Method and system for reconstructing an image from electron diffraction data

Abstract

This paper provides a method for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - receiving electron diffraction data of the test sample (100), the electron diffraction data including a plurality of test event-detection counts (210a, 210b, 210c); - for each test event-detection count (210a, 210b, 210c), selecting a nuclear image (e.g., 310b, e, h) from a nuclear image library (300) including a plurality of nuclear images (e.g., 310a to i) (selected nuclear images (e.g., 310b, e, h)); wherein each nuclear image (e.g., 310a to i) in the nuclear image library (300) is a pre-calculated mosaic for adding to a reconstruction grid (400) of the reconstructed image (500), wherein the plurality of selected added mosaics are capable of stitching together and rendering the reconstructed image (500); and - Generate the reconstructed image (500), which includes adding each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500).

Description

Technical Field

[0001] This invention is in the field of image reconstruction based on electron diffraction data. Background Technology

[0002] Recent trends in the field of scanning transmission electron microscopy (STEM) have led to powerful methods for directly retrieving the electrostatic potential of solid-state samples. These methods include centroidal integrals of far-field intensity (iCoM) and ptychography, which were first pioneered in the field of X-ray microscopy. Electron ptychography exists in various forms, including iterative reconstruction schemes and analytical methods. Analytical ptychography itself includes sideband integrals (SBI), alternatively known as single-sideband (SSB) reconstruction, and Wigner distribution deconvolution (WDD).

[0003] The focus of those potential retrieval methods is their inherent dose efficiency. STEM experiments are often associated with a high probability of beam damage, which makes imaging dose-sensitive materials difficult. In that context, performing measurements with the most efficient use of each detected electron is crucial for extending known experimental methods to new scientific fields. One such field is structural biology, where less sophisticated TEM methods are now common, and their main experimental limitation is the high probability of beam damage. In that context, any breakthrough leading to an increase in the information recovered from each input electron could lead to significant advances.

[0004] Typically, both iCoM and stacked imaging require pre-recording a set of diffraction patterns corresponding to the transverse scanning grating of the converging electron probe. This experimental setup is often referred to as 4D-STEM because it results in resolution in both real and reciprocal spaces, and thus in four recording dimensions. One of the main challenges of such experiments is the readout speed of the detectors used. Even in 1-bit mode, the frame time is at most 80 µs for a Medipix3-type detector. Therefore, 4D-STEM is often time-intensive, making it susceptible to sample drift and beam distortion. This is partly why it is important to introduce faster options for both data acquisition and processing, even toward real-time imaging, as other groups have explored. The second challenge is the sheer volume of data generated as part of the experiment, which can reach tens of gigabytes in just a few seconds. Transmitting this data for further processing implies heavy demands in terms of data bandwidth and computing power, resulting in very expensive, specialized, and difficult-to-maintain hardware.

[0005] This invention aims to provide a novel framework for performing analytical overlay imaging, including SBI and WDD methods, but also extended to iCoM methods, with minimal input electron costs, shorter acquisition times, and lower data storage and transmission requirements. For this purpose, recent advances in event-driven electron detection are utilized. Summary of the Invention

[0006] This paper provides a method for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c). - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from a kernel image library (300) that includes multiple kernel images (e.g., 310b, e, h), where: - The kernel image (e.g., 310b, e, h) is pre-computed and is a mosaic of the reconstruction mesh (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to be mosaicking and rendering the reconstructed image (500). as well as - Generate reconstructed images (500), including: - Add each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500).

[0007] This paper provides a method for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c). - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from a kernel image library (300) including multiple kernel images (310a to i) - the selected kernel image (310b, 310e, 310h), where: - Each kernel image (e.g., 310a to i) in the kernel image library (300) is pre-computed and is a mosaic of the reconstruction grid (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). as well as - Generate reconstructed images (500), including: - Add each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500).

[0008] According to the preferred aspect, each nuclear image (310a to i) in the nuclear image library (300): - Has a dimension that is part of the size of the reconstructed mesh (400); - is based on the simulated counting grid vector position on the simulated counting grid of the simulated counting detector ( The simulation is based on a single simulated event-detection count.

[0009] Based on the preferred aspects: - The electron diffraction data also includes multiple different scan grid vector positions (100a, 100b, 100c, ...). Each scan grid vector position is the vector location of the scan grid (120). The position of the test electron probe that interacts with the test sample (100) at the test sample (100) and thereby generates the outgoing wave (120a) from the test sample (100); - Each of the multiple test event-detection counts (210a, 210b, 210c) has a test count grid vector position on the count grid (205) of the event-driven detector (200). ); - The test count grid vector position in the electron diffraction data ( The scan grid vector positions (100a, 100b, 100c) linked to the generated outgoing wave (120a) are... The emitted wave (120a) is located at the test count grid vector position ( The test event is triggered at (210a, 210b, 210c). - when( )and( When matching, select kernel images (310b, 310e, 310h). - Adding each selected kernel image (310b, 310e, 310h) to the reconstruction grid (400) is done at the reconstruction grid vector position ( At the set of ), the reconstructed grid vector position ( The set of ) is used to generate the scan grid vector position of the outgoing wave (120a) Centered on ), the outgoing wave (120a) causes nuclear images (310b, 310e, 310h) to be selected by the test event - detection count (210a, 210b, 210c).

[0010] This paper also provides a method for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive electron diffraction data of a test sample (100), the electron diffraction data including multiple test event-detection counts (210a, 210b, 210c), wherein the electron diffraction data includes: - Multiple different scan grid vector locations (100a, 100b, 100c, Each scan grid vector position is the vector location of the scan grid (120). The position of the test electron probe that interacts with the test sample (100) at the test sample (100) and thereby generates the outgoing wave (120a) from the test sample (100); - For each scan grid vector position ( At least one test event-detection count (210a, 210b, 210c) is generated by the interaction of the emitted wave (120a) with the event-driven detector (200), wherein each test event-detection count (210a, 210b, 210c) has a test count grid vector position on the count grid (205) of the event-driven detector (200). ); - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from the kernel image library (300), where: - The kernel image library (300) includes several pre-computed kernel images (310a to i), wherein: - Each of the plurality of nuclear images (310a to i): - is a mosaic used to add to the reconstruction mesh (400) of the reconstructed image (500), where multiple selected mosaics can be stitched together and rendered into the reconstructed image (500). - Has a dimension that is part of the size of the reconstructed mesh (400); - is based on the simulated counting grid vector position on the simulated counting grid of the simulated counting detector ( The simulation is based on a single simulated event-detection count; - when( )and( When matching, select kernel images (310b, 310e, 310h). - Generate reconstructed images (500), including: - Add each selected kernel image (310b, 310e, 310h) to the reconstruction grid (400) at the reconstruction grid vector location ( The set of reconstructed grid vector positions ( The set of data revolves around the scan grid vector positions of the generated test events - detection counts (210a, 210b, 210c) (on which kernel images (310b, 310e, 310h) are selected). Centered on ), the image (500) is gradually pieced together and rendered.

[0011] According to a preferred aspect, adding each kernel image (310b, 310e, 310h) to the reconstruction grid (400) includes the following steps: for each reconstruction grid vector location ( The individual pixel intensities of the added kernel images (310b, 310e, 310h) at the same reconstructed grid vector location are compared with those at the same location. The summation of the previous pixel intensities at () is performed.

[0012] According to the preferred aspect, at the location of the reconstructed grid vector ( At point ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400) if they are adjacent, overlapping, or superimposed, where: - Adjacent additions include positioning the selected kernel images (310b, 310e, 310h) next to the existing kernel images on the reconstructed mesh (400); - The addition of overlap involves positioning the selected kernel images (310b, 310e, 310h) such that one or more of their edges overlap with a portion of one or more existing kernel images on the reconstructed mesh (400); - The addition of overlays involves precisely positioning the selected kernel image (310b, 310e, 310h) on top of one or more existing kernel images on the reconstructed mesh (400).

[0013] The existing kernel image on the reconstruction mesh (400) is a selected kernel image previously added to the reconstruction mesh (400) during the gradual stitching and rendering of the reconstruction image (500).

[0014] According to the preferred aspect, after the selected kernel images (310b, 310e, 310h) are added to the reconstruction mesh (400), the reconstruction mesh becomes updated, and only a portion of the (selected) kernel images (310b, 310e, 310h) in the reconstruction mesh (400) becomes updated.

[0015] According to a preferred aspect, the size of each kernel image (310a to i) in the kernel image library (300) is equal to or less than 50% of the size of the reconstructed mesh (400) in terms of area.

[0016] According to a preferred aspect, based on the simulated counting grid vector position ( The simulation of each nuclear image (310a to i) in the image library (300) includes analysis of the stacked imaging.

[0017] Based on the preferred aspects: - Analysis of stacked imaging involves using integral centroid (iCOM), sideband integral (SBI), or Wigner distribution deconvolution (WDD); and - (The simulation of) nuclear images (310a to i) in the image library (300) represents the interaction constant ( Weighted electrostatic potential ( The diagram is composed of the same interaction constant ( Weighted charge density ( The real or imaginary part of the graph, or the transmission function (T) graph, is measured, where the transmission function is determined by the interaction constant (T). Weighted electrostatic potential ( The complex index of ).

[0018] According to the preferred aspect, the generation of the reconstructed image (500) is performed in real time.

[0019] A method is further provided for interactively generating a reconstructed image (500) based on electron diffraction data of a test sample (100), including the method described herein, wherein: - The method also includes sending instructions to an electron microscope (EM) to direct the test electron probe to the scanning grid vector positions (100a, 100b, 100c, ...). ), which causes the electron probe to interact with the test sample (100) and generate an outgoing wave (120a). - In the following steps: Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test event-detection counts (210a, 210b, 210c) caused by the generated outgoing wave (120a). The electron diffraction data of the test sample (100) were generated by electron microscopy (EM); - In the following steps: Each selected kernel image (310b, 310e, 310h) is added to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500). At the location of reconstructing the grid vector ( At the set of ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position ( The set of ) is used to generate the scan grid vector position of the outgoing wave (120a) Centered on ), the outgoing wave (120a) causes nuclear images (310b, 310e, 310h) to be selected by the test event - detection count (210a, 210b, 210c). - The method also includes: Repeat the sending, receiving, selecting, and adding process, wherein for at least one repetition, different scan grid vector positions (100a, 100b, 100c, ...) are used. ), thereby gradually stitching together and rendering the reconstructed image (500).

[0020] Furthermore, an (interactive) method is provided for interactively generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Send instructions to the electron microscope (EM) to guide the test electron probe to the scanning grid vector positions (100a, 100b, 100c, ...). ), which causes the electron probe to interact with the test sample (100) and generate an outgoing wave (120a). - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c) caused by the generated outgoing wave (120a). - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from a kernel image library (300) that includes multiple kernel images (310a to i), where: - Each kernel image (e.g., 310a to i) in the kernel image library (300) is a pre-computed mosaic of a reconstruction grid (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). - At the location of reconstructing the grid vector ( At the set of ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position ( The set of ) is used to generate the scan grid vector position of the outgoing wave (120a) Centered on ), the outgoing wave (120a) causes nuclear images (310b, 310e, 310h) to be selected by the test event-detection count (210a, 210b, 210c). - Repeat the sending, receiving, selecting, and adding process, wherein for at least one repetition, different scan grid vector positions (100a, 100b, 100c, ...) are used. ), thereby gradually stitching together and rendering the reconstructed image (500).

[0021] According to a preferred aspect, the (interactive) method is based on any aspect of an additional subject matter incorporated into the method described herein for generating a reconstructed image (500) from electron diffraction data of a test sample (100).

[0022] The method described here is a computer-implemented method.

[0023] A computing device or system configured to perform a method according to the (computer-implemented) method described herein is further provided.

[0024] A computer program or computer program product having instructions that, when executed by a computing device or system, cause the computing device or system to perform the (computer-implemented) method as described herein.

[0025] A further provision provides a computer-readable medium having a computer program stored thereon, the computer program having instructions that, when executed by a computing device or system, cause the computing device or system to perform the (computer-implemented) method as described herein. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the methods and systems described in this article. Detailed Implementation

[0027] Before describing the present systems and methods, it should be understood that the invention is not limited to the specific systems, methods, or combinations described, as such systems, methods, and combinations can certainly vary. It should also be understood that the terminology used herein is not intended to be limiting, as the scope of the invention will be limited only by the appended claims.

[0028] As used herein, the singular forms “a,” “one,” and “the” include both the singular and plural indicators, unless the context clearly indicates otherwise.

[0029] As used herein, the terms “comprising,” “comprises,” and “comprised of” are synonymous with “including,” “includes,” “containing,” and “contains,” and are inclusive or open-ended, not excluding additional, uncited components, elements, or method steps. It should be understood that, as used herein, the terms “comprising,” “comprises,” and “comprised of” encompass the terms “consisting of,” “consists,” and “consists of.”

[0030] The reference to the numerical range given by the endpoints includes all numbers and fractions covered in the corresponding range, as well as the referenced endpoints.

[0031] The terms “approximately” or “about” used herein when referring to measurable values ​​such as parameters, quantities, and durations of time are intended to cover deviations from a specified value of + / -10% or less, preferably + / -5% or less, more preferably + / -1% or less, and even more preferably + / -0.1% or less, provided that such deviations are suitable for implementation in the disclosed invention. It should be understood that the values ​​referred to by the modifiers “approximately” or “about” are themselves specifically and preferably disclosed.

[0032] While the terms “one or more” or “at least one” such as “one or more of a set of components” are self-explanatory, by way of further example, the term in particular covers a reference to any one of the components, or to any two or more of the components, such as, for example, any ≥3, ≥4, ≥5, ≥6 or ≥7 of the components, up to all of the components.

[0033] All references cited in this specification are incorporated herein by reference in their entirety. In particular, the teachings of all references specifically mentioned herein are incorporated herein by reference.

[0034] Unless otherwise defined, all terms (including technical and scientific terms) used in disclosing this invention have the meanings commonly understood by one of ordinary skill in the art to which this invention pertains. Further guidance, including terminology definitions, is provided to better understand the teachings of this invention.

[0035] In the following paragraphs, different aspects of the invention are defined in more detail. Each aspect thus defined may be combined with any other one or more aspects unless expressly indicated otherwise. In particular, any feature indicated as preferred or advantageous may be combined with any other one or more features indicated as preferred or advantageous.

[0036] Throughout this specification, references to "an embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of the invention. Therefore, the phrases "in one embodiment" or "in an embodiment" appearing in various places throughout this specification do not necessarily all refer to the same embodiment, but may refer to the same embodiment. Furthermore, as will be apparent to those skilled in the art based on this disclosure, particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Moreover, while some embodiments described herein include some, but not all, of the other features included in other embodiments, it is understood by those skilled in the art that combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, any claimed embodiment may be used in any combination in the appended claims.

[0037] In the present description of the invention, reference is made to the accompanying drawings, which form a part of this document, and which illustrate specific embodiments in which the invention may be practiced only by way of example. Parenthetical or bold reference numerals appended to corresponding elements illustrate elements only by way of example and are not intended to limit the corresponding elements. Unless otherwise indicated, all figures and drawings in this document are not drawn to scale and have been selected for the purpose of illustrating different embodiments of the invention. In particular, the dimensions of various components are described in illustrative terms only, and unless so indicated, relationships between the dimensions of the various components should not be inferred from the drawings.

[0038] It should be understood that other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the invention. Therefore, the following detailed description should not be construed in a limiting sense, and the scope of the invention is defined by the appended claims.

[0039] &&core

[0040] This article provides a method or system for generating a reconstructed image (500) based on electron diffraction data of a test sample (100). Figure 1A flowchart illustrating the method is presented. The method or system receives electron diffraction data of a test sample (100) collected by an event-driven detector (200) and progressively stitches together and renders a reconstructed image (500) on a reconstruction grid (400) using multiple pre-computed nuclear images (310b, e, h), wherein each pre-computed nuclear image (310b, e, h) is selected from a nuclear image library (300) based on the count grid vector position of detected events (210a, 210b, 210c) on the count grid (210) of the event-driven detector (200). ). The vector positions of the selected kernel image (310 b, e, h) on the reconstructed grid (400) ) is based on the position of the scan grid vector ( The location of the scan grid vector is determined by ( ). ) is the location where the test electron probe (105) interacts with the test sample (100) on the scanning grid (120).

[0041] Therefore, according to one aspect, a method is provided for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c). - For each test event - detection count (210a, 210b, 210c), a kernel image (e.g., 310b, e, h) is selected from a kernel image library (300) that includes multiple kernel images (e.g., 310a to i), where: - The kernel image (e.g., 310a to i) of the kernel image library (300) is pre-computed and is a mosaic of the reconstruction grid (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). - Generate reconstructed images (500), including: - Add each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500).

[0042] Based on the preferred aspects: - Each kernel image (310a to i) in the kernel image library (300): - Has a dimension that is part of the size of the reconstructed mesh (400); - is based on the simulated counting grid vector position on the simulated counting grid of the simulated counting detector ( The simulation is based on a single simulated event-detection count; Based on the preferred aspects: - The electron diffraction data also includes multiple different scan grid vector positions (100a, 100b, 100c, ...). Each scan grid vector position is the vector location of the scan grid (120). The position of the test electron probe that interacts with the test sample (100) at the test sample (100) and thereby generates the outgoing wave (120a) from the test sample (100); - Each of the multiple test event-detection counts (210a, 210b, 210c) has a test count grid vector position on the count grid (205) of the event-driven detector (200). ); - The test count grid vector position in the electron diffraction data ( The scan grid vector positions (100a, 100b, 100c) linked to the generated outgoing wave (120a) are... The emitted wave (120a) is located at the test count grid vector position ( The test event is triggered at (210a, 210b, 210c). - when( )and( When matching, select kernel images (310b, 310e, 310h). - Adding each selected kernel image (310b, 310e, 310h) to the reconstruction grid (400) is done at the reconstruction grid vector position ( At the set of ), the reconstructed grid vector position ( The set of ) is used to generate the scan grid vector position of the outgoing wave (120a) Centered on ), the outgoing wave (120a) causes nuclear images (310b, 310e, 310h) to be selected by the test event - detection count (210a, 210b, 210c).

[0043] In more detail, a method is provided for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive the electron diffraction data of the test sample (100), wherein the electron diffraction data includes: - Multiple different scan grid vector locations (100a, 100b, 100c, Each scan grid vector position is the vector location of the scan grid (120). The position of the test electron probe that interacts with the test sample (100) at the test sample (100) and thereby generates the outgoing wave (120a) from the test sample (100); - For each scan grid vector position ( At least one test event-detection count (210a, 210b, 210c) is generated by the interaction of the emitted wave (120a) with the event-driven detector (200), wherein each test event-detection count (210a, 210b, 210c) has a test count grid vector position on the count grid (205) of the event-driven detector (200). ); - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from the kernel image library (300), where: - The kernel image library (300) includes several pre-computed kernel images (310a to i), wherein: - Each of the plurality of nuclear images (310a to i): - is a mosaic used to add to the reconstruction mesh (400) of the reconstructed image (500), where multiple selected mosaics can be stitched together and rendered into the reconstructed image (500). - Has a dimension that is part of the size of the reconstructed mesh (400); - is based on the simulated counting grid vector position on the simulated counting grid of the simulated counting detector ( The simulation is based on a single simulated event-detection count; - when( )and( When matching, select kernel images (310b, 310e, 310h). - Generate reconstructed images (500), including: - At the location of reconstructing the grid vector ( At the set of ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position ( The set of ) is used to generate test events - detection counts (210a, 210b, 210c) (on which kernel images (310b, 310e, 310h) are selected) scan grid vector positions ( Centered on ), the image (500) is gradually pieced together and rendered.

[0044] exist Figure 1The image shows a portion of the reconstructed mesh (400). This portion is represented as a dashed box on the reconstructed image (500).

[0045] In typical existing technology methods or systems, diffraction patterns are recorded using a 2D full-frame imaging sensor that reads out a full-frame 2D image of the diffraction pattern (2D-FF-DP, 2D full-frame diffraction pattern). This is done to obtain the diffraction pattern at different scan grid vector positions ( The results of multiple 2D-FF-DP transformations obtained at [location] can determine the reconstructed image. However, this determination is typically performed offline using multiple stored 2D-FF-DP transformations and requires one or more high-power processors to perform the necessary burden of transforming each 2D-FF-DP transformation and combining them into the reconstructed image. Processing time typically ranges from less than an hour to a full day. Typical storage requirements are high (e.g., 2 to 32 GB) due to the full-frame readout.

[0046] By creating a library (300) of pre-computed kernel images (310a to i), the reconstructed image can be created in real time during the acquisition process, since most of the necessary computations have been performed in advance. Therefore, computation time is significantly reduced, and the final reconstructed image becomes available at the end of the acquisition.

[0047] The advantage of this type of real-time reconstruction is that it allows the test electron probe (105) to be redirected to areas of the scan grid (120) that produce poor-quality data (e.g., due to sample thickness) for local improvement. Therefore, interactivity or responsiveness allows for real-time adjustments that are impossible in offline scenarios.

[0048] This method and system also allow for reduced beam impairment because acquisition can be stopped once the reconstructed image has sufficient quality, and / or only local regions of samples requiring additional data can be exposed to longer beam dwell times. In contrast, in existing technologies, samples may be illuminated across all regions for longer than necessary to avoid data shortages.

[0049] Since reconstruction primarily involves adding images / mosaic elements, simplified processors such as Field Programmable Gate Arrays (FPGAs) can be used. This reduces hardware costs, and the high processing speed of FPGAs enables the real-time image reconstruction mentioned above, providing valuable real-time feedback to the user.

[0050] Because the reconstruction grid (400) effectively acts as a storage device for each selected kernel image (e.g., 310b, e, h), there is no requirement to store any data associated with the event-driven detector (200). In particular, there is no need to store records of detected events (210a, 210b, 210c). Because an event-driven detector is used, regular full-frame images are not generated, which significantly reduces data handling requirements. Since each detected event (210a, 210b, 210c) is used to select a pre-computed kernel image (e.g., 310b, e, h from 310a to i) added to the reconstruction grid (400), there is no need to store any intermediate reconstructed images; as mentioned elsewhere, the reconstruction grid (400) effectively acts as a storage device. Therefore, data storage requirements, apart from the reconstruction grid (400) and / or the reconstructed images, are reduced or essentially eliminated.

[0051] According to one aspect, a method is provided for interactively generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Send commands to the EM to direct the test electron probe to the scan grid vector positions (100a, 100b, 100c, ... ), which causes the electron probe to interact with the test sample (100) and generate an outgoing wave (120a). - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c) caused by the generated outgoing wave (120a). - For each test event - detection count (210a, 210b, 210c), a kernel image (e.g., 310b, e, h) is selected from a kernel image library (300) that includes multiple kernel images (e.g., 310a to i), where: - The kernel image (e.g., 310b, e, h) is a pre-computed mosaic of the reconstruction mesh (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). - At the location of reconstructing the grid vector ( At the set of ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position ( The set of ) is used to generate the scan grid vector position of the outgoing wave (120a) Centered on ), the outgoing wave (120a) causes nuclear images (310b, 310e, 310h) to be selected by the test event-detection count (210a, 210b, 210c). - Repeat (multiple times) the sending, receiving, selecting, and adding, wherein for at least one repetition (preferably for multiple repetitions), different scan grid vector positions (100a, 100b, 100c, ...) are used or sent. ), thereby gradually stitching together and rendering the reconstructed image (500).

[0052] The comparative steps described in this paper for the method of generating reconstructed images (500) can be applied to the method of interactively generating reconstructed images (500) after making necessary modifications.

[0053] Optionally, the reconstructed mesh (400) is displayed on a display (such as a computer monitor, a display of a smart device). Optionally, the method can receive different scan mesh vector positions (100a, 100b, 100c, ...). The input is displayed and the display allows the user or machine to interactively monitor the progress of the reconstructed mesh (400) generation.

[0054] If defects or areas requiring improvement are observed at one or more locations on the reconstructed grid (400), the test electron probe (205) can be moved to the corresponding scan grid vector locations (100a, 100b, 100c, ...). This adds additional kernel images or mosaics to the defect locations or areas requiring improvement in the reconstructed grid (400). This means that only the areas requiring additional data collection are illuminated, thereby reducing beam damage. In the prior art, beam scanning of the entire sample area is performed when defects are shown in the area of ​​the reconstructed grid (400), which equally damages all areas and causes distortion of the results acquired subsequently.

[0055] According to one aspect, the reconstructed grid (400) identified from the display shows the location of defects or areas requiring improvement, and the provided input directs the test electron probe (205) to the scan grid vector position (100a, 100b, 100c) corresponding to the defect location or area requiring improvement in the reconstructed grid (400). It should be understood that the reconstructed mesh (400) showing locations of defects or areas needing improvement can be identified automatically (e.g., using machine learning protocols) or based on user input.

[0056] Electron diffraction data are acquired using an electron microscope (EM). An EM can be any device that emits an electron beam (test electron probe (105)) capable of generating ED data for a test sample. Examples of EMs are scanning electron microscopes (SEMs) or scanning transmission electron microscopes (STEMs). Examples of EM providers include Zeiss, Tescan, JEOL, Hitachi, and Thermo Fisher Scientific. Standard methods for acquiring ED data for a sample are known in the art, for example, using standard protocols for EMs, and / or as taught in textbooks such as Williams DB and Carter CB, “Transmission Electron Microscopy,” Springer, New York, NY, 2009.

[0057] The test electron probe (105) is preferably a convergent beam, thereby generating a convergent beam electron diffraction (CBED) pattern on the counting grid (205) of the event-driven detector.

[0058] Typically, EM includes: - Electronic source; - Focusing unit, configured to form electrons into a test electron probe (105); - A deflector unit is configured to adjust the position of the test electron probe (105) on the test sample (100) or the scan grid (120); - One or more event-driven detectors are configured to detect electrons generated by the interaction between the test electron probe (105) and the test sample on the scanning grid (120). The one or more event-driven detectors (200) are positioned behind the test sample for collecting ED data; - Controller, configured to control the EM; - Memory, configured to store computer-readable instructions for instructing the controller to operate the EM in accordance with this disclosure; - A test sample stage for adjusting the position of the sample or scanning grid (120) relative to the beam; and - Vacuum chamber and attached vacuum pump and vacuum sensor.

[0059] An event-driven detector (200) detects events generated by the interaction between the test electron probe (105) and the test sample (110) on the scan grid (120), thereby generating a test event-detection count (TEDC). TEDC events originate from the scan grid vector position of the test electron probe (105) and the sample (100) on the scan grid (120). The interaction at ( ). More specifically, TEDC is the interaction of the outgoing wave of electrons emitted after the test sample (100) at (one or more) counting grid vector positions ( ). The collision counting grid (210) is generated at the location. TEDC is created only when an event exists, in which case it is assigned the counting grid vector position ( ). TEDC does not have a gradient amplitude and can be represented as "one" or "yes" or other binary indications. As understood in the art, the event-driven detector (200) does not output frame-based readouts. Regions of the counting grid (210) not struck by the outgoing wave do not form part of the output of the event-driven detector (200). The event-driven detector (200) outputs event-based readouts. Only regions of the counting grid (210) struck by the outgoing wave form part of the output of the event-driven detector (200). The counting grid (210) is located after the test sample (110). Any suitable current or future event-driven detector can be used. Event-driven detectors are known in the art. Examples of providers include Advacam, Amsterdam Scientific Instruments, and QuantumDetectors.

[0060] The test electron probe (205) is a converging focused beam that creates a converged beam electron diffraction (CBED) pattern on the counting grid (205) of the event-driven detector.

[0061] The test electron probe moves over the scanning grid (120) in a scanning pattern, and at each scanning grid vector position ( (100a, 100b, 100c) stayed for a period of time, and during the period of time, one or more test events - detection counts (TEDC) were detected by an event-driven detector (200).

[0062] After obtaining multiple different scanning locations ( When analyzing ED data at a location, the coils of the deflector unit controlling the lateral positioning of the test electron probe (105) are typically positioned within the scan pattern. Types of scan patterns include raster scans, serpentine pattern scans, random positioning scans, or user-controlled scans. A raster scan means that the test electron probe (105) is controlled to move across the sample in a parallel sweep array. A user-controlled scan means that the user is free to move the scan test electron probe (105) in any direction. User-controlled scans allow the user to "paint" onto the reconstructed grid in real-time sweeps; areas with poor resolution can receive repeated sweeps by the user-controlled test electron probe (105). The test electron probe (105) is preferably positioned at two spatially separated (non-adjacent) different scan grid vector locations (100a, 100b, 100c, ...). The hidden area was eliminated during the repositioning between the two.

[0063] The ED data includes multiple test event-detection counts (TEDCs), each TEDC being represented as at least one count grid vector position on the count grid (210) of the event-driven detector (210). (x,y coordinates) and the location of the scan grid vector where TEDC occurred ( The TEDC does not have a gradient amplitude and can be represented as "one" or "yes" or other binary indicators, or simply by the presence of at least one x,y coordinate in the ED data. The arrival time of the event on the event-driven detector (200) does not need to be recorded or exist in the ED data. The TEDC is created only when the event exists, in which case it is assigned a counting grid vector position ( ).

[0064] Detection events are typically located on the counting grid (210) at the counting grid vector position ( The activation of a pixel at position ( ). The anticipated detection event could be on the counting grid (210) including the counting grid vector position ( ). A cluster of adjacent (next) pixels.

[0065] ED data is typically generated at a beam energy suitable for generating ED data. The beam energy can have an accelerating voltage of 1 to 300 keV, preferably 30 to 300 keV. Since this disclosure allows for the collection of ED data using lower beam energies, EMs can be able to provide EM beams with energies of 1 to 30 keV. Each sample region can be used... Receive 50 to 10 6 Meaningful ED datasets are obtained by using the beam dose of individual electrons. This beam dose is typically achieved through a dwell time of 1 µs to 1 ms per scan point.

[0066] Scan grid vector positions (100a, 100b, 100c, The grid vectors (100a, 100b, 100c, ...) can be different and are only accessed once by the test electron probe (105). The grid vector positions (100a, 100b, 100c, ...) are scanned. The points can be different and visited at least twice by the test electron probe (105). Multiple visits at each visited point at a lower dose result in a reduction in beam damage compared to only one beam visit per point and a higher dose per visited point.

[0067] The reconstruction grid vector position of each selected kernel image (310b, 310e, 310h) on the reconstruction grid (400) The set of ) is added to the reconstructed mesh (400). Reconstructed mesh vector position ( The vector position of the set represents a continuous group of pixels (e.g., a square box) with dimensions corresponding to (e.g., the same or scaled up or down) the selected kernel image (310b, 310e, 310h). Adding or supplementing means combining the pixel intensity from the selected kernel image (310b, 310e, 310h) with the existing pixel intensity on the reconstruction grid (400) at the position where the selected kernel image (310b, 310e, 310h) is placed on the reconstruction grid (400) (e.g., by summation).

[0068] Add each selected kernel image (310b, 310e, 310h) progressively, stitching together and rendering the reconstructed image (500). Stitching, or performing stitching, refers to stitching together different and sometimes the same reconstructed grid vector positions over time (...). The process involves adding multiple selected kernel images (310b, 310e, 310h) (mosaic) to the reconstructed image (500) and gradually rendering it. The kernel image (or mosaic) has a size that is a fraction (smaller part) of the size of the reconstructed grid (400). For example, the kernel image (or mosaic) has an area that is 50% or less of the area of ​​the reconstructed grid (400). This area ratio depends on both the kernel image size and the number of scan points accessed (e.g., the total size of the scan window).

[0069] The nuclear image size is preferably user-defined. Preferably, it represents the region in the beam-exposed sample considered to participate in the scattering of electrons by the electron probe. Therefore, it is a measure of the choice of the real-space extension of the electron probe. It can be expressed in nanometers or pixels. The nuclear image size is typically from 9x9 to 41x41 pixels.

[0070] The addition to the reconstruction grid (400) can be initial, which means that the selected kernel image (310b, 310e, 310h) or mosaic on the reconstruction grid (400) occupies a position that the existing kernel image does not occupy and is not adjacent to, overlap with or superimpose on the existing kernel image.

[0071] The existing kernel image on the reconstruction mesh (400) is a selected kernel image previously added to the reconstruction mesh (400) during the gradual stitching and rendering of the reconstruction image (500).

[0072] Adding to the reconstruction grid (400) can be adjacent, meaning that the selected kernel image (310b, 310e, 310h) or mosaic is positioned next to an existing kernel image on the reconstruction grid (400) (without overlap).

[0073] The additions to the reconstructed mesh (400) may be additionally or alternatively overlapping, meaning that the selected kernel image (310b, 310e, 310h) or mosaic is positioned such that one or more edges of its edges overlap with a portion of one or more existing kernel images on the reconstructed mesh (400).

[0074] The addition to the reconstruction grid (400) can be additionally or alternatively overlaid, meaning that the selected kernel image (310b, 310e, 310h) or mosaic is precisely positioned on top of one or more existing kernel images on the reconstruction grid (400). For example, in Figure 1 In the process, three kernel images (310b, 310e, 310h) are added to the reconstructed mesh (400) by overlay.

[0075] After a selected kernel image (310b, 310e, 310h) or mosaic is added to the reconstruction mesh (400), the reconstruction mesh becomes updated. Only a portion of the reconstruction mesh (400) containing the added kernel image (310b, 310e, 310h) or mosaic can become updated. Multiple selected added mosaics can be stitched together and rendered into a reconstructed image (500).

[0076] The selected kernel images (310b, 310e, 310h) are added to the reconstruction grid vector positions of the reconstruction grid (400). The location is based on the scan grid vector position. The location of the scan grid vector is determined by ( ). ) is the location where the test electron probe (105) interacts with the test sample (100) on the scan grid (120). Typically, the reconstructed grid vector location ( ) is based on the position of the scan grid vector ( The proportional relationship is determined by the scale, and optionally allows for boundaries (e.g., blank or black) within the reconstructed mesh (400). This proportional relationship can be expressed according to Equation 1: [Equation 1] in: ( The location of the reconstructed mesh vector to be determined; ( ) is the scan grid vector position; It is a scaling factor constant (scaling can be either magnification or reduction). It is one when scaling is not present.

[0077] It is an offset constant, which optionally allows for the existence of boundaries. It is zero when no boundaries exist.

[0078] By adding pixels and stitching together multiple selected kernel images (310b, 310e, 310h), the signal-to-noise ratio of the reconstructed mesh (400) and the reconstructed image (500) is increased.

[0079] The kernel image library (300) includes multiple pre-computed kernel images (310a to i). Each kernel image (310a to i) is a mosaic of the reconstruction grid (400) used to add to the reconstructed image (500). Each kernel image (310a to i) is based on a simulated counting grid position on a simulated counting grid of a simulated counting detector. The simulation is based on a single simulated event-detection count.

[0080] Each kernel image (310a to i) in the kernel image library (300) is compared with the position of the simulated counted grid vector of the kernel image (310a to i) or mosaic. (320a to 320i) are linked. Therefore, by using the test count grid vector position ( ) Query the kernel image library (300), and you can select the link to the equivalent simulated counting grid vector location ( The kernel image (310a to i) of ). Therefore, when ( )and( When matching, select kernel images (310b, 310e, 310h).

[0081] Nuclear images (310a to i) can be represented by the interaction constant ( Weighted electrostatic potential ( The diagram is composed of the same interaction constant ( Weighted charge density ( The real or imaginary part of the transmission function (T) plot is measured. This transmission function is determined by the interaction constant (T). Weighted electrostatic potential ( The complex index of ).

[0082] Pre-calculation means that the nuclear image library (300) is generated before the acquisition of electron diffraction data or before the data collection session. "Before" could be, for example, within a few days (e.g., 1 to 3) days, hours, or minutes prior to the acquisition of the electron diffraction data. The nuclear image library (300) is sample-independent and therefore can be retrieved from a database of stored nuclear image libraries (300). "Before" could be, for example, longer than a few days (e.g., 1 to 3) days, hours, or minutes prior to the acquisition of the electron diffraction data.

[0083] Each nuclear image (310a to i) in the nuclear image library (300) is based on a simulated counting grid vector position ( The simulation is based on a single simulation count. This simulation includes the use of analytical overlay imaging, the definition of which is extended herein to include the iCoM method. The simulation does not require scanning grid vector locations (100a, 100b, 100c, ...). ).

[0084] Analyzing stacked imaging preferably involves one of the following: integral centroid (iCoM), sideband integral (SBI) (alternatively referred to as single-sideband (SSB) reconstruction), or Wigner distribution deconvolution (WDD). These methods are known in the art.

[0085] iCOM's first time in Lazi In Ultramicroscopy, Vol. 160, January 2016, pp. 265-280 (…). https: / / doi.org / 10.1016 / j.ultramic.2015.10.011 It is described in “Phase contrast STEM for thinsamples: Integrated differential phase contrast”. It is important to note that this method differs from integrated differential phase contrast (iDPC), which is also being described for the first time in this publication.

[0086] SBI first appeared in Rodenburg et al., Ultramicroscopy, Vol. 48, No. 3, March 1993, pp. 304-314. https: / / doi.org / 10.1016 / 0304-3991(93)90105-7 As described in "Experimental tests on double-resolution coherent imaging via STEM".

[0087] WDD was first published in Rodenburg and Bates, Philosophical Transactions of the Royal Society of London. Series A: Physical and Engineering Sciences, June 15, 1992, Vol. 339, No. 1655. https: / / doi.org / 10.1098 / rsta.1992.0050 ) described in "The theory of super-resolution electron microscopy via Wigner-distribution deconvolution".

[0088] As defined elsewhere in this document, each nuclear image (310a to i) can be represented by the interaction constant ( Weighted electrostatic potential ( The diagram is composed of the same interaction constant ( Weighted charge density ( The real or imaginary part of the graph, or the transmission function (T) graph, is measured. Accordingly, the reconstructed image (500) can be derived from the interaction constant ( Weighted electrostatic potential ( ) diagram, based on the same interaction constant ( Weighted charge density ( The real or imaginary part of the transmission function (T) plot of a sample or a graph of the transmission function (T).

[0089] To generate nuclear images (310a to i) in the nuclear image library (300), the usual formal system for iCoM, SBI, and WDD methods is extended by describing the simulated diffraction data as a sum of Dirac functions (one for each simulation). The existence of those Dirac functions allows for mathematical simplification of integrals and Fourier transforms that would otherwise have to be performed explicitly on the complete four-dimensional electron diffraction dataset. Thus, each analytical stack-up imaging process is simplified to the summation of nuclear images, each pre-computed and reduced in real space by a given nuclear image size, thereby allowing for fast processing and saving memory bandwidth.

[0090] The simulation of each nuclear image preferably uses one or more of the following (preferably all of them): - One or more recording condition parameters (accelerating voltage) U (in kV), half convergence angle (in mrad), scan step size (in nm); - The counting grid (205) on the event-driven detector (200) Coordinate calibration.

[0091] - Desired core size (in nm); - Wiener number (For example, smaller numbers used to avoid division by zero, typically in the range of 10) -4 Up to 10 -8 (It takes values ​​between 0 and 1, and serves as a noise filtering parameter). Other optional parameters can be selected for bandpass filtering, aberration correction, incoherence compensation, contrast transfer function (CTF) compensation, and modulation transfer function (MTF) compensation.

[0092] The image kernel library (300) is independent of the sample.

[0093] The kernel image library (300) is stored in memory accessible to a standard computer system configured to generate a reconstructed image (500). The standard computer system is configured to perform one or more methods described herein. The standard computer system may be, for example, a computer system 2 based on the Intel IA-32 architecture. (One or more) methods may be implemented as programming instructions for one or more software modules stored on a non-volatile (e.g., hard disk or solid-state drive) storage device associated with the corresponding computer system. However, as mentioned elsewhere herein, at least some steps, particularly selecting the kernel image and / or generating the reconstructed image, may alternatively, in part or in whole, be implemented as, for example, one or more dedicated hardware components, such as gate configuration data for one or more field-programmable gate arrays (FPGAs), or implemented as application-specific integrated circuits (ASICs).

[0094] After an event has been detected by the event-driven detector (200), the test count grid vector location of the event is queried against the kernel image library (300). Select the location of the equivalent simulated counting grid vector from the kernel image library (300). The kernel images (310a to i) are selected. The selected kernel images (310a to i) are added to the reconstructed mesh (400).

[0095] The method or system described herein can be used to generate reconstructed images (500) in real time based on electron diffraction data. Real time means that the detection of events (210a, 210b, 210c) on the counting grid (210) of an event-driven detector (200) triggers the selection of a nuclear image (e.g., 310b, e, h) from a nuclear image library (300). The selection of the nuclear image (e.g., 310b, e, h) occurs before the test electron probe (105) has fully scanned the scanning grid (120). Preferably, the selection of the nuclear image (e.g., 310b, e, h) occurs when the test electron probe (105) moves to the next vector position on the scanning grid (120). Before that. Real-time can mean that the ED data generated during the data collection session is not stored in a storage device such as a hard drive. It is anticipated that multiple detected events (210a, 210b, 210c) during real-time processing can be buffered, for example, held in a temporary storage device with limited capacity. Real-time generation allows the identification of one or more locations in the reconstructed grid (400) that show defects or need improvement during the acquisition of ED data, thereby allowing the test electron probe (205) to be moved to the corresponding scan grid vector location (100a, 100b, 100c, ... ), used to add additional kernel images or mosaics to defect locations or desired locations.

[0096] As mentioned elsewhere in this document, each kernel image (310a to i) is based on the position of the simulated count grid vector on the simulated count grid of the simulated count detector ( The simulation is based on a single simulated event-detection count. The following is a detailed description of generating pre-computed kernel images (310a to i) for the kernel image library (300).

[0097] The simulated diffraction dataset contains a set of simulated event-detection counts, which are determined by the spatial frequencies located on the simulated counting detector. The pixel at that location is positioned in the simulated scan. For example, the set of indices d for each index c was detected.

[0098] This indicates the aperture inserted into the focal plane of the microscope. For conditions satisfying... spatial frequency It equals 1, otherwise it equals 0. It is by The given spatial frequency limit, where It is the semi-convergence angle of the illumination, and It is the electron wavelength, which itself depends on the accelerating voltage U.

[0099] This is the kernel constraint, which includes the Hann window. The constraint radius, or kernel radius, is selected by the user and expressed in nm. This distance is directly converted to the number of pixels for the kernel image size.

[0100] This is the bandpass filter window. The frequency limit is selected by the user. This bandpass filter window can alternatively be replaced by a frequency weighting function, for example, by assigning it a value between 0 and 1 and directly dependent on the spatial frequency. The value of .

[0101] This is the Wiener parameter, which is selected by the user.

[0102] p is the CTF compensation parameter, which is between 0 and 1 and is selected by the user.

[0103] The effective source size to be compensated (spatial incoherence correction). The caustics that need to be compensated (time incoherence correction). and Provided by the user as a single quantity (expressed in nm).

[0104] This is the aberration function to be corrected. It is given by the following formula:

[0105] Where parameters (expressed in nm) and (Expressed in radians) A specific pair of integers provided by the user. supply. It is spatial frequency The angle is facing, and It is its model.

[0106] This represents the camera's modulation transfer function. It is provided by the user if correction is desired. Otherwise, set it to 1.

[0107] and These represent the Fourier transform and the inverse Fourier transform, respectively. Arbitrary function. The Fourier transform is given by the following equation:

[0108] And any function The inverse Fourier transform is given by the following equation:

[0109] Integral centroid (iCOM)

[0110] electrostatic potential:

[0111] Charge density:

[0112] in:

[0113] CTF Compensation Items:

[0114] Sideband integral (SBI)

[0115] electrostatic potential:

[0116] Charge density:

[0117] According to one aspect:

[0118] More preferably: and It is a double-overlapping function, given by the following equation: and The irradiation correction term can be given by the following formula:

[0119] More preferably, and The irradiation correction term is given by the following formula:

[0120] CTF Compensation Items:

[0121] Wigner distribution deconvolution (WDD)

[0122] Virtual part:

[0123] Real part:

[0124] in

[0125] Irradiation convolution term:

[0126] Irradiation correction items can be:

[0127] More preferably, the irradiation correction item is:

[0128] After the initial calculation using the formula given above, the library... , , , , and It can be used to generate images through processes described elsewhere in this document. Spatial frequency grid ( ) and the simulated counting grid vector position ( They correspond to the spatial frequency coordinates of the counting grid of the event-driven detector and have been calibrated before computation. Real space grid ( (This refers to) reconstructing the mesh. The previously used notation method... The instructions will first be passed through ( )and( The nuclear image selected based on the correspondence between the scan location and the target location. Centered on.

[0129] ED data is collected during a data collection session. During the data collection session, the samples are held in the same position relative to the sample holder.

[0130] This document provides an EM (and a computing device or system) configured to perform the methods described herein. The EM may be able to provide an EM beam with energy ranging from 1 keV to 300 keV, preferably 30 to 300 keV. Since this disclosure allows for the collection of ED data using lower beam energy, the EM may be able to provide an EM beam with energy ranging from 1 keV to 30 keV.

[0131] The method described here is a computer-implemented method. This method can be executed by the standard (general-purpose) computer systems described elsewhere in this document.

[0132] A computing device or system configured to perform the methods or a portion thereof as described herein is further provided.

[0133] A computer program or computer program product having instructions that, when executed by a computing device or system, cause the computing device or system to perform the methods or a portion thereof as described herein.

[0134] A further provision provides a computer-readable medium having a computer program (product) stored thereon, the computer program (product) having instructions that, when executed by a computing device or system, cause the computing device or system to perform (each step) or a portion thereof of the method described herein.

[0135] A data stream is further provided, which represents a computer program or computer program product having instructions that, when executed by a computing device or system, cause the computing device or system to perform each step of the method described herein, or a portion thereof.

[0136] It should be understood that the methods described herein can alternatively be performed by an electron microscope combined with a standard computer system.

[0137] Furthermore, an electron microscope configured to perform the methods or a portion thereof as described herein is provided.

[0138] A computer program or computer program product having instructions that, when executed by an electron microscope, cause the electron microscope to perform the methods or a portion thereof as described herein.

[0139] A further provision provides a computer-readable medium having a computer program (product) stored thereon, the computer program (product) having instructions that, when executed by an electron microscope, cause the electron microscope to perform (each step) or a portion thereof the methods described herein.

[0140] A data stream is further provided, which represents a computer program or computer program product having instructions that, when executed by an electron microscope, cause the electron microscope to perform each step of the method described herein, or a portion thereof.

Claims

1. A method for generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c). - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from a kernel image library (300) including multiple kernel images (e.g., 310a to i) - the selected kernel image (310b, 310e, 310h), where: - Each kernel image (e.g., 310a to i) in the kernel image library (300) is pre-computed and is a mosaic of the reconstruction grid (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). as well as - Generate reconstructed images (500), including: - Add each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500).

2. The method according to claim 1, wherein each nuclear image (310a to i) of the nuclear image library (300): - Has a dimension that is part of the size of the reconstructed mesh (400); - is based on the simulated counting grid vector position on the simulated counting grid of the simulated counting detector ( The simulation is based on a single simulated event-detection count.

3. The method according to claim 1 or 2, wherein: - The electron diffraction data also includes multiple different scan grid vector positions (100a, 100b, 100c), each scan grid vector position being a vector location within the scan grid (120). The position of the test electron probe that interacts with the test sample (100) and thereby generates an outgoing wave (120a) from the test sample (100); - Each of the plurality of test event-detection counts (210a, 210b, 210c) has a test count grid vector position on the count grid (205) of the event-driven detector (200). ; - Position the test count grid vector in the electron diffraction data Linked to the scan grid vector positions (100a, 100b, 100c) of the generated outgoing wave (120a), which is at the test count grid vector position. The test event was triggered at the location - the detection count (210a, 210b, 210c). - Selected kernel images (310b, 310e, 310h) in and Selected during matching; Adding each selected kernel image (310b, 310e, 310h) to the reconstruction grid (400) is done at the reconstruction grid vector position. At the set location, the reconstructed grid vector position The set is centered on the scan grid vector position that generates the outgoing wave (120a), which causes the selected nuclear images (310b, 310e, 310h) to be selected by the test event-detection count (210a, 210b, 210c).

4. The method of claim 3, wherein adding each selected kernel image (310b, 310e, 310h) to the reconstruction mesh (400) comprises the following steps: For each reconstructed grid vector position The intensity of each pixel in the selected kernel image (310b, 310e, 310h) added at the location is the same as the intensity of the reconstructed grid vector. The previous pixel intensities at that location are summed.

5. The method according to claim 3 or 4, wherein at the location of the reconstructed grid vector Each selected kernel image (310b, 310e, 310h) added to the reconstruction grid (400) is adjacent, overlapping, or superimposed, where: - Adjacent additions include positioning the selected kernel images (310b, 310e, 310h) next to the existing kernel images on the reconstructed mesh (400); - The addition of overlap involves positioning the selected kernel images (310b, 310e, 310h) such that one or more of their edges overlap with a portion of one or more existing kernel images on the reconstructed mesh (400); - The overlay involves precisely positioning the selected kernel image (310b, 310e, 310h) over one or more existing kernel images on the reconstructed mesh (400); and - The existing kernel image on the reconstruction mesh (400) is a selected kernel image previously added to the reconstruction mesh (400) during the gradual stitching and rendering of the reconstruction image (500).

6. The method according to any one of claims 1 to 5, wherein after the selected kernel images (310b, 310e, 310h) are added to the reconstruction mesh (400), the reconstruction mesh becomes updated, and only a portion of the selected kernel images (310b, 310e, 310h) in the reconstruction mesh (400) becomes updated.

7. The method according to any one of claims 1 to 6, wherein the size of each kernel image (310a to i) of the kernel image library (300) is equal to or less than 50% of the size of the reconstructed mesh (400) in terms of area.

8. The method according to any one of claims 2 to 7, wherein the position of the simulated counting grid vector is determined based on... The simulation of each nuclear image (310a to i) in the image library (300) includes analysis of layered imaging.

9. The method according to claim 8, wherein: - Analysis of stacked imaging involves integral centroid (iCOM), sideband integration (SBI), or Wigner distribution deconvolution (WDD); and - The simulated representation of the nuclear images (310a to i) in the image library (300) is determined by the interaction constant ( Weighted electrostatic potential ( The diagram is composed of the same interaction constant ( Weighted charge density ( The real or imaginary part of the graph, or the transmission function (T) graph, is measured, where the transmission function is determined by the interaction constant (T). Weighted electrostatic potential ( The complex index of ).

10. The method according to any one of claims 1 to 9, wherein the generation of the reconstructed image (500) is performed in real time.

11. A method for interactively generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising the method according to any one of claims 1 to 10, wherein: - The method further includes sending instructions to an electron microscope (EM) to direct the test electron probe to the scanning grid vector position (100a, 100b, 100c), such that the electron probe interacts with the test sample (100) and generates an outgoing wave (120a). - In the following steps: Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test event-detection counts (210a, 210b, 210c) caused by the generated outgoing wave (120a). The electron diffraction data of the test sample (100) were generated by electron microscopy (EM); - In the following steps: Each selected kernel image (310b, 310e, 310h) is added to the reconstruction mesh (400), thereby gradually stitching together and rendering the reconstructed image (500). At the location of reconstructing the grid vector ( At the set of ), each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position ( The set of ) is centered on the scan grid vector position that generates the outgoing wave (120a), which causes the selected nuclear images (310b, 310e, 310h) to be selected by the test event - detection count (210a, 210b, 210c). - The method further includes: Repeat the sending, receiving, selecting, and adding process, wherein for at least one repetition, different scan grid vector positions (100a, 100b, 100c) are used, thereby progressively stitching together and rendering the reconstructed image (500).

12. A method for interactively generating a reconstructed image (500) based on electron diffraction data of a test sample (100), comprising: - Send instructions to the electron microscope (EM) to guide the test electron probe to the scanning grid vector position (100a, 100b, 100c), so that the electron probe interacts with the test sample (100) and generates an outgoing wave (120a). - Receive electron diffraction data of test sample (100), the electron diffraction data including multiple test events - detection counts (210a, 210b, 210c) caused by the generated outgoing wave (120a). - For each test event - detection count (210a, 210b, 210c), select a kernel image (e.g., 310b, e, h) from a kernel image library (300) including multiple kernel images (e.g., 310a to i) - select a kernel image (e.g., 310b, 310e, 310h), where: - Each kernel image (e.g., 310a to i) in the kernel image library (300) is a pre-computed mosaic of a reconstruction grid (400) to be added to the reconstructed image (500), wherein multiple selected added mosaics are able to stitch together and render the reconstructed image (500). - At the location of reconstructing grid vectors At the set location, each selected kernel image (310b, 310e, 310h) is added to the reconstruction grid (400), the reconstruction grid vector position The set is centered on the scan grid vector position that generates the outgoing wave (120a), which causes the nuclear images (310b, 310e, 310h) to be selected as test events - detection counts (210a, 210b, 210c). - Repeat the sending, receiving, selecting and adding process, wherein for at least one repetition, different scan grid vector positions (100a, 100b, 100c) are used, thereby gradually stitching together and rendering the reconstructed image (500).

13. The method of claim 12, in conjunction with the subject matter of any one of claims 2 to 10.

14. The method according to any one of claims 1 to 13, wherein the method is a computer-implemented method.

15. A computing device or system configured to perform the method of claim 14.

16. A computer program or computer program product having instructions that, when executed by a computing device or system, cause the computing device or system to perform the method according to claim 14.

17. A computer-readable medium having a computer program stored thereon, the computer program having instructions that, when executed by a computing device or system, cause the computing device or system to perform the method according to claim 14.

Images