Method and system for determining sample composition from spectral data

Abstract

To provide a method and system for determining sample composition from spectral data acquired by a charged particle microscopy system.SOLUTION: Chemical elements in a sample are identified by processing spectral data with a trained neural network (NN). If the identified chemical elements do not match a known elemental composition of the sample, the trained NN is retrained with the spectral data and the known elemental composition of the sample. The retrained NN can then be used to identify chemical elements within other samples.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] This specification generally relates to methods and systems for determining a sample composition from spectral data, and more particularly, to methods and systems for determining a sample composition based on spectral data obtained in response to irradiating a sample with a charged particle beam.

Background Art

[0002] Charged particle microscopes are well-known and are increasingly important technologies for imaging microscopic objects. Multiple types of emissions from a sample in response to charged particle irradiation can provide structural and compositional information about the sample. For example, based on the energy spectrum of X-ray emissions in response to electron beam irradiation, energy-dispersive X-ray spectroscopy (EDS or EDX) can be used for elemental analysis or chemical property evaluation.

[0003] One method for determining chemical elements in a sample is to compare each detected energy spectrum with known spectra of chemical elements. For example, chemical elements can be identified by comparing the peak locations within the detected spectrum with the known peak locations of each chemical element. However, the applicant recognizes that the identified chemical elements can potentially include a large number of false positives and / or false negatives, particularly when the detected spectrum is sparse or has interfering peaks.

Summary of the Invention

[0004] In one embodiment, the method includes irradiating a first sample with a charged particle beam; detecting a first type of emission from the first sample; forming one or more first type spectra from the detected first type of emission; identifying one or more first chemical elements in the first sample by processing the first spectra with a trained neural network; detecting a second type of emission from the first sample; displaying a sample image generated based on the detected second type of emission; selecting one or more pixels in the sample image; displaying one or more chemical elements corresponding to the selected pixels; retraining a neural network trained with the selected pixels and spectra corresponding to known elemental compositions in response to displayed chemical elements that differ from known elemental compositions; obtaining one or more second spectra by irradiating one or more locations of a second sample with a charged particle beam; detecting a first type of emission from the second sample; and identifying one or more second chemical elements in the second sample by processing the second spectra with a retrained neural network. In this way, chemical elements can be identified quickly and with high accuracy.

[0005] It should be understood that the above summary is provided in a simplified manner to introduce a selection of concepts that will be further described in the modes for carrying out the invention. It is not intended to identify the main or essential features of the claimed subject matter, and its scope is uniquely defined by the claims that follow the modes for carrying out the invention. Furthermore, the claimed subject matter is not limited to implementations that solve any of the defects mentioned above or in any part of this disclosure. [Brief explanation of the drawing]

[0006] [Figure 1] This shows a charged particle microscope. [Figure 2] This is a flowchart illustrating an exemplary method for determining sample composition from spectral data. [Figure 3]This is a flowchart illustrating an exemplary method for retraining a neural network trained for elemental analysis. [Figure 4] This section displays elemental analysis results and provides an exemplary user interface for retraining a trained neural network. [Figure 5] Figure 2 illustrates an example of the data flow used in this method. [Figure 6] This is a flowchart illustrating an exemplary method for generating a neural network trained for elemental analysis.

[0007] Similar reference numbers refer to the corresponding parts across several figures in the drawing. [Modes for carrying out the invention]

[0008] The following description relates to a method and system for determining the composition of a sample based on spectral data, such as an energy spectrum, obtained by detecting emissions in response to irradiating the sample with a charged particle beam. In U.S. Patent No. 9048067, Owen discloses a method for identifying minerals by sequentially decomposing spectral data from mineral definitions to elements. The decomposition determines the proportion of chemical elements in an unknown mineral by fitting known elemental spectra from an elemental list to the detected X-ray spectrum, and the elemental list is obtained by selecting a mineral definition in a mineral database. The applicant recognizes that accurate compositional analysis requires an accurate elemental list. When the X-ray spectrum of an unknown sample obtained is sparse (i.e., the total number of counts per spectrum is small), the elemental list may be inaccurate. In sparse spectra, it is difficult to accurately position peaks or distinguish peaks that are close to or interfering with each other. Increasing the number of X-ray counts (and thus having a denser spectrum) can improve the accuracy of elemental identification. However, this can also significantly increase the total data acquisition time. Furthermore, because the detected X-ray spectra of the same mineral may differ from microscope to microscope due to factors including detector aging and different operating conditions, known elemental spectra may not adequately match the detected X-ray spectra.

[0009] To address the above issues, elemental analysis is performed using a trained neural network (NN) adapted to the specific operating conditions of the microscope system. In one example, after loading a first sample into the microscope, a charged particle beam irradiates one or more locations on the first sample. A first type of emission, such as X-rays, from each location is acquired in the form of an energy spectrum, with each energy spectrum corresponding to one sample location. The trained NN receives each of the acquired energy spectra as input and outputs the chemical elements identified in the spectrum and the probability of each identified chemical element. After processing the acquired energy spectra, a list of the first chemical elements contained in the first sample can be generated. A second type of emission from the sample may be detected simultaneously or separately from one or more locations on the sample. A sample image may be generated based on the detected second type of emission. The sample image shows the structure of the sample. The sample image may be an optical image, a scanning electron microscope (SEM) image, a backscattered electron (BSE) image, or a transmission electron microscope (TEM) image. In some examples, the sample image also includes chemical element information determined from the trained neural network (NN). One or more pixels are selected within the sample image. The chemical elements corresponding to the selected pixels are displayed. In some examples, the first sample is a reference sample with a known elemental composition. If the displayed chemical elements do not match the known elemental composition at the corresponding sample location, the trained NN is retrained with the spectrum corresponding to the selected pixels. The retrained NN is then used for elemental analysis of spectral data obtained from the microscope, from another sample. In this way, the retrained NN is adapted to microscope-specific parameters and operating conditions. For example, the retrained NN can be adapted to energy shifts induced by the detector in the spectral data. Furthermore, the retrained NN can be adapted for specific uses. For example, through retraining, the NN can be adapted to distinguish interference peaks characteristic of certain minerals, thus allowing for more accurate identification of chemical elements.

[0010] In another example, after loading a first sample into a microscope, a charged particle beam, such as an electron beam, irradiates one or more locations on the sample. A first type of emission, such as X-rays, from each location is detected and used to form an energy spectrum. A second type of emission, such as scattered electrons, is detected simultaneously with or separately from the first type of emission. A sample image is generated based on the second type of emission. From the sample image, one or more pixels are selected, for example, by an operator. In one example, one or more spectra corresponding to the selected pixels are selected from the energy spectrum formed based on the first type of emission. In another example, the sample region corresponding to the selected pixels is irradiated with a charged particle beam, and the spectrum corresponding to the selected pixels is recollected. The recollected spectrum may be denser than the first spectrum. The chemical element corresponding to the selected pixel is identified by processing the selected spectrum or the recollected spectrum using a trained neural network (NN). If the identified chemical element differs from the known elemental composition in the selected sample region, the trained NN is retrained with the selected spectrum and the known composition.

[0011] In another example, a sample image showing the structure of the sample is displayed to the operator. The sample image may be an optical image, SEM image, BSE image, or TEM image. One or more pixels in the sample image are selected automatically or by the operator. The sample region corresponding to the selected pixels is then irradiated with a charged particle beam, and spectral data is collected by detecting a first type of emission, such as X-rays. The chemical elements corresponding to the selected pixels are identified by processing the collected spectral data using a trained neural network (NN). If the identified chemical elements differ from the known elemental composition in the selected sample region, the trained NN is retrained with the spectral data and the known composition.

[0012] A retrained neural network (NN) may be stored in the non-temporary memory of the microscope where the spectral data for retraining is acquired, in order to process spectral data acquired by the microscope. The retrained NN may replace the trained NN for future elemental analysis. In some examples, both the trained and retrained NNs are stored in a library. The NNs in the library may serve as baseline trained NNs for further retraining in the same or different microscopes. The microscope may irradiate one or more locations of a second sample with a charged particle beam and obtain an energy spectrum from each of the sample locations. The retrained NN processes each of the obtained spectra and outputs the chemical elements and their corresponding probabilities. Chemical elements with probabilities higher than a threshold probability may be identified as the second chemical element of the second sample. A composition map may be generated based on the second chemical element. The composition map shows the spatial distribution of chemical elements or components in the sample. In one example, the composition map is an element map showing the spatial distribution of the second chemical element. In another example, the obtained spectrum of a sample can be further decomposed into multiple spectral components and / or abundances based on a second chemical element. Each component may correspond to a chemical element, a mixture of chemical elements, or a chemical phase. Each component corresponds to one spectral component and abundance. A spectral component can be the spectrum of an element. An abundance can be the quantity of an element. In one example, the abundance is the ratio of the elements at the sample site. A composition map may show the spatial distribution of components within the sample.

[0013] In some examples, a first spectrum of several first locations in a sample can be obtained by scanning the sample with a charged particle beam. The chemical elements of the sample are identified based on the first spectrum using a trained NN. The first spectrum can then be decomposed into various components based on the identified chemical elements. In addition, or alternatively, the abundance of each component can be determined through the decomposition process. In some examples, after identifying the chemical elements in the sample, a second spectrum of several second locations in the sample is obtained by scanning the sample with a charged particle beam. The first spectrum may have a higher density than the second spectrum; that is, the count number will be higher for each of the first spectra than for each of the second spectra. Alternatively, the second sample locations may differ from the first sample locations. For example, the second sample locations may cover a wider area and / or have higher spatial resolution than the first sample locations. The second spectrum can be decomposed based on the identified chemical elements to generate a composition map.

[0014] In some examples, after the trained neural network (NN) identifies the chemical elements of a reference sample, the identified chemical elements are displayed to the operator or user, for example, in the periodic table. The operator may input the known composition / chemical elements of the reference sample into the microscope to retrain the trained NN. In another example, a sample image is displayed simultaneously with the identified chemical elements. The sample image may show the structure of the reference sample. Alternatively or additionally, the sample image may show the spatial distribution of the identified chemical elements in the reference sample. The sample image may contain all scanned sample locations or a subset of sample locations. The operator may select one or more regions within the sample image and retrain the NN with the integrated spectrum and the known elemental composition of the selected regions. The regions may be lines, areas, or points within the sample image. In yet another example, the integrated spectrum is displayed together with the sample image and identified chemical elements, for example, in the same window. The integrated spectrum is the sum of all spectra within the selected region. By selecting a region of the sample and displaying the integrated spectrum, the operator can select a region that can accurately reflect the known elemental composition. For example, the operator may select a region where the distribution of chemical elements does not change abruptly, or a region excluding the sample boundaries. In another example, the operator may select a region where peaks overlap and retrain the neural network to better identify or differentiate the peaks.

[0015] In some examples, a trained neural network (NN) can be automatically retrained with the known elemental composition of a reference sample after obtaining spectral data from the reference sample. Instead of the operator selecting a region within the sample image, the region within the sample image is automatically selected. The trained NN can then be retrained based on the spectral data from the selected region. In some examples, a sample image showing the sample structure may be displayed to the operator. The operator can select a location within the sample image for collecting spectral data. The sample image may be an optical image, SEM image, BSE image, or TEM image.

[0016] In some examples, a trained neural network (NN) is generated by training a naive NN on training data, which may include simulated and / or experimental data. Simulated training data may be generated based on multiple chemical elements and their known spectra. Simulated training data may include individual chemical elements and their spectra, as well as combinations of chemical elements and their corresponding combined spectra. For example, a combination of chemical elements might be elements in a mineral, and a combined spectrum might be the spectrum of that mineral. Experimental data may be generated by imaging a reference sample with a known composition using one or more microscopes. Different trained NNs may be generated for different microscope operating conditions. Operating conditions may include one or more of the following: acceleration voltage, beam current, takeoff angle, working distance, and detector settings (such as bias voltage).

[0017] A neural network (NN) may be trained offline to identify many chemical elements, such as 100, or combinations of chemical elements. However, a trained NN may not be sensitive to specific combinations of chemical elements, especially when the peaks of individual chemical elements interfere with each other. Retraining a trained NN can adjust or adapt it to become more sensitive to identifying specific types of materials / minerals of interest to the operator, and therefore more accurate.

[0018] In some examples, charged particle beams include electron beams and ion beams. Energy spectra include X-ray spectra and electron energy loss spectra. In other examples, elemental chemical analysis may be applied to the analysis of other types of spectral data, such as Raman spectra.

[0019] Referring to Figure 1, which is a very schematic depiction of an embodiment of a charged particle microscope (CPM) in which the present invention is implemented, more specifically, an embodiment of a scanning electron microscope (SEM) system. The system axis is shown as axis 110. The microscope 100 comprises a particle optical column 1 that creates a beam 3 (in this case, an electron beam) of charged particles propagating along the particle optical axis 101. The particle optical axis 101 may be aligned with the Z axis of the system. The column 1 is mounted in a vacuum chamber 5 which comprises a sample holder 7 and associated actuator 8 for holding / positioning a sample 6. The vacuum chamber 5 is evacuated using a vacuum pump (not shown). Vacuum ports 9 which may be opened to introduce / remove items (components, samples) into / from the inside of the vacuum chamber 5 are also shown. The microscope 100 may have multiple such ports 9 as needed.

[0020] Column 1 (in this case) comprises an electron source 10 and an illuminator 2. The illuminator 2 comprises lenses 11 and 13 for focusing the electron beam 3 onto the sample 6, and a deflection unit 15 (for performing beam steering / scanning of the beam 3). The microscope 100 further comprises, among other things, a controller / computer processing unit 26 for controlling the deflection unit 15, lenses 11 and 13, and detectors 19 and 21, and for displaying information collected from the detectors 19 and 21 on a display unit 27. The display unit may also function as an input unit for receiving operator input.

[0021] Detectors 19 and 21 are selected from a variety of potential detector types that can be used to inspect different types of "induced" radiation emitted from the sample 6 in response to irradiation by the (acting) beam 3. Detector 19 may be a solid-state detector (such as a photodiode) used to detect cathodoluminescence emitted from the sample 6. Alternatively, it may be an X-ray detector such as a silicon drift detector (SDD) or a silicon lithium (Si(Li)) detector. Detector 21 may be an electron detector in the form of a solid-state photomultiplier tube (SSPM) or a vacuum photomultiplier tube (PMT), which may be used to detect backscattered and / or secondary electrons emitted from the sample 6. Those skilled in the art will understand that many different types of detectors can be selected in a configuration such as the one shown, including, for example, annular / segmented detectors. By scanning beam 3 over the sample 6, stimulated radiation, including, for example, X-rays, infrared / visible / ultraviolet rays, secondary electrons (SE), and / or backscattered electrons (BSE), is emitted from the sample 6. Since such stimulated emission is position-sensitive (due to the scanning motion), the information obtained from detectors 19 and 21 will also be position-dependent. This fact makes it possible to use the signal from detector 21 (for example) to create a BSE image of (part of) the sample 6, which is essentially a map of the signal as a function of the scanning path position on the sample 6.

[0022] Signals from detectors 19 and 21 pass through control lines (buses) 25, are processed by the controller 26, and are displayed on the display unit 27. Such processing may include operations such as combining, integrating, subtracting, false coloring, edge enhancement, and other processes known to those skilled in the art. Furthermore, an automated recognition process (such as that used in particle analysis) may be included in such processing. The controller includes a processor 28 and non-temporary memory 29 for storing computer-readable instructions. The methods disclosed herein can be implemented by executing computer-readable instructions stored in the non-temporary memory 29 within the processor 28.

[0023] Note that many improvements and alternatives to such settings are known to those skilled in the art, such as the use of a controlled environment within the microscope 100 (relatively large amounts), for example, maintaining a background pressure of a few millibars (used in environmental SEM or low-pressure SEM).

[0024] FIG. 2 shows a method 200 for extracting sample composition from spectral data. In one example, the spectral data can be an EDS spectrum obtained using the microscope 100 of FIG. 1. The trained NN performed elemental analysis to identify chemical elements in the sample. The trained NN can be adapted to the microscope 100 and a particular sample type by retraining (or updating) the NN trained with measurements from samples having known elemental compositions.

[0025] At 202, the sample is loaded into the microscope. At 204, the system parameters are set. The system parameters include operating conditions such as one or more of acceleration voltage, beam current, working distance, and detector settings. The detector settings can include integration time at each sample location. The system parameters may also include scanning parameters such as a scanning path and a scanning area.

[0026] In some examples, a sample image indicating the structure or dimensions of the sample can be displayed to the operator. The sample image can be an EM image. The operator can select a scanning area within the sample image. The operator may also move the sample relative to the charged particle beam to move the sample into the current field of view (FOV) to facilitate selection of the scanning area.

[0027] In 206, one or more locations on the sample are scanned with a charged particle beam. The scanned locations on the sample are determined in 204. At each location on the sample, one or more types of charged particles emitted from the sample are acquired by one or more detectors. The acquired charged particles may include selected and X-rays. Energy spectra, such as X-ray spectra, are acquired for each location on the sample. The acquired charged particles may also include backscattered electrons or secondary electrons.

[0028] In step 208, method 200 determines whether elemental analysis should be performed on the spectral data obtained in step 206. In one example, elemental analysis is required if the current sample is a reference sample for retraining a trained NN. In another example, elemental analysis is not required if elemental analysis has been previously performed on the sample. In yet another example, elemental analysis is not required if the elemental composition of the sample is known, such as when a trained NN is being retrained on a reference sample having a similar composition to the current sample. If elemental analysis is not required, method 200 proceeds to step 216 to decompose the spectral data based on the known elemental composition. If elemental analysis is required, method 200 proceeds to step 210 to perform elemental analysis on the spectral data.

[0029] In step 210, chemical elements in the sample are selectively identified by a trained neural network (NN). The trained NN may be selected from a group of trained NNs, such as a baseline NN, based on the operating conditions set in step 204. The trained NN may be a convolutional NN or a network for natural language processing. The trained NN analyzes one or more of the obtained spectra and outputs a list of chemical elements and their corresponding probabilities for each spectrum. Chemical elements identified at a particular sample location are those with a probability greater than or equal to a threshold probability. For example, the threshold probability is 0.75. In some examples, the threshold is any number between 0 and 1, including 1. In some examples, elemental analysis may be performed on a subset of the obtained spectra.

[0030] In step 212, method 200 determines whether the NN requires retraining. If the trained NN has not been retrained with spectral data obtained from the current microscope, retraining may be required. If the trained NN has been retrained with spectral data obtained from a microscope with a reference sample of similar composition, retraining may not be required. If retraining is not required, method 200 proceeds to step 216. If retraining is required, method 200 proceeds to step 214 to retrain the trained NN. Details of the retraining process are shown in Figure 3.

[0031] After retraining the trained NN in 214, the retrained NN can be saved. The retrained NN can be used to analyze a sample with similar compositional or spectral data obtained under similar operating conditions. In some examples, the spectral data used to retrain the trained NN can be further decomposed in 216.

[0032] In 216, the spectral data obtained in 206 is decomposed based on elemental information. The elemental information may be known or predetermined based on the sample type and loaded into the controller. Alternatively, the elemental information may be the chemical elements identified in 210. In one example, each obtained spectrum is decomposed into one or more spectral components and / or abundances of each component. Each component corresponds to one spectral component and one abundance. Each component may be an element, a mixture of elements, or a chemical phase. The components may include various combinations of known or identified elements. One method for decomposing spectral data is described in U.S. Patent Application No. 17 / 166,885 by Petr Hlavenka et al., filed 3 February 2021, which is incorporated herein by reference in its entirety.

[0033] In step 218, one or more compositional images are generated. The compositional image includes an elemental map showing the distribution of identified chemical elements in the sample. The compositional image also includes an image generated based on spectral components and / or abundances obtained in step 216. In the compositional image, compositional data may be superimposed on a structural image of the sample. The structural image can be generated based on scattered charged particles, such as scattered electrons, obtained in step 206.

[0034] In step 220, method 200 determines whether additional scanning of the sample is necessary. Additional scanning may be necessary to obtain more spectral data for retraining. In some examples, in the first scan, multiple first sample locations may be scanned to determine the elemental composition and / or to retrain the trained NN. The sample may be rescanned in a second scan, and multiple second sample locations may be scanned to generate a compositional image. The system parameters in the second scan may differ from those in the first scan. In one example, the second scan may have higher spatial resolution than the first scan. In another example, multiple first sample locations may differ from the second sample locations. The number of sample locations in the second scan may be greater than the number of sample locations in the first scan. In yet another example, the count of each spectrum in the second scan may be less than that in the first scan. If additional scanning is necessary, method 200 proceeds to 204 to reset the system parameters. Otherwise, method 200 terminates.

[0035] Thus, elemental analysis is performed using a trained neural network (NN). The trained NN can be specifically retrained with respect to the microscope, operating conditions, and sample type, based on measurements of one or more reference samples with known elemental compositions using the microscope. Once the NN is retrained, it is memorized and can be used for elemental analysis of other similar samples. Based on the chemical elements identified from the elemental analysis, spectral data can be decomposed to obtain spectral components and abundances at each sample location, which can then be used to generate additional compositional images to provide more insight into the sample composition.

[0036] Figure 3 shows method 300 for retraining a trained NN. The trained NN can be retrained with an integrated spectrum generated from acquired spectral data, and the known elemental composition corresponding to the integrated spectrum. Method 300 can interact with an operator through a user interface for receiving known elements.

[0037] In 302, a sample image showing the sample structure is displayed. The sample image may be generated based on scattered electrons obtained during scanning in 206 of Figure 2. Alternatively, the sample image may be obtained separately from the scanning in 206 of Figure 2. The sample image may be an optical image, SEM image, BSE image, or TEM image. In one example, if elemental analysis is performed in 210 of Figure 2, the sample image may be an elemental map showing the spatial distribution of identified chemical elements in the sample. The elemental map may be generated based on both the obtained spatially resolved scattered electrons and the identified chemical elements. For example, signals from scattered selections are displayed in grayscale to show structural information of the sample. Identified elements may be color-coded and superimposed on the grayscale image.

[0038] In some examples, the chemical elements identified in Figure 2, item 210, are displayed along with the sample image. In one example, the identified chemical elements may be displayed within the periodic table. In another example, the identified chemical elements are displayed as a list.

[0039] In step 304, one or more regions within the sample image are selected. The region may be selected by the operator after observing the elemental map. Alternatively, the region may be selected automatically. For example, the central region of the sample is selected automatically. The selected region contains one or more pixels of the sample image. In one example, the selected region may be an area that reflects a known composition of the sample. In another example, the selected region corresponds to a characteristic spectrum, such as a spectrum with overlapping peaks.

[0040] In 308, the charged particle beam is optionally directed to a sample region corresponding to a selected region (or pixel) in the sample image to collect additional spectral data. This additional spectral data is analyzed using a trained neural network (NN) to identify chemical elements within the selected region. System parameters can be adjusted to collect additional spectral data. In one example, the detector integration time is longer compared to the integration time in 206, thus obtaining a denser spectrum. In another example, the scanning step is adjusted.

[0041] In 310, the chemical elements identified within the selected region are displayed in 310. Chemical elements may be identified in either 210 or 310. In one example, chemical elements are displayed by updating the identified chemical elements displayed in 306. In one example, the chemical elements identified in 310 may be displayed as a periodic table or list. One or more of the integrated spectrum, identified chemical elements, and sample images may be selectively displayed individually or together on the display unit.

[0042] In 312, the integrated spectrum is generated based on spectra obtained within the selected region. For example, one or more spectra obtained from sample locations within the selected region are summed or averaged to generate the integrated spectrum. In some examples, the integrated spectrum is displayed to the operator.

[0043] In step 314, the chemical elements identified within the selected region are compared with the known elemental composition within the selected region. The comparison may be performed by the microscope controller or by the operator. If the identified chemical elements and the known elements do not match, method 300 proceeds to step 316 to receive the known elemental composition from user input. Otherwise, method 300 proceeds to step 322.

[0044] In 316, the known elemental composition from the selected region is received via user input. For example, the operator can select or deselect the chemical elements displayed in 310.

[0045] In 318, the trained neural network (NN) is retrained with the integrated spectrum and received known compositional information. In some examples, the retrained NN may be validated with spectral data not used for retraining the NN. In one example, the trained NN is replaced and stored by the retrained NN. In some examples, one or more retrained NNs are stored as baseline NNs. The baseline NNs may be used, for example, as a starting point for retraining for different samples or different microscopy systems.

[0046] In step 320, the acquired spectral data can optionally be reprocessed with a retrained neural network (NN) to generate an updated elemental map. The updated elemental map can then be used to evaluate the accuracy of the retrained NN.

[0047] In step 322, method 300 checks whether additional retraining of the NN is needed with the current spectral data. For example, the operator may decide whether to retrain the NN based on the updated elemental map. If retraining is needed, method 300 proceeds to step 306, where the NN may be retrained with spectral data from different regions of the sample. If no additional retraining is needed, or if additional spectral data is needed for retraining, method 300 terminates.

[0048] By receiving the chemical elements selected by the operator, new training data not used for offline training of the NN can be acquired to fine-tune the NN, making it more sensitive to specific user cases and / or compositions of particular interest.

[0049] Figure 4 shows an example of a user interface for displaying results from elemental analysis and receiving user input for retraining the trained neural network (NN). An elemental map 402, a periodic table 403 showing the identified chemical elements, and an integrated spectrum 404 are displayed simultaneously within a window 401 on the display unit. The elemental map shows the identified elements, color-coded and superimposed on an electron microscope image of the sample. The identified elements Fe, Zn, and Sn are highlighted in the periodic table. The integrated spectrum 404 can be generated based on all spectral data received within the sample region shown in the elemental map 402. Peaks in the integrated spectrum corresponding to the identified elements can be shown in the integrated spectrum.

[0050] The operator may select one or more regions within the element map 402. These regions may be lines, areas, or points within the sample image. During or after selection, the periodic table 402 and the integrated spectrum 404 are updated with the identified elements and the updated integrated spectra of the selected regions. If the identified elements differ from known elemental compositions, the operator may input known elemental compositions by selecting and / or deselecting elements within the periodic table 403. The operator may then retrain the NN with the input information by clicking the "Learn" button 405 in the window.

[0051] Figure 5 illustrates an exemplary data flow for extracting sample composition from spectral data using a trained neural network (NN). The acquired spectrum 501 is processed by the trained NN 502. For each chemical element in each processed spectrum, the trained NN generates a list of chemical elements 504 and probabilities 505. An exemplary list of chemical elements and probabilities is shown in 503. The list may contain all chemical elements in the periodic table. The probability values ​​are in the range of 0 to 1. Chemical elements 506 in the sample contain all chemical elements identified from spectrum 501. For example, chemical elements 506 contains chemical element 504 with a probability greater than a threshold probability. Each spectrum of spectrum 501 can then be decomposed in 508 into spectral components and / or abundances based on the identified chemical elements 506. An example of components and abundances at one sample location is shown in 511. In this example, there are two components at the sample location. Each component is a combination of individual chemical elements. The first component contains Fe at a concentration of 0.4 and oxygen at a concentration of 0.6, and the second component contains Si at 0.33 and oxygen at 0.66. Each component has an abundance of 0.5. In other words, the composition of the sample location is half of the first component and half of the second component. An elemental map 511 can be generated based on the chemical elements 506 identified at each sample location. A compositional image 510 can be generated based on the spectral components and / or abundances at the scanned sample locations.

[0052] Figure 6 shows a method 600 for generating a trained neural network offline. The trained neural network can then be retrained online, as shown in method 200 in Figure 2.

[0053] In 602, various operating conditions are defined. These operating conditions may include one or more of the following: acceleration voltage, beam current, working distance, and detector settings. A trained neural network (NN) can be generated for each operating condition.

[0054] In 604, the training dataset is determined based on the operating conditions. The training dataset includes the training spectrum and the corresponding elemental composition. The training spectrum may include simulated data 606 and / or experimental data 608 obtained from one or more microscopes. The training spectrum may include characteristic spectra of a single chemical element and / or spectra of combined chemical elements.

[0055] In 610, the training data is augmented by introducing noise, nonlinearity, and tuned gain. In one example, noise may be introduced into the training data by shifting one or more peaks within a specific energy range of peak locations. The energy range may be less than 100 eV. In one example, the energy range is less than 50 eV. In another example, noise may be added to the amplitude of the training spectrum.

[0056] In 612, a naive NN is trained on the training dataset to produce a trained NN. In some examples, the trained NN is produced for each operating condition defined in 602.

[0057] In 614, the validation data includes a validation spectrum and generates corresponding elemental compositions, and the trained NN is validated against the validation data. The validation spectrum is different from the training spectrum. For example, the validation spectrum may include experimental spectra obtained from different microscopes. Elemental analysis is performed on the validation spectrum with the trained NN. The performance of the trained NN is determined by comparing the elemental compositions in the validation data with the identified chemical elements through the trained NN.

[0058] If the performance of the trained neural network (NN) is acceptable at step 616, the trained NN is saved at step 620. Otherwise, an additional training dataset is generated at step 618 to continue training the NN.

[0059] The technical benefit of using neural networks (NNs) for elemental analysis is that they can perform elemental analysis quickly and accurately on spectral data, especially sparse spectral data. The technical benefit of retraining NNs trained by measuring reference samples is that offline-trained NNs can be adapted to specific use cases, systems, and materials. The technical benefit of selecting a region of the reference sample for retraining is that a sample region reflecting a known elemental composition can be selected. The technical benefit of performing elemental analysis before decomposing spectral data is that sample components can be identified quickly and accurately.

[0060] In representation, a non-temporary computer-readable medium, when executed by the controller's processor, instructs the controller to irradiate one or more locations of a first sample with a charged particle beam, to obtain one or more first spectra by detecting first emissions from the first sample in response to the irradiation of the first sample, to determine one or more first chemical elements in the first sample by processing the first spectra with a trained neural network stored in the computer-readable medium, to display one or more first chemical elements, and to display the known chemical elemental composition of the first sample. The instructions include causing a received and trained neural network to retrain with one or more of the first spectrum and known elemental compositions of the first sample; to irradiate one or more locations of the second sample with a charged particle beam; to obtain one or more second spectra by detecting first emissions from the second sample in response to the irradiation of the second sample; to determine one or more second chemical elements in the second sample by processing the second spectra with the retrained neural network; and to generate a composition map of the second sample based on the second chemical elements.

Claims

1. It is a method, Irradiating the first sample with a charged particle beam, The first type of emission from the first sample is detected, and one or more first spectra are formed from the detected first type of emission. By processing the first spectrum with a trained neural network, one or more first chemical elements in the first sample can be identified, The system detects a second type of emission from the first sample and displays a sample image generated based on the detected second type of emission. Selecting one or more pixels in the aforementioned sample image, Displaying one or more chemical elements corresponding to the selected pixels, In response to the displayed chemical elements which differ from known elemental compositions, the trained neural network is retrained with the selected pixels and the spectra corresponding to the known elemental compositions. Irradiating one or more locations of the second sample with the charged particle beam, Obtaining one or more second spectra by detecting the first type of emission from the second sample, A method comprising identifying one or more second chemical elements in the second sample by processing the second spectrum with the retrained neural network.

2. Identifying one or more first chemical elements in the first sample by processing the first spectrum with a trained neural network, Inputting one or more of the first spectrum into the trained neural network, The trained neural network outputs one or more chemical elements and probabilities, wherein each chemical element corresponds to one probability. The method according to claim 1, comprising selecting a chemical element having a probability greater than a threshold probability as the first chemical element.

3. The method according to claim 1, further comprising generating an element map showing the spatial distribution of the second chemical element in the second sample.

4. The method according to claim 1, further comprising decomposing each of the second spectra into one or more spectral components and abundances based on the second chemical element, wherein each component corresponds to one abundance and each spectral component is the spectrum of a component containing one or more of the second chemical elements.

5. The method according to claim 1, wherein the sample image is further displayed based on the first chemical element, and the sample image shows the spatial distribution of the first chemical element.

6. The method according to claim 1, further comprising generating the trained neural network by training the neural network with training spectra of a plurality of chemical elements, wherein the training spectra include simulated spectra and / or experimental spectra.

7. The method according to claim 1, further comprising receiving the known elemental composition of the first sample via user input, and retraining the trained neural network with spectra corresponding to the selected pixels and the known elemental composition, comprising generating an integrated spectrum from the spectra corresponding to the selected pixels and retraining the trained neural network with the integrated spectrum and the known elemental composition.

8. The method according to claim 7, further comprising displaying the integrated spectrum simultaneously with one or more chemical elements from the first chemical element corresponding to the selected pixel.

9. The method according to claim 7, wherein displaying the first chemical element on a display includes displaying the first chemical element in the periodic table, and the user input includes selecting or deselecting the first chemical element displayed in the periodic table.

10. A charged particle microscope system, A sample holder for positioning the sample, A charged particle source for irradiating the sample with a charged particle beam, A first detector for detecting a first type of emission from the sample in response to irradiation with the charged particle beam, A second detector for detecting a second type of emission from the sample in response to irradiation with the charged particle beam, A controller including a non-temporary computer-readable medium for storing instructions, wherein when executed by the processor of the controller, the controller has Irradiating the first sample with the charged particle beam, The first type of emission is detected from the first sample, and one or more first spectra are formed from the detected first type of emission. The system detects the second type of emission from the first sample and displays a sample image generated based on the detected second type of emission. Receiving the selection of one or more pixels in the aforementioned sample image, Obtaining one or more spectra corresponding to the selected pixels, By processing the acquired spectrum with a trained neural network, one or more first chemical elements in the first sample are identified, Retraining the trained neural network with the selected spectrum and the known elemental composition in response to the first chemical element which has a different elemental composition from known elements, A charged particle microscope system comprising a controller that causes the retrained neural network to be stored in the non-temporary computer-readable medium.

11. When the non-temporary computer-readable medium is executed by the processor of the controller, the controller: Irradiating one or more locations of the second sample with the charged particle beam, Obtaining one or more second spectra by detecting the first type of emission from the second sample, The charged particle microscope system according to claim 10, further comprising instructions to cause the retrained neural network to process the second spectrum to identify one or more second chemical elements in the second sample.

12. The charged particle microscope system according to claim 11, further comprising instructions to cause the controller to detect the second type of emitters from the second sample and to generate a composition map showing the distribution of the second chemical element in the second sample based on the detected second type of emitters, when the non-temporary computer-readable medium is executed by the processor of the controller.

13. A charged particle microscope system according to any one of claims 10 to 12, wherein the charged particle beam is an electron beam, the first type of emitter is X-rays, and the second type of emitter is scattered electrons.

14. The charged particle microscope system according to claim 10, wherein the trained neural network receives a spectrum and outputs one or more chemical elements and probabilities for each chemical element.

15. The charged particle microscope system according to claim 14, wherein determining one or more first chemical elements in the first sample by processing the first spectrum with a trained neural network includes selecting the chemical element output by the trained neural network as the first chemical element with a probability higher than a threshold probability.

16. A non-temporary computer-readable medium that, when executed by a processor, is accessible to a computing device. Accessing first detector data corresponding to a first type of emission from a sample, wherein the first detector data is acquired in response to irradiation of the sample with a charged particle beam. This causes the display of a sample image generated based on the first pixel data, Receiving the selection of one or more pixels in the aforementioned sample image, Accessing second detector data corresponding to a second type of emission from one or more sample regions corresponding to the selected pixels, wherein the second detector data is acquired in response to irradiation of the sample region with the charged particle beam. By processing the second detector data with a trained neural network, one or more chemical elements in the sample can be identified. To cause the display of the identified chemical element, In response to the displayed chemical elements, which differ from the known elemental composition, the trained neural network is retrained with the second detector data and the known elemental composition. A non-temporary computer-readable medium containing instructions for storing the retrained network within the non-temporary computer-readable medium.

Images