In-situ tracking three-dimensional detection system

Abstract

The invention provides an in-situ tracking three-dimensional detection system. The system comprises: an electron microscope, wherein the electron microscope comprises an electron microscope cavity and a sample stage mounted in the electron microscope cavity; the detection device comprises a driving assembly and a probe assembly; wherein the probe assembly is movably arranged in the electron microscope cavity and is used for collecting back scattering electronic signals of a to-be-detected sample from multiple directions; wherein a to-be-tested sample is placed on the sample table; the driving assembly is used for driving the probe assembly to reciprocate between a preset initial position and a detection position; wherein the detection position corresponds to the sample table; and the three-dimensional reconstruction device is electrically connected with the probe assembly and is used for reconstructing the three-dimensional shape of the to-be-detected sample according to the acquired back scattering electronic signals in the multiple directions. According to the invention, non-destructive detection of the to-be-detected sample is realized, so that the morphology change of the sample in the reaction process can be dynamically tracked.

Description

Technical Field

[0001] This invention relates to the field of surface science and technology, and in particular to an in-situ tracking three-dimensional detection system. Background Technology

[0002] Traditional 3D imaging techniques often employ multi-view imaging, which involves using multiple cameras to photograph the same object from different directions and reconstructing the object's surface 3D structure based on the positional differences of the same point in different images. In the field of electron microscopy, transmission electron microscopy achieves multi-angle imaging through dual tilting rods, but this method requires separate image capture for each tilt angle, resulting in slow image acquisition and cumbersome operation, making it difficult to meet the 3D imaging needs in dynamic processes. While scanning electron microscopy (SEM) offers advantages in surface imaging, it requires refocusing when photographing tilted samples, a complex and difficult-to-repeat process. Currently, multi-view reconstruction technology is not mature. 3D structure reconstruction in SEM mainly relies on focused ion beam cutting imaging technology, which achieves 3D reconstruction through layer-by-layer peeling and image stacking. However, this method is highly destructive and cannot be used for in-situ tracking, making it unsuitable for in-situ dynamic tracking in complex environments such as high temperatures and atmospheres. Therefore, there is a need for an in-situ tracking 3D detection system. Summary of the Invention

[0003] This invention provides an in-situ tracking three-dimensional detection system to improve the technical problem that existing scanning electron microscopes require the sample to be in a static state when imaging the sample, and cannot dynamically track the morphological changes of the sample during the reaction process.

[0004] This invention provides an in-situ tracking three-dimensional detection system, comprising: an electron microscope, including an electron microscope cavity and a sample stage installed within the electron microscope cavity; a detection device, including a drive assembly and a probe assembly; wherein the probe assembly is movably disposed within the electron microscope cavity for acquiring backscattered electron signals of the sample under test from multiple directions; wherein the sample under test is placed on the sample stage; the drive assembly is used to drive the probe assembly to reciprocate between a preset initial position and a detection position; wherein the detection position corresponds to the sample stage; and a three-dimensional reconstruction device, electrically connected to the probe assembly, for reconstructing the three-dimensional morphology of the sample under test based on the acquired backscattered electron signals from multiple directions.

[0005] In one embodiment of the present invention, the probe assembly includes a clamping member and a multi-azimuth probe detachably mounted on the clamping member.

[0006] In one embodiment of the present invention, the probe is a metal electrode with a carbon coating.

[0007] In one embodiment of the present invention, the driving component includes: a cylinder connected to the probe assembly; and a first driving mechanism connected to the cylinder for driving the probe assembly to reciprocate between an initial position and a detection position via the cylinder, so as to adjust the distance between the probe assembly and the detection position by a first amplitude.

[0008] In one embodiment of the present invention, the driving component further includes: a second driving mechanism connected to the cylinder, used to drive the cylinder to perform a second amplitude adjustment of the probe component in multiple directions within a plane with the first driving mechanism as the normal vector; wherein the second amplitude is smaller than the first amplitude.

[0009] In one embodiment of the present invention, a data cable sealing cavity for accommodating the data cable is provided inside the cylinder, and the probe assembly and the three-dimensional reconstruction device transmit backscattered electronic signals through the data cable.

[0010] In one embodiment of the present invention, the three-dimensional reconstruction device includes: a data acquisition module, used to acquire backscattered electron signals from multiple directions of the sample to be tested and form corresponding backscattered electron images; wherein the backscattered electron signals are obtained by scanning the sample to be tested with an electron beam using a probe assembly; a feature extraction module, used to input the backscattered electron images from multiple directions into the feature extraction network of the three-dimensional reconstruction model, and extract the local morphological features of the corresponding backscattered electron images by progressive downsampling based on the shadow information of the backscattered electron images; wherein the feature extraction network is a convolutional neural network; a fusion module, used to input each local morphological feature into the fusion network of the three-dimensional reconstruction model, and perform weighted fusion of each local morphological feature based on an attention mechanism to generate global morphological features for characterizing the overall morphology of the sample to be tested; and a three-dimensional reconstruction module, used to input the global morphological features into the prediction network of the three-dimensional reconstruction model, and reconstruct the three-dimensional morphology of the sample to be tested by progressive upsampling; wherein the structure of the feature extraction network is symmetrical to the structure of the prediction network.

[0011] In one embodiment of the present invention, the electron microscope further includes a gas supply unit connected to the electron microscope cavity, the gas supply unit being used to provide gas supply to the electron microscope cavity.

[0012] In one embodiment of the present invention, the electron microscope further includes a laser heating stage disposed within the electron microscope cavity for heating the sample to be tested.

[0013] In one embodiment of the present invention, the detection system further includes: a vacuum camera disposed on the sample stage, used to observe the position and state of the probe assembly and guide the probe assembly to align with the detection position.

[0014] The beneficial effects of this invention are as follows: The in-situ tracking three-dimensional detection system proposed in this invention, by setting a movable probe assembly within the electron microscope cavity, allows the probe assembly to reciprocate between an initial position and a detection position under the drive of a driving component. This facilitates both rapid retrieval of the probe assembly and precise sampling of the sample. Furthermore, the movable probe assembly can collect backscattered electron signals emitted by the sample from multiple directions, not only compensating for the information loss caused by single-angle acquisition and improving the accuracy of subsequent three-dimensional reconstruction, but also achieving non-destructive detection of the sample, thereby dynamically tracking morphological changes during the sample's reaction process. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0016] In the attached diagram:

[0017] Figure 1 This is a schematic diagram of the structure of an in-situ tracking three-dimensional detection system provided in an embodiment of the present invention;

[0018] Figure 2 This is a cross-sectional view of a detection device provided in an embodiment of the present invention;

[0019] Figure 3 This is a cross-sectional view of a probe assembly provided in an embodiment of the present invention.

[0020] Component designation explanation:

[0021] 100. Electron microscope; 110. Electron microscope cavity; 120. Sample stage; 130. Gas supply unit; 140. Laser heating stage; 150. Polar shoe; 160. Coiled pipe; 200. Detection device; 210. Probe assembly; 211. Clamping component; 212. Probe; 220. Drive assembly; 221. Cylinder; 2211. Data cable sealing cavity; 222. First drive mechanism; 223. Second drive mechanism. Detailed Implementation

[0022] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0023] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. The drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0024] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0025] This invention provides an in-situ tracking three-dimensional detection system. It simultaneously acquires backscattered electron signals from different azimuth angles using probes in multiple directions to obtain multi-angle shadow information of the sample under test, thereby providing a more comprehensive characterization of the microstructure features of the sample surface. By combining the directional response characteristics of backscattered electrons, scanning image simulation, and a three-dimensional reconstruction model based on deep learning technology, the three-dimensional height information of the sample surface can be accurately reconstructed, enabling real-time and quantitative monitoring of the three-dimensional morphology of the target material during in-situ chemical reactions. Furthermore, this three-dimensional detection system can operate normally under large-span vacuum and high-temperature conditions, meeting the environmental conditions required for different in-situ experiments. This allows for dynamic tracking of morphological changes during the reaction process of the target material under various experimental environments, demonstrating good versatility.

[0026] This invention is based on the following inventive concept: Unlike multi-view imaging, in an electron microscope, the electron beam acts as a single camera, and the detector can be approximated as a light source. The object's shape can be reconstructed based on the shadows cast on the object by light sources at different angles. Because scanning electron microscopes typically require samples with flat surfaces or thin sheets or powder samples, the amount of shadow information required for reconstruction is much smaller than that for ordinary three-dimensional objects, thus requiring fewer detectors. Considering the limitations on the number and size of probes within the scanning electron microscope cavity, this invention places a multi-azimuth probe directly below the microscope's pole piece. Its zenith angle remains constant relative to the sample, uniformly occupying four azimuth angles, allowing simultaneous collection of backscattered electron signals from four directions, and withstanding temperatures up to 1500℃. Since, for the same substance, under the same electron microscope parameter settings, the signal intensity at a point on the material surface is related to the tilt angle θ at that point, and is proportional to... The varying intensity signals related to the tilt angle are the shadows required for morphology reconstruction. This method does not require changing the zenith angle to capture images; it can directly reconstruct the three-dimensional morphology of the sample by generating multiple backscattered electron images in a single scan, thus enabling morphology monitoring of dynamic chemical reactions.

[0027] like Figure 1 As shown, the in-situ tracking three-dimensional detection system includes: an electron microscope 100, a detection device 200, and a three-dimensional reconstruction device (not shown in the figure). The electron microscope 100 of this application is an Environmental Scanning Electron Microscope (ESEM), a prerequisite for in-situ chemical reactions. It can provide the required temperature and atmosphere for in-situ chemical reactions, with a maximum temperature of up to 1500℃ and a maximum pressure of up to 1000Pa, suitable for various in-situ experimental conditions. The electron beam can perform high-resolution scanning of the sample surface to obtain atomic-level images, allowing for precise observation and control of the reaction process. Specifically, the electron microscope includes an electron microscope chamber 110, a sample stage 120 installed within the electron microscope chamber 110, and pole shoes 150. The electron microscope chamber 110 forms a sealed space to create the vacuum environment required for electron beam scanning. The composition and gas pressure within the electron microscope chamber 110 can also be adjusted by a gas supply unit 130 to meet the experimental requirements of variable atmospheres in in-situ high-temperature reactions. The sample stage 120 is installed in the center of the electron microscope cavity 110 to hold the sample to be tested and is aligned with the electron beam scanning path to ensure scanning accuracy.

[0028] The aforementioned detection device 200 includes a drive assembly 220 and a probe assembly 210. The probe assembly 210 is movably disposed within the electron microscope cavity 110 and is used to acquire backscattered electron signals from the sample under test from multiple directions. The sample under test is placed on the sample stage 120. The drive assembly 220 drives the probe assembly 210 to reciprocate between a preset initial position and a detection position. The detection position corresponds to the sample stage 120. Specifically, the probe assembly 210 is disposed within the electron microscope cavity 110 and can move along a predetermined path under the action of the drive assembly 220. The drive assembly 220 is disposed outside the electron microscope cavity 110 and connected to the probe assembly 210, used to drive the probe assembly 210 to reciprocate between the preset initial position and the detection position. The detection position is the working position where the probe assembly 210 is aligned with the sample stage 120 for acquiring backscattered electron signals, and the initial position is the position of the probe assembly 210 when it is in the retracted state.

[0029] like Figures 1 to 3 As shown, in an optional embodiment of the present invention, the probe assembly 210 includes a clamping member 211 and a multi-azimuth probe 212 detachably mounted on the clamping member 211. Specifically, the clamping member 211 is mounted at the end of the drive assembly 220 facing the electron microscope cavity 110, for fixing the multi-azimuth probe 212 and ensuring that each probe 212 remains stable during the acquisition of backscattered electron signals. The probe 212 is divided into multiple backscattering receiving plates to acquire backscattered electron signals from multiple directions, providing multi-azimuth shadow information for three-dimensional topography reconstruction. To balance the diversity of angle acquisition and the compactness of the structure, preferably, the probe 212 has four backscattering receiving plates, making it a high-temperature four-quadrant probe to adapt to the high-temperature environment of the experiment. The four backscattering receiving plates are evenly mounted on the clamping member 211 and are centrally symmetrically distributed around the azimuth of the electron beam scanning area, so that the backscattered electron signals generated by the sample under test at different azimuth angles can be acquired separately. Furthermore, the probe 212 is detachably mounted on the clamp 211, allowing for the replacement of the required type of probe 212 according to experimental needs. For example, when the probe 212 is a four-quadrant probe, the type of four-quadrant probe can be changed based on experimental conditions. It is understood that the detachable method includes, but is not limited to, plug-in connections such as snap-fit ​​or spring clips, or screw connections, etc., and is not specifically limited. It should be noted that there are various types of clamps 211; any clamp capable of supporting the probe 212 is applicable, and its specific structure is not limited.

[0030] Continue as Figures 1 to 3As shown, in an optional embodiment of the present invention, probe 212 is a metal electrode with a carbon coating. Preferably, the metal electrode is a fan-shaped metal electrode. The surface of each metal electrode is coated with a cleanable and recoated conductive carbon layer. In high-temperature in-situ conditions, higher accelerating voltages are typically used, and the carbon coating optimizes the absorption efficiency of backscattered electrons under high accelerating voltages. Furthermore, the carbon coating also exhibits good thermal stability and electrical signal response performance, making it suitable for high-temperature in-situ experimental conditions. Further, considering that high-temperature in-situ experiments generate a large number of hot electrons and stray light signals, which can interfere with the absorption of backscattered electron signals, a preset bias voltage range (e.g., -10V to 10V) can be applied to probe 212 to filter the hot electrons and stray light signals, thereby improving the quality of the subsequently generated backscattered electron image.

[0031] Preferably, considering that during the heating of the sample, the light and hot electrons emitted by the hot sample can easily saturate electronic components, and that unavoidable contamination under in-situ conditions limits the lifespan of sensitive sensors, traditional detectors cannot adapt to in-situ high-temperature experiments. This invention uses a high-temperature resistant four-quadrant probe, which can apply a bias voltage to the electrodes, repelling secondary electrons and hot electrons. Furthermore, the high-temperature resistant four-quadrant probe is light-blind; light emitted by the hot sample or scattered by the laser heater will not generate a signal. The electrodes can be coated with different materials to optimize absorption efficiency in certain energy ranges, such as carbon coatings for higher accelerating voltages, which can be easily removed, cleaned, and recoated.

[0032] like Figure 1 and Figure 2 As shown, in an optional embodiment of the present invention, the driving component 220 includes: a cylindrical body 221 connected to the probe component 210; and a first driving mechanism 222 connected to the cylindrical body 221, used to drive the probe component 210 to reciprocate between an initial position and a detection position via the cylindrical body 221, so as to adjust the distance between the probe component 210 and the detection position by a first amplitude. Specifically, the cylindrical body 221 is connected to the probe component 210 and is used to support and position the probe component 210. The first driving mechanism 222 drives the cylindrical body 221 to move along the axial direction of the electron microscope cavity 110, thereby driving the probe component 210 to reciprocate between the initial position and the detection position. Through the above-mentioned movement process, the forward and backward stroke of the probe component 210 inside the electron microscope cavity 110 can be adjusted by a large amplitude (i.e., the first amplitude), so as to control the axial travel direction of the probe component 210. Preferably, a spiral adjustment mechanism (not shown in the figure) can be provided at the tail end of the cylinder 221 to more precisely control the extension range of the probe assembly 210, thereby realizing the displacement adjustment of the probe assembly 210 from the inner wall of the electron microscope cavity 110 to directly below the pole shoe 150, with a stroke range of 0–20 cm and a positioning accuracy of 0.1 cm.

[0033] like Figures 1 to 3 As shown, to ensure precise alignment of the central opening of the probe assembly 210 with the electron beam irradiation area directly below the pole piece 150, in an optional embodiment of the present invention, the driving assembly 220 further includes a second driving mechanism 223 connected to the cylinder 221. This second driving mechanism 223 drives the cylinder 221 to perform a second amplitude adjustment of the probe assembly 210 in multiple directions within a plane with the first driving mechanism 222 as the normal vector; wherein the second amplitude is smaller than the first amplitude. Specifically, the driving assembly 220 further includes a second driving mechanism 223 connected to the cylinder 221 and configured to, within a plane with the movement direction of the first driving mechanism 222 as the normal vector, drive the cylinder 221 to adjust its attitude in all angular directions, thereby enabling the probe assembly 210 to perform a second amplitude fine adjustment within the aforementioned plane, thus improving the accuracy of subsequent three-dimensional topography reconstruction. Compared to the large-amplitude forward and backward displacement of the first driving mechanism 222, the second driving mechanism 223 can perform small-amplitude, high-precision attitude adjustment of the probe 212. Preferably, the second drive mechanism 223 is a full-angle spiral adjustment device set in three directions to perform all-round attitude fine adjustment of the probe 212 in the plane with the first drive mechanism 222 as the normal vector. By adjusting the distance between the probe 212 and the pole shoe 150, it can adapt to the working distance requirements of different experiments.

[0034] Continue as Figure 1 and Figure 2 As shown, in an optional embodiment of the present invention, a data cable sealing cavity 2211 for accommodating the data cable is provided inside the cylindrical body 221. The probe assembly 210 and the three-dimensional reconstruction device transmit backscattered electron signals through the data cable. Specifically, the data cable sealing cavity 2211 for accommodating the data cable is formed inside the cylindrical body 221, and the data cable sealing cavity 2211 extends through the cylindrical body 221. The probe assembly 210 and the three-dimensional reconstruction device transmit signals through this data cable, thereby transmitting the backscattered electron signals detected by the probe assembly 210 to the three-dimensional reconstruction device at the rear end. Further, the data cable sealing cavity 2211 is a sealed structure, and its outlet end is provided with a sealing connector to isolate it from the external environment of the electron microscope cavity 110, preventing leakage of the vacuum environment inside the electron microscope cavity 110 due to the data cable being led out, so as to meet the vacuum requirements of in-situ detection. Furthermore, the data cable sealing cavity 2211 can also be adapted to the replacement operation of the probe assembly 210, and can provide an unobstructed cable receiving channel during the extension, retraction and movement of the probe assembly 210, which greatly improves the flexibility of the in-situ tracking three-dimensional detection system.

[0035] In an optional embodiment of the present invention, the detection system further includes a gas supply unit 130 connected to the electron microscope cavity 110, the gas supply unit 130 being used to provide gas to the electron microscope cavity 110. The gas supply unit 130 is used to provide the required atmospheric environment inside the electron microscope cavity 110 through a coiled pipe 160, thereby realizing the construction and adjustment of in-situ reaction conditions. The gas supply unit 130 can operate within a preset gas pressure range (e.g., 10...). -6 Pa to 10 3 Pa) Adjusts the air pressure inside the electron microscope cavity 110, and works with the PID control system to perform closed-loop regulation of the air pressure inside the electron microscope cavity 110 to maintain the air pressure in a stable state.

[0036] In an optional embodiment of the present invention, the detection system further includes a laser heating stage 140 disposed within the electron microscope cavity 110, used to heat the sample to be tested, with a power range of 1W to 100W and a temperature control range up to 1500℃. During the laser heating process, the temperature is measured in real time by a thermocouple sensor, and combined with a PID temperature control system, the temperature error is controlled within a preset temperature difference range (e.g., ±0.1℃) to ensure the stability of the chemical reaction environment.

[0037] In an optional embodiment of the present invention, the detection system further includes a vacuum camera (not shown in the figures) mounted on the sample stage 120, used to observe the position of the probe assembly 210 in real time under vacuum conditions and guide the probe assembly 210 to align with the detection position. The vacuum camera is miniaturized and designed to withstand high vacuum, and is pluggably connected to the sample stage 120 for easy replacement. Image signals are transmitted via a USB through flange mounted on the electron microscope chamber 110, ensuring that the vacuum environment inside the electron microscope chamber 110 is not disrupted. When adjusting the position of probe 212, the vacuum camera can capture a real-time image of the relative position of probe assembly 210 and sample stage 120 to determine whether probe 212 is aligned with the detection position. If not aligned, the position of probe assembly 210 can be adjusted by drive assembly 220 to ensure that the central opening on probe assembly 210 is aligned with the electron beam axis from pole shoe 150 to the observation point of the sample to be tested. This makes the multi-azimuth probe 212 symmetrically distributed with respect to the plane normal vector of the sample to be tested, so as to collect the backscattered electron signal excited by the sample in equal amounts, thereby achieving the purpose of uniform contrast of multiple backscattered electron images and preventing distortion and deformation of the reconstructed morphology caused by uneven contrast.

[0038] The aforementioned 3D reconstruction device is electrically connected to the probe assembly 210 and is used to reconstruct the 3D morphology of the sample under test based on backscattered electron signals acquired from multiple directions. Specifically, the 3D reconstruction device is located outside the electron microscope 100 and can be deployed as an independent module or integrated into an image processing terminal; the specific method is not limited. The 3D reconstruction device performs joint analysis of backscattered electron images from different directions, extracting local morphological features of the sample surface at different angles based on the differences in signal intensity and corresponding scanning positions in each direction. These local morphological features from various angles are then fused, and the fused features are used to reconstruct the 3D morphology of the sample under test.

[0039] In an optional embodiment of the present invention, the three-dimensional reconstruction device includes a data acquisition module, a feature extraction module, a fusion module, and a three-dimensional reconstruction module. The data acquisition module is used to acquire backscattered electron signals from multiple directions of the sample under test and form corresponding backscattered electron images; wherein, the backscattered electron signals are obtained by scanning the sample under test with an electron beam using the probe assembly 210.

[0040] Specifically, such as Figure 1 and Figure 3 As shown, when scanning the sample to be tested is required, the pole piece 150 emits an electron beam to scan the sample point by point. The probe assembly 210 is located directly below the pole piece 150, and its multi-azimuth probe 212 is divided into multiple backscattering receivers. These backscattering receivers are centrally symmetrically distributed around the scanning area of ​​the electron beam, preferably a four-quadrant probe. Since the emission intensity of backscattered electrons is affected by the tilt angle of the sample surface and has a significant angle dependence, the backscattering receivers distributed at different azimuth angles can collect backscattered electron signals reflecting the morphological characteristics of the sample to be tested in different directions. The collected backscattered electron signals are amplified by the preamplifiers of their respective backscattering receivers and then transmitted to the external data acquisition module via a data line. For the backscattered electron signals in each direction: the data acquisition module reconstructs the received backscattered electron signals into grayscale values ​​of the corresponding pixel positions based on the azimuth angle of the corresponding backscattering receiver and the electron beam scanning position, generating a backscattered electron image in that direction.

[0041] Preferably, the data acquisition module includes a multi-channel signal amplifier and a digital imaging scanning system. The multi-channel signal amplifier simultaneously processes backscattered electronic analog signals from the multi-azimuth probe 212, performing synchronous amplification, gain and offset calibration, real-time signal mixing, and automatic control on multiple analog signals for integration with the probe 212. Furthermore, the amplifier integrates power supply and bias control, supports four differential input / output channels and two auxiliary input interfaces for external synchronization or control signals, thus meeting the needs of complex signal acquisition scenarios. The digital imaging scanning system generates high-resolution image data, supporting traditional rectangular (raster) scanning and point-based (vector) scanning, and can simultaneously acquire 4 analog inputs and 12 digital inputs. It integrates a MICS amplifier, expanding the number of analog input channels to 16. It supports a minimum pixel dwell time of 10 ns and a maximum image resolution of 5 megapixels.

[0042] The aforementioned feature extraction module is used to input backscattered electron images from multiple directions into the feature extraction network of the three-dimensional reconstruction model. Based on the shadow information of the backscattered electron images, the local morphological features of the corresponding backscattered electron images are extracted through stepwise downsampling. The feature extraction network is a convolutional neural network.

[0043] The feature extraction network has multiple independent branches, each processing a backscattered electron image in one direction to extract local morphological features representing the sample under test in that direction. Specifically, since each backscattered electron image carries shadow information formed by the surface microstructure and tilt angle of the sample under test, each backscattered electron image is input into the feature extraction network of a pre-trained 3D reconstruction model. Through multi-layer convolution and downsampling operations, local features representing the microstructure of the sample, such as edges, textures, and brightness distribution, are extracted from the image to obtain local morphological features carrying the shadow information and surface texture structure of the sample under test in that direction.

[0044] The aforementioned fusion module is used to input various local morphological features into the fusion network of the 3D reconstruction model, and to perform weighted fusion of various local morphological features based on the attention mechanism to generate global morphological features that characterize the overall morphology of the sample under test.

[0045] After the backscattered electron images from each direction are processed by the aforementioned feature extraction network, local topographic features in that direction are generated. Considering the differences in acquisition angles between the backscattered electron images from different directions, the local topographic features they reflect have a certain viewpoint bias. Therefore, it is necessary to fuse the local topographic features from multiple directions to form a complete global topographic feature. Specifically, the fusion network aligns the local topographic features from each direction to map them to the same spatial dimension. Based on an attention mechanism, it comprehensively considers the local topographic features from different directions, adaptively assigns corresponding fusion weights to each local topographic feature, and performs weighted fusion of the local topographic features from each direction based on the fusion weights to obtain the global topographic feature representing the overall features of the sample under test. The global topographic feature obtained through the attention mechanism not only comprehensively fuses effective information from different directions but also suppresses redundant and noise signals, thereby improving the accuracy of subsequent 3D reconstruction.

[0046] It is understandable that the fusion network can be any network with an attention mechanism, as long as it can effectively fuse local morphological features from multiple directions; there are no specific limitations. Preferably, considering the strong spatial dependence of local morphological features in each direction, and in order to improve the ability to model complex spaces, the fusion network is a Transformer.

[0047] The aforementioned 3D reconstruction module is used to input global topographic features into the prediction network of the 3D reconstruction model, and reconstruct the 3D topography of the sample to be tested through progressive upsampling; wherein, the structure of the feature extraction network is symmetrical to the structure of the prediction network.

[0048] Specifically, the prediction network employs a layer-by-layer reconstruction structure that combines deconvolution or upsampling with convolution, symmetrical to the feature extraction network structure. This gradually restores the fused global topography feature map to the resolution of the original backscattered electron image in the spatial dimension. Specifically, the prediction network, through step-by-step deconvolution or upsampling operations, progressively recovers the detailed features of the image while maintaining feature consistency, thereby reconstructing the three-dimensional height value of each scan position and generating a predicted three-dimensional height map corresponding to the sample under test. This predicted three-dimensional height map is used to characterize the grayscale information such as brightness and shadow in the two-dimensional backscattered electron image as the height change of the sample surface in the Z-axis direction, and through rendering operations, reconstructs the surface topography of the sample under test. Preferably, the feature extraction network and the prediction network are the encoder and decoder modules of the UNet network, respectively. During feature extraction, residual connections are used to retain more spatial detail information in the prediction network, thereby improving the prediction network's ability to restore local details during reconstruction.

[0049] It should be further explained that the aforementioned 3D reconstruction model is trained. During training, backscattered electron images from multiple directions are input into the 3D reconstruction model to generate a predicted 3D height map. This predicted height map is then compared point by point with the actual 3D height map, and the mean square error between the two is calculated. Based on this, the parameters of the 3D reconstruction model are updated to ensure the geometric accuracy of the final reconstructed shape.

[0050] Furthermore, the 3D detection system also includes a 3D automatic calibration system to improve the geometric accuracy of 3D topography reconstruction. This 3D automatic calibration system, through pre-calibration of a standard sample, can eliminate reconstruction errors such as field of view error, height error, and sphericity error, thereby improving the accuracy of the subsequent 3D reconstructed structure. The standard sample is fabricated by focused ion beam deposition and includes three independent tertiary pyramid arrays, a spherical element, and multiple ring-shaped nanomarkers. The three pyramid arrays are used to construct a spatial reference framework, the spherical element is used for multi-angle response calibration of the backscattering probe 212 in four directions, and the ring-shaped nanomarkers constitute a reference zero point, the center position of which has been measured and calibrated using high-precision instruments and can serve as a reference point for X, Y, and Z axis calibration. It should be noted that, in another embodiment of the invention, in order to achieve accurate 3D reconstruction, in addition to the backscattered electron signal acquired by the probe 212, it is also necessary to know the electron beam voltage, current, working distance, probe shape and size, and the distance from the probe to the sample, etc. The specific details can be adapted to the experimental scenario by those skilled in the art and are not limited here.

[0051] The following describes the usage of the three-dimensional detection system of this invention, demonstrating the application of an in-situ tracking 3D detector under high-temperature atmospheric conditions in catalysis scientific research: (e.g.) Figures 1 to 2As shown, when performing 3D reconstruction of the sample to be tested, probe alignment is required first. The vacuum camera is inserted into the sample stage 120 and connected to the USB straight-through flange. The first drive mechanism 222 of the rotary drive assembly 220 inserts the four-quadrant probe 212 below the pole piece 150. Then, the second drive mechanism 223 uses fine-tuning screws to raise each probe 212 to a position 0.5 mm from the pole piece 150. Using the real-time image transmitted from the vacuum camera, the position of the probe 212 is adjusted so that the center of the four-quadrant probe 212 is perfectly aligned with the bottom of the pole piece 150. After alignment, the vacuum camera is removed. Next, the 3D calibration standard is prepared and calibrated. The 3D calibration standard is placed on the sample stage 120 of the electron microscope cavity 110 and heated to 800°C using the laser heating stage 140. A mixture of H2 and CO gas is introduced, maintaining a pressure of 100 Pa. The image is focused on the 3D calibration standard, and the brightness and contrast are adjusted to a reasonable range to ensure the image is neither overexposed nor underexposed. Error elimination and automatic calibration are performed using a ring-shaped reference zero-point calibration image scale to ensure accurate coordinate axis calibration of the measuring instrument; the electron beam is deflected to remove noise and improve signal purity. The sample is moved to a flat area of ​​the standard sample for spherical calibration. Finally, the sample is moved to the standard pyramid, and the standard file is imported for calibration, ensuring that the field of view error and height error are less than 1%. Then, experimental sample installation and condition settings are performed. Catalyst samples such as Pt or Al2O3 are placed on the sample stage 120, maintaining the same experimental conditions as the calibration standard. Data acquisition is then performed. By activating probe 212 and adjusting the resolution and pixel dwell time (e.g., 10 ns), the backscattered electron signals of the sample under test in multiple directions are recorded. A scan is performed every 10 seconds, generating multiple backscattered electron images, which are then transmitted to the 3D reconstruction device. The 3D reconstruction device preprocesses each backscattered electron image and extracts the local topographic features of the corresponding backscattered electron images. These local topographic features are then fused, and an attention mechanism is used to enhance the integrity of the topographic information to obtain a global topographic feature that characterizes the overall topography of the sample under test. By progressively upsampling the global topographic feature, a 3D height map of the sample surface is generated, thus achieving 3D reconstruction of the sample. By sequentially arranging the 3D reconstructed images corresponding to each sampling time, a video image reflecting the real-time dynamic topographic changes of the sample under test can be generated.

[0052] In summary, the in-situ tracking three-dimensional detection system proposed in this invention possesses high-temperature tolerance and a controllable atmosphere environment, making it suitable for in-situ experiments under extreme conditions. Using multi-azimuth probes (such as high-temperature four-quadrant probes), backscattered electron signals from multiple directions (e.g., four directions) of the sample can be simultaneously acquired during a single electron beam scan, generating backscattered electron images from multiple angles. This enables the capture of nanoscale morphological changes during chemical reactions, making it particularly suitable for short-cycle, high-dynamic nanomorphological monitoring in in-situ chemical reactions. Furthermore, this three-dimensional detection system can complete three-dimensional structural reconstruction without relying on traditional methods such as sample stage tilting or focused ion beam cutting. Whether it is a flat sample, thin sheet, or powder material, high-precision three-dimensional morphological reconstruction can be achieved using only limited shadow information, allowing for non-destructive in-situ analysis of samples. Moreover, traditional mathematical methods for three-dimensional reconstruction typically rely on rigorous geometric optical models and physical assumptions, limiting their ability to handle complex scenarios such as non-uniform illumination, noise interference, and complex surface textures. This 3D detection system also introduces a deep learning-driven 3D reconstruction mechanism. Compared with traditional mathematical methods for 3D reconstruction, this invention uses a neural network to learn from massive training data, thereby learning the nonlinear mapping relationship between shadow information in the image and the reconstructed 3D morphology. Even under conditions such as non-uniform lighting, noise interference, or complex textures, this 3D reconstruction model can still accurately extract the required local morphological features to achieve 3D reconstruction of the sample morphology. Therefore, this 3D reconstruction model has strong robustness and generalization ability, and does not require explicit modeling of the physical process, reducing errors caused by model assumptions. After training, the computation speed in the inference stage is very fast, and optimizing the network structure and hardware acceleration is relatively easy, making it particularly suitable for real-time or dynamic 3D reconstruction tasks. Therefore, this invention, through a high-temperature resistant multi-quadrant probe and deep learning technology, achieves real-time monitoring of the dynamic morphological changes of catalyst surfaces during high-temperature reactions. This invention has advantages such as high-temperature adaptability, high precision, non-destructive nature, and efficient computation, providing a powerful tool for surface science and catalysis research. The in-situ tracking three-dimensional detection system proposed in this invention can be widely used in fields such as surface science and catalysis science. It can achieve atomic-level high resolution and is used for in-situ monitoring of nanoscale morphological changes on surfaces during dynamic chemical reactions. It belongs to an innovative real-space in-situ high-precision sample surface morphology analysis device.

[0053] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. An in-situ tracking three-dimensional detection system, characterized in that, The system includes: An electron microscope, comprising an electron microscope cavity and a sample stage mounted within the electron microscope cavity; The detection device includes a drive assembly and a probe assembly; wherein, The probe assembly is movably disposed within the electron microscope cavity for acquiring backscattered electron signals of the sample under test from multiple directions; wherein the sample under test is placed on the sample stage; The driving component is used to drive the probe assembly to reciprocate between a preset initial position and a detection position; wherein the detection position corresponds to the sample stage; A three-dimensional reconstruction device, electrically connected to the probe assembly, is used to reconstruct the three-dimensional morphology of the sample under test based on backscattered electron signals collected from multiple directions.

2. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The probe assembly includes a clamp and a multi-azimuth probe detachably mounted on the clamp.

3. The in-situ tracking three-dimensional detection system according to claim 2, characterized in that, The probe is a metal electrode with a carbon coating.

4. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The driving component includes: The cylindrical body is connected to the probe assembly; A first driving mechanism, connected to the cylinder, is used to drive the probe assembly to reciprocate between the initial position and the detection position via the cylinder, so as to make a first magnitude adjustment to the distance between the probe assembly and the detection position.

5. The in-situ tracking three-dimensional detection system according to claim 4, characterized in that, The driving component also includes: The second driving mechanism is connected to the cylinder and is used to drive the cylinder to make a second amplitude adjustment of the probe assembly in multiple directions within a plane with the first driving mechanism as the normal vector; wherein the second amplitude is smaller than the first amplitude.

6. The in-situ tracking three-dimensional detection system according to claim 4, characterized in that, The cylinder is provided with a data cable sealing cavity for accommodating the data cable, and the probe assembly and the three-dimensional reconstruction device transmit backscattered electronic signals through the data cable.

7. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The three-dimensional reconstruction device includes: The data acquisition module is used to acquire backscattered electron signals from multiple directions of the sample under test and form corresponding backscattered electron images; wherein, the backscattered electron signals are obtained by scanning the sample under test with an electron beam using the probe assembly; The feature extraction module is used to input backscattered electron images from multiple directions into the feature extraction network of the three-dimensional reconstruction model. Based on the shadow information of the backscattered electron images, the module extracts the local morphological features of the corresponding backscattered electron images through stepwise downsampling. The feature extraction network is a convolutional neural network. The fusion module is used to input various local morphological features into the fusion network of the three-dimensional reconstruction model, and to perform weighted fusion of various local morphological features based on the attention mechanism to generate global morphological features that characterize the overall morphology of the sample under test. The 3D reconstruction module is used to input the global topography features into the prediction network of the 3D reconstruction model, and reconstruct the 3D topography of the sample to be tested by upsampling step by step; wherein the structure of the feature extraction network is symmetrical with the structure of the prediction network.

8. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The electron microscope further includes a gas supply unit connected to the electron microscope cavity, the gas supply unit being used to provide gas supply to the electron microscope cavity.

9. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The electron microscope further includes a laser heating stage disposed within the electron microscope cavity for heating the sample to be tested.

10. The in-situ tracking three-dimensional detection system according to claim 1, characterized in that, The detection system further includes a vacuum camera mounted on the sample stage, used to observe the position of the probe assembly and guide the probe assembly to align with the detection position.

Images