Image processing method and electron microscope system

Abstract

Provided is an image processing method for determining a structure boundary position between a first phase and a second phase in a microscope image of a sample in which the first phase and the second phase are in contact with each other in a structure boundary region, wherein a luminance profile of the microscope image is obtained in a direction intersecting the longitudinal direction of the structure boundary region, a moving standard deviation profile is obtained from the luminance profile using a plurality of calculation widths, the calculation width of the moving standard deviation profile in which a convex waveform of the moving standard deviation profile has a peak and the half-value width of the convex waveform is narrowest is specified, and the peak position of the moving standard deviation profile calculated using the specified calculation width is determined as the structure boundary position. Thus, there is provided an image processing method in which the effect of the noise in a microscope image is small and boundary information detection is highly accurate.

Description

Image processing method and electron microscope system

[0001] The present invention relates to an image processing device and an electron microscope system.

[0002] In processing microscopic images, there are several methods for extracting edges from noisy images, but general noise removal processing requires the selection of a processing method and parameter adjustment, and also reduces positional accuracy. For example, Patent Document 1 is a document related to edge extraction of image data, and discloses a method for calculating the difference in luminance variance between a first pixel of image data and a second pixel surrounding the first pixel, and determining edge pixels of the image data based on the difference in luminance variance.

[0003] Semantic segmentation using machine learning has also become widely recognized in recent years, but pre-training requires a lot of effort and time.

[0004] Japanese Patent Application Laid-Open No. 2017-163407

[0005] Transmission electron microscopes (TEM) and scanning transmission electron microscopes (STEM) use thin-film samples to observe electrons transmitted through or diffracted from the sample. Proper TEM / STEM observations require sample orientation, aligning the optical axis of the electron beam with the sample's three-dimensional structure. Sample orientation is typically achieved using the diffraction pattern of the single crystal. However, when the sample being observed is a semiconductor device without a single crystal, the orientation must be aligned with the structure of the thin-film sample. For example, if the sample being observed is a thin-film sample containing phases A and B, the current practice is for the operator to visually determine the sample angle so that the structural boundary between phases A and B is parallel to the electron beam by visually judging the TEM / STEM image. Therefore, sample orientation is highly dependent on the individual and highly difficult task, making it difficult for operators unfamiliar with TEM / STEM operation to perform sample orientation with high accuracy.

[0006] An image processing method according to one embodiment of the present invention is an image processing method for determining a structural boundary position between a first phase and a second phase in a microscope image of a sample, in which the first phase and the second phase contact each other at a structural boundary region. The image processing method includes the following steps: a first step of specifying, in the microscope image, a second direction that includes the structural boundary region and intersects with a first direction that is the longitudinal direction of the structural boundary region; a second step of obtaining an intensity profile along the second direction; a third step of calculating a moving standard deviation of the intensity profile using a plurality of calculation widths to obtain a moving standard deviation profile; a fourth step of calculating, for each moving standard deviation profile calculated using the plurality of calculation widths, a half-width of a convex waveform included in the moving standard deviation profile; a fifth step of specifying, among the plurality of moving standard deviation profiles, a calculation width of a moving standard deviation profile in which the convex waveform has a peak and the half-width of the convex waveform is narrowest; and a sixth step of determining, as the structural boundary position, the position of the peak in the moving standard deviation profile calculated using the specified calculation width.

[0007] The present invention provides an image processing method that is less affected by noise in microscope images and has high accuracy in detecting boundary information, and an electron microscope system that uses the same. Other objects and novel features will become apparent from the description of this specification and the accompanying drawings.

[0008] 1 is a diagram showing a flow for determining a structure boundary position; FIG. 2 is a diagram showing a flow for determining a structure boundary width; FIG. 3 is a diagram specifically showing a process for determining a structure boundary position / width for a model image; FIG. 4 is a diagram specifically showing a process for determining a structure boundary position / width for a model image; FIG. 5 is a diagram specifically showing a process for determining a structure boundary position / width for a model image; FIG. 6 is a diagram specifically showing a process for determining a structure boundary position / width for a model image; FIG. 7 is a diagram specifically showing a process for determining a structure boundary position / width for a model image; FIG. 8 is a diagram specifically showing a process for determining a structure boundary position / width for a TEM image; FIG. 9 is a diagram specifically showing a process for determining a structure boundary position / width for a TEM image; FIG. 10 is a diagram specifically showing a process for determining a structure boundary position / width for a TEM image; FIG. 1 is a diagram schematically illustrating the basic configuration of a scanning transmission electron microscope system. FIG. 2 is an example of the hardware configuration of a computer. FIG. 3 is a diagram for explaining the relationship between a structure boundary surface and a structure boundary region. FIG. 4 is a diagram illustrating a sample tilt adjustment flow for aligning the sample orientation. FIG. 5 is a diagram illustrating the relationship between a sample orientation angle and a structure boundary width. FIG. 6 is a diagram illustrating the relationship between a sample orientation angle and a structure boundary width. FIG. 7 is a diagram for explaining a method for determining a structure boundary position and a structure boundary width according to a modified example.

[0009] FIG. 4A shows a schematic diagram of the basic configuration of a scanning transmission electron microscope (STEM) system. The STEM system includes a STEM 51 and a control system 42. The STEM 51 includes an electron optical system and a detector 31, whose main components include an electron beam source 1, an extraction electrode 2, a condenser lens 11, a condenser aperture 12, an axis adjustment deflector 13, an astigmatism corrector 14, a scan deflector 15, an objective lens 20, an objective aperture 23, an axis adjustment deflector 24, a selected area aperture 25, and an imaging lens 30. A sample 21 is placed on a sample stage 22 and inserted between the upper and lower magnetic poles of the objective lens 20. Note that the imaging lens 30 is not necessarily required for the STEM to function.

[0010] A primary electron beam (probe) generated by the electron beam source 1 is extracted by an extraction electrode 2 and directed toward a sample 21. A condenser lens 11 adjusts the focusing conditions of the primary electron beam. A condenser aperture 12 controls the aperture angle of the primary electron beam. An axis adjustment deflector 13 adjusts the position of the primary electron beam relative to the objective lens 20. An astigmatism corrector 14 adjusts the beam shape of the primary electron beam incident on the sample 21. A scan deflector 15 raster-scans the primary electron beam incident on the sample 21. The objective lens 20 adjusts the focus position of the primary electron beam relative to the sample 21. An objective aperture 23 transmits transmitted electrons and / or specific diffracted electrons. An axis adjustment deflector 24 adjusts the position of the transmitted electron beam that has passed through the sample 21 relative to a detector 31. A selected area aperture 25 limits the area of ​​the sample when displaying an electron diffraction pattern. An imaging lens 30 forms an image of the transmitted electron beam. The detector 31 detects the imaged transmitted electron beam.

[0011] The control system 42 includes a control device 40 and a computer 41. The computer 41 stores a control program and an image processing program for the STEM 51, and by executing these programs, the computer 41 functions as a control unit that controls the STEM 51 or an image processing unit that processes the STEM images generated by the STEM 51. The control device 40 receives control signals from the computer 41 and controls each component of the STEM 51. The control device 40 includes an AD converter, a DA converter, a memory, and an arithmetic device such as an FPGA or a microprocessor.

[0012] The process by which the STEM 51 generates a STEM image will be described. This process is controlled by the control system 42. The electron beam source 1 irradiates the sample 21 on the sample stage 22 with a primary electron beam. The focus of the primary electron beam is roughly adjusted by adjusting the Z position of the sample stage 22 or adjusting the control parameters of the objective lens 20. Then, the XY movement mechanism of the sample stage 22 is used to select a field of view for adjusting the electron optical system. The deviation, focus, and astigmatism of the optical system are corrected in the field of view for adjusting the electron optical system. Specifically, the adjustment parameters of the axis adjustment deflector 13, the astigmatism corrector 14, and the objective lens 20 are adjusted.

[0013] Next, the specimen stage 22 is used to move the observation field to the photographing field, and the focus of the objective lens 20 is fine-tuned by an operator or adjusted to an appropriate focus position using the focus tracking function so that a sharp image can be observed. When the primary electron beam is incident on the specimen 21, most of the electrons are transmitted through the specimen 21. The imaging lens 30 directs the transmitted electron beam onto the detector 31 at an appropriate angle, thereby obtaining an STEM image. The magnification of the STEM image is set by the scan deflector 15.

[0014] 4B shows an example of the hardware configuration of the computer 41. The computer 41 includes a processor 411, a memory (main storage device) 412, an auxiliary storage device 413, an output device 414, an input device 415, and a communication interface (I / F) 416. These components are connected to each other via a bus 417 so that they can communicate with each other.

[0015] The memory 412 is composed of, for example, a semiconductor memory, and is mainly used to store programs and data currently being executed. The processor 411 executes various processes in accordance with the programs stored in the memory 412. The processor 411 operates in accordance with the programs, thereby realizing various functional units. The auxiliary storage device 413 is composed of, for example, a large-capacity storage device such as a hard disk drive or solid-state drive, and is used to store programs and data for a long period of time. The programs and data stored in the auxiliary storage device 413 are loaded into the memory 412, and the processor 411 executes the programs, thereby performing various processes of the computer 41.

[0016] The input device 415 is a hardware device that allows an operator to input instructions, information, etc. to the computer 41. The output device 414 is a hardware device that presents various images for input and output, such as a display device or a printing device. The communication I / F 416 is an interface for connection to a network.

[0017] Fig. 1A shows a structural boundary position determination flow for determining a structural boundary position in a microscope image, and Fig. 1B shows a structural boundary width determination flow for determining a structural boundary width in a microscope image. These flows are processes realized by the computer 41 executing a program stored in the auxiliary storage device 413. Figs. 2A to 2F show how the structural boundary position and structural boundary width of the model image 101 are determined according to the flows in Figs. 1A and 1B. The flows in Figs. 1A and 1B will be described below with reference to Figs. 2A to 2F. First, the flow in Fig. 1A will be described.

[0018] First, a microscope image is acquired by the STEM 51 (S01). The model image 101 shown in Fig. 2A is a model image created assuming a TEM image of a thin film sample having a structure in which three phases are arranged (hereinafter, images acquired by a TEM or STEM are collectively referred to as TEM images (transmission electron microscope images)). An example of determining the structural boundary position and structural boundary width at the boundary between the first phase 103 and the second phase 104 will be described below.

[0019] Next, an evaluation area is specified on the acquired microscope image (S02). For example, an operator can specify an evaluation area using the input device 415 for the microscope image displayed on the output device 414. The evaluation area 102 shown in FIG. 2A is set as a strip-shaped area extending in the y direction and the x direction perpendicular thereto, where the direction in which the boundary between phase 103 and phase 104 extends is the y direction. The size of the evaluation area 102 is set to one pixel or more in the y direction and to a size in the x direction that includes at least the phase boundary and the phases 103 and 104 adjacent to each other at the boundary. In the example of FIG. 2A, the size of the evaluation area 102 in the x direction is set to the entire range in the x direction of the model image 101.

[0020] Next, the luminance profile of the evaluation area 102 is calculated (S03). FIG. 2B shows the luminance profile calculated for the evaluation area 102. The horizontal axis indicates the position (pixel position) in the x direction, and the vertical axis indicates the luminance. Actual microscope images contain noise. The influence of noise can be reduced by setting the evaluation area 102 to a size of k pixels in the y direction and, for example, setting the average luminance of the k pixels as the luminance at the pixel position in the x direction.

[0021] Next, a calculation width is set (S04), and the moving standard deviation of luminance at each pixel position in the x direction of the evaluation region is calculated using the set calculation width, thereby obtaining a moving standard deviation profile (S05). For example, if the calculation width is set to 7 pixels, the standard deviation at the i-th pixel position is obtained for each pixel position as the standard deviation of luminance at seven pixel positions from (i-3) pixel to (i+3) pixel. Multiple calculation widths are set, and the moving standard deviation profile of luminance calculated for each calculation width is displayed (S06). Figure 2C shows the moving standard deviation profile of luminance calculated by setting five calculation widths A to E (where A < B < C < D < E) for the evaluation region 102. As such, a convex waveform appears in the moving standard deviation profile at the position where the phase boundary exists. Based on this convex waveform, the structural boundary position and structural boundary width are identified.

[0022] First, peaks of the convex waveform are extracted (S07). In the graph of FIG. 2C, the convex waveforms for calculation widths A and B have flat tops and no peaks, whereas peaks are observed for the convex waveforms for calculation widths C, D, and E. This difference in the shape of the convex waveform occurs because, for calculation widths A and B, the calculation width is short relative to the structure boundary width W, and the calculation width at multiple pixel positions is included within the structure boundary region 105. In contrast, when the calculation width is sufficiently long relative to the structure boundary width W (for calculation widths C, D, and E), peak p appears in the convex waveform.

[0023] Next, the half-width of each convex waveform is calculated, and the half-width of the convex waveform for each calculation width is displayed graphically (S08). Figure 2D is a graph showing the half-width of the convex waveform for each calculation width calculated from the moving standard deviation profile of Figure 2C. Of the calculation widths C, D, and E where peak p appears in the convex waveform, the half-width of the convex waveform for calculation width C is the smallest, and the graph in Figure 2D also shows that the convex waveform obtained is the steepest. When the calculation width is approximately the same as the structural boundary width W, the convex waveform becomes a steep waveform with a peak. As the calculation width increases beyond the structural boundary width W, the proportion of the phase portion with little brightness fluctuation in the calculation width increases, causing the base of the convex waveform to broaden. From the above, in this example, it can be determined that the convex waveform of the moving standard deviation profile of brightness calculated for calculation width C most accurately represents the structural boundary region. Therefore, the calculation width for determining the structure boundary position and structure boundary width is specified as the calculation width where a peak appears and a convex waveform with the smallest half-width is obtained (calculation width C in this case) (S09). The position where the peak of the convex waveform of the moving standard deviation profile of brightness calculated using the specified calculation width appears is displayed on the microscope image as the structure boundary position (S10). That is, FIG. 2F shows the position where the peak of the convex waveform exists in the case of calculation width C as the structure boundary position 112 on the model image 101. In this way, by using the moving standard deviation of brightness as an index, the structure boundary position can be determined even from a noisy microscope image without image processing for noise removal.

[0024] Next, the flow of FIG. 1B will be described. In this flow, processing is performed on the moving standard deviation profile of luminance calculated using the calculation width (calculation width C in this example) determined in the flow of FIG. 1A. First, element differences in the moving standard deviation profile of luminance between adjacent pixel positions are calculated to obtain a moving standard deviation difference profile (S11) and display it (S12). FIG. 2E shows the moving standard deviation difference profile of the moving standard deviation profile of luminance calculated using calculation width C. Note that, although an example is described here in which difference values ​​are calculated by treating the waveform of the standard deviation of luminance as discrete values ​​for each pixel position, similar processing can also be performed by treating the waveform of the standard deviation of luminance as continuous values ​​to calculate differential values ​​at pixel positions and creating a moving standard deviation differential profile of luminance.

[0025] Next, a pair of maximum and minimum points in the standard moving deviation difference profile is extracted (S13). The moving standard deviation difference graph in FIG. 2E shows a maximum point 110 and a minimum point 111. The region sandwiched between the positions where the difference in moving standard deviation reaches an extreme value is identified as the structure boundary region 105. Therefore, a structure boundary width W is calculated from the difference between the positions where the standard moving deviation difference profile reaches an extreme value (S14), and the structure boundary width W is displayed on the microscope image (S15). FIG. 2F shows the positions where the standard moving deviation difference profile reaches an extreme value as boundaries 113, 114 of the structure boundary region 105 on the model image 101.

[0026] As will be described in detail later, in the model image 101 of a thin film sample shown in FIG. 2A , the appearance of the structural boundary surface changes depending on the relationship between the structural boundary surface between phases 103 and 104 and the optical axis of the electron beam. When the structural boundary surface between phases 103 and 104 coincides with the optical axis of the electron beam, the structural boundary width W of the structural boundary region 105 is minimized. The greater the degree of mismatch, the larger the structural boundary width W of the structural boundary region 105. In this embodiment, the structural boundary position and structural boundary width of the microscope image can be determined based on numerical values, thereby providing information for highly accurate sample orientation without relying on the operator's TEM / STEM operation skills. Furthermore, while both the structural boundary position and structural boundary width can be determined by performing the processes shown in FIGS. 1A and 1B , it is also possible to determine only one of the structural boundary position and structural boundary width. In this case, the steps related to determining only the structural boundary position or structural boundary width, which are not required, can be omitted.

[0027] In this example, the evaluation area 102 is set to calculate the brightness at a pixel position as the average brightness of multiple pixels in order to reduce the influence of noise, but if such averaging processing is not performed, it is not necessary to set the evaluation area 102, and it is sufficient to specify the direction in which to obtain the brightness profile in step S02. The specified direction must intersect with the direction in which the boundary between phases 103 and 104 extends, but it does not have to be perpendicular.

[0028] The above is an example of processing for a simple model image 101, but similar processing can be performed on actual microscope images. Figure 3A shows an example of an actual TEM image. An evaluation area 122 is set for TEM image 121. TEM image 121 is a TEM image of a thin film sample having a structure consisting of a first phase 123, a second phase 124, and a third phase 125. Below, an example is shown of determining the structural boundary position and width at the boundary between phases 124 and 125. Figure 3B shows the brightness profile calculated for evaluation area 122. Figure 3C shows the moving standard deviation profile of brightness calculated for evaluation area 122 by setting four calculation widths A to D (where A < B < C < D). Depending on the size of the calculation width, differences similar to those in model image 101 appear in the convex waveform shape for each calculation width. In the case of calculation width D, a waveform with multiple peaks is observed in a single convex waveform, which was not present in the example of model image 101. This is thought to be because the calculation width D was too large, and the moving standard deviation was calculated including brightness information for three phases. Based on the moving standard deviation profile of FIG. 3C , calculation width B is specified as the calculation width for identifying the structure boundary position and structure boundary width. FIG. 3D shows only the moving standard deviation profile of brightness calculated with calculation width B extracted from FIG. 3C . The structure boundary position at the boundary between phase 124 and phase 125 is determined as the position of peak p in the moving standard deviation graph of FIG. 3D . FIG. 3E shows the moving standard deviation difference profile of brightness calculated with calculation width B. The moving standard deviation difference profile shows a pair of maximum point 126 and minimum point 127. The structure boundary width W can be calculated from the difference between the positions that give extreme values ​​to the standard moving deviation difference profile. FIG. 3F shows the position where the peak of the convex waveform exists in the case of calculation width B as structure boundary position 128 on TEM image 121, and also shows the positions where extreme values ​​are given to the standard moving deviation difference profile as boundaries 129 and 130 of the structure boundary region on TEM image 121.

[0029] The relationship between the structural boundary surface in the sample structure and the structural boundary region that appears in a TEM image will be explained using FIG. 5 . FIG. 5 shows a cross section of a thin film sample 201. The thin film sample 201 is composed of three phases: an A phase 202, a B phase 203, and a C phase 204. A structural boundary surface 205 between the A phase 202 and the B phase 203 and a structural boundary surface 206 between the B phase 203 and the C phase 204 are formed in the thin film sample 201. A TEM image 211 is obtained by irradiating the surface of the thin film sample 201 with an electron beam in a direction perpendicular to the surface (z direction). In the TEM image 211, a structural boundary region 215 with a structural boundary width W1 is observed between the A phase 212 and the B phase 213, and a structural boundary region 216 with a structural boundary width W2 is observed between the B phase 213 and the C phase 214. In the example of Figure 5, the structure boundary width W1 is greater than the structure boundary width W2, which indicates that the greater the inclination of the structure boundary surface relative to the electron beam incident angle, the wider the structure boundary width in the microscope image. Utilizing this relationship, the thin film sample is tilted so that the structure boundary width W is minimized, thereby aligning the sample orientation.

[0030] FIG. 6 shows the flow for adjusting the sample tilt for sample orientation alignment using the structure boundary width W in the microscope image. First, the sample 21 is placed on the sample stage 22 (S21). In the initial position, the surface of the sample 21 is placed perpendicular to the optical axis of the primary electron beam of the STEM 51. A microscope image is acquired in this state (S22). The boundary and direction of the structure to be analyzed are specified (S23). Hereinafter, the longitudinal direction of the structure boundary region used for sample orientation alignment is defined as the y-axis direction, the direction parallel to the surface of the sample 21 and perpendicular to the y-axis direction is defined as the x-axis direction, and the direction perpendicular to the surface of the sample 21 is defined as the z-axis direction. Next, the flow described with reference to FIGS. 1A and 1B is applied to calculate the structure boundary width for the specified structure boundary region (S24). Then, the sample 21 is tilted by the sample stage 22 (S25), a microscope image is acquired (S26), and the structure boundary width for the specified structure boundary region is calculated for the acquired microscope image (S24). This process is repeated while changing the sample tilt angle.

[0031] FIG. 7 shows the information on the structure boundary area (structure boundary width W ａｉ7 (i = 1 to 7) is shown. The sample has a first phase 302 and a second phase 303, and the structural boundary between the first phase 302 and the second phase 303 is the target of analysis. As shown by the coordinate axes 301, the sample is tilted around the y-axis, which is parallel to the longitudinal direction of the structural boundary region. Here, the sample tilt angle ai is defined as positive when the z-coordinate sign is +, as zero when the z-coordinate sign is 0, and as negative when the z-coordinate sign is -, with the relationship a1 < a2 < a3 < a4 < a5 < a6 < a7. Thus, the structural boundary width changes depending on the sample tilt angle. Note that, for ease of understanding, the sample is tilted around the y-axis, which is parallel to the longitudinal direction of the structural boundary region. However, even if the central axis for changing the sample tilt angle is not parallel to the longitudinal direction of the structural boundary region, the relationship shown in FIG. 7 can be extracted.

[0032] Returning to the explanation of FIG. 6, the sample tilt angle ai and the structure boundary width W ａｉ The relationship between the sample tilt angle ai and the structure boundary width W corresponding to FIG. 7 is graphed and displayed on the output device 414 (S27). ａｉ The graph is shown in Figure 8. Sample tilt angle ai / structure boundary width W ａｉ The point that gives the minimum value to the structure boundary width is extracted from the graph (S28). In the example of FIG. 8, the structure boundary width W ａ５ If the value of a is sufficiently small, the sample tilt angle a5 is extracted as the point that gives the minimum value. ａｉ From the graph, fitting is performed to determine the sample tilt angle ai as an explanatory variable, and the structural boundary width W ａｉ A function with the objective variable σ can be obtained, and explanatory variables that minimize the objective variable can be identified from the function. The sample tilt angle ai thus obtained is determined as the tilt parameter of the sample (S29), and the sample is tilted (S30). This completes the sample orientation adjustment.

[0033] In this way, whether the structural boundary surface is perpendicular to the optical axis of the primary electron beam, i.e., whether the structural boundary width in the microscope image is narrow, can be determined numerically rather than by human eyes, so that the variation in observation conditions that depends on the TEM / STEM operator can be reduced, and automation becomes possible. Furthermore, the adjustment method of this embodiment makes it possible to adjust the structural orientation of the target portion not only for samples that have single crystal regions and samples that do not have single crystal regions as exemplified, but also for samples with curved thin film surfaces.

[0034] (Modification) In the above examples, examples were shown in which the structural boundary position and structural boundary width were calculated based on a one-dimensional luminance profile. However, it is also possible to calculate the structural boundary position and structural boundary width by expanding the calculation width to two dimensions. When calculating the moving standard deviation of luminance at each pixel position in the evaluation area, for example, if the calculation width is set to 7 pixels, the standard deviation of pixel position (i, j) is calculated for each pixel position as the standard deviation of luminance in an area including a total of 7 x 7 pixel positions, from (i-3) pixel to (i+3) pixel in the x direction and from (j-3) pixel to (j+3) pixel in the y direction. Since a peak in the moving standard deviation of luminance appears two-dimensionally at the position where the phase boundary exists, the two-dimensional distribution of the structural boundary region in the evaluation area can be determined. Note that, although an example is shown in which the calculation width in the x direction and the calculation width in the y direction are the same, different calculation widths may also be used. Figure 9 schematically shows the structural boundary position 901 and structural boundary width W calculated in an evaluation area 900 when expanded two-dimensionally. Here, the vertical and horizontal directions of the microscope image are represented as the y and x directions, respectively, and the evaluation area 900 is the area surrounded by coordinates (x1, y1), (x2, y1), (x1, y2), and (x2, y2).

[0035] The structural boundary position 901 is determined as follows: In the section (y1-y2) of the evaluation area 900, a moving standard deviation profile along the x direction is obtained, and the peak of the convex waveform contained in each moving standard deviation profile is extracted. Furthermore, in the section (x1-x2) of the evaluation area 900, a moving standard deviation profile along the y direction is obtained, and the peak of the convex waveform contained in each moving standard deviation profile is extracted. When determining the structural boundary position 901, the size of the region used to calculate the moving standard deviation for each pixel position may be any size as long as a waveform in which the peak of the convex waveform is clear is obtained. For both the moving standard deviation profile along the x direction and the moving standard deviation profile along the y direction, the pixel position (x, y) that gives the peak of the convex waveform contained in them is the structural boundary position 901. The line obtained by connecting the structural boundary positions 901 is referred to as the structural boundary line.

[0036] Next, the structure boundary width W is calculated as follows. The calculation method is the same as in the embodiment, but is characterized in that a brightness profile, a moving standard deviation profile, a moving standard deviation difference profile, or a moving standard deviation differential profile along a direction perpendicular to the structure boundary line is used. Specifically, the method is as follows.

[0037] First, for an arbitrary structural boundary position, an intensity profile along a direction perpendicular to the structural boundary line is obtained, and a moving standard deviation of the intensity profile is calculated using multiple calculation widths to obtain a moving standard deviation profile. In Figure 9, for a structural boundary position 905, the direction perpendicular to the structural boundary line is shown as direction 902 (direction o, although this direction differs for each structural boundary position 901 on the structural boundary line). Next, for each moving standard deviation profile calculated using multiple calculation widths, the half-width of the convex waveform included in the moving standard deviation profile is calculated. Next, of the multiple moving standard deviation profiles, the calculation width of the moving standard deviation profile in which the convex waveform has a peak and the half-width of the convex waveform is narrowest is identified.

[0038] Next, for each position along the o direction, an element difference or derivative of the moving standard deviation profile calculated using the specified calculation width is calculated to obtain a moving standard deviation difference profile or a moving standard deviation differential profile. Next, a pair of a first position 903 at which a maximum value is achieved and a second position 904 at which a minimum value is achieved is identified from the moving standard deviation difference profile or the moving standard deviation differential profile. Next, the distance between the first position 903 and the second position 904 is determined as the structure boundary width W. This process is performed for each structure boundary position 901 on the structure boundary line.

[0039] The present invention is not limited to the above-described embodiment and includes various modifications. The above-described embodiment has been described in detail to clearly explain the present invention, and is not necessarily limited to a configuration including all of the described components. For example, for simplicity of explanation, in FIG. 2A and the like, the direction in which the intensity profile is obtained is the x direction of the microscope image, but this direction is not limited. If the direction in which the intensity profile is obtained is different from the x or y direction of the microscope image, including the o direction in the modified example, the intensity at the position where the intensity profile is obtained can be obtained by interpolating the intensity of neighboring pixels. Furthermore, some of the components of the embodiment can be added, deleted, or replaced with other components. Furthermore, the image processing method of the present invention can be applied not only to TEM images but also to images acquired by a scanning electron microscope (SEM).

[0040] 1: electron beam source, 2: extraction electrode, 11: condenser lens, 12: condenser aperture, 13: axis adjustment deflector, 14: astigmatism corrector, 15: scan deflector, 20: objective lens, 21: sample, 22: sample stage, 23: objective aperture, 24: axis adjustment deflector, 25: selected area aperture, 30: imaging lens, 31: detector, 40: control device, 41: computer, 42: control system, 51: STEM, 101: model image, 102: evaluation area, 103: first phase, 104: second phase, 105: structure boundary area, 110: maximum point, 111: minimum point, 112: structure boundary position, 113, 114: boundary, 121: TEM image, 122: evaluation area, 123: first phase, 1 24: second phase, 125: third phase, 126: maximum point, 127: minimum point, 128: structural boundary position, 129, 130: boundary, 201: thin film sample, 202: phase A, 203: phase B, 204: phase C, 205, 206: structural boundary surface, 211: TEM image, 212: phase A, 213: phase B, 214: phase C, 215, 216: structural boundary area, 301: coordinate axis, 302: first phase, 303: second phase, 411: processor, 412: memory, 413: auxiliary storage device, 414: output device, 415: input device, 416: communication interface, 417: bus, 900: evaluation area, 901, 905: structural boundary position, 902: direction, 903: first position, 904: second position.

Claims

1. An image processing method for determining a structural boundary position between a first phase and a second phase in a microscope image of a sample, in which a first phase and a second phase contact each other at a structural boundary region, comprising: a first step of specifying, in the microscope image, a second direction that includes the structural boundary region and intersects with a first direction that is a longitudinal direction of the structural boundary region; a second step of obtaining a brightness profile along the second direction; a third step of calculating a moving standard deviation of the brightness profile using a plurality of calculation widths to obtain a moving standard deviation profile; a fourth step of calculating, for each of the moving standard deviation profiles calculated using the plurality of calculation widths, a half-width of a convex waveform included in the moving standard deviation profile; a fifth step of specifying, among the plurality of moving standard deviation profiles, a calculation width of a moving standard deviation profile in which the convex waveform has a peak and the half-width of the convex waveform is narrowest; and a sixth step of determining, as the structural boundary position, a peak position of the moving standard deviation profile calculated using the specified calculation width.

2. An image processing method as claimed in claim 1, wherein in the first step, the second direction is defined so as to be perpendicular to the first direction, and a band-shaped evaluation area extending in the first direction and the second direction is specified, and in the second step, the average value of the luminance at a position for which the luminance profile is to be obtained is used relative to the position in the evaluation area along the first direction.

3. An image processing method according to claim 1, wherein the determined structure boundary position is superimposed on the microscope image and displayed.

4. An image processing method for determining a structure boundary width, which is the width of a structure boundary region in a microscope image of a sample where a first phase and a second phase contact each other at the structure boundary region, comprising: a first step of specifying, in the microscope image, a second direction which includes the structure boundary region and intersects with a first direction which is a longitudinal direction of the structure boundary region; a second step of obtaining an intensity profile along the second direction; a third step of calculating a moving standard deviation of the intensity profile using a plurality of calculation widths to obtain a moving standard deviation profile; a fourth step of calculating, for each of the moving standard deviation profiles calculated using the plurality of calculation widths, a half-width of a convex waveform included in the moving standard deviation profile; a fifth step of specifying, among the plurality of moving standard deviation profiles, a calculation width of a moving standard deviation profile in which the convex waveform has a peak and in which the half-width of the convex waveform is narrowest; and a sixth step of calculating an element difference or derivative of the moving standard deviation profile calculated using the specified calculation width to obtain a moving standard deviation difference profile or a moving standard deviation differential profile. a seventh step of identifying a pair of a first position having a maximum value and a second position having a minimum value from the moving standard deviation difference profile or the moving standard deviation differential profile; and an eighth step of determining a distance between the first position and the second position as the structure boundary width.

5. An image processing method as claimed in claim 4, wherein in the first step, the second direction is defined so as to be perpendicular to the first direction, and a band-shaped evaluation area extending in the first direction and the second direction is specified, and in the second step, the average value of the luminance at a position for which the luminance profile is to be obtained is used relative to the position in the evaluation area along the first direction.

6. An image processing method according to claim 4, wherein the first position and the second position are displayed superimposed on the microscope image.

7. An image processing method as claimed in claim 4, wherein a plurality of transmission electron microscope images are taken as the microscope images by a transmission electron microscope or a scanning transmission electron microscope while changing the tilt angle of the sample relative to the optical axis of a primary electron beam, and the tilt angle that minimizes the structure boundary width is determined as the tilt angle when the sample is observed with the transmission electron microscope or the scanning transmission electron microscope based on the structure boundary width determined for each of the plurality of transmission electron microscope images.

8. An electron microscope system comprising: an electron microscope that irradiates a sample with a primary electron beam and forms an image of electrons transmitted through or diffracted by the sample on a detector; and a control system that controls the electron microscope, wherein the control system uses the electron microscope to obtain a plurality of electron microscope images by changing the tilt angle of the sample with respect to the optical axis of the primary electron beam, wherein each of the plurality of electron microscope images shows a first phase and a second phase in contact at a structure boundary region, and the control system determines a structure boundary width, which is the width of the structure boundary region, for each of the plurality of electron microscope images, and the control system determines the tilt angle that minimizes the structure boundary width based on the structure boundary width determined for each of the plurality of electron microscope images, as the tilt angle when observing the sample with the electron microscope.

9. The method according to claim 8, wherein the control system determines the structure boundary width by executing steps 1 to 8 for each of the plurality of electron microscope images, the first step being a step of specifying a second direction in the electron microscope image that includes the structure boundary region and intersects with a first direction that is a longitudinal direction of the structure boundary region, the second step being a step of obtaining a luminance profile along the second direction, the third step being a step of calculating a moving standard deviation of the luminance profile using a plurality of calculation widths to obtain a moving standard deviation profile, the fourth step being a step of calculating a half-width of a convex waveform included in each of the moving standard deviation profiles calculated using the plurality of calculation widths, and the fifth step being a step of specifying a calculation width of a moving standard deviation profile in which the convex waveform has a peak and in which the half-width of the convex waveform is narrowest, among the plurality of moving standard deviation profiles, The sixth step is a step of calculating an element difference or derivative of the moving standard deviation profile calculated using the identified calculation width to obtain a moving standard deviation difference profile or a moving standard deviation differential profile; the seventh step is a step of identifying a pair of a first position having a maximum value and a second position having a minimum value from the moving standard deviation difference profile or the moving standard deviation differential profile; and the eighth step is a step of determining the distance between the first position and the second position as the structure boundary width.

10. An electron microscope system as claimed in claim 9, wherein the control system, in the first step, defines the second direction to be perpendicular to the first direction, specifies a band-shaped evaluation area extending in the first direction and the second direction, and, in the second step, uses, as the brightness of the position for which the brightness profile is to be obtained, an average value of brightnesses of multiple positions in the evaluation area along the first direction relative to that position.

11. An electron microscope system as described in claim 8, wherein the control system determines a function with the tilt angle as an explanatory variable and the structural boundary width as a target variable based on the structural boundary width determined for each of the plurality of electron microscope images, and determines the tilt angle that minimizes the structural boundary width as the tilt angle when observing the sample with the electron microscope.

12. An electron microscope system according to claim 8, wherein the electron microscope scans the primary electron beam on the sample.

13. An image processing method for determining a structural boundary position between a first phase and a second phase in a microscope image of a sample, in which a first phase and a second phase contact each other at a structural boundary region, comprising: a first step of specifying, in the microscope image, an evaluation region including the structural boundary region and extending in a first direction and a second direction perpendicular to the first direction; a second step of calculating, for each position included in the evaluation region, a moving standard deviation of a region specified by a first calculation width in the first direction and a second calculation width in the second direction, to obtain a first moving standard deviation profile along the first direction and a second moving standard deviation profile along the second direction; and a third step of determining, as the structural boundary position, a position giving a peak of a first convex waveform included in the first moving standard deviation profile and a peak of a second convex waveform included in the second moving standard deviation profile.

14. The method according to claim 13, further comprising: a fourth step of determining a luminance profile along a third direction that passes through the structure boundary positions and is perpendicular to a structure boundary line obtained by connecting the structure boundary positions, and calculating a moving standard deviation of the luminance profile using a plurality of third calculation widths to determine a third moving standard deviation profile; a fifth step of calculating a half-width of a third convex waveform included in each of the third moving standard deviation profiles calculated using the plurality of third calculation widths; a sixth step of specifying, among the plurality of third moving standard deviation profiles, a third calculation width of a moving standard deviation profile in which the third convex waveform has a peak and the half-width of the third convex waveform is narrowest; a seventh step of calculating an element difference or differential of the moving standard deviation profile calculated using the specified third calculation width to determine a moving standard deviation difference profile or a moving standard deviation differential profile; and an eighth step of specifying a set of a first position at which a maximum value is obtained and a second position at which a minimum value is obtained from the moving standard deviation difference profile or the moving standard deviation differential profile. and a ninth step of determining a distance between the first position and the second position as a structure boundary width.

15. An image processing method according to claim 13, wherein the area specified by the first calculation width and the second calculation width is set with the position included in the evaluation area as the center.

Images