JP2005134181A - Data processing method and analyzing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To overcome the problem wherein data, such as processed measurement data, are influenced by noise. <P>SOLUTION: In a data processing method for obtaining average value by using a surface analyzing apparatus, measuring the same region of a sample at least twice and accumulating the measurement data, the values of the measurement data are compiled in the order of the magnitude, the predetermined number of the measurement data are accumulated from the median, and the average value is obtained. Alternatively, the measured data are compiled, the data out of a predetermined range are excluded by a noise filter, the measured data are accumulated, and the average data are obtained. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a data processing method in an analyzer such as a transmission electron microscope, a scanning electron microscope, an electron probe microanalyzer, an Auger electron spectrometer, and a photoelectron spectrometer.

  When image data and spectral data are measured by scanning an electron beam or the like on the sample surface to obtain a scanning electron microscope image or mapping, high sensitivity and high S / N data are obtained by integrating measurement data in the same visual field. Sometimes. Conventionally, as shown in FIG. 6, the measurement data is added by the number of integrations.

  However, since only the measurement data of the same visual field is added, the finally obtained integrated image is an image of the average value of each pixel. Since the measurement data is randomly affected by noise and accidental sample drift, if this method includes abnormal data on one of the integration planes, the pixels containing the abnormal data may differ from the surrounding values ( FIG. 7). In an analyzer, the influence of noise or the like is very large while high resolution is required. When a large number of data can be measured, the influence of the abnormal value is small. However, the number of times of measurement of the surface treatment apparatus is limited due to the change with time of the sample and the influence of drift.

  As a conventional technique, there is a spectroscope peak search method for accurately and easily obtaining a peak position by a short time measurement (for example, Patent Document 1).

JP-A-8-31367

  The problem to be solved is that data obtained by processing measurement data is affected by noise or the like.

  According to a first aspect of the present invention, there is provided a data processing method in which the same region of a sample is measured at least twice using a surface analyzer and the measurement data is integrated to obtain an average value. An average value is obtained by integrating a predetermined number of measurement data from the values.

  According to a second aspect of the present invention, there is provided a data processing method in which the same region of a sample is measured at least twice using a surface analyzer and the measurement data is integrated to obtain an average value. A predetermined number is deleted from each of the value and the maximum value, and the measurement data is integrated to obtain an average value.

  The invention of claim 3 is a data processing method in which the same region of a sample is measured at least twice using a surface analyzer and the measurement data is integrated to obtain an average value. Data other than the predetermined range is excluded, and the measurement data is integrated to obtain an average value.

  The invention of claim 4 is characterized in that the measurement data is image data.

  The invention of claim 5 is characterized in that the measurement data is spectrum data.

  The invention according to claim 6 is a surface analysis apparatus including the data processing means.

  According to the present invention, it is possible to obtain processed data with less influence of noise or the like.

  Aggregate the measurement data, apply statistical methods and mathematical filters to remove outliers, and then integrate the measurement data.

  FIG. 19 shows a scanning electron microscope. The scanning electron microscope is equipped with an electron gun 1 for generating electrons, a condenser lens 2 and an objective lens 4 for obtaining a thin electron beam bundle, and a scanning coil 5 for scanning an electron beam. The surface of the sample 6 is scanned with an electron beam having a diameter of several nm. The sample 6 is attached to a sample moving table, and an electron beam is irradiated to the sample position to be observed. Various signals such as secondary electrons, reflected electrons, X-rays, and light are emitted from the portion irradiated with the electron beam according to the form of the sample. The scanning electron microscope detects secondary electrons having the largest signal amount among these, and displays the difference in the amount of secondary electrons on the CRT 11 as the difference in brightness. The measured screen is stored as image data by a computer (not shown).

FIG. 1 shows image data measured by the scanning electron microscope of FIG. One screen is obtained by one measurement, and N integrated planes are obtained by N times. In one integrated surface, pixels represented by the data _a n (i, j).

When integrating the data of each integration plane, pixel data a ₁ (i, j) to a _N (i, j) corresponding to coordinates (i, j) on the integration plane 1 to N times as shown in FIG. The intensity values are rearranged in descending order, and the median (sample median) in that order or the average value of a predetermined number of data near the median is employed.

  The sample median is the median value when the measured values of the sample are arranged in order of size. However, when the number of samples N is an even number, there is no central value, so an average value of two values corresponding to the center counted from both ends is taken. If the distribution is close to symmetry, then average = arithmetic mean = sample median and should be nearly consistent. When the distribution is not symmetric, this difference is large. The most important point that the sample median differs from the average is that it is determined by the order of the magnitude rather than the magnitude of the measurement. Therefore, even if there is a particularly large or small abnormal value in the data, it does not depend on it. On the other hand, the simple average value depends on the abnormal value.

  FIG. 2 is an example when there are 19 integration planes. As shown in the left diagram of FIG. 2, the twelfth integration plane includes abnormal data. The right side of FIG. 2 shows the data of the left side of FIG. 2 arranged according to the magnitude of the intensity value. From the median, 9 data are adopted for each of the upper and lower data and averaged to obtain the average value according to the present invention. Therefore, as shown in the right diagram of FIG. 2, two abnormal values from the bottom and two abnormal values from the top are excluded. As a result, as shown in the left diagram of FIG. 2, the average value according to the present invention is closer to a probable value than the simple average including abnormal values. In other words, even if there is a pixel that contains abnormal data due to noise or sample drift on one of the integration planes, the probable value for each pixel is always near the median, so the influence of the abnormal value Can be cut. In particular, the influence of spike noise in the scanning electron microscope image and the shift in the mapping data for a long time can be reduced, and the edge portion that is a problem in general image smoothing does not occur.

  FIG. 8 is a scanning electron microscope image with vibration noise and acoustic noise. FIG. 9 is an accumulation of the screen with noise shown in FIG. The left column of FIG. 9 is a conventional simple average, and the right column is an integration screen according to the present invention. In the CRT screen, it can be clearly confirmed that the integration screen according to the present invention is clearer.

  In addition, this invention is not limited to said form, A various deformation | transformation is possible. For example, you may apply to the data processing method in analyzers, such as a transmission electron microscope, an electron probe microanalyzer, and a photoelectron spectrometer. Further, instead of taking an average of a predetermined number near the median value, it is also possible to take the average of the remaining data by excluding the predetermined number from the top and bottom of the data arranged according to the magnitude of the intensity value.

  3 and 4 show another data processing method in the scanning electron microscope of FIG. The description of the same components as those in the first embodiment is omitted. The second embodiment is a method for processing data obtained by the same apparatus as that of the first embodiment, and the difference is that for each of the corresponding image data groups obtained from the same field of view, N pixels are to be integrated. Is to apply a mathematical filter. For example, a filter that performs statistical processing and removes a value far from the standard deviation is used.

  FIG. 3 is an example when there are 19 integration planes. As shown in the left diagram of FIG. 3, the twelfth integration plane includes abnormal data. A standard deviation of this data is obtained, and a mathematical filter is applied to remove data greatly deviating from the 12th standard deviation. As a result, as shown in the left diagram of FIG. 2, the average value according to the present invention is closer to a probable value than the simple average including abnormal values.

  FIG. 5 is a flowchart of the statistical processing. In the computer that controls the apparatus, condition input is performed, measurement is started, and data of 19 integrated planes are captured and stored in the memory. Next, 19 pieces of data are read out, a standard deviation is calculated, and data greatly deviating from the standard deviation is removed for each pixel. The data that has not been removed is integrated and an image is output.

  As a result, even if image data including abnormal data is obtained on one of the integration planes due to noise or sample drift, the influence of the abnormal data can be reduced for each pixel.

  FIG. 20 shows an Auger electron spectrometer. The electron beam irradiated from the electron gun 18 is irradiated onto the surface of the sample 13. At this time, an electron beam is scanned on the surface of the sample 13 using a deflection coil (not shown), and an Auger image of a predetermined region on the sample surface can be observed. Auger electrons that have jumped out of the sample surface are applied to an electrostatic hemispherical analyzer 20 that is an apparatus for analyzing energy via an input lens unit 19. When the voltage to be applied to the electrostatic hemispherical analyzer 20 is determined, only electrons having energy corresponding to the voltage pass through the electrostatic hemispherical analyzer 20 and are detected by the detector 21.

  Thus, by examining the amount of Auger electrons detected while manipulating the voltage applied to the electrostatic hemispherical analyzer 20, it is possible to determine what elements are present on the sample surface. If the energy corresponding to Auger electrons is recorded on the horizontal axis and the amount corresponding to the amount of Auger electrons is recorded on the vertical axis, the spectrum of Auger electrons can be obtained. Further, an Auger image is obtained by scanning an electron beam on the sample surface, measuring the amount of Auger electrons at each point on the sample, and displaying the amount corresponding to the color.

  FIG. 10 shows the measured spectral data. For spectral data groups under the same measurement conditions, the energy value is plotted on the horizontal axis, and the corresponding intensity value of each spectrum is plotted on the vertical axis. The right side of FIG. 11 is a diagram in which the intensity values are rearranged in descending order. Regardless of the intensity value, the median (median) in that order or the average value of several points near the median is obtained and used as the value of the new spectrum data. If this process is performed on the entire horizontal axis of the spectrum, even if one of the measured spectrum groups is measured including abnormal data such as noise, an average spectrum with its influence cut off can be obtained. By applying this method, it is possible to cancel irregular changes due to irregular acoustic / vibration noise or accidental drift, and a stable spectrum can be obtained. Further, even if this method is applied, no sneak in the small peak portion, which is a problem in general smoothing, is generated.

    FIG. 17 shows 10 spectra with electrical noise. FIG. 18 is obtained by integrating the spectrum with noise shown in FIG. The middle part of FIG. 18 is a conventional simple average spectrum, and the lower part is a spectrum integrated according to the present invention. Clearly, it can be confirmed that the integration according to the present invention is less affected by noise.

  12 and 13 show another data processing method in the Auger electron spectrometer. The description of the same components as those in the third embodiment is omitted. The fourth embodiment is a method of processing data obtained by the same apparatus as that of the third embodiment. The difference is that a mathematical filter is applied to the intensity value of the spectral data group obtained in the same field of view. It is to be. For example, a filter that performs statistical processing and removes a value far from the standard deviation is used. Averaging is performed using the values that have not been removed, and the entire spectrum is reconstructed to obtain a new average spectrum. As a result, even if measurement is performed including abnormal data such as noise in one of the measured spectrum groups, an average spectrum in which the influence is cut can be obtained. By applying this method, it is possible to cancel irregular changes due to irregular acoustic / vibration noise or accidental drift, and a stable spectrum can be obtained. FIG. 14 is a flowchart of the statistical processing.

It is the image measured with the scanning electron microscope, and the image processed by this invention. (Example 1) It is a graph of the pixel data in FIG. (Example 1) It is the image measured with the scanning electron microscope, and the image processed by this invention. (Example 2) It is a graph of the pixel data in FIG. (Example 2) It is a flowchart of the data processing by a statistical processing method. (Example 2) It is the image measured with the scanning electron microscope, and the image processed by the prior art. (Example 1) It is a graph of the pixel data in FIG. (Example 1) It is a scanning electron microscope image with vibration noise and acoustic noise. (Example 1) It is the figure which processed the image in FIG. (Example 1) It is the spectrum measured with the Auger electron spectrometer and the spectrum processed by this invention. Example 3 It is a graph of the spectrum of the predetermined energy in FIG. Example 3 It is the spectrum measured with the Auger electron spectrometer and the spectrum processed by this invention. Example 3 It is a graph of the spectrum of the predetermined energy in FIG. Example 3 It is a flowchart of the data processing by a statistical processing method. (Example 4) It is the spectrum measured by the Auger electron spectrometer and the spectrum processed by the prior art. Example 3 It is a graph of the spectrum of the predetermined energy in FIG. Example 3 The spectrum includes electrical noise. Example 3 It is the figure which processed the spectrum of FIG. Example 3 It is a schematic diagram of the scanning electron microscope used for data measurement. (Examples 1 and 2) It is a schematic diagram of the Auger electron spectrometer used for data measurement. (Examples 3 and 4)

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Electron gun 2 Condenser lens 3 Objective lens aperture 4 Objective lens 5 Scanning coil 6 Sample 7 Secondary electron detector 8 Scanning power supply 9 Image signal amplifier 10 Deflection coil 11 Braun tube 12 Analysis chamber 13 Sample 14 Sample holder 15 Sample stage 16 Sample stage Adjustment mechanism 17 Vacuum pump 18 Electron gun 19 Input lens unit 20 Electrostatic hemisphere analyzer 21 Detector 22 Computer

Claims (6)

  In a data processing method in which the same area of a sample is measured at least twice using a surface analyzer and the measurement data is integrated to obtain an average value. A data processing method that obtains the average value by integrating.   In a data processing method in which the same area of a sample is measured at least twice using a surface analyzer and the measurement data is integrated to obtain an average value, the measurement data is aggregated in order of magnitude, and the minimum value and the maximum value are respectively predetermined. A data processing method that removes numbers and integrates measurement data to obtain an average value.   In a data processing method that uses a surface analyzer to measure the same region of a sample at least twice and integrates the measurement data to obtain an average value, the measured data is aggregated, and data outside a predetermined range is obtained by a statistical method. A data processing method that excludes and integrates measurement data to obtain an average value.   The data processing method according to claim 1, wherein the measurement data is image data.   The data processing method according to claim 1, wherein the measurement data is spectrum data. A surface analysis apparatus comprising the data processing means according to claim 1.

Images