Image reconstruction methods, apparatus, computer equipment, and storage media
By acquiring high-angle annular dark-field images and energy dispersive spectral datasets of semiconductor devices, grayscale and elemental boundary analysis is performed to generate fusion boundary confidence images, solving the problem of destructive reconstruction of samples in existing technologies and achieving efficient and accurate physical property failure analysis.
Patent Information
- Application Number
- CN202611391325.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-09
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies for semiconductor failure analysis require destroying the original structural information of the sample to reconstruct the image, resulting in permanent loss and high costs.
By acquiring high-angle annular dark-field images and energy dispersive spectra of semiconductor devices, gray-level histogram analysis and elemental composition boundary determination are performed. Combined with transformation matrix and weighted fusion, a fusion boundary confidence image is generated, and image reconstruction is performed to avoid physical cutting operations.
It enables image reconstruction without destroying sample structural information, reducing machine time and labor costs, avoiding positioning errors, and improving analytical accuracy.
Smart Images

Figure CN123118683A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of semiconductor technology, and in particular to an image reconstruction method, apparatus, computer device, and storage medium. Background Technology
[0002] In semiconductor failure analysis, the current standard workflow is as follows: first, planar or cross-sectional samples are prepared using a focused ion beam (FIB), and failure points are located by imaging with a transmission electron microscope (TEM). The sample must then be removed and subjected to a "planar-to-cross" or "cross-to-planar" operation within the FIB system to observe the morphology of the defects in the longitudinal or planar direction. However, this workflow destroys the original structural information of the sample, causing permanent damage. Summary of the Invention
[0003] Therefore, it is necessary to provide an image reconstruction method, apparatus, computer device, and storage medium that can reconstruct the original structural information of the sample without destroying the above-mentioned technical problems.
[0004] In a first aspect, this application provides an image reconstruction method, the method comprising:
[0005] Acquire high-angle annular dark-field images and energy-dispersive spectra of the target region in a semiconductor device;
[0006] Gray-level histogram analysis was performed on the high-angle annular dark-field image to determine the initial gray-level boundary in the high-angle annular dark-field image; the elemental composition boundary was determined based on the spectral imaging dataset.
[0007] The initial grayscale boundary and element composition boundary are aligned and fused to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional or planar images.
[0008] The reconstructed image is obtained by reconstructing the confidence image of the fusion boundary; the spatial orientation of the reconstructed image is opposite to that of the confidence image of the fusion boundary.
[0009] In one embodiment, grayscale histogram analysis is performed on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image, including:
[0010] Gray-level histogram analysis was performed on high-angle annular dark field images to determine the reference average gray value and reference gray standard deviation of the reference region in the semiconductor device.
[0011] The grayscale recognition window for each candidate material is determined based on the reference average grayscale value and the reference grayscale standard deviation; the candidate materials are those included in semiconductor devices.
[0012] Pixel material classification is performed on high-angle annular dark field images based on grayscale recognition windows to determine the initial grayscale boundary.
[0013] In one embodiment, the grayscale recognition window for each candidate material is determined based on a reference average grayscale value and a reference grayscale standard deviation, including:
[0014] The theoretical relative gray value of each candidate material is determined based on the reference average gray value and the atomic number of each candidate material.
[0015] The grayscale recognition window for each candidate material is determined based on the reference average grayscale value, the reference grayscale standard deviation, and the theoretical relative grayscale values.
[0016] In one embodiment, determining elemental composition boundaries based on a spectral imaging dataset includes:
[0017] Based on the spectral imaging dataset, quantitative corrections are performed on each pixel to determine the elemental content of each pixel; the quantitative corrections include atomic number correction, absorption correction, and fluorescence correction.
[0018] Based on the element content of each pixel, the target element set is filtered out using a preset element content threshold.
[0019] Based on the target element set and the elemental composition of each candidate material, pixel material annotation is performed on each pixel;
[0020] Connectivity analysis is performed based on the pixel material annotations of each pixel to determine the elemental composition boundaries.
[0021] In one embodiment, the initial grayscale boundary and element composition boundary are aligned and fused to obtain a fused boundary confidence image, including:
[0022] Based on the transformation matrix between the initial grayscale boundary and the element composition boundary, coordinate mapping is performed on the element composition boundary to obtain the intermediate element composition boundary.
[0023] The initial grayscale boundary and the intermediate element component boundary are weighted and fused to obtain the fused boundary confidence image.
[0024] In one embodiment, if the fusion boundary confidence image is a cross-sectional image and the reconstructed image is a planar image, reconstructing the fusion boundary confidence image to obtain the reconstructed image includes:
[0025] Based on the fusion boundary confidence image, the boundary continuity and connected components of each layer in the semiconductor device in the horizontal direction are extracted;
[0026] Based on boundary continuity, connected domains, and the horizontal extensibility of each layer, a mapping extrapolation is performed to obtain the reconstructed image.
[0027] In one embodiment, if the fusion boundary confidence image is a planar image and the reconstructed image is a cross-sectional image, reconstructing the fusion boundary confidence image to obtain the reconstructed image includes:
[0028] The location of the cutting line is determined based on the confidence image of the fusion boundary, and the gray-level distribution or element distribution at the location of the cutting line is extracted to obtain the sequence signal.
[0029] The reconstructed image is obtained by reconstructing the image based on the preset layer thickness model and layer sequence signal.
[0030] Secondly, one embodiment of this application provides an image reconstruction apparatus, the apparatus comprising:
[0031] The acquisition module is used to acquire high-angle annular dark-field images and energy dispersive spectra of the target region in a semiconductor device, forming a spectral imaging dataset.
[0032] The determination module is used to perform grayscale histogram analysis on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image; and to determine the elemental composition boundary based on the spectral imaging dataset.
[0033] The fusion module is used to align and fuse the initial grayscale boundaries and element composition boundaries to obtain a fused boundary confidence image. The spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional or planar images.
[0034] The reconstruction module is used to reconstruct the fusion boundary confidence image to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to that of the fusion boundary confidence image.
[0035] Thirdly, one embodiment of this application provides a computer device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the method provided in the first aspect above.
[0036] Fourthly, one embodiment of this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method provided in the first aspect above.
[0037] The aforementioned image reconstruction method, apparatus, computer equipment, and storage medium acquire a high-angle annular dark-field image and an energy dispersive spectrum spectral imaging dataset of a target region in a semiconductor device; perform gray-level histogram analysis on the high-angle annular dark-field image to determine the initial gray-level boundary in the high-angle annular dark-field image; determine the elemental composition boundary based on the spectral imaging dataset; align and fuse the initial gray-level boundary and the elemental composition boundary to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark-field image, both being cross-sectional or planar images; reconstruct the fused boundary confidence image to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to the spatial orientation of the fused boundary confidence image. In this embodiment, the initial grayscale boundary determined based on the high-angle annular dark field image and the elemental composition boundary determined based on the spectral imaging dataset are combined to determine the fusion boundary confidence image, which can improve the accuracy of the determined fusion boundary confidence image. The reconstructed image is obtained based on the fusion boundary confidence image, so there is no need to perform plane-to-section or section-to-plane operations on the semiconductor device, and no permanent loss of semiconductor device information in the original direction is caused. This can reduce machine time and labor costs, and there is no positioning error, thereby enabling the analysis of the physical property failure of semiconductor devices. Attached Figure Description
[0038] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0039] Figure 1 This is a schematic diagram of the FIB system in one embodiment;
[0040] Figure 2 This is a flowchart illustrating the steps of an image reconstruction method in one embodiment;
[0041] Figure 3 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0042] Figure 4 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0043] Figure 5 This is a schematic diagram of grayscale histogram analysis of a high-angle annular dark field image in one embodiment;
[0044] Figure 6 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0045] Figure 7 This is a schematic diagram illustrating EDS analysis in one embodiment;
[0046] Figure 8 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0047] Figure 9 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0048] Figure 10 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0049] Figure 11 This is a flowchart illustrating the steps of an image reconstruction method in another embodiment;
[0050] Figure 12 This is a schematic diagram illustrating the mutual reconstruction of a cross-sectional image and a planar image in one embodiment;
[0051] Figure 13 This is a schematic diagram of the structure of an image reconstruction device in one embodiment;
[0052] Figure 14 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0054] It should be noted that the terms "first," "second," etc., used in this application may be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion.
[0055] First, before introducing the technical solutions of the embodiments disclosed in this application, the background technology or technological evolution on which the embodiments of this application are based will be introduced. In the field of semiconductor technology, the current standard workflow for performing physical property failure analysis is to first prepare planar or cross-sectional samples using a focused ion beam (FIB) system, and then use a transmission electron microscope (TEM) to find the failure points of the semiconductor device. Subsequently, the sample must be removed, and the FIB system must be used to perform the planar-to-cross-section or cross-section-to-planar operation again to observe the morphology of the defects in the longitudinal direction or in the plane, so as to realize the analysis of semiconductor physical property failure. However, the aforementioned sample conversion process has several drawbacks, including: firstly, its destructive and irreversible nature: both planar-to-section and section-to-planar conversion operations require cutting and thinning, which permanently loses the sample information in the original direction; secondly, its time-consuming and costly nature: re-transferring and preparing samples in the FIB system typically takes several hours, increasing both machine and labor costs; and thirdly, the risk of positioning errors: during repeated transfers and sample preparation, the failure point position can easily shift due to ion beam damage or mechanical drift, leading to sample preparation failure. For semiconductor property failure analysis, although it is recognized that substrate images from TEM images can be used to distinguish different materials, current applications largely rely on subjective judgment by observers, failing to convert them into a quantitatively reconstructable objective data source. Furthermore, component analysis using Energy-Dispersive X-ray Spectroscopy (EDS) is mostly qualitative, lacking a quantitative method for structural reconstruction that combines contrast grayscale and EDS component analysis without destroying the original structural information of the sample. Therefore, this application provides an image reconstruction method.
[0056] The image reconstruction method provided in this application can be applied to FIB systems. For example... Figure 1As shown, the FIB system includes an electron beam tube 11, a scanning transmission electron microscopy-high-angle Annular Dark-Field (STEM-HAADF) detector 12, an EDS spectrometer 13, and a control device 14. The control device 14 is connected to the electron beam tube 11, the STEM-HAADF detector 12, and the EDS spectrometer 13, and is used to control the electron beam tube 11 to generate a scanning incident electron beam, the STEM-HAADF detector 12 to acquire the transmission electron signal, and the EDS spectrometer 13 to acquire the energy-dispersive X-ray spectrum. Based on the acquired transmission electron signal and energy-dispersive X-ray spectrum, the image reconstruction method provided in the following embodiments is executed. This embodiment does not limit the specific structure of the electron beam tube 11, the STEM-HAADF detector 12, the EDS spectrometer 13, and the control device 14, as long as their functions can be achieved.
[0057] The technical solution of this application and how the technical solution of this application solves the technical problem are described in detail below with specific embodiments.
[0058] In one embodiment, such as Figure 2 As shown, an image reconstruction method is provided, which can be applied to... Figure 1 The method will be illustrated using the FIB system as an example. In this embodiment, the method includes the following steps:
[0059] Step 200: Obtain the high-angle annular dark field image and energy dispersive spectrum spectral imaging dataset of the target region in the semiconductor device.
[0060] In semiconductor devices, the target region refers to the area where the failed circuit has been identified through electrical testing. An electron beam tube in the FIB system provides a scanning incident electron beam to the target region of the semiconductor device. A STEM-HAADF detector acquires the transmitted electron signal, while an EDS spectrometer acquires the energy dispersive spectrum (energy dispersive X-ray spectrum) of the target region. The control equipment generates a HAADF image of the target region based on the acquired transmitted electron signal and determines the spectral imaging dataset based on the energy dispersive spectrum.
[0061] Step 210: Perform grayscale histogram analysis on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image; determine the elemental composition boundary based on the spectral imaging dataset.
[0062] The grayscale intensity of HAADF images is approximately proportional to the square of the atomic number (Z-contrast), clearly reflecting the hierarchical structure of different materials. The relationship between grayscale intensity and atomic number can be expressed as: Where I is the gray intensity, Z is the atomic number of the material, and N is an empirical exponent, typically between 1.7 and 2, with the specific value depending on the geometric parameters of the STEM-HAADF detector, such as the interior and exterior angles. If the material in the semiconductor device is a compound, then Z represents the average atomic number of the compound. For example, for compound A... m B n The average atomic number can be expressed as: , among which, T A and T B Z represents the relative atomic masses of elements A and B, respectively. A and Z B These are the atomic numbers of elements A and B, respectively. Based on this, the control device can identify different materials in the high-angle annular dark-field image by performing grayscale histogram analysis, thereby determining the initial grayscale boundaries between different materials. The elemental analysis of the target region using the EDS spectral imaging dataset enables material identification of the semiconductor device and determines the elemental composition boundaries of the EDS.
[0063] Step 220: Align and fuse the initial grayscale boundary and element composition boundary to obtain the fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional or planar images.
[0064] After obtaining the initial grayscale boundary and elemental composition boundary, the control device aligns and fuses these boundaries in spatial coordinates to obtain a fused boundary confidence image. Essentially, the initial grayscale boundary and elemental composition boundary are first unified to the same spatial coordinate system before fusion. This unification can be done to the spatial coordinate system of the high-angle annular dark field image or the spatial coordinate system of the spectral imaging dataset.
[0065] When acquiring a high-angle annular dark field image, if the acquisition orientation of the high-angle annular dark field image is that the electron beam is incident perpendicular to the chip wafer surface, then the acquired high-angle annular dark field image is a planar image, and the fusion boundary confidence image is also a planar image; if the acquisition orientation of the high-angle annular dark field image is that the electron beam is incident parallel to the wafer surface, then the acquired high-angle annular dark field image is a cross-sectional image, and the fusion boundary confidence image is also a cross-sectional image.
[0066] Step 230: Reconstruct the fusion boundary confidence image to obtain the reconstructed image; the spatial orientation of the reconstructed image is opposite to that of the fusion boundary confidence image.
[0067] When performing semiconductor property failure analysis, it is necessary to analyze the longitudinal and planar morphology of semiconductor devices using planar and cross-sectional images to analyze property failures. After acquiring the fusion boundary confidence image, the control device can reconstruct the image based on the fusion boundary confidence image to obtain a reconstructed image. If both the high-angle annular dark field image and the fusion boundary confidence image are planar images, the reconstructed image is the corresponding cross-sectional image; if both the high-angle annular dark field image and the fusion boundary confidence image are cross-sectional images, the reconstructed image is the corresponding planar image. Different spatial orientations of the fusion boundary confidence image require different image reconstruction methods. This embodiment does not impose restrictions on this, as long as the function can be achieved.
[0068] The image reconstruction method provided in this application involves acquiring a high-angle annular dark-field image and an energy dispersive spectrum spectral imaging dataset of a target region in a semiconductor device; performing gray-level histogram analysis on the high-angle annular dark-field image to determine the initial gray-level boundary in the high-angle annular dark-field image; determining the elemental composition boundary based on the spectral imaging dataset; aligning and fusing the initial gray-level boundary and the elemental composition boundary to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark-field image, both being cross-sectional or planar images; reconstructing the fused boundary confidence image to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to the spatial orientation of the fused boundary confidence image. In this embodiment, the initial grayscale boundary determined based on the high-angle annular dark field image and the elemental composition boundary determined based on the spectral imaging dataset are combined to determine the fusion boundary confidence image, which can improve the accuracy of the determined fusion boundary confidence image. The reconstructed image is obtained based on the fusion boundary confidence image, so there is no need to perform plane-to-section or section-to-plane operations on the semiconductor device, and no permanent loss of semiconductor device information in the original direction is caused. This can reduce machine time and labor costs, and there is no positioning error, thereby enabling the analysis of the physical property failure of semiconductor devices.
[0069] In one embodiment, such as Figure 3 As shown, this involves performing grayscale histogram analysis on a high-angle annular dark field image to determine the initial grayscale boundary in the image. The steps of this implementation include:
[0070] Step 300: Perform grayscale histogram analysis on the high-angle annular dark field image to determine the reference average grayscale value and reference grayscale standard deviation of the reference region in the semiconductor device.
[0071] In a high-angle annular dark-field image, the value of each pixel is called its grayscale value. This grayscale value is obtained by analog-to-digital conversion of the intensity of scattered electrons received by the STEM-HAADF detector. Typically, in chip interconnect structures, holes / wires filled with heavy metals such as tungsten or copper will appear bright, while low-dielectric-constant dielectric layers such as silicon or oxygen will appear dark. In other words, different materials correspond to different grayscale values.
[0072] After obtaining the high-angle annular dark field image, the control device performs grayscale histogram analysis on the image to statistically analyze the grayscale value of each pixel. The reference average grayscale value and reference grayscale standard deviation of the reference region can be calculated using the statistically analyzed grayscale values of each pixel. The reference region can be preset by the observer. Optionally, the reference region can be a single-crystal silicon contrast region in a semiconductor device, and the reference average grayscale value and reference grayscale standard deviation are the average grayscale value and standard deviation of the single-crystal silicon contrast region.
[0073] Step 310: Determine the grayscale recognition window for each candidate material based on the reference average grayscale value and the reference grayscale standard deviation; the candidate materials are the materials included in the semiconductor device.
[0074] When fabricating semiconductor devices, a variety of candidate materials are used. Different candidate materials have different gray values. For each candidate material, the gray value recognition window can be determined based on the reference average gray value and the reference gray standard deviation.
[0075] In one embodiment, such as Figure 4 As shown, this involves an implementation method for determining the grayscale recognition window of each candidate material based on a reference average grayscale value and a reference grayscale standard deviation. The steps of this implementation method include:
[0076] Step 400: Determine the theoretical relative gray value of each candidate material based on the reference average gray value and the atomic number of each candidate material.
[0077] If the candidate material is a compound, the atomic number is the average atomic number. For each candidate material, the control device acquires the atomic number of the reference material in the reference region, calculates the ratio of the atomic number of the candidate material to the atomic number of the reference material, and calculates the product of this ratio and the average gray value to determine the theoretical relative gray value of the candidate material. Thus, the theoretical relative gray value of each candidate material can be obtained.
[0078] In an optional embodiment, the reference average gray value is represented as The theoretical relative gray value of candidate material k It can be represented as ,in, This represents the atomic number of candidate material k. This indicates the atomic number of the reference material in the reference region.
[0079] Step 410: Determine the grayscale recognition window for each candidate material based on the reference average grayscale value, the reference grayscale standard deviation, and each theoretical relative grayscale value.
[0080] For each candidate material, after obtaining the theoretical relative gray value of the candidate material, the gray standard deviation of the candidate material can be determined first based on the reference average gray value, the reference gray standard deviation, and the theoretical relative gray value. Based on the gray standard deviation and the relative theoretical gray value of the candidate material, the gray recognition window of the candidate material can be determined.
[0081] In an optional embodiment, the gray standard deviation of candidate material k It can be represented as ,in, To reference the grayscale standard deviation, the grayscale recognition window for candidate material k is: .
[0082] Step 320: Perform pixel material classification on the high-angle annular dark field image based on the grayscale recognition window to determine the initial grayscale boundary.
[0083] After obtaining the grayscale recognition windows for each candidate material, the control device determines which grayscale recognition window each pixel in the high-angle annular dark-field image falls into, and then identifies that pixel as the candidate material corresponding to that grayscale recognition window. If a pixel does not fall into any of the grayscale recognition windows for candidate materials, it is identified as another material or a mixed signal. This allows for the determination of the material to which each pixel in the high-angle annular dark-field image belongs. Based on the grayscale gradient changes corresponding to different materials, the initial grayscale boundary can be determined; that is, the boundary between different materials is the initial grayscale boundary.
[0084] In this embodiment, the reference average gray value and reference gray standard deviation of the reference region in the high-angle annular dark field image are used as benchmarks, eliminating the need for additional external standards and reducing the processing and testing costs of semiconductor devices. By classifying pixel materials through a defined gray-level recognition window before generating the initial gray-level boundary, false boundaries caused by image noise and local gray-level fluctuation noise can be suppressed, thereby improving the accuracy of the defined initial gray-level boundary.
[0085] In an optional embodiment, grayscale histogram analysis is performed on the high-angle annular dark field image as follows: Figure 5 As shown. Figure 5The image on the left is a high-angle annular dark-field image, and the image on the right is the image after grayscale analysis. Pixels within the red-framed area of the high-angle annular dark-field image were not identified, and the boundaries between other regions with different grayscale values are the initial grayscale boundaries. Different materials may have similar average atomic numbers, resulting in indistinct grayscale differences in the high-angle annular dark-field image, making accurate differentiation impossible. Therefore, a differentiation method is needed, as provided in the following embodiments.
[0086] In one embodiment, such as Figure 6 As shown, this relates to an implementation method for determining elemental composition boundaries based on a spectral imaging dataset. The steps of this implementation method include:
[0087] Step 600: Based on the spectral imaging dataset, perform quantitative correction on each pixel to determine the elemental content of each pixel; the quantitative correction includes atomic number correction, absorption correction and fluorescence correction.
[0088] Atomic number correction can correct backscattering loss of electron beams in semiconductor devices and suppress power differences; absorption correction can correct absorption attenuation of characteristic X-rays as they travel from the excitation depth to the surface of the semiconductor device; fluorescence correction can deduct the contribution of secondary fluorescence signals generated by high-energy X-rays exciting low-energy elements.
[0089] The elemental content of each pixel is related to the intensity of the characteristic X-rays in the spectral imaging dataset. After obtaining the spectral imaging dataset from the EDS, the control device acquires the intensity of the characteristic X-rays at each pixel location in the spectral imaging dataset, performs quantitative correction on it, and thus obtains the elemental content of each pixel.
[0090] In an optional embodiment, the content C of element i i It can be represented as: ,in, As a comprehensive correction factor, , , , , Let i be the standard instrument constant for element i. Let i be the characteristic X-ray intensity of element i in the spectral imaging dataset. Let be the characteristic X-ray intensity corresponding to the standard sample of element i. The atomic number correction factor for element i. Let be the RR composite factor of the standard sample, where RR is the composite function of the backscattering factor and the stopping power. Let be the RR composite factor of element i; Let be the absorption correction factor for element i. Let X be the X-ray absorption attenuation term corresponding to the standard sample of element i. The X-ray absorption attenuation term of element i, The combined parameters are calculated from the X-ray exit angle, the density of element i, and the mass coefficient of element i. The fluorescence correction factor for element i. The characteristic X-rays of other high-energy elements j, and the additional intensity of the secondary fluorescent X-rays generated by exciting element i. This represents the summation of all elements j that can excite fluorescence in element i.
[0091] Step 610: Based on the element content of each pixel, filter out the target element set using a preset element content threshold.
[0092] The preset element content threshold can be set by the observer. For the element content of each pixel, the control device compares it with the preset element content threshold to determine whether the element exists. If the element exists, it is added to the target element set. Specifically, if the element content is greater than the element content threshold, the element is determined to exist; if the element content is less than or equal to the preset content threshold, the element is determined not to exist.
[0093] Step 620: Based on the target element set and the elemental composition of each candidate material, perform pixel material annotation for each pixel.
[0094] Based on the elemental composition of each candidate material in the semiconductor device and the elements in the target element set, pixel material standards are established for each pixel. It can be understood that if all elements in the candidate materials are also in the target element set, then each pixel corresponding to all those elements is designated as a candidate material, thus achieving material labeling for each pixel.
[0095] Step 630: Perform connected component analysis based on the pixel material annotations of each pixel to determine the element composition boundaries.
[0096] The control device, based on the pixel material standard of each pixel, groups spatially adjacent pixels with the same material label into a connected region to obtain the element composition boundary.
[0097] In this embodiment, ZAF quantitative correction is introduced to reduce the absorption effect error of light elements (such as oxygen and nitrogen) in EDS analysis, thereby improving the accuracy of subsequent material determination. Furthermore, filtering the target element set by element content threshold can eliminate invalid interfering elements, reducing subsequent computational overhead. Based on the target element set and the elemental composition of each candidate material, pixel-level material labeling is performed on each pixel, achieving pixel-level mapping from element content to material category. Connectivity analysis can reduce false boundaries caused by misidentification of single pixels, thus improving the accuracy of the determined elemental composition boundaries.
[0098] In an optional embodiment, the image for EDS analysis is as follows: Figure 7 As shown. Figure 7 The image on the left is the HAADF image, and the image on the right is the EDS image. Figure 7 As can be seen, EDS can identify materials that were not identified in HAADF-based images.
[0099] In one embodiment, such as Figure 8 As shown, this invention relates to an implementation method for aligning and fusing initial grayscale boundaries and elemental composition boundaries to obtain a fused boundary confidence image. The steps of this implementation method include:
[0100] Step 800: Based on the transformation matrix between the initial grayscale boundary and the element composition boundary, perform coordinate mapping on the element composition boundary to obtain the intermediate element composition boundary.
[0101] The transformation matrix between the initial grayscale boundary and the elemental composition boundary can be calculated based on reference feature points. Specifically, an elemental surface distribution image can be generated based on the spectral imaging dataset of the acquired energy dispersive spectrum. A first reference feature point (e.g., interface intersection point, contact hole corner, boundary sharp corner, etc.) in the high-angle annular dark field image and the corresponding second reference feature point in the elemental surface distribution image are selected. The transformation matrix between the high-angle annular dark field image and the elemental surface distribution image, i.e., the transformation matrix between the initial grayscale boundary and the elemental composition boundary, is calculated based on the first and second reference feature points.
[0102] The control device transforms the element component boundary to the spatial coordinates of the high-angle annular dark field image based on the transformation matrix, thus obtaining the intermediate element component boundary. At this time, the intermediate element component boundary and the initial grayscale boundary are in the same spatial coordinates.
[0103] Step 810: Weighted fusion of the initial grayscale boundary and the intermediate element component boundary to obtain the fused boundary confidence image.
[0104] The control device performs weighted fusion of the initial grayscale boundary and the intermediate element component boundary, which are in the same spatial coordinates, to obtain a fusion boundary confidence image.
[0105] In an optional embodiment, the fused boundary confidence image can be determined based on a weighted consensus function, specifically, the weighted consensus function is expressed as: ,in, Represents pixel coordinates, Represents pixels The initial grayscale boundary information, Represents pixels The boundary information of the intermediate element components, These represent the weights corresponding to the initial grayscale boundary information and the intermediate element component boundary information, respectively. It can be pre-set by the observer.
[0106] In this embodiment, by performing coordinate mapping on the element component boundaries based on the transformation matrix between the initial grayscale boundaries and the element component boundaries, intermediate element component boundaries are obtained. This eliminates coordinate deviations and ensures that the two types of boundaries are in the same spatial coordinates, providing a reliable data foundation for subsequent weighted fusion. Furthermore, by combining the dual physical information of the initial grayscale boundaries and intermediate element component boundaries in a weighted manner, the accuracy and reliability of the obtained fused boundary confidence image can be improved.
[0107] In one embodiment, if the fused boundary confidence image is a cross-sectional image, the reconstructed image is a planar image. Based on this, such as Figure 9 As shown, this paper describes an implementation method for reconstructing a reconstructed image from a fusion boundary confidence image. The steps of this implementation method include:
[0108] Step 900: Based on the fusion boundary confidence image, extract the boundary continuity and connected components of each layer in the semiconductor device in the horizontal direction.
[0109] Based on the fusion boundary confidence image, the control device scans the fusion boundary confidence information layer by layer with the surface direction of the semiconductor device as the horizontal direction, records the position coordinates of the boundary pixels of each layer in the horizontal direction, obtains the boundary continuity, and determines the connected domain based on the area enclosed by the same connected boundary in the same layer.
[0110] Step 910: Based on the boundary continuity and connected domains, as well as the horizontal extensibility of each layer, perform mapping extrapolation to obtain the reconstructed image.
[0111] The horizontal extension of each layer refers to the width of each layer in the horizontal direction. Based on boundary continuity and connected components, the control device calculates the left and right endpoints of continuous boundary segments in the horizontal direction, and calculates the coordinate difference between the left and right endpoints to obtain the horizontal extension of that boundary segment. For each boundary segment in a layer, it is mapped and extrapolated horizontally according to the horizontal extension of the boundary segments in that layer, forming a horizontal region in the planar image. The mapping results of each layer are then superimposed layer by layer to obtain a reconstructed image of the planar image.
[0112] In this embodiment, by extracting the boundary continuity and connected components of each layer in the horizontal direction in the fused boundary confidence image, and using the horizontal extensibility of each layer as a parameter, a mapping relationship between the cross-sectional layer width and the planar layout outline is established, realizing the conversion from the cross-sectional image to the planar image. This eliminates the need to perform a cross-sectional to planar operation on the semiconductor device, avoids permanent loss of semiconductor device information in the original direction, reduces machine time and labor costs, and eliminates the risk of positioning error.
[0113] In one embodiment, if the fused boundary confidence image is a planar image, the reconstructed image is a cross-sectional image, such as... Figure 10 As shown, this relates to an implementation method for reconstructing a reconstructed image from a fused boundary confidence image. The steps of this implementation method include:
[0114] Step 1001: Determine the position of the cutting line based on the confidence image of the fusion boundary, and extract the gray-level distribution or element distribution at the position of the cutting line to obtain the sequence signal.
[0115] The control device responds to the cutting line determined by the observer based on the fusion boundary confidence information, acquiring the position of the cutting line, i.e., acquiring the coordinate information of the cutting line. Based on this coordinate information, it maps back to the high-angle annular dark-field image, extracting the grayscale distribution pixel by pixel along the cutting line to obtain one-dimensional grayscale distribution information, i.e., the toroidal signal. Alternatively, based on the coordinate information of the cutting line, it maps back to the energy dispersive spectrum spectral imaging dataset, reading the ZAF-corrected elemental distribution pixel by pixel along the cutting line to obtain one-dimensional elemental distribution information, i.e., the toroidal signal.
[0116] Step 1002: Reconstruct the image based on the preset layer thickness model and layer sequence signal.
[0117] The preset layer thickness model can be pre-generated based on the layout and process parameters of the semiconductor device. The preset layer thickness model includes the standard layer thickness value of each material layer in the vertical direction. The control device uses the position of each pixel in the layer sequence signal as the horizontal axis, the material type at that position, and the standard thickness of that material type as the vertical axis, and stacks the layers upwards from the substrate surface of the semiconductor device to form a reconstructed image of the cross-sectional image.
[0118] In this embodiment, the cutting line is located by fusing the boundary confidence image, and the gray-scale distribution or element distribution at the cutting line position is extracted to obtain the layer sequence signal; the reconstructed image is obtained by combining the preset layer thickness model, realizing the conversion from planar image to cross-sectional image. There is no need to perform planar to cross-sectional operation on the semiconductor device, and the semiconductor device information in the original direction will not be permanently lost. It can reduce machine time cost and labor cost, and there is no risk of positioning error.
[0119] Please see Figure 11 One embodiment of this application provides an image reconstruction method, the steps of which include:
[0120] Step 1101: Obtain the high-angle annular dark field image and energy dispersive spectrum spectral imaging dataset of the target region in the semiconductor device;
[0121] Step 1102: Perform grayscale histogram analysis on the high-angle annular dark field image to determine the reference average grayscale value and reference grayscale standard deviation of the reference region in the semiconductor device;
[0122] Step 1103: Determine the theoretical relative gray value of each candidate material based on the reference average gray value and the atomic number of each candidate material;
[0123] Step 1104: Determine the grayscale recognition window for each candidate material based on the reference average grayscale value, the reference grayscale standard deviation, and each theoretical relative grayscale value;
[0124] Step 1105: Perform pixel material classification on the high-angle annular dark field image based on the grayscale recognition window to determine the initial grayscale boundary;
[0125] Step 1106: Based on the spectral imaging dataset, perform quantitative correction on each pixel to determine the elemental content of each pixel; quantitative correction includes atomic order correction, absorption correction and fluorescence correction;
[0126] Step 1107: Based on the element content of each pixel, filter the target element set using a preset element content threshold;
[0127] Step 1108: Based on the target element set and the elemental composition of each candidate material, perform pixel material annotation for each pixel;
[0128] Step 1109: Perform connected component analysis based on the pixel material annotations of each pixel to determine the element composition boundaries;
[0129] Step 1110: Based on the transformation matrix between the initial grayscale boundary and the element composition boundary, perform coordinate mapping on the element composition boundary to obtain the intermediate element composition boundary;
[0130] Step 1111: Perform weighted fusion of the initial grayscale boundary and the intermediate element boundary to obtain the fused boundary confidence image.
[0131] If the fusion boundary confidence image is a cross-sectional image, then based on the fusion boundary confidence image, the boundary continuity and connected domains of each layer in the semiconductor device in the horizontal direction are extracted, and the reconstructed image is obtained by mapping extrapolation according to the horizontal extensibility of each layer.
[0132] If the fusion boundary confidence image is a planar image, the position of the cutting line is determined based on the fusion boundary confidence image, the gray-level distribution or element distribution at the cutting line position is extracted to obtain the layer sequence signal, and the reconstructed image is obtained by combining it with the layer thickness model.
[0133] In an optional embodiment, the mutual reconstruction of the cross-sectional image and the planar image is as follows: Figure 12 As shown, Figure 12The image on the left is a planar image, and the image on the right is a cross-sectional image.
[0134] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.
[0135] Based on the same inventive concept, this application also provides an image reconstruction apparatus for implementing the image reconstruction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more image reconstruction apparatus embodiments provided below can be found in the limitations of the image reconstruction method described above, and will not be repeated here.
[0136] In one embodiment, such as Figure 13 As shown, an image reconstruction apparatus is provided, comprising: an acquisition module 20, a determination module 21, a fusion module 22, and a reconstruction module 23, wherein:
[0137] The acquisition module 20 is used to acquire high-angle annular dark field images and energy dispersive spectra of the target region in the semiconductor device, which are spectral imaging datasets.
[0138] The determination module 21 is used to perform grayscale histogram analysis on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image; and to determine the elemental composition boundary based on the spectral imaging dataset.
[0139] The fusion module 22 is used to align and fuse the initial grayscale boundary and the element composition boundary to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional images or planar images;
[0140] The reconstruction module 23 is used to reconstruct the fusion boundary confidence image to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to that of the fusion boundary confidence image.
[0141] In one embodiment, the determining module 21 includes a first determining unit and a second determining unit. The first determining unit is used to perform grayscale histogram analysis on the high-angle annular dark-field image to determine the reference average grayscale value and reference grayscale standard deviation of the reference region in the semiconductor device. The second determining unit is used to determine the grayscale recognition window of each candidate material based on the reference average grayscale value and reference grayscale standard deviation; the candidate materials are materials included in the semiconductor device; and the high-angle annular dark-field image is classified into pixels based on the grayscale recognition window to determine the initial grayscale boundary.
[0142] In one embodiment, the second determining unit is specifically used to determine the theoretical relative gray value of each candidate material based on the reference average gray value and the atomic number of each candidate material; and to determine the gray value recognition window of each candidate material based on the reference average gray value, the reference gray standard deviation, and each theoretical relative gray value.
[0143] In one embodiment, the determining module 21 further includes a correction unit, a filtering unit, a labeling unit, and a third determining unit. The correction unit is used to quantitatively correct each pixel based on the spectral imaging dataset to determine the elemental content of each pixel; the quantitative correction includes atomic number correction, absorption correction, and fluorescence correction. The filtering unit is used to filter out a target element set based on the elemental content of each pixel using a preset elemental content threshold. The labeling unit is used to label the pixel material of each pixel based on the target element set and the elemental composition of each candidate material. The third determining unit is used to perform connected component analysis based on the pixel material labels of each pixel to determine the elemental composition boundaries.
[0144] In one embodiment, the fusion module 22 is specifically used to perform coordinate mapping on the element component boundary based on the transformation matrix between the initial gray-level boundary and the element component boundary to obtain the intermediate element component boundary; and to perform weighted fusion on the initial gray-level boundary and the intermediate element component boundary to obtain the fused boundary confidence image.
[0145] In one embodiment, the reconstruction module 23 is specifically used to extract the boundary continuity and connected domains of each layer in the semiconductor device in the horizontal direction based on the fusion boundary confidence image; and to perform mapping extrapolation based on the boundary continuity and connected domains, as well as the horizontal extensibility of each layer, to obtain the reconstructed image.
[0146] In one embodiment, the reconstruction module 23 is further configured to determine the position of the cutting line based on the fusion boundary confidence image, extract the gray-scale distribution or element distribution at the position of the cutting line to obtain the layer sequence signal, and reconstruct the reconstructed image based on the preset layer thickness model and the layer sequence signal.
[0147] Each module in the aforementioned image reconstruction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0148] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 14 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an image reconstruction method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0149] Those skilled in the art will understand that Figure 14 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0150] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method embodiments.
[0151] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method embodiments.
[0152] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the above-described method embodiments.
[0153] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0154] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0155] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. An image reconstruction method, characterized in that, The method includes: Acquire high-angle annular dark-field images and energy-dispersive spectra of the target region in a semiconductor device; Gray-level histogram analysis is performed on the high-angle annular dark-field image to determine the initial gray-level boundaries in the high-angle annular dark-field image; elemental composition boundaries are determined based on the spectral imaging dataset. The initial grayscale boundary and the elemental composition boundary are aligned and fused to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional or planar images; The fusion boundary confidence image is reconstructed to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to that of the fusion boundary confidence image.
2. The method according to claim 1, characterized in that, The step of performing grayscale histogram analysis on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image includes: Perform grayscale histogram analysis on the high-angle annular dark field image to determine the reference average grayscale value and reference grayscale standard deviation of the reference region in the semiconductor device; The grayscale recognition window for each candidate material is determined based on the reference average grayscale value and the reference grayscale standard deviation; the candidate material is a material included in the semiconductor device. The high-angle annular dark field image is classified into pixels based on the grayscale recognition window to determine the initial grayscale boundary.
3. The method according to claim 2, characterized in that, The step of determining the grayscale recognition window for each candidate material based on the reference average grayscale value and the reference grayscale standard deviation includes: The theoretical relative gray value of each candidate material is determined based on the reference average gray value and the atomic number of each candidate material. The grayscale recognition window for each candidate material is determined based on the reference average grayscale value, the reference grayscale standard deviation, and each of the theoretical relative grayscale values.
4. The method according to any one of claims 1-3, characterized in that, Determining the elemental composition boundaries based on the spectral imaging dataset includes: Based on the spectral imaging dataset, each pixel is quantitatively corrected to determine the elemental content of each pixel; the quantitative correction includes atomic number correction, absorption correction and fluorescence correction. Based on the element content of each pixel, a target element set is selected using a preset element content threshold. Based on the target element set and the elemental composition of each candidate material, pixel material annotation is performed on each pixel; Connectivity analysis is performed based on the pixel material annotations of each pixel to determine the element composition boundaries.
5. The method according to any one of claims 1-3, characterized in that, The step of aligning and fusing the initial grayscale boundary and the elemental composition boundary to obtain a fused boundary confidence image includes: Based on the transformation matrix between the initial grayscale boundary and the element composition boundary, coordinate mapping is performed on the element composition boundary to obtain the intermediate element composition boundary; The initial grayscale boundary and the intermediate element component boundary are weighted and fused to obtain the fused boundary confidence image.
6. The method according to claim 5, characterized in that, If the fusion boundary confidence image is a cross-sectional image and the reconstructed image is a planar image, the process of reconstructing the fusion boundary confidence image to obtain the reconstructed image includes: Based on the fusion boundary confidence image, the boundary continuity and connected components of each layer in the semiconductor device in the horizontal direction are extracted; The reconstructed image is obtained by mapping extrapolation based on the boundary continuity, the connected domain, and the horizontal extensibility of each layer.
7. The method according to claim 5, characterized in that, If the fusion boundary confidence image is a planar image and the reconstructed image is a cross-sectional image, the process of reconstructing the fusion boundary confidence image to obtain the reconstructed image includes: The position of the cutting line is determined based on the fusion boundary confidence image, and the gray-level distribution or element distribution at the position of the cutting line is extracted to obtain the sequence signal. The reconstructed image is obtained by reconstructing the image based on the preset layer thickness model and the layer sequence signal.
8. An image reconstruction apparatus, characterized in that, The device includes: The acquisition module is used to acquire high-angle annular dark-field images and energy dispersive spectra of the target region in a semiconductor device, forming a spectral imaging dataset. The determination module is used to perform grayscale histogram analysis on the high-angle annular dark field image to determine the initial grayscale boundary in the high-angle annular dark field image; and to determine the elemental composition boundary based on the spectral imaging dataset. The fusion module is used to align and fuse the initial grayscale boundary and the element composition boundary to obtain a fused boundary confidence image; the spatial orientation of the fused boundary confidence image is consistent with the acquisition orientation of the high-angle annular dark field image, both being cross-sectional or planar images. The reconstruction module is used to reconstruct the fusion boundary confidence image to obtain a reconstructed image; the spatial orientation of the reconstructed image is opposite to the spatial orientation of the fusion boundary confidence image.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.
