Charged particle beam device
The sample images of the charged particle beam device are processed through multiple learned models, and the problem of detecting the mixture of different types of structures is solved, and effective identification of multiple structures is achieved.
Patent Information
- Application Number
- CN201980099996.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-09-04
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2039-09-04
AI Technical Summary
Existing charged particle beam devices are difficult to handle efficiently in the detection of the presence of different kinds of structural mixing, especially the learned models are not suitable for detecting particles of some shapes or non-particles in the presence of other shapes.
The sample image is processed using multiple learned models, and the detection results are obtained from the first and second learned models, and the location or region of each structure is determined based on these results, and the comprehensive result information is output to identify multiple structures.
By utilizing multiple learned models, multiple structures can be effectively detected, which improves the recognition ability of different types of structures, especially in the presence of mixed structures.
Smart Images
Figure CN114341631B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a charged particle beam device. Background Art
[0002] In order to detect and / or evaluate an object of interest in a sample, a charged particle beam device is used. The charged particle beam device irradiates a sample with a charged particle beam and detects and / or evaluates the object of interest using a signal caused by the irradiation. Patent document 1 discloses an example of such a charged particle beam device.
[0003] Prior art literature
[0004] Patent Literature
[0005] Patent Document 1: Japanese Patent Application Publication No. 2019-60741 Summary of the invention
[0006] Problems to be solved by the invention
[0007] However, the prior art is effective in detecting a specific structure, but has the problem of difficulty in processing when different types of structures are mixed. For example, a learned model learned to detect particles of a certain shape is sometimes not suitable for detecting mixed particles of other shapes, non-particle foreign matter, etc.
[0008] The present invention has been made to solve such a problem, and an object of the present invention is to provide a charged particle beam device capable of appropriately detecting different types of structures.
[0009] Means for solving problems
[0010] An example of the charged particle beam device of the present invention has:
[0011] an irradiation unit for irradiating the sample with a charged particle beam;
[0012] a particle detection unit configured to detect particles caused by irradiating the sample with the charged particle beam;
[0013] an image generating unit that generates an image of the sample based on the output from the particle detecting unit; and
[0014] Structural testing department,
[0015] The structure detection unit performs the following processing:
[0016] inputting an image of the specimen into a first learned model for detecting a first structure;
[0017] inputting the image of the specimen into a second learned model for detecting a second structure;
[0018] Obtaining a first detection result related to the first structure from the first learned model;
[0019] Obtaining a second detection result related to the second structure from the second learned model;
[0020] determining a position or region of the first structure according to the first detection result;
[0021] Determining the position or area of the second structure according to the second detection result; and
[0022] The integrated result information indicating the position or area of the first structure and the position or area of the second structure is output.
[0023] Effects of the Invention
[0024] The charged particle beam device of the present invention can detect the corresponding multiple structures by using a plurality of learned models. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 This is a front view schematically showing the charged particle beam device according to the first embodiment of the present invention.
[0026] Figure 2 This is an example of a GUI displayed on the display unit and showing an example of an integrated result image.
[0027] Figure 3 This is an example of a GUI for selecting a learned model, which is displayed on the display unit.
[0028] Figure 4 This is a diagram for explaining the processing of each learned model.
[0029] Figure 5 This is an example of an integrated result image using different colors for each structure.
[0030] Figure 6 This is an example of a composite result image using the same representation for multiple structures.
[0031] Figure 7 This is an example of a comprehensive result image using image processing.
[0032] Figure 8 This is an example of a comprehensive result image showing a rectangular frame for the overlapping structure.
[0033] Fig. 9 It means in Figure 8 An example of correction processing performed when overlapping is performed in the integrated result image and a diagram showing the state of the new structure.
[0034] Fig.10 It means in Figure 8 An example of correction processing performed when overlapping in the comprehensive result image, a diagram showing a state in which the structure of one side is given priority.
[0035] Fig.11 It means in Figure 8 An example of correction processing performed when overlapping is performed in the integrated result image of FIG. 1 , and a diagram showing a state in which an operation of selecting any one of the structures is accepted.
[0036] Fig.12 FIG. 4 is a diagram schematically showing an example of training data generated according to a selected structure. DETAILED DESCRIPTION
[0037] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.
[0038] Implementation Method 1
[0039] <Regarding the Structure of the Charged Particle Beam Device 100 >
[0040] Figure 1 1 is a front view schematically showing a charged particle beam device 100 according to Embodiment 1. The charged particle beam device 100 is a scanning electron microscope. However, the charged particle beam device may also be other charged particle beam devices such as a transmission electron microscope and an ion beam device. It should be noted that the structure of the charged particle beam device 100 is only an example. In other words, the specific structure of each part of the charged particle beam device 100 can vary depending on the type and structure of the charged particle beam device 100.
[0041] The charged particle beam device 100 has an irradiation unit 110 for irradiating a sample S with a charged particle beam. The irradiation unit 110 of this example has an electron source 111 and a lens barrel 112. The electron source 111 releases electrons that become the source of the charged particle beam (in this example, an electron beam). The lens barrel 112 has a focusing lens, a scanning coil, an objective lens, etc., and guides the charged particle beam to the sample S.
[0042] The irradiation unit 110 is connected to a sample chamber 120. Typically, the sample chamber 120 is evacuated by a vacuum pump (not shown).
[0043] In this example, a detection unit 130 that outputs a signal caused by irradiation of the sample S with a charged particle beam is provided in the sample chamber 120. The detection unit 130 may detect secondary electrons, reflected electrons, X-rays, Auger electrons, etc. In addition, a plurality of detection units 130 may be provided.
[0044] The detection unit 130 includes a particle detection unit that detects particles caused by irradiation of the sample S with a charged particle beam. The particle detection unit is, for example, a first detection unit 130A that detects secondary electrons. In addition, the detection unit 130 may also include an electromagnetic wave detection unit that detects electromagnetic waves caused by irradiation of the sample S with a charged particle beam. The electromagnetic wave detection unit is, for example, a second detection unit 130B that detects X-rays. The detection unit 130 is not limited to being disposed in the sample chamber 120. As one example, in a scanning electron microscope, the detection unit 130 is disposed inside the lens barrel 112. As another example, in a transmission electron microscope, the detection unit 130 is disposed at a position downstream of the flow of the electron beam than the sample S to detect the electron beam that has transmitted the sample S.
[0045] The charged particle beam device 100 has a stage 140. The stage 140 may be a movable stage. Typically, the stage 140 may be movable in the X direction and / or the Y direction (a direction in a plane perpendicular to the optical axis of the charged particle beam): Figure 1 The stage 140 is movable in the Z direction (direction of the arrow). Furthermore, the stage 140 can move in the Z direction (direction perpendicular to the optical axis of the charged particle beam). Furthermore, the stage 140 can rotate (rotate with the Z-axis direction as the rotation axis). Furthermore, the stage 140 can tilt (rotate with the X direction or the Y direction as the rotation axis). The stage 140 can be configured to support a sample holder 141. In this example, the sample S is placed on the sample holder 141.
[0046] The charged particle beam device 100 can be connected to the control unit 151, the input unit 152, the storage unit 153, and the display unit 154. The control unit 151, the input unit 152, the storage unit 153, and the display unit 154 can be part of the charged particle beam device 100, or can be independent of the charged particle beam device 100. The connection between each unit can be a wired connection or a wireless connection. Therefore, Figure 1 The connection lines shown in the figure are merely examples. In addition or in lieu thereof, a connection via a communication line such as the Internet may also be adopted. For example, the storage unit 153 may also be a storage unit 153 on an intranet, the Internet, or a cloud service. When the charged particle beam device 100 is connected to a plurality of levels of control units 151 or storage units 153, the control unit 151 or storage unit 153 at the upper level is sometimes referred to as an upper device.
[0047] In a typical charged particle beam device 100, the control unit 151 that receives the signal from the detection unit 130 can generate an image or a spectrum. Alternatively, the control unit 151 can analyze or evaluate the place or area (hereinafter referred to as "area" in this specification) of the sample S that is irradiated with the charged particle beam. In this example, the control unit 151 can generate an SEM image (here, a secondary electron image) based on the signal output by the first detection unit 130A. Similarly, the control unit 151 can generate an X-ray analysis image or spectrum based on the signal output by the second detection unit 130B. Therefore, the control unit 151 functions as an image generation unit that generates an image of the sample S based on the output from the detection unit 130. Furthermore, the control unit 151 can analyze or evaluate, for example, the surface shape or surface elements of the sample S based on these images or signals.
[0048] The input of information to the charged particle beam device 100 or the output of information from the charged particle beam device 100 can be performed via a UI (User Interface). In this example, the UI is a GUI (Graphical User Interface) and is displayed on the display unit 154 .
[0049] The control of the charged particle beam device 100, the generation of an image or spectrum, or the analysis or evaluation of the sample S by the control unit 151 can be performed based on the learned model stored in the storage unit 153. The storage unit 153 stores a plurality of learned models. In the present embodiment, the storage unit 153 stores a model M1 (a first learned model), a model M2 (a second learned model), and a model M3 (a third learned model), but the number of stored learned models may be 2 or more than 4.
[0050] Each learned model is configured to detect different structures by performing machine learning using different training data. More specifically, model M1 is a model for detecting a first structure, model M2 is a model for detecting a second structure, and model M3 is a model for detecting a third structure. The structure refers to, for example, the shape of a particle (spherical, a specific crystal shape, etc.), the shape of a foreign body, or the shape on a sample S (scratches, etc.). In addition, it is not limited to the shape in appearance, and may also include the structure of a portion having a specific composition.
[0051] In addition, each structure does not need to be dedicated to different types of substances, etc., and can also be configured so that the corresponding models can detect multiple structures that the same type of substance can have. For example, when a substance appears in the form of round particles or angular particles on the sample S, if a certain model is made to learn the round structure and another model is made to learn the angular structure, the same type of particles can be detected by multiple models.
[0052] The control unit 151 may also control the overall operation of the charged particle beam device 100. The control unit 151 may also implement the control by executing a program. The program may be stored in the storage unit 153 or in a storage unit in the control unit 151 that is independent of the storage unit 153.
[0053] The charged particle beam device 100 may further include a navigation image capturing device 160. The navigation image capturing device 160 captures a navigation image 206 (described later). Figure 1 In the example of FIG. 1 , the navigation image capture device 160 is provided in the sample chamber 120. Figure 1 In the embodiment, the stage 140 is located at Figure 1 When the navigation image capture device 160 is at the right end of the arrow, it is possible to capture at least a portion of the portion where the sample holder 141 is estimated to be present. Figure 1 In the example of FIG. 1 , the navigation image capturing device 160 is an optical camera and can capture images of all portions where the sample holder 141 is expected to be present. In addition, a light source for the navigation image capturing device 160 may be provided in the sample chamber 120 .
[0054] and Figure 1 Unlike the example of FIG. 1 , an image acquisition unit other than an optical camera may be used as the navigation image capture device 160. Figure 1 Different from the example, a navigation image capture device 160 that is independent of the charged particle beam device 100 may be used. For example, an optical microscope that is independent of the charged particle beam device 100, that is, an optical microscope that can fix the sample holder 141 at a certain position and in a certain direction, may be used as the navigation image capture device 160. In addition, the navigation image capture device 160 may be connected to the control unit 151, the input unit 152, the storage unit 153, and the display unit 154.
[0055] The control unit 151 also functions as a structure detection unit for detecting a specific structure. As the structure detection unit, the control unit 151 detects the structure using the learned model based on the image of the sample S. More specifically, the control unit 151 inputs the image of the sample S to the model M1, obtains the detection result of the first structure (first detection result) from the model M1, and determines the position or area of the first structure based on the first detection result.
[0056] Similarly, the control unit 151 inputs the image of the sample S to the model M2, obtains the detection result of the second structure (second detection result) from the model M2, and determines the position or area of the second structure based on the second detection result. In addition, the control unit 151 inputs the image of the sample S to the model M3, obtains the detection result of the third structure (third detection result) from the model M3, and determines the position or area of the third structure based on the third detection result.
[0057] Furthermore, the control unit 151 functions as a structure detection unit to integrate information indicating the positions or regions of the respective structures (eg, the first structure, the second structure, and the third structure), and generates and outputs integrated result information indicating these positions or regions.
[0058] <Information Regarding Output of Charged Particle Beam Device 100>
[0059] Figure 2 This is an example of a GUI displayed on the display unit 154. Such a GUI can be implemented as a screen 200 displayed on the display unit 154. The comprehensive result information can be configured as an image displayed in such a GUI. However, the output form of the comprehensive result information is not limited to an image, and it can also be expressed in a table or other forms.
[0060] The screen 200 includes a table portion 201 and an overlapping portion 202. The table portion 201 displays the evaluation results of the sample S in a table format. Figure 2 In the example, for each detected structure, the image number ("Image") corresponding to the area containing the structure, the number ("No") identifying the structure among the structures contained in the area, the evaluation result related to the structure ("Score"), the position of the structure ("X" and "Y"), and the area of the structure ("Area") are displayed.
[0061] In the overlapping portion 202, the evaluation result is displayed overlapping the image of the sample S. In this example, a plurality of integrated result images 203 are displayed in the overlapping portion 202. In each integrated result image 203, the position of the first structure is indicated by a black circle, and the position of the second structure is indicated by a gray pentagon, overlapping the image of the sample S. In this example, the third structure is not detected, but if the third structure is detected, the position of the third structure may also be indicated by a corresponding graphic.
[0062] For each integrated result image 203 , the evaluation result related to each structure included in the image is displayed as a numerical value.
[0063] In this example, only the position of each structure is shown, but when each structure is a region having a shape, the region may be displayed overlapping with the integrated result image 203. For example, the color of the pixels constituting each structure may be changed to a specific color (e.g., red or green). Additionally or alternatively, the pixels constituting each structure (the region constituted) may be changed to a specific brightness, or to have a specific hatching, and / or to flash in a specific pattern.
[0064] The screen 200 may also include a mapping unit 204. The symbols 205 corresponding to the respective integrated result images 203 are displayed in the mapping unit 204. The display mode of the symbols 205 changes according to the information associated with the corresponding integrated result images 203. The change in the display mode includes a change in shape, color, hue, shade, size, position, etc. In addition, the change in the display mode includes a change in the brightness, hatching, flashing pattern, etc., which can distinguish two or more display modes by visual confirmation.
[0065] For example, the shape of the symbol 205 (a rectangle in this example) can be determined based on the shape of the comprehensive result image 203, for example, it can be the same as the shape of the comprehensive result image 203. In addition, the color or shade of the symbol 205 can also be determined based on the evaluation result related to the structure of interest included in the comprehensive result image 203. In addition, the size of the symbol 205 can be determined based on the size of the comprehensive result image 203 in the sample S, for example, the relative size of each symbol 205 can be determined based on the relative size of each comprehensive result image 203. In addition, the position of the symbol 205 in the mapping unit 204 can also correspond to the position of the comprehensive result image 203 in the sample S.
[0066] In addition, the screen 200 may also include a navigation image 206. The navigation image 206 is an image for the user to visually specify a search range in the sample S placed on the sample holder 141, and is typically captured by the navigation image capture device 160. However, as long as the observation magnification of the charged particle beam device 100 can be set to be sufficiently low, an image such as a SEM image may be used as the navigation image 206. The navigation image 206 can be used, for example, to set or display a search area that the charged particle beam device 100 should search for the sample S.
[0067] <About the specific processing of the learned model>
[0068] Figure 3 is another example of a GUI displayed on the display unit 154. This GUI is used to select a learned model. Figure 3 In the example of , information for specifying three models (eg, models M1, M2, and M3) is input. The control unit 151 inputs an image of the sample S to each of the models specified here.
[0069] In addition, you can enter information that specifies the macros used for each model. Figure 3 In the example of , a macro is specified for model M1, but no macro is specified for models M2 and M3. A macro is, for example, a program for performing image processing, and is used to receive information output by a learned model as an image and perform specific image processing on the image.
[0070] use Figure 4 , the processing of each learned model is explained. Figure 4 (a) is an image of the sample S. The control unit 151 can generate such an image based on the signal from the detection unit 130 . Figure 4 (b) shows the position or area of the first structure 401 determined based on the detection result of the model M1. Figure 4 (c) shows the position or area of the second structure 402 determined based on the detection result of the model M2. Although the third structure is not shown in the figure, the processing is performed in the same manner.
[0071] The output of each model does not need to be directly Figure 4 (b) or Figure 4 (c) (a form indicating an area in an image). For example, a certain model may output a heat map based on an image. A heat map indicates, for example, the possibility of a structure existing by pixel (for example, reliability, probability, or likelihood). In this case, the control unit 151 performs an operation including, for example, binary processing based on the heat map, thereby being able to Figure 4 (b) or Figure 4 The position or area of each structure is determined as shown in (c).
[0072] <About Display Example of Comprehensive Result Image>
[0073] Figure 5 An example of an integrated result image 203 as an output example of integrated result information is shown. The integrated result image 203 shows the position or area of the structure (for example, the first structure 401 and the second structure 402) detected by each learned model in the image of the sample S.
[0074] Each structure can be represented using different expressions. For example, the control unit 151, as a structure detection unit, displays a first expression at a position or region of the first structure 401 and displays a second expression at a position or region of the second structure 402 in the image of the sample S, thereby generating and outputting the comprehensive result image 203. For example, the first expression is displayed at a position or region of the first structure 401 and the second expression is displayed at a position or region of the second structure 402, thereby generating the comprehensive result image 203. The first expression and the second expression can be displayed in superposition with the image of the sample S, or can be displayed by replacing a part of the image of the sample S.
[0075] Different expressions can be achieved using different colors, different tones, different shades, different coating patterns (vertical stripes, horizontal stripes, diagonal stripes, dot patterns, gradients, etc.), different flashing states, etc. As a more specific example, in Figure 5In the example, the first representation of the first structure 401 is represented by the first color (i.e., gray), and the second representation of the second structure 402 is represented by the second color (i.e., white). In the case where color display is possible, the first structure 401 may be represented by red, for example, and the second structure 402 may be represented by green, for example.
[0076] In this way, the charged particle beam device 100 can detect the corresponding types of structures by using a plurality of learned models. In addition, by using different expressions for different structures, the user of the charged particle beam device 100 can easily identify each structure.
[0077] Figure 6 203 shows another example of the integrated result image 203. It is not necessary to use different representations for each structure detected by each learned model, and the same representation may be used for multiple or all learned models. Figure 6 In the example of , the first structure 401 detected by the model M1 and the third structure 403 detected by the model M3 are represented by the same representation (gray). In this way, the integrated result image 203 can also be generated by further displaying the first representation or the second representation at the position or region of the third structure 403 in the image of the sample S.
[0078] In the case where the structure of an object is not limited to one, it is useful to use the same representation for multiple structures detected by multiple learned models. For example, if a certain type of particle can have various shapes, it is easy to identify the particle if different models are used to learn the assumed shapes and the same representation is used for all structures detected by these models.
[0079] Figure 7 FIG. 2 shows another example of the integrated result image 203. Before displaying the detection result of any structure on the image of the sample S, image processing may be performed on the detection result. Figure 7 In the example, for the detection result of the first structure 401, use Figure 3 The macro specified in the above manner performs image processing, and as a result, the range of the first structure 401 is larger than Figure 5 The range shown is narrow.
[0080] The specific content of the image processing can be appropriately designed by a person skilled in the art, for example, it can be set to a process for determining the range (or area) of the structure based on the image of the sample S and the position of each structure. For example, if the position where a certain structure exists is shown in the image of the sample S, the range (or area) of the structure can be determined by performing appropriate image processing on the range (or nearby area) containing the position of the sample S. In addition, in this example, the range (or area) of the structure is determined by image processing, but image processing does not need to determine the range (or area) of the structure, and can also be used to determine only the position.
[0081] According to such image processing, the process of detecting the position or area of each structure can be separated from the process of detecting the range of the structure, and the best process can be used for each, so that the appropriate processing result can be obtained as a whole. For example, when the model M1 can appropriately detect the existence of the first structure 401, but its range (or area) tends to be larger than the actual detection, the range (or area) is detected by image processing, thereby obtaining a result with appropriate position and size. In addition, the specific content of the image processing can also be different according to the structure (i.e., according to the learned model). In addition, the image to be processed can be the result image extracted by machine learning as the object, or the image of the sample S before extraction can be used.
[0082] In this way, the control unit 151, as a structure detection unit, determines the position or area of the first structure 401 through image processing based on the image of the sample S and the first detection result. Similarly, the control unit 151 can also determine the position or area of the second structure 402 through image processing based on the image of the sample S and the second detection result. In addition, when the third structure is detected, the control unit 151 can also determine the position or area of the third structure through image processing based on the image of the sample S and the third detection result.
[0083] Based on the detection results of each structure, the charged particle beam device 100 may also perform additional processing on the sample S. For example, the sample S may be analyzed at a position or area corresponding to the position or area of each structure (the first structure, the second structure, or the third structure). The analysis of the sample S may include shape analysis and composition analysis. For example, the position or area where the first structure 401 is detected (as an example, the position or area corresponding to the image of Figure 4 (a)) may be further irradiated with a charged particle beam via the irradiation unit 110, and the signal caused by the irradiation may be detected via the detection unit 130.
[0084] The analyzed signal may be, for example, a signal based on electrons generated when a charged particle beam is irradiated onto the sample S, or a signal based on X-rays generated when a charged particle beam is irradiated onto the sample S. In particular, if the detection of each structure is performed by a signal based on electrons and the analysis of each structure is performed by a signal based on X-rays, the search can be performed efficiently and the analysis can be performed with high precision. In the analysis, for example, high-magnification photography using energy dispersive X-ray spectroscopy (EDS) or electron backscatter diffraction (EBSD) can be utilized. In addition, a highly accelerated or high-current charged particle beam may also be used in the photography.
[0085] <Specific Example of Signal Acquisition Based on X-Rays>
[0086] It is also possible to select whether to perform X-ray analysis based on the detection results of each structure based on each model. For example, when model M1 is a model for detecting foreign matter whose element is to be determined, and model M2 is a model for detecting other than foreign matter, by performing X-ray analysis only on model M1, unnecessary X-ray analysis can be eliminated, and analysis can be performed efficiently.
[0087] Figure 8 Another example of the integrated result image 203 is shown. Inside the rectangular frame 801, the position or area of the first structure 401 overlaps with the position or area of the second structure 402. Here, those skilled in the art can appropriately interpret the meaning of "position or area overlap", for example, including a case where the positions of the structures in the image of the sample S are consistent, a case where the areas of the structures in the image of the sample S partially or completely overlap, etc.
[0088] The control unit 151 may function as a structure detection unit and, when the positions or regions of the structures overlap, execute overlap correction processing based on the structures. Figure 8 The example shows a case where the position or region of the first structure 401 overlaps with the position or region of the second structure 402, and correction processing is performed based on the first structure 401 and the second structure 402 when overlapping.
[0089] By appropriately performing the overlap correction process, when a plurality of models detect the same object, a more appropriate integrated result image 203 can be obtained.
[0090] <Specific example of processing when overlapping>
[0091] The overlap correction process may also include a process of outputting information indicating the overlap of positions or regions for each structure where the positions or regions overlap. Figure 8In the example of FIG. 4 , a rectangular frame 801 indicating overlap is displayed so as to surround the portion where the positions or regions of the first structure 401 and the second structure 402 overlap. The rectangular frame 801 overlaps with the image of the sample S, for example.
[0092] The information indicating overlap may also be expressed in a form other than the rectangular frame 801. For example, it may be a circle, or a form other than a graphic (such as a message). Figure 8 In the example of FIG. 4 , the information indicating the overlap is displayed in association with the positions or regions of the first structure 401 and the second structure 402 , but the display position of the information indicating the overlap is not limited thereto.
[0093] Fig. 9 Indicated in Figure 8 The state of an example of overlapping correction processing performed in the comprehensive result image 203. In this example, the overlapping correction processing includes processing to determine the position or area of the new structure (fourth structure 404) based on the first structure 401 and the second structure 402 whose positions or regions overlap. The comprehensive result image 203 indicates the position or area of the fourth structure 404. The fourth structure 404 can be obtained as a logical OR or a logical AND of the areas of the overlapping structures, for example. In addition, for example, in the image of the sample S, the fourth representation is displayed at the position or area of the fourth structure 404.
[0094] The fourth structure 404 is detected as a structure having two characteristics. That is, when a certain structure is detected as both the first structure 401 and the second structure 402, the structure may be a specific structure having these two characteristics. The overlap correction process may also be a process of detecting the fourth structure 404 as a new type of structure different from both the first structure 401 and the second structure 402.
[0095] By detecting a new structure in accordance with the overlap in this way, it is possible to perform detection processing under more detailed conditions.
[0096] Fig.10 Indicated in Figure 8 Another example of a state in which the overlap correction process is performed in the integrated result image 203 of FIG. In this example, the overlap correction process is performed using the priority rule between the structures. Fig.10 In the example, as a result of giving priority to the first structure 401, the first structure 401 is displayed in the image of the sample S (gray), and the second structure 402 is not displayed in the image of the sample S (white) (or Figure 8 The comprehensive result image 203 of the state is subtracted).
[0097] Such overlap correction processing can be said to be processing for the first structure 401 and the second structure 402 whose positions or regions overlap, so that the position or region of the first structure 401 or the position or region of the second structure 402 is not included in the integrated result information.
[0098] Priority rules can also be defined between more than three structures. For example, consider the case where the first structure is set to the first position, the second structure is set to the second position, and the third structure is set to the third position. When all three structures overlap, only the first structure is displayed in the image of sample S. When the first structure and the second structure overlap and when the first structure and the third structure overlap, only the first structure is displayed in the image of sample S. When the second structure and the third structure overlap, only the second structure is displayed in the image of sample S.
[0099] The specific content of the priority rule (such as the priority of each structure, etc.) can be found in Figure 3 The GUI shown or other GUI is input to the charged particle beam device 100. In addition, the priority rule is not limited to the order, and can also be realized by a process of calculating the priority score of each structure.
[0100] By using the priority rule in this way, more appropriate detection processing can be performed.
[0101] Fig.11 Indicated in Figure 8 The state of another example of the overlap correction process performed in the integrated result image 203 of FIG. In this example, the overlap correction process includes a process of accepting an operation of selecting one of the structures for each structure whose position or region overlaps. For example, for the first structure 401 and the second structure 402 whose position or region overlaps, an operation of selecting one of the first structure 401 and the second structure 402 is accepted.
[0102] exist Fig.11 In the example, the user of the charged particle beam device 100 can select the first structure 401 by operating the radio button “A” and the “OK” button in the selection window 1101 , and can select the second structure 402 by operating the radio button “B” and the “OK” button.
[0103] In this way, by allowing the user to select an appropriate structure, appropriate detection processing can be performed based on human judgment.
[0104] The processing after the selection operation is received can be appropriately designed by those skilled in the art, for example, Fig.10 As shown, only the representation of the selected structure is displayed in the image of the sample S.
[0105] In addition, training data for the learning model may be generated based on the selected structure. That is, the overlap correction process may also include the process of generating training data used in machine learning. Fig.11 In the state of , the training data generated when the first structure 401 is selected is training data indicating that the unselected second structure 402 does not appear in the image of the specimen S. On the other hand, the training data generated when the second structure 402 is selected is training data indicating that the unselected first structure 401 does not appear in the image of the specimen S.
[0106] Fig.12 An example of such training data is schematically shown. In this example, the training data is shown in a form such as the integrated result image 203, but the form of the training data can be appropriately designed according to the learning model. Fig.12 The example corresponds to Figure 8 The first structure 401 is selected in the integrated result image 203. The unselected structures in the second structure 402 (the structures within the rectangular frame 801) are not shown in the image of the sample S.
[0107] The generated training data can be used for new learning models, or can be used as feedback for relearning models in use (such as models M1, M2, and M3).
[0108] In this way, training data is generated according to the user's selection, thereby, a more appropriate learned model can be generated. In particular, when the structures of two particles are similar to each other, it is sometimes difficult to identify them based on the learned model, but in such a case, training data for more accurately identifying the two particles can be generated.
[0109] In addition, any of the structures with overlapping positions or regions may be classified by a classifier. For example, the control unit 151 may also function as the classifier. That is, the overlap correction process may also include the following process: for the first structure 401 and the second structure 402 with overlapping positions or regions, one of the first structure 401 and the second structure 402 is selected according to the position or region of the first structure 401 and the position or region of the second structure 402.
[0110] The specific processing content of the classifier can be appropriately designed by those skilled in the art, for example, it can be selected according to the area of the region of each structure, more specifically, it is also possible to select a structure with a larger area. Alternatively, when the charged particle beam device 100 performs some analysis or evaluation on each structure, it is also possible to select according to the analysis result or the evaluation result. The classifier can be constructed using image processing technology, or it can be constructed using a learned model prepared for the classifier.
[0111] The processing after the selection based on the classifier can be appropriately designed by those skilled in the art, for example, Fig.10 Only the selected structures are shown in the image of the sample S.
[0112] By performing selection using a classifier in this way, an appropriate structure can be selected efficiently.
[0113] Explanation of symbols
[0114] 100…Charged particle beam device
[0115] 110…Irradiation Department
[0116] 111…Electron source
[0117] 112…Eyepiece
[0118] 120…Sample Room
[0119] 130…Detection Department
[0120] 140…Stage
[0121] 141…Specimen holder
[0122] 151 ...control unit (image generation unit, structure detection unit)
[0123] 152…Input
[0124] 153…Storage
[0125] 154…Display unit
[0126] 160…Navigation image capture machine
[0127] 200…screen
[0128] 201…Formal Department
[0129] 202…Overlapping
[0130] 203…Comprehensive result image
[0131] 204…Mapping Department
[0132] 205…Symbol
[0133] 206…Navigation Image
[0134] 401…First Structure
[0135] 402…Second Structure
[0136] 403…The third structure
[0137] 404…The fourth structure
[0138] 801…Rectangular frame
[0139] 1101…Select Window
[0140] S...Sample
[0141] M1…model (first learned model)
[0142] M2…model (second learned model)
[0143] M3…model (the third learned model).
Claims
1. A charged particle beam device, characterized in that have: an irradiation unit for irradiating the sample with a charged particle beam; a particle detection unit configured to detect particles caused by irradiating the sample with the charged particle beam; an image generating unit that generates an image of the sample based on the output from the particle detecting unit; and Structural testing department, The structure detection unit performs the following processing: inputting an image of the specimen into a first learned model for detecting a first structure; inputting the image of the specimen into a second learned model for detecting a second structure; Obtaining a first detection result related to the first structure from the first learned model; Obtaining a second detection result related to the second structure from the second learned model; determining a position or region of the first structure according to the first detection result; determining a position or region of the second structure according to the second detection result; as well as The integrated result information indicating the position or region of the first structure and the position or region of the second structure in different expressions is output.
2. The charged particle beam device according to claim 1, characterized in that The comprehensive result information is represented by a comprehensive result image showing the position or region of the first structure and the position or region of the second structure in the image of the sample, The structure detection unit generates the integrated result image by displaying a first representation at a position or region of the first structure and a second representation at a position or region of the second structure in the image of the sample.
3. The charged particle beam device according to claim 2, characterized in that The first expression is realized by a first color, The second representation is achieved by a second color.
4. The charged particle beam device according to claim 2, characterized in that The structure detection unit performs the following processing: inputting the image of the specimen into a third learned model for detecting a third structure; Obtaining a third detection result related to the third structure from the third learned model; and Determine the position or area of the third structure according to the third detection result, The integrated result image is generated by further displaying the first representation or the second representation at a position or region of the third configuration in the image of the specimen.
5. The charged particle beam device according to claim 1, characterized in that The structure detection unit performs the following processing: Determining the position or area of the first structure by image processing based on the image of the sample and the first detection result; and The position or region of the second structure is determined by image processing based on the image of the sample and the second detection result.
6. The charged particle beam device according to claim 1, characterized in that The charged particle beam device analyzes the sample in a region corresponding to the position or region of the first structure or the position or region of the second structure.
7. The charged particle beam device according to claim 1, characterized in that When the position or region of the first structure overlaps with the position or region of the second structure, the structure detection unit performs overlapping correction processing based on the first structure and the second structure.
8. The charged particle beam device according to claim 7, characterized in that The overlap correction process includes a process of determining the position or area of a fourth structure based on the first structure and the second structure whose positions or areas overlap. The integrated result information also indicates the location or area of the fourth structure.
9. The charged particle beam device according to claim 7, characterized in that: The overlap correction process includes the following process: for the first structure and the second structure whose positions or regions overlap, the position or region of the first structure or the position or region of the second structure is not included in the integrated result information.
10. The charged particle beam device according to claim 7, characterized in that The overlap correction process includes a process of outputting information indicating that the positions or regions overlap with each other, with respect to the first structure and the second structure whose positions or regions overlap with each other.
11. The charged particle beam device according to claim 10, characterized in that: The overlap correction process further includes a process of accepting an operation of selecting one of the first structure and the second structure whose positions or regions overlap.
12. The charged particle beam device according to claim 7, characterized in that The overlap correction process also includes the following process: for the first structure and the second structure whose positions or regions overlap, one of the first structure and the second structure is selected according to the position or region of the first structure and the position or region of the second structure.
13. A charged particle beam device, characterized in that: have: an irradiation unit for irradiating the sample with a charged particle beam; a particle detection unit configured to detect particles caused by irradiating the sample with the charged particle beam; an image generating unit that generates an image of the sample based on the output from the particle detecting unit; and Structural testing department, The structure detection unit performs the following processing: inputting an image of the specimen into a first learned model for detecting a first structure; inputting the image of the specimen into a second learned model for detecting a second structure; Obtaining a first detection result related to the first structure from the first learned model; Obtaining a second detection result related to the second structure from the second learned model; determining a position or region of the first structure according to the first detection result; determining a position or region of the second structure according to the second detection result; as well as outputting comprehensive result information indicating the position or region of the first structure and the position or region of the second structure in different expressions, When the position or area of the first structure overlaps with the position or area of the second structure, the structure detection unit performs overlapping correction processing based on the first structure and the second structure. The overlap correction process includes the following processes: For the first structure and the second structure whose positions or regions overlap, outputting information indicating that these positions or regions overlap; For the first structure and the second structure whose positions or regions overlap, accepting an operation of selecting one of the first structure and the second structure; as well as The process of generating training data used in machine learning, When the first structure is selected, the training data is training data indicating that the second structure that is not selected does not appear in the image of the sample. When the second structure is selected, the training data is training data indicating that the first structure that is not selected does not appear in the image of the sample.