Surface analysis device

Abstract

To avoid detecting false clusters when automatically clustering points on a scatter diagram and to improve the accuracy of clustering.SOLUTION: One aspect of a surface analysis instrument according to the present invention includes: a measuring unit (1-2, 4-8) for acquiring signals reflecting the amounts of a plurality of components or elements to be analyzed at a plurality of positions on a sample (3); a scatter diagram creation unit (97) that creates a ternary scatter diagram based on the measurement results by the measuring unit; a data point selection unit (98) for using distribution information of a sum signal value obtained by adding up signal values of three components or elements corresponding to data points in the ternary scatter diagram, and excluding data points in a predetermined signal value range in which the sum signal value is relatively small from all data points present on the ternary scatter diagram; and a cluster analysis unit (94) that clusters data points in the ternary scatter diagram that were not excluded in the data point selection unit using a density-based cluster analysis method.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to a surface analyzer for examining the distribution of components and elements present in a one-dimensional or two-dimensional measurement area on a sample. This surface analyzer includes an electron probe micro analyzer (EPMA), a scanning electron microscope (SEM), an X-ray fluorescence analyzer, etc. [Background technology]

[0002] Elemental mapping analysis using EPMA allows the type and amount of elements contained in each of many microscopic regions within a two-dimensional area on a sample to be determined. When analyzing the results of such elemental mapping analysis, a scatter diagram (a diagram with the relative intensities of elements plotted on two or three axes) of characteristic X-ray intensities for two or three elements or element concentrations calculated from those intensities is created, and the types and content ratios of compounds contained in the sample are identified from the distribution of plotted points on the diagram, i.e., phase analysis (see Patent Documents 1 and 2). For example, Figure 10 of Patent Document 2 shows an example of a two-way scatter diagram, and Figure 11 of the same document shows an example of a three-way scatter diagram.

[0003] Each point on a scatter plot (hereinafter, a point plotted on a scatter plot is referred to as a "data point") corresponds to a single point (a small area) on the sample. Therefore, a region where data points are densely concentrated on a scatter plot is presumed to correspond to a location on the sample where the elements are contained in similar proportions. In phase analysis, an analyst typically recognizes a region on a scatter plot where data points are densely concentrated as a cluster, i.e., a collection of related data points. Then, using a pointing device such as a mouse, the analyst encloses the region with an appropriate shape, such as a polygon, and then assigns a different display color to each region. After performing these operations, a phase map is displayed on the EPMA display screen, in which the locations on the sample corresponding to each data point in one or more cluster regions are colored in the designated colors.

[0004] In recent years, with the rapid advancement of AI (artificial intelligence) technology, attempts have been made to utilize such technology to automatically allocate a large number of data points on a scatter plot, such as the one described above, to multiple sets. Cluster analysis, a representative method of unsupervised machine learning, is suitable for such processing. While various algorithms are known for cluster analysis, density-based clustering, as disclosed in, for example, Non-Patent Documents 1 and 2, is a useful method for dividing data points on a scatter plot into multiple clusters based on their density. Figure 12 shows an example of clusters automatically extracted using density-based cluster analysis from a two-way scatter plot obtained by actual measurements. In this example, six clusters are extracted. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-125952 [Patent Document 2] Japanese Patent Application Laid-Open No. 2011-153858 [Non-patent literature]

[0006] [Non-Patent Document 1] Ester M. and 3 others, “A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise,” Proceedings of 2nd International Conference on Knowledge Discovery and Data Mining (KDD-96), pp. 226-231, 1996. [Non-patent document 2] Ricardo JGB Campello and 2 others, "Density-Based Clustering Based on Hierarchical Density Estimates," Springer, pp. 160-172, 2013 Summary of the Invention [Problem to be solved by the invention]

[0007] However, in scatter plots created based on data obtained by EPMA, various factors can cause uneven distribution or peculiar distribution of data points, which can lead to the detection of false clusters when applying existing cluster analysis methods such as those described above.

[0008] For example, Figure 4 shows the results of automatic clustering of data points on a two-way scatter plot created based on actual measurement results. In the figure, the area enclosed by a polygonal line is one automatically detected cluster area. At the two locations indicated by arrows in the figure, many small clusters extending vertically in a linear fashion have been detected, but these should actually be detected as one large cluster at each location. In other words, the small clusters extending vertically in a linear fashion are false clusters.

[0009] Figure 8 shows the results of automatic clustering of data points on a 3-way scatter plot created based on actual measurement results. In the figure, the area enclosed by a polygonal line is one cluster area that was automatically detected. As is clear from Figure 8, in a 3-way scatter plot, a set of linearly connected data points may appear radially, but multiple small clusters containing part of that set of data points have been detected. Some of these are not necessarily suitable as clusters, and many are false clusters.

[0010] The present invention has been made to solve the above-mentioned problems, and its main object is to provide a surface analysis instrument that can suppress the detection of false clusters when automatically clustering data points on a scatter plot, thereby improving the accuracy of the clustering. [Means for solving the problem]

[0011] In order to solve the above problems, one aspect of the surface analysis device according to the present invention is to a measuring unit for acquiring signals reflecting the amounts of a plurality of components or elements to be analyzed at a plurality of positions on the sample; a scatter diagram creation unit that creates a three-way scatter diagram based on the measurement results by the measurement unit; a data point selection unit that uses distribution information of an added signal value obtained by adding up the signal values ​​of three components or elements corresponding to data points in the ternary scatter diagram to exclude data points in a predetermined signal value range in which the added signal value is relatively small from all data points present on the ternary scatter diagram; a cluster analysis unit that clusters the data points in the three-way scatter plot that were not excluded by the data point selection unit using a density-based cluster analysis technique; Equipped with.

[0012] The surface analysis device according to the above aspect of the present invention is, for example, an analysis device such as an EPMA SEM, an X-ray fluorescence analysis device, etc. Such an analysis device repeats measurements while changing the position on the sample to which an excitation beam (such as an electron beam or an X-ray) is irradiated, thereby obtaining signals that reflect the abundance of multiple elements at each of multiple positions within a two-dimensional or one-dimensional region on the sample.

[0013] Furthermore, in the surface analysis instrument according to the above aspect of the present invention, the cluster analysis method can be hierarchical density-based spatial clustering of applications with noise (DBSCAN), which is an improved version of the general density-based clustering of applications with noise (DBSCAN). [Effects of the Invention]

[0014] When creating a ternary scatter diagram such as that shown in FIG. 8, as described in Patent Document 1, it is necessary to normalize each data point in the ternary scatter diagram by the sum of the intensities of the three elements. Therefore, when there are many data points with similar intensity ratios for the three elements, a radially extending linear distribution of data points appears on the ternary scatter diagram. The present inventors have found that when a false cluster containing a linear distribution of points is detected as shown in FIG. 8, there are many data points with low intensities for all three elements, and that these many data points significantly contribute to the occurrence of the radially extending linear distribution of data points.

[0015] In the above-described aspect of the surface analysis instrument according to the present invention, the data point selection unit creates a histogram showing the relationship between the sum signal value classes and frequencies as distribution information for the sum signal value obtained by adding the signal values ​​of the three components or elements corresponding to each data point in the ternary scatter diagram. As described above, if there are many data points where the intensities of all three elements are low, a relatively large peak will appear in the histogram where the sum signal value is low. Therefore, the data point selection unit excludes the data points that form this peak from all the data points. The cluster analysis unit performs clustering on only the remaining data points. This eliminates the radially extending linear distribution of data points in the ternary scatter diagram and also avoids the detection of the associated false clusters.

[0016] As described above, the surface analysis instrument according to the above aspect of the present invention can suppress the detection of false clusters when automatically clustering data points on a scatter plot, thereby improving the accuracy of clustering of the data points, i.e., micro-regions on a sample based on the concentrations of multiple components or elements, thereby enabling a user to accurately and efficiently perform phase analysis based on the clustering results of data points on a scatter plot, for example. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a configuration diagram of a main part of an EPMA according to a first embodiment of the present invention; [Figure 2] 4 is a flowchart showing an example of a cluster analysis parameter adjustment process in the EPMA of the first embodiment. [Figure 3] FIG. 10 is a diagram showing an example of a histogram of intensity values ​​of elements with a narrow measurement range. [Figure 4] This figure shows an example of false clusters detected as a result of automatic clustering of data points on a two-way scatter plot. [Figure 5] This figure shows the results of clustering the data points on the two-way scatter plot shown in Figure 4 after adjusting the cluster analysis parameters. [Figure 6] 1 is a configuration diagram of a main part of an EPMA according to an embodiment of the present invention. [Figure 7] 10 is a flowchart showing an example of an added intensity value data selection process in the EPMA of the present embodiment. [Figure 8] This figure shows an example of false clusters detected as a result of automatic clustering of data points on a 3-way scatter plot. [Figure 9] FIG. 10 is a diagram showing an example of a histogram of added intensity values. [Figure 10] FIG. 10 is a diagram showing a histogram of the added intensity values ​​shown in FIG. 9 after excluding outliers on the larger intensity value side. [Figure 11] FIG. 9 is a diagram showing the results of clustering performed on the data points on the three-way scatter diagram shown in FIG. 8 after performing an addition intensity value data selection process. [Figure 12] FIG. 10 shows an example of the results of automatic clustering performed on a two-way scatter plot. DETAILED DESCRIPTION OF THE INVENTION

[0018] [First embodiment] An EPMA according to a first embodiment of the surface analysis instrument of the present invention will be described with reference to the accompanying drawings, in which: Figure 1 is a diagram showing the configuration of the main parts of the EPMA according to the first embodiment;

[0019] As shown in Fig. 1, the electron beam irradiation unit 1 includes an electron gun 100 and a deflection coil (not shown), and irradiates a very small diameter electron beam onto a sample 3 placed on a sample stage 2. In response to this electron beam, characteristic X-rays having wavelengths specific to the element are emitted from the surface of the sample 3. Secondary electrons and the like are also emitted from the surface of the sample 3.

[0020] Characteristic X-rays emitted from the sample 3 are wavelength-dispersed by the analyzing crystal 4, and diffracted X-rays of specific wavelengths are detected by the X-ray detector 5. The electron beam irradiation position on the sample 3, the analyzing crystal 4, and the X-ray detector 5 are always positioned on the Rowland circle, and the analyzing crystal 4 is moved linearly and tilted by a driving mechanism (not shown), and the X-ray detector 5 is rotated in conjunction with this movement. This achieves wavelength scanning of the X-rays to be analyzed so as to satisfy Bragg's diffraction condition, that is, while maintaining the incident angle of the characteristic X-rays and the exit angle of the diffracted X-rays with respect to the analyzing crystal 4 equal. The detection signal of the X-ray intensity by the X-ray detector 5 is input to the data processing unit 9.

[0021] The sample stage 2 can be moved in two mutually perpendicular axial directions, the X-axis and the Y-axis, by a sample stage drive unit 7, and this movement two-dimensionally scans the irradiation position of the electron beam on the sample 3. Alternatively, instead of moving the sample stage 2, the irradiation position of the electron beam on the sample 3 can be scanned by deflecting the emission direction of the electron beam in the electron beam irradiation unit 1.

[0022] The data processing unit 9 includes, as functional blocks, an element intensity calculation unit 90, a data storage unit 91, a scatter plot creation unit 92, a cluster analysis parameter adjustment unit 93, a cluster analysis unit 94, a cluster region detection unit 95, and a display processing unit 96. In order to perform analysis on the sample 3, the analysis control unit 8 controls the operation of the sample stage drive unit 7 and the drive mechanisms that move the dispersing crystal 4 and the X-ray detector 5. The central control unit 10 controls the entire apparatus and handles input / output processing, and is connected to an operation unit 11, which includes a keyboard and a mouse (or other pointing device), and a display unit 12.

[0023] For example, all or part of the central control unit 10, analysis control unit 8, and data processing unit 9 may be configured as a personal computer, and each function may be achieved by executing dedicated control and processing software installed on the computer.

[0024] When element mapping analysis is performed with the EPMA of this embodiment, the analysis control unit 8 fixes the position of the analyzing crystal 4 in accordance with the characteristic X-ray wavelength of the element of interest, and operates the sample stage driving unit 7 and other units to repeatedly detect the characteristic X-rays and secondary electrons while changing the irradiation position (microscopic area) of the electron beam in a predetermined order within a predetermined two-dimensional area (usually designated by the analyst) on the sample 3. Then, once acquisition of the intensity distribution for one element has been completed, similar measurements are performed for other target elements.

[0025] The element intensity calculation unit 90 acquires the intensity (concentration) of the target element for each micro-region on the sample 3. This intensity data is stored in the data storage unit 91. When an energy dispersive X-ray spectrometer is used, the element intensity calculation unit 90 creates an X-ray spectrum for each micro-region in the two-dimensional region, detects a peak of a specific wavelength corresponding to the target element in the X-ray spectrum, and calculates the intensity of the peak, thereby calculating the intensity (concentration) of the target element.

[0026] When measurements have been completed for all micro-regions within the two-dimensional area on the sample 3 and the analyst performs a predetermined operation on the operation unit 11, the scatter diagram creation unit 92 reads out the predetermined data from the data storage unit 91 and creates a two-way scatter diagram showing the relationship between the intensities of two predetermined elements. Each data point on the two-way scatter diagram corresponds to a micro-region on the sample 3. Therefore, for example, if measurements are performed on 1,000 micro-regions on the sample 3, the number of data points plotted on the scatter diagram will be 1,000.

[0027] The cluster analysis unit 94 performs clustering on all data points on the created scatter plot according to a predetermined algorithm, and labels each data point as belonging to one or more clusters, or not belonging to any cluster.

[0028] Various cluster analysis techniques are known. Generally, clustering of data points on such scatter plots is performed using the distance between data points. Scatter plots obtained by surface analysis such as EPMA often have areas where data points are extremely dense and areas where they are sparsely populated. In areas where data points are densely populated, if the distance between data points is relatively short, they will be separated into separate clusters, resulting in the formation of clusters with an extremely large number of data points. Conversely, in areas where data points are sparsely populated, if the distance between data points is relatively long, they will be included in the same cluster, resulting in the formation of many clusters with extremely small numbers of data points. To address this issue, we employ the hierarchical density-based cluster analysis technique disclosed in Non-Patent Document 2. This technique is an improvement over the general density-based cluster analysis disclosed in Non-Patent Document 1, and according to the inventor's research, it is capable of clustering data points in scatter plots obtained by EPMA with considerable success.

[0029] In density-based cluster analysis, including the above-mentioned hierarchical density-based cluster analysis, the following two parameters must be determined in advance in order to determine that a set of data points on a scatter plot is a cluster: (1) Minimum cluster size: The minimum number of data points required to determine a cluster (the minimum number of data points that make up one cluster). (2) Distance threshold ε: The distance threshold for determining that two adjacent clusters are different clusters. Multiple clusters that are closer than this threshold are automatically merged. To effectively detect clusters on a scatter plot, the above parameters must be set to appropriate values. However, since it is tedious for the user (analyst) to set these parameters one by one, the manufacturer sets experimentally determined values ​​as default values ​​for each parameter, and the user can manually change these values.

[0030] When the intensity ranges of two elements reflected in a two-way scatter plot are relatively close and the measurement ranges (intensity ranges) are similar, appropriate clustering is generally possible even when the default value of the distance threshold ε is used. However, as described above, when there is a relatively large difference in the abundance (concentration) of two elements and the measurement range of the less abundant element is extremely narrow, characteristic false clusters resulting from this are easily detected. Unless the distance threshold ε is adjusted to an appropriate value based on the spatial distribution of the data points, multiple false clusters cannot be merged into one. Therefore, the cluster analysis parameter adjustment unit 93 adjusts the distance threshold ε as follows before actually performing the cluster analysis. Figure 2 is a flowchart showing an example of the cluster analysis parameter adjustment process.

[0031] First, the cluster analysis parameter adjustment unit 93 creates a histogram of the intensity values ​​of the element with the narrower measurement range (Mn in the example of Figure 4) to grasp the distribution of intensity values ​​of that element (step S1). At this time, the number of classes T of the intensity values ​​is determined appropriately. Figure 3 shows a histogram corresponding to the data points on the two-way scatter diagram shown in Figure 4. Here, the number of classes T is set to 1000. As shown in Figure 3, a histogram reflecting data points that exist only discretely due to a narrow measurement range shows extremely high frequencies in certain classes, with comb-like gaps between multiple specific classes. Due to this characteristic tendency in the frequency distribution of intensity values, the two-way scatter diagram becomes as described above, and false clusters are detected.

[0032] In density-based cluster analysis, clusters are more likely to form in areas of the scatter plot where data points are more concentrated, i.e., where the frequency is higher. Therefore, the cluster analysis parameter adjustment unit 93 detects the maximum value, i.e., the peak, in the histogram and finds the class corresponding to the peak (step S2), and then identifies the class with the highest frequency among the classes showing the maximum value (step S3). In Figure 3, the class with the highest frequency among the classes showing the maximum value is indicated by an underlined arrow.

[0033] Next, the cluster analysis parameter adjustment unit 93 calculates the number of consecutive classes with a frequency of 0, NL, between the class showing the largest maximum value and the class showing the most recent maximum value below it (the class with a smaller intensity value) in the histogram (step S4). Similarly, the cluster analysis parameter adjustment unit 93 calculates the number of consecutive classes with a frequency of 0, Nu, between the class showing the largest maximum value and the class showing the most recent maximum value above it (the class with a larger intensity value) in the histogram (step S5). That is, the number of consecutive classes with a frequency of 0 is calculated on both sides of the class showing the largest maximum value.

[0034] Thereafter, the cluster analysis parameter adjustment unit 93 compares the consecutive number Nu and the consecutive number NL calculated in steps S4 and S5, determines the larger one as the consecutive class number N (step S6), and determines whether the consecutive class number N is larger than the distance threshold ε at that time (step S7). If the consecutive class number N is equal to or smaller than the distance threshold ε, there is no need to modify the threshold ε, so the value is maintained (step S9), and the process ends. On the other hand, if the consecutive class number N is larger than the distance threshold ε, the cluster analysis parameter adjustment unit 93 modifies the value of the threshold ε using the following equation (1) (step S8), and the process ends. ε = (number of consecutive classes N / total number of classes T) + correction constant K (1) The correction constant K may be any value determined experimentally, but here it is set to 0.002.

[0035] When formula (1) is used, the distance at which data points in the highest frequency class are recognized as separate clusters increases in a two-way scatter plot, reflecting the length of the intervals with a frequency of 0 before and after the class with the highest frequency. As a result, multiple clusters with small distances between them are more likely to be merged into a single cluster.

[0036] The cluster analysis unit 94 then performs clustering on the data points on the two-way scatter plot using the modified parameters as conditions. As a result, each data point on the two-way scatter plot is labeled as belonging to one or more clusters, or none. In this state, each data point is simply labeled, making it difficult to treat it as a cluster area on the scatter plot. Therefore, the cluster area detection unit 95 uses an appropriate method, such as the convex hull method, to define polygonal cluster areas that include all or most of the data points belonging to each cluster. Note that the vertically extending linear cluster areas and the rectangular area containing a large number of data points depicted in the lower area of ​​the two-way scatter plot shown in Figure 4 were also obtained by the cluster area detection process described above.

[0037] Figure 5 is a two-way scatter plot showing the results of cluster analysis and cluster region detection performed on the same data points as in Figure 4 after the cluster analysis parameter adjustment process described above. Figure 5 shows that the multiple linear cluster regions in Figure 4 have been merged into single large cluster regions. Meanwhile, the shape of the cluster regions remains unchanged in the lower part of the scatter plot, where the data point density is not particularly high. Thus, the EPMA of this embodiment can avoid the detection of false clusters in the two-way scatter plot and perform accurate clustering, even when there is a large difference in the abundance of two target elements, especially when one element is present in a small amount and its variance is small, resulting in a narrow measurement range.

[0038] Although the EPMA of the above embodiment focuses only on the areas before and after the class showing the maximum frequency in the parameter adjustment process shown in Fig. 2, it is also possible to find the number of consecutive classes with a frequency of 0 for all classes in the histogram (i.e., all classes before and after the positions of the maximum values) and use the maximum of these consecutive numbers as the number of consecutive classes N. However, in many cases, even if such a process is performed, the number of consecutive classes N selected will be the same as the result of the process shown in Fig. 2, and therefore, performing the process shown in Fig. 2 is preferable in terms of calculation processing time, etc.

[0039] [Second embodiment] Next, a second embodiment of an EPMA, which is one embodiment of a surface analysis instrument according to the present invention, will be described with reference to the accompanying drawings. Fig. 6 is a configuration diagram of the main parts of the second embodiment of the EPMA. In Fig. 6, components that are the same as or correspond to those of the instrument shown in Fig. 1 are designated by the same reference numerals.

[0040] The basic configuration of the EPMA of the second embodiment is the same as that of the EPMA of the first embodiment, except that the data processing unit 9 is provided with a 3-way scatter plot creation unit 97 instead of the scatter plot creation unit 92, and with an added intensity value data selection processing unit 98 instead of the cluster analysis parameter adjustment unit 93.

[0041] In the EPMA of this embodiment, similarly to the EPMA of the first embodiment, analysis is performed on a large number of microregions within a two-dimensional region on the sample 3 under the control of the analysis control unit 8. An element intensity calculation unit 90 acquires intensity data reflecting the abundance of the target element for each microregion within the two-dimensional region on the sample 3. This intensity data is stored in a data storage unit 91.

[0042] When the analyst performs a predetermined operation on the operation unit 11, the ternary scatter diagram creation unit 97 reads the predetermined data from the data storage unit 91 and creates a ternary scatter diagram showing the relationship between the intensities of the three specified elements. Each data point on the ternary scatter diagram corresponds to a small area on the sample 3. The display processing unit 96 displays the created ternary scatter diagram on the screen of the display unit 12. As shown in Figure 8, if a set of radially extending linear data points is observed in the ternary scatter diagram, accurate clustering is likely to be impossible even with automatic cluster analysis. Therefore, when the analyst performs a predetermined operation on the operation unit 11, the sum intensity value data selection processing unit 98 executes the following data selection process. Figure 7 is a flowchart of this data selection process.

[0043] The added intensity value data selection processing unit 98 calculates the added value of the intensities (hereinafter referred to as "added intensity value") of the three elements (Fe, Mg, and K in the example of FIG. 8) shown in the ternary scatter diagram for each minute region, and excludes a predetermined percentage of data as outliers in descending order of added intensity value (step S11). The reason for excluding the outliers and the specific method thereof are as follows.

[0044] Figure 9 is a histogram of the summed intensity values ​​for all data points (i.e., a small region on sample 3) plotted in the ternary scatter diagram shown in Figure 8. As shown, a significant proportion of the data is concentrated in a fairly narrow intensity range near the summed intensity value = 0, forming a peak. It is assumed that these large numbers of data with small summed intensity values ​​are the cause of the radially extending linear collection of data points in the ternary scatter diagram. To resolve this, it is necessary to remove all or part of this data. However, in the histogram shown in Figure 8, the data has a low frequency and extends to a large summed intensity value. Therefore, the peak indicating a high frequency appears compressed in the horizontal direction, making it difficult to determine a threshold value for selecting data to be removed.

[0045] In a histogram, it is assumed that the frequency of data with large added intensity values ​​is low, and that such data exists discretely on the horizontal axis. Therefore, if such data with large added intensity values ​​is temporarily excluded and the histogram is recreated, the state of peaks in the area with small added intensity values, that is, the distribution state of data in the histogram, can be grasped in more detail. Therefore, as an example, here, outliers with large added intensity values ​​are excluded using a method of outlier detection using quartiles, which is often used in statistics.

[0046] Generally, when detecting outliers using quartiles, outliers are found using the interquartile range (IQR), which is the value obtained by subtracting the first quartile (Q1), which represents 25%, from the third quartile (Q3), which represents 75% of the total number, when all data is sorted in ascending order. Specifically, the lower and upper boundaries are usually found using the following formula, and data outside these boundaries are considered to be outliers. Lower boundary=Q1-IQR×1.5 Upper boundary = Q3 + IQR × 1.5 However, here, small outliers are not necessary. Therefore, to remove data with large values, data with intensities above the upper boundary are excluded. Note that the method for detecting large outliers is not limited to the above method; other outlier detection methods, such as the Smirnoff-Grubbs test, can also be used.

[0047] Next, the added intensity value data selection processing unit 98 creates a histogram from the added intensity value data after excluding outliers as described above (step S12). Then, a predetermined algorithm is used to detect the positions (classes) of maximum and minimum values ​​in the histogram (step S13). FIG. 10 shows a histogram created based on the added intensity value data from which the histogram shown in FIG. 9 was created after excluding outliers. In FIG. 10, detected maximum values ​​are indicated by black circles, and minimum values ​​are indicated by white circles. Note that a fixed detection width is set here so that increases and decreases in frequency within a narrow intensity range are not recognized as maximums or minimums. Therefore, for example, the peak closest to intensity = 0 is not detected as a maximum.

[0048] The added intensity value data selection processing unit 98 identifies the nearest minimum value above the maximum value indicating the highest frequency (step S14), and excludes from all added intensity value data (including the outliers excluded in step S11) the added intensity value data that falls within the intensity range from the minimum intensity, i.e., intensity = 0, to the identified minimum value (step S15). In the example shown in Fig. 10, all added intensity value data with an added intensity value of 156 or less is excluded. This results in excluding all data that form the peak with the highest frequency in the histogram shown in Fig. 10.

[0049] In the summed intensity value data after the outliers have been removed in step S11, the maximum intensity may change significantly (become smaller) compared to the original data. In the examples shown in Figures 9 and 10, the number of classes of summed intensity values ​​used to create the histogram is set to 256, but if the maximum intensity of the data after the outliers have been removed is less than 256, it is necessary to adjust the number of classes of intensity values ​​accordingly. This is because, if the number of classes is not adjusted in this way, classes with a frequency of 0 will appear in a comb-like pattern in the created histogram, making it impossible to accurately determine the positions of the minimum values.

[0050] The summed intensity value data selection processing unit 98 selects data to be clustered by excluding data with small summed intensity values ​​and high frequencies, as described above. The cluster analysis unit 94 performs clustering, for example, using hierarchical density-based cluster analysis, on the data points on the ternary scatter plot after selection. As a result, each data point on the ternary scatter plot is labeled as belonging to one or more clusters, or none. The cluster region detection unit 95 uses an appropriate method, for example, the convex hull method, to define polygonal cluster regions that include all or most of the data points belonging to each cluster.

[0051] Figure 11 is a ternary scatter plot showing the results of clustering after the data points on the ternary scatter plot shown in Figure 8 were reduced by performing the above-mentioned sum intensity value data selection process. In Figure 11, the radially extending linear distribution of data points that appeared in the ternary scatter plot shown in Figure 8 has disappeared, and no clusters that are expected to be false clusters have been detected. From these facts, it can be assumed that clustering has been performed appropriately.

[0052] In this way, the EPMA of the second embodiment can exclude noise-like data points that appear in the ternary scatter diagram from the clustering process, thereby suppressing the detection of erroneous cluster regions. As a result, the accuracy of cluster detection on the ternary scatter diagram can be improved, and the accuracy and efficiency of phase analysis using this can be improved.

[0053] In the above explanation, data included in the intensity range from intensity=0 to the nearest minimum value above the maximum value indicating the highest frequency in the histogram shown in Fig. 10 was excluded, but the intensity range of the data to be excluded can be changed as appropriate depending on the data distribution status, etc. For example, the user may be allowed to select the intensity range of the data to be excluded, or clustering may be performed on each of the data groups after excluding data in multiple different intensity ranges, to obtain multiple clustering results, which are then presented to the user together.

[0054] Furthermore, since the data selection process described above does not always need to be performed, the data selection process may be performed in response to user operation as described above, or the data selection process may be performed automatically in response to clustering results, etc.

[0055] Furthermore, although the first and second embodiments are EPMAs, the present invention is applicable to a variety of analytical instruments in general that can acquire signals reflecting the amounts of elements or components (compounds, etc.) in multiple micro-regions within a one-dimensional or two-dimensional region on a sample, such as SEMs and X-ray fluorescence analyzers. In other words, the present invention is not particularly limited to the measurement method or analytical method itself, and any analytical instrument capable of performing mapping analysis may be used.

[0056] Furthermore, the second embodiment described above is merely one example of the present invention, and it goes without saying that any appropriate modifications, alterations, additions, etc. made within the spirit of the present invention will also fall within the scope of the claims of the present application.

[0057] [Various aspects] It will be apparent to those skilled in the art that the above-described exemplary embodiments are examples of the following aspects.

[0058] (Item 1) One aspect of the surface analysis instrument according to the present invention is a measuring unit for acquiring signals reflecting the amounts of a plurality of components or elements to be analyzed at a plurality of positions on the sample; a scatter diagram creation unit that creates a three-way scatter diagram based on the measurement results by the measurement unit; a data point selection unit that uses distribution information of an added signal value obtained by adding up the signal values ​​of three components or elements corresponding to data points in the ternary scatter diagram to exclude data points in a predetermined signal value range in which the added signal value is relatively small from all data points present on the ternary scatter diagram; a cluster analysis unit that clusters the data points in the three-way scatter plot that were not excluded by the data point selection unit using a density-based cluster analysis technique; Equipped with.

[0059] For example, if there are many data points where the signal value ratios of three elements contained in a sample are approximately the same, a radially extending linear distribution of data points may appear on a 3-way scatter plot, which may cause the detection of false clusters. The surface analysis instrument described in paragraph 1 can eliminate the characteristic distribution of data points that causes such false clusters. This suppresses the detection of false clusters when automatically clustering data points plotted on a 3-way scatter plot, improving the accuracy of clustering of those data points, i.e., micro-regions on a sample based on the abundances and concentrations of multiple components or elements. This allows users to accurately perform phase analysis based on the clustering results, for example.

[0060] (Item 2) In the surface analysis instrument described in item 1, the data point selection unit can create a histogram of the added signal values ​​and determine the signal value range of the data to be excluded using classes indicating the maximum and / or minimum values ​​detected in the histogram.

[0061] (Item 3) In the surface analysis instrument described in item 2, the data point selection unit can identify the class indicating the nearest minimum value above the class indicating the maximum value in the histogram, and exclude data included in the intensity range from intensity=0 to the class of the identified minimum value.

[0062] The surface analysis instruments described in paragraphs 2 and 3 can accurately exclude data points with relatively small sum signal values ​​and high frequencies that are likely to cause false clusters. This not only eliminates data points that are likely to cause false clusters, but also avoids the undesired exclusion of data points that are not likely to cause false clusters, making it possible to accurately define cluster regions on a ternary scatter plot.

[0063] (Item 4) In the surface analysis instrument described in any one of items 1 to 3, the cluster analysis section may perform hierarchical density-based cluster analysis.

[0064] The surface analysis instrument described in paragraph 4 can effectively cluster data points on a three-way scatter diagram created based on data collected by, for example, an EPMA, etc. This allows the user to accurately perform phase analysis based on the clustering results. [Explanation of symbols]

[0065] 1...Electron beam irradiation unit 100...Electron gun 2...Sample stage 3...Sample 4...Analyzing crystal 5...X-ray detector 7...Sample stage drive unit 8...Analysis control unit 9...Data processing section 90...Element intensity calculation section 91...Data storage section 92...Scatter plot creation section 93...Cluster analysis parameter adjustment section 94...Cluster Analysis Department 95...Cluster region detection section 96...Display processing unit 97...3-way scatter plot creation section 98...Addition intensity value data selection processing section 10...Central control unit 11...Operation unit 12...Display section

Claims

1. a measuring unit for acquiring signals reflecting the amounts of a plurality of components or elements to be analyzed at a plurality of positions on the sample; a scatter diagram creation unit that creates a three-way scatter diagram based on the measurement results from the measurement unit; a data point selection unit that uses distribution information of an added signal value obtained by adding the signal values ​​of three components or elements corresponding to data points in the ternary scatter diagram to exclude data points in a predetermined signal value range in which the added signal value is relatively small from all data points present on the ternary scatter diagram; a cluster analysis unit that clusters the data points in the three-way scatter plot that were not excluded by the data point selection unit using a density-based cluster analysis technique; wherein the data point selection unit creates a histogram of the sum signal values, and determines data points to be excluded using classes indicating maximum and minimum values ​​detected in the histogram, identifies classes indicating minimum values ​​that are above classes indicating maximum values ​​in the histogram, and excludes data points whose sum signal values ​​fall within a range from sum signal value=0 to the identified class of minimum values.

2. 2. The surface analysis instrument according to claim 1, wherein the cluster analysis unit performs hierarchical density-based cluster analysis.

Images