-dimensional structure of the wafer to be inspected and the optical parameters of each layer of the wafer. This involves defining the three-dimensional structure of the wafer to be inspected, including its thickness and shape, and determining the optical parameters of each layer of the wafer (such as refractive index and extinction coefficient).

[0053] RCWA calculation and near-field electric field distribution acquisition: Based on the defined wafer three-dimensional structure and optical parameters, the RCWA algorithm is used to calculate the near-field electric field distribution near the wafer surface.

[0054] Far-field imaging calculation and reference image generation utilize near-field electric field distribution combined with imaging system modeling to perform far-field imaging calculations, thereby generating an ideal, defect-free reference image for subsequent comparison. Specifically, using the near-field electric field distribution obtained in the previous step, combined with imaging system modeling (including parameters such as numerical aperture and aberrations), far-field imaging calculations are performed to generate the reference image. The reference image represents an ideal, defect-free wafer image, used for subsequent defect detection and comparison.

[0055] Image pairing processing involves registering the image to be detected with a reference image and performing brightness compensation to ensure consistency in position and brightness between the two images, facilitating subsequent difference analysis.

[0056] Difference map acquisition involves performing difference processing on the image to be detected and the reference image to obtain a difference map. The difference map highlights the regions of difference between the image to be detected and the reference image, which may contain potential defect information.

[0057] Thresholding algorithms and convolution kernel processing are applied to the difference map to further extract and enhance defect features. Thresholding algorithms help distinguish between signals and noise, improving the accuracy and reliability of defect detection.

[0058] Defect location and signal strength information acquisition: Through the above processing, defect feature information including defect location and signal strength is obtained.

[0059] Post-processing involves performing operations such as classification, sampling, and clustering on the detection results to further refine and organize the defect information.

[0060] The inspection report is generated based on the processed defect information. The report includes key information such as the location, type, quantity, and severity of the defects, providing a scientific basis for quality control and process improvement in wafer manufacturing.

[0061] Finally, the solution proposed in this application can be used not only for wafer inspection, but also for similar photomask inspection.

[0062] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wafer inspection device based on an improved RCWA algorithm, comprising an illumination system (10), a wafer motion stage (20), a microscopic imaging system (30), an image processor (40), and an imaging simulation system (50); characterized in that: The microscopic imaging system (30) is positioned toward the wafer stage (20) to acquire raw images of the wafer; The imaging simulation system (50) calculates the near-field electric field distribution on the wafer surface based on the improved RCWA algorithm, and calculates remote imaging in combination with the imaging system simulation model to produce reference images; The image processor (40) is used for preprocessing of the original wafer image to obtain the image to be detected, and compares the image to be detected with the reference image to obtain and output defect detection information.

2. The wafer inspection apparatus of claim 1, wherein: The lighting system (10) is used to provide multiple types of switchable lighting light to the wafer motion stage (20), including circular light, ring light, dipole light and fan light.

3. The wafer inspection apparatus of claim 1, wherein: The imaging simulation system (50) includes a light source model and a wafer three-dimensional structure model. The light source model simulates illumination light of different types and different light source parameters. The wafer three-dimensional structure model includes a near-field light field sub-model, a far-field light field sub-model and an imaging surface sub-model. The near-field light field sub-model is used to solve the near-field electromagnetic scattering characteristics of the wafer surface. The far-field light field sub-model is used for far-field calculation of the far-field imaging process of the optical system. The imaging surface sub-model is used to calculate the simulation results of the simulated optical image and intermediate process simulation information.

4. The wafer inspection apparatus according to claim 3, characterized in that: The light source parameters of the light source model include the wavelength, shape, and polarization of the light source.

5. The wafer inspection apparatus according to claim 3, characterized in that: The wafer parameters of the three-dimensional wafer structure model include thickness, shape, refractive index of each layer, and extinction coefficient.

6. The wafer inspection apparatus according to claim 3, characterized in that: The hardware of the imaging simulation system (50) adopts a GPU computing cluster module with a distributed cluster architecture. The distributed cluster architecture is achieved by the master node and multiple computing nodes working together, and by parallel computing based on the RCWA improved algorithm to obtain simulated optical images of single-layer or multi-layer structures on the wafer surface.

7. The wafer inspection apparatus according to claim 1, characterized in that: The improved RCWA algorithm includes: S1. Calculate the matrix exponent E of the transmission matrix T using the fast matrix exponent approximation method; S2. Calculate the transmission matrix T after stacking vacuum dielectric layers on adjacent sides; S3. Using the layer scattering matrix transformation formula, calculate the scattering matrix S corresponding to the transmission matrix T; S4. Traverse all layers of the multilayer structure and calculate the scattering matrix of each layer. j=1,2,…,L; S5. Obtain the global scattering matrix by using Redheffer star product operation. .

8. The wafer inspection apparatus according to claim 7, characterized in that: The fast matrix exponent approximation method in step S1 includes: approximating the matrix exponent by summing an infinite series as follows: The definition of matrix R is... , This represents the layer thickness of the target structure to be simulated. Let represent the wavelength of the incident light, and P and Q be the electromagnetic field transformation matrices; first, calculate the reduced matrix exponents. The original matrix exponent is then recovered by repeated squaring operations, i.e. Where the integer m takes the following values: That is, the choice of integer m is based on the 1-norm of matrices P and Q.

9. A wafer optical inspection method based on the wafer inspection apparatus according to any one of claims 1-8, characterized in that, Wafer optical inspection methods include: Wafer image acquisition: The wafer inspection device acquires wafer images and obtains raw image data; Image preprocessing is performed to obtain the image to be inspected. The wafer image is preprocessed, including noise reduction and background elimination, to obtain the image to be inspected for subsequent comparative analysis and defect identification. Light source modeling involves constructing a light source model with parameters including wavelength, shape, and polarization. Material three-dimensional structure and optical parameters definition: Define the three-dimensional structure of the wafer to be tested and the optical parameters of each layer of the wafer; RCWA calculation and near-field electric field distribution acquisition: Based on the defined wafer three-dimensional structure and optical parameters, the RCWA algorithm is used to calculate the near-field electric field distribution near the wafer surface. Far-field imaging calculation and reference image generation: Far-field imaging calculation is performed by combining the near-field electric field distribution with the modeling of the imaging system, thereby generating a defect-free reference image under ideal conditions for subsequent comparison. Image pairing processing involves registering the image to be detected with a reference image and performing brightness compensation to ensure consistency in position and brightness between the two images, facilitating subsequent difference analysis. Difference map acquisition involves performing difference processing on the image to be detected and the reference image to obtain a difference map; Thresholding algorithm and convolution kernel processing: Apply thresholding algorithm to difference map to further extract and enhance defect features; Defect location and signal strength information acquisition: Through the above processing, defect feature information including defect location and signal strength is obtained; Post-processing involves performing operations such as classification, sampling, and clustering on the detection results to further refine and organize the defect information. Inspection report generation: Based on the processed defect information, an inspection report on wafer defects is generated.

Images