TW201508697A - Pattern dimension measurement method and device - Google Patents

Abstract

In order to reliably detect hole patterns and carry out pattern measurement even under the influence of noise, distortion, and surrounding patterns in an image in which concentrated patterns having a regular arrangement are imaged, an image of patterns on a substrate is imaged, pattern arrangement detection processing is carried out in which the regular arrangement of the patterns within an image area is detected, template matching is carried out in which the positions of the patterns within the image area are detected, the position of each pattern is detected through the integration of the results of the pattern arrangement detection and the template matching, measurement cursors are set, and pattern dimension measurement processing is carried out for each measurement cursor.

Description

Pattern size measuring method and device thereof

The present invention relates to a method for measuring the size of a plurality of patterns formed on a semiconductor wafer, and an apparatus therefor, and more particularly to a pattern size measuring method suitable for measuring a size in which a plurality of fine patterns are formed densely to a substrate and Device.

In the measurement of the semiconductor pattern using the SEM (Scanning Electron Microscope), in recent years, the measurement of the hole pattern has been greatly increased due to the miniaturization and high density of the hole pattern. Further, the variation in the shape is also greatly increased, and the conventional pattern matching used to set the pattern measurement portion in an automatic manner is insufficient in performance, and the number of man-hours for which the user manually measures the cursor setting is increased.

Non-Patent Document 1 discloses a technique for improving the performance of the template matching itself. Further, Patent Document 1 discloses a method of creating a statistical model image and improving the accuracy of template matching. Patent Document 2 discloses a method of analyzing the repeatability of a pattern image, and detecting a field having the same pattern to automatically generate a template.

Moreover, in Non-Patent Document 2, it is revealed that there is a use in the The characteristic portion of the pattern in the image is called the feature extraction method of SIFT to perform matching.

Further, Non-Patent Document 3 discloses a method of obtaining the size or the surface end of a hole pattern by elliptical matching. Further, Patent Document 3 describes that the distance between patterns is measured using a threshold method.

[Previous Technical Literature] [Patent Literature]

[Patent Document 1] Japanese Patent Laid-Open Publication No. 2010-276487

[Patent Document 2] Japanese Patent Laid-Open Publication No. 2011-70602

[Patent Document 3] Japanese Patent No. 4262592

[Non-patent literature]

[Non-Patent Document 1] Murakami, 2 other, "Intensive Symbol Correlation and Robust Image Control", Letter Society, 2000

[Non-Patent Document 2] David G. Lowe, "Distinctive image features from scale-invariant keypoints", Journal of Computer Vision, 60, 2, pp. 91-110, 2004

[Non-Patent Document 3] Yamada, Kimi, "High-precision calculation method for ellipse and its performance comparison" Research Report of the Information Processing Society, 2006-CVIM-154-36, 2006/5/18, 19, pp.339-346

When the size of the pattern is measured by photographing the images of the regularly arranged patterns, it is necessary to perform the measurement algorithm in setting the measurement area (measurement cursor) for one pattern. In order to automatically set the measurement cursor, it is known that the image obtained by the pre-acquisition is registered as a template in the pattern area of the measurement target, and the measurement cursor is automatically set via the template matching. However, when the deformation of the shape is increased and the unstable pattern is targeted, sufficient performance cannot be satisfied under the conventional model matching.

The methods disclosed in Non-Patent Documents 1 and 2 and Patent Documents 1 and 2 are patterned by either a template matching method or a technique of detecting a regular arrangement of patterns in an image with a focus on repeatability. Detection. The template matching is proportional to the matching accuracy due to the instability of the pattern shape of the matching end. The method of detecting the regular arrangement of patterns is to cause a malfunction when the regularity cannot be locally established or when there is no overall regularity.

The present invention solves the above-described problems of the prior art, and provides a case where a pattern of a pattern is unstable in a plurality of fine patterns formed in a dense substrate, or the regularity of arrangement of the patterns cannot be partially established. In the case where there is no overall regularity, the position of the pattern can be accurately detected, and the pattern size measuring method for measuring the size of the pattern and the device thereof can be used.

In order to solve the above problems of the prior art, in the present invention, In the method of measuring the size of a pattern, a plurality of patterns having the same shape formed on a sample are photographed to obtain a plurality of patterns; and a pattern extracted by using a template matching method for images of a plurality of patterns is used. The information is integrated with the information of the pattern extracted from the periodic information of the plurality of patterns of the image using the plurality of patterns, and the measurement cursor is set, and the image of the plurality of patterns obtained is used. The measurement cursor is set, and the size measurement area is set; the image of the plurality of patterns is processed in the image of the pattern of the measurement area set by the measurement cursor, and the size of the pattern is measured.

In order to solve the above problems, in the present invention, the apparatus for measuring the size of the pattern is provided with an image acquisition means for photographing a plurality of patterns having the same shape formed on the sample to obtain a plurality of patterns. The image of the pattern; the measurement cursor setting means is a method of extracting the information of the pattern by the template matching method for the image of the plurality of patterns obtained by the image acquisition means, and using the periodic information of the plurality of patterns of the image of the plurality of patterns Extracting the information of the pattern, integrating the information of the pattern extracted by the template matching method with the information of the pattern extracted by the periodic information of the plurality of patterns, and integrating the results, and using the obtained integrated result setting measurement The cursor; the size measurement area setting means sets the measurement measurement field set by the measurement cursor setting means for the image of the plurality of patterns obtained by the image acquisition means, and sets the size measurement area; and the size measurement means uses the image acquisition means The image of the plurality of patterns obtained is present in In the measurement cursor setting means set the cursor to set the measurement area of the image pattern measured good dimensional It is processed to measure the size of the pattern.

In the present invention, two methods of detecting the pattern matching and the regular arrangement of the patterns are performed. The arrangement pattern detection is a pattern in which the arrangement of the patterns is statistically detected from the periodicity of the entire image, and the pattern cannot be detected because the performance of the template matching is insufficient. By integrating the two results, the detection pattern can be stabilized, and the automatic setting of the measurement cursor can be performed.

According to the present invention, even in the case of a plurality of fine patterns densely formed on the substrate, even if the shape of the pattern is unstable, or the regularity of the arrangement of the patterns cannot be partially established, or the overall regularity is not obtained. In this case, the position of the pattern can be accurately detected and the size of the pattern can be measured.

100‧‧‧Measurement length SEM

102‧‧‧Electronic gun

103‧‧‧Concentrating lens

104‧‧‧ bias coil

105‧‧‧object lens

106‧‧‧ platform

107‧‧‧Measurement Wafer (Sample)

108‧‧‧Detector

109‧‧‧A/D converter

110‧‧‧Image Processing Department

111‧‧‧ platform controller

112‧‧‧Electron Optics Control Department

113‧‧‧The overall control department of the device

114‧‧‧Control terminal

200‧‧‧Import into I/F

201‧‧‧Image Processing Control Department

202‧‧‧ Calculation Department

203‧‧‧ memory

204‧‧‧ Busbar

205‧‧‧Data input I/F

206‧‧‧ memory device

Fig. 1 is a block diagram showing a hardware configuration of a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a video processing unit according to Embodiment 1 of the present invention.

Fig. 3 is an image of a pattern of a measurement target according to the first embodiment of the present invention.

Fig. 4 is an image of a pattern of a measurement target according to the first embodiment of the present invention.

Fig. 5 is an image of a pattern of a measurement target according to the first embodiment of the present invention.

Fig. 6 is a flowchart showing a detection algorithm of a measurement cursor according to the first embodiment of the present invention.

Fig. 7 is a view for explaining an image of a pattern matching pattern according to Embodiment 1 of the present invention.

FIG. 8 is a graph showing a distribution of correlation coefficients of a template image and a pattern image obtained by the template matching according to the first embodiment of the present invention. FIG.

Fig. 9 is an image obtained by photographing an actual pattern relating to the template matching in the first embodiment of the present invention by SEM.

Fig. 10 is a graph showing the distribution of the correlation coefficient between the template image obtained by the template matching in the first embodiment of the present invention and the image obtained by imaging the actual pattern by SEM.

Fig. 11 is an image of a measurement target pattern according to the first embodiment of the present invention.

FIG. 12 is a spectrum image of the measurement target pattern of FIG. 12. FIG.

Fig. 13 is an image of a measurement target pattern according to the first embodiment of the present invention.

Fig. 14 is a spectrum image of a measurement target pattern according to the first embodiment of the present invention.

Fig. 15 is a view showing an image of a measurement target pattern according to the first embodiment of the present invention.

Fig. 16 is a view showing a measurement target pattern according to Embodiment 1 of the present invention. Fourier spectrum diagram.

Fig. 17 is a view showing an image of a measurement target pattern according to the first embodiment of the present invention.

Fig. 18 is a view showing an image of a measurement target pattern according to the first embodiment of the present invention.

19 is an image of a measurement target pattern according to the first embodiment of the present invention, which is an image in a case where a part of the pattern arrangement is disordered.

Fig. 20 is a view showing an image of a measurement target pattern according to the first embodiment of the present invention, which is an image in the case where a terminal having a repeating pattern is present in the image.

Fig. 21 is a view showing an image of a measurement target pattern according to the first embodiment of the present invention in a case where a part of the pattern arrangement is disordered.

Fig. 22 is a correlation coefficient image obtained by performing template matching on the measurement target pattern according to the first embodiment of the present invention.

[Fig. 23] Fig. 23 is an image showing the result of detecting the arrangement pattern of the measurement target pattern in the first embodiment of the present invention.

FIG. 24 is an image showing the result of detecting the arrangement pattern of the measurement target pattern in the first embodiment of the present invention.

Fig. 25 is an image of a pattern of a measurement target according to the first embodiment of the present invention.

Fig. 26 is a view showing an image of a symmetrical pattern according to the first embodiment of the present invention.

FIG. 27 is a view showing a distribution of an image of a measurement target pattern according to the first embodiment of the present invention and an autocorrelation function in the X direction and the Y direction of the measurement target pattern.

Fig. 28 is a view showing the positional relationship between the image and the original image shifted in the case of the autocorrelation function according to the first embodiment of the present invention.

[Fig. 29] A diagram for explaining the arrangement pattern detection of the autocorrelation function of the direction of the orthogonal arrangement of the image of the measurement target pattern and the measurement target pattern in the first embodiment of the present invention.

Fig. 30 is a flow chart showing the processing procedure of the first embodiment of the present invention.

Fig. 31 is a flow chart showing the flow of processing at the time of measurement in the first embodiment of the present invention.

Fig. 32 is a front view of a GUI relating to Embodiment 1 of the present invention.

Fig. 33 is a flow chart showing the operational sequence of the apparatus at the time of formula setting in the first embodiment of the present invention.

Fig. 34 is a flow chart showing the processing procedure of the second embodiment of the present invention.

Fig. 35 is a flow chart showing the processing procedure at the time of measurement in the second embodiment of the present invention.

Fig. 36 is a flow chart showing the processing procedure of the third embodiment of the present invention.

Hereinafter, the first embodiment of the present invention will be described with reference to the overall configuration, and the contents of the respective processes will be described in order.

[Example 1]

Fig. 1 is a view showing the overall configuration of a fine pattern measuring apparatus according to the present invention.

The length measuring SEM 100 of the present embodiment is configured by: a stage 106 on which a measuring wafer (sample) 107 is placed, an irradiation optical system that controls the electron beam 101 discharged from the electron gun 102, and detection is emitted from the sample. The second electronic detector 108 and the signal processing system for detecting signals. The illumination optical system is configured by an electron gun 102, a condensing lens 103 positioned on the path of the electron beam harness 101, a deflection coil 104, and a counter lens 105. The electron beam harness 101 is collected by the optical system: there is a field in which the pattern of the measurement target on the wafer 107 is designated.

The secondary electrons detected by the detector 108 are converted into digital signals by the A/D converter 109. The converted digital signal is sent to the image processing unit 110, and the image processing unit 110 takes out a digital signal stored in the memory, performs image processing, and performs pattern size measurement and the like. 111 is a platform controller, 112 is an electro-optical system control unit, 113 is a control unit of the entire apparatus, and 114 is a control terminal connected to the control unit.

The video processing unit 110, the overall control unit 113, and the control terminal 114 are configured to be connectable to a recording medium (not shown), and the program executed by the video processing unit 110 is read from the recording medium and can be loaded. Go to the image processing unit 110.

FIG. 2 is a view showing the configuration of the image processing unit 110. The second electronic signal that is converted into a digital signal by the A/D converter 109 is The data input I/F 205 is sent to the memory 203, and the image data is stored in the memory 203 to be read. The video processing program is read from the memory 203 or the memory medium via the video processing control unit 201. The video processing control unit 201 controls the calculation unit 202 to process the intermediate processing data obtained by processing the image data stored in the memory 203 or the image data in accordance with the read video processing program, and measures the pattern.

The pattern measurement result is sent to the overall control unit 113 through the I/F 200, and the measurement result is indicated by the control terminal 114 shown in FIG. Further, the operation command of the image processing unit 110 is input from the overall control unit 113 to the image processing control unit 201 via the input/output I/F 200. The transmission and reception of data in the image processing unit 110 is performed through the bus bar 204.

3, 4, and 5 show an example of a pattern image to be detected. The pattern image 301 of FIG. 3 is a pattern in which the holes 302 to be measured are arranged in a checkered shape. The pattern image 401 of FIG. 4 is a pattern in which the arrangement of the holes of the pattern image 301 is inclined by 45 degrees. Further, as in the pattern image 501 of FIG. 5, there are two patterns in which a plurality of holes are arranged as a regular arrangement. In the measurement, the hole pattern of the single image is detected from the entire image, and the hole field 303 is set, and the field 303 is processed by the length measurement algorithm. In the field of performing the length measurement algorithm, for example, a field surrounded by a broken line as shown by 303 or 304 in Fig. 3 is referred to as a measurement cursor.

Figure 6 shows an overview of the detection method of the measurement cursor. Want. The input image 601 is a setting target image of the measurement cursor, and the input image 602 is a template image of the measurement target pattern. First, the arrangement pattern detection processing 603 performs periodic analysis in the image to obtain a result 605 in which the hole position is detected. Moreover, a template match 604 is performed on the image 601 using the template image 602, and a result 606 of detecting the position of the hole is obtained.

The method of matching the pattern detection method and the template is to detect the presence of a tricky pattern; in the case of the example of Fig. 6, in the case where the hole position is detected by the arrangement pattern detection method, the field 609 having no hole actually exists. The hole is detected by mistake. On the one hand, in the case where the template matching is used, the field 609 is correctly recognized as the field without holes, but a part of the missing hole pattern 611 is missed, and the field 610 spanning the plurality of holes is regarded as Mistaken detection is performed in the field of the hole pattern. The missing pattern 611 is due to distortion or electrification affecting the hole in the image where the contour is not closed. This pattern 611 is detected as a hole area in the image 605 detected by the arrangement pattern detecting method. Here, in a manner in which the two results are integrated by the integration process 607, the detection result 608 in which the length measurement field is set can be obtained by finally using the measurement cursor without missing or false detection.

The template matching 604 of FIG. 6 will be described in detail with reference to FIGS. 3 and 4. For the pattern image 301 of FIG. 3, the field 303 is used as a template image, and the height of the template image is N, the banner is M, and the pixel image of the coordinates (i, j) on the template image has a pixel value of I (i, i. j) When the pixel value of the template image 303 is T (i, j), the template matching is performed using the normalization correlation expressed by (Formula 1).

At this time, the field 304 surrounded by the hole pattern is a peak having a correlation coefficient RNCC which is high, and is a cause of erroneous detection. Similarly, in the video 401 of FIG. 4, when 402 is used as the template image, the correlation coefficient having a high peak is formed in the field 403 of the hole pattern.

The image 701 of FIG. 7 is an image in which a part of the field of the image 401 is enlarged, and is an ideal image without the influence of noise or the like. The vertical axis 801 of the graph of Fig. 8 is a correlation coefficient obtained by normalizing the correlation (Equation 1) while shifting the center of the template image 402 along the arrow 702, and the horizontal axis 806 represents the x on the arrow 702. coordinate. The correlation coefficient represents peaks 802, 804 at points 703, 705 where the holes are present. Further, the point 704 is surrounded by the edge indicated by 706 and has a composition of a pixel close to the shape of the hole, and the correlation coefficient has a peak 803. If there is no influence of noise or the like, the correlation coefficient of the point 704 is sufficiently lower than the correlation coefficient of the points 703 and 705, and only the hole pattern can be detected by setting the threshold 805.

However, the image obtained by photographing the actual pattern by SEM is as shown in the image 901 of FIG. 9, and the distortion due to noise or imaging conditions or the hole pattern existing in the periphery affects the matching. For this reason, if the graph of the correlation coefficient on the arrow 902 is shown in Fig. 10, the peak value 1001 of the point 903 between the holes occurs, and the correlation coefficient of the peak value 1002 of the point 904 of the center position of the hole is higher. In this case, set threshold 1003 When the hole of the detection point 904 is detected, the point 903 having no hole is erroneously detected; if the threshold is set to 1004 so that the point 903 is not detected, the hole 904 is omitted. Here, in the processing sequence shown in FIG. 6, unlike the template matching 604 which is susceptible to noise or distortion, the regularity of the arrangement of the patterns is regarded as periodic analysis of the target, and statistical stability is performed. The arrangement pattern is detected 603.

The arrangement pattern detection 603 of FIG. 6 is explained. The image 1101 of FIG. 11 in which the hole patterns are arranged in a lattice pattern is subjected to discrete Fourier transform to calculate the spectrum image 1201 of FIG. The spectrum image shows a large frequency from the center toward the outside, and a grid-like peak occurs in response to the periodicity existing in the pattern image. The horizontal interval 1102 of the pattern of FIG. 11 corresponds to the peak point 1203 on the spectral image of FIG. 12; the vertical interval 1103 of FIG. 11 corresponds to the peak point 1204 of the spectral image of FIG. These peaks can be detected as the first and second largest peak points of the peak 1202 of the center.

In Fig. 12, the frequency 1205 in the x-axis direction of the peak point 1203 is p, and the frequency 1206 in the y-axis direction of the peak point 1204 is q. In Fig. 11, the horizontal interval 1102 of the pattern image is P, and the vertical interval 1103 When it is Q, P and Q can be calculated by (Formula 2) and (Formula 3). However, N is the banner of the pattern image, and M is the vertical width of the pattern image.

In the video 1101 of FIG. 11, the center position of the template field 1104 is used as a base point, and the grid 1105 is arranged at the obtained intervals P and Q, so that the pattern where the pattern exists periodically can be used as the intersection of the grids. take. Further, the base point of the grid can also be obtained from the phase component of the peak frequency obtained by Fourier transform.

FIG. 13 and FIG. 14 show an example of the arrangement pattern detection in the case where the hole pattern is skewed. The image 1401 of FIG. 14 is a Fourier spectrum image of the lattice pattern 1301 in a lattice shape arranged as shown in FIG. The grid 1305 on the pattern image 1301 of FIG. 13 corresponds to the peak point 1402 of the spectrum image 1401 of FIG. The peak point 1402 is the first or second largest peak point except the center point 1403. At this time, in FIG. 13, the angle 1304 of the line 1303 perpendicular to the grid 1305 and the X-axis is the same as the line 1404 connected to the peak point 1402 from the center point 1403 of the spectrum image of FIG. 14 and the angle 1404 of the X-axis.

In FIG. 14, when the coordinate of the peak point 1402 is (u, v), the straight line connected from the center point 1403 to the peak point 1402 and the angle 1404 of the X-axis are obtained as θ of (Expression 4).

The period of the peak point 1402 is represented by a distance 1405 from the center point 1403 to the peak point 1402, which can be derived from (mathematical formula 2) Find the interval 1302 of the grid.

Next, as shown in the pattern image 1501 of FIG. 15, an example of the arrangement pattern detecting method in the case where the pattern 1502 of the two holes is skewed is shown. The Fourier spectrum image of the hole pattern image 1501 is shown to 1601 of FIG. The arrangement pattern of the pattern image 1501 is divided into: a pattern in which one hole is arranged on the grid 1701 and the grid 1702 of FIG. 17, and a grid 1801 and a grid 1802 of FIG. 18 in which two groups of 1502 are skewed. Configuration pattern.

The grid 1701 of FIG. 17 corresponds to the peak 1602 in the spectral image of FIG. 16, and the grid 1702 corresponds to the peak 1603 in the spectral image. The grating 1801 of FIG. 18 corresponds to the spectral image of FIG. The peak 1604 occurs, and the grid 1802 occurs corresponding to the peak 1605 in the spectral image. In Fig. 16, these four peaks become peaks of four bits above the center in the spectrum image. When the threshold is set here and a plurality of peaks which are equal to or greater than the threshold value are detected, the upper four peaks can be detected. In this case, in order to prevent detection of the peak 1606 or the peak 1607 other than the 1602, 1603, 1604, and 1605 of FIG. 16, the maximum frequency of the detected peak can be determined by the parameter.

The arrangement pattern detection 603 detects the pattern by statistical processing from the arrangement of the plurality of hole patterns, and even if the pattern shape collapses due to the influence of noise or distortion, the influence is less than the template matching 604. However, for the pattern images 1901 and 2101 in which the regularity of the arrangement as shown in FIGS. 19, 20, and 21 is partially collapsed, or the image 2001 having the repeated terminal, the pattern detection cannot be performed correctly. Here, as shown in Figure 6. In general, an integration process 607 for complementing the shortness of the template match 604 with the arrangement pattern detection 603 is performed.

The integration process 607 of the flow of the process shown in FIG. 6 will be described using FIG. 22 and FIG. 23 in detail. The image 2201 of FIG. 22 is a correlation coefficient image obtained by performing template matching on the hole pattern image 301 shown in FIG. The pixel value of the pixel 2204 of the correlation coefficient image 2201 is a correlation coefficient when the center of the template image 303 is integrated to the pixel 2204, and corresponds to the wave height value of the waveform shown in the graph of FIG. 8 or FIG. . In the correlation coefficient image 2201, as described with reference to FIGS. 7 and 8, there are a peak 2202 occurring at the center position of the hole and a peak 2203 occurring in the field between the hole and the hole.

The image 2301 of FIG. 23 is a result image of the arrangement pattern detection of the hole pattern image 301 shown in FIG. The strip-shaped field 2303 is a field in which the grid 2302 obtained by the array detection is the center and the width is only 2304. The same strip-shaped field 2308 is also obtained in the grid 2305. The field 2306 where the fields of the respective grids intersect is a field in which there is a high possibility that the hole exists as a result of the arrangement pattern detection. In this example, the position of the hole of the pattern image 301 is actually obtained.

Here, as an example in which the correlation coefficient between the holes affected by the noise is higher than the correlation coefficient of the hole center position, considering that the correlation coefficient 2202 of the hole center position of FIG. 22 is 0.6, the correlation coefficient 2203 between the holes is 0.7. The occasion. In the video 2301 of FIG. 23, the area 2306 in which the bands overlap is 1.0, and the area other than this is 0.6, and the weighting factor corresponding to the probability that the pattern exists in each position is set. Put these images with 2201 The correlation coefficient between the pixel values of the same coordinates of the image 2301 and the value obtained by multiplying the weighting factor (integrated value) is used as an integration result.

The correlation coefficient 2202 of the center position of the hole of Fig. 22 corresponds to the field 2306 in which the possibility of the hole of Fig. 23 is high, and the result of the multiplication is 0.6; the correlation coefficient 2203 of the hole of Fig. 22 corresponds to the presence of the figure. The probability of the hole of 23 is low for the field 2307, and the result of the multiplication is 0.42. By this integration result, the integrated value of the hole position is higher than the integrated value between the holes, and the threshold value can be set to the integrated value so that the hole can be detected without erroneous detection.

In FIG. 24, the image 2401 is an image indicating the detection result of the hole pattern image 1501 shown in FIG. For the grids 2402, 2403, 2404, 2405 in all directions, the strip-shaped fields are defined by the webs 2406, 2407, 2408, 2409, and all overlapping regions 2410 can be defined as fields with holes.

An example of the integration process 607 is applied to the image shown in FIGS. 19, 20, and 21 for the possibility of failure to detect the pattern 604 by the arrangement pattern. The lines 1904, 2003, and 2104 in each pattern image are the grids obtained by the arrangement pattern detection.

The hole pattern image 1901 of Fig. 19 is arranged in a lattice shape with holes, but in the field 1902, there is a pattern image without holes. In the case where the image is subjected to the template matching, the field 1902 has a small influence on the template matching compared with the area 1903 which is easy to be erroneously detected, and the correlation coefficient is a value close to 0.0. Further, in the detection result of the arrangement pattern, the field 1902 is the place where the grid 1904 intersects. The reason weighting factor is 1.0, but the integrated value multiplied by the correlation coefficient is a value close to 0.0, and the part will cause false detection. On the other hand, in Fig. 19, the field in which the hole exists is equivalent to the processing result of the pattern image 301, and the pattern can be correctly detected by the integration process.

The hole pattern image 2001 of Fig. 20 is the end of the repeating pattern. When the pattern image is subjected to template matching, the field in the field 2002 after the pattern end is repeated, and there is no field surrounded by the hole, and the correlation coefficient is a value close to 0.0. For this reason, in the field 2002, the result of the arrangement pattern detection is not related, and the integrated value is also a value close to 0.0, and erroneous detection does not occur. The field other than the field 2002 has the same result as the processing result of the pattern image 301, and the pattern can be correctly detected by the integration process.

In the hole pattern image 2101 of Fig. 21, the hole holes are arranged in a lattice shape, and there is a pattern image of the hole 2102 which has no relationship with the repeating pattern. The correlation coefficient due to the template matching of the holes 2102 is 0.8. In the result of the arrangement pattern detection, the hole 2102 is not present at the intersection of the grating 2104, and the weighting factor is 0.6, and the integration value is 0.8 × 0.6 = 0.48, which is a smaller value than the result of only the template matching. However, the hole 2103 in which the repeating pattern of the pattern image 2101 is formed has a margin between the holes in which only the other holes 2102 can be disposed, and the correlation coefficient does not become high in places other than the holes. For this purpose, the integrated value of the hole 2102 has a larger value than the field in which the hole does not exist, and the correct pattern can be detected by the threshold processing of the integrated value.

The detection algorithm of the measurement cursor shown in FIG. 6 can also be Used in other pattern images. The image 2501 of FIG. 25 is a pattern image in which lines are regularly arranged. In the case where the template matching is performed in the field of the field to be measured by the field 2502, there is a high possibility that the field 2503 having a component similar to the horizontal edge of the field 2502 is erroneously detected.

Further, the image 2601 of FIG. 26 is a regular horizontal arrangement, and a part of the lines are longitudinal image images. Field 2602 is the interior of the pattern. When the line terminal of the field 2603 is matched with the terminal as a measurement target, the possibility of erroneous detection is high in the field 2604 having an approximate shape. That is, it is difficult to perform pattern detection using the template matching method for the image 2601-like pattern of FIG.

However, when the detection algorithm of the measurement cursor described in FIG. 6 is used, the target images 2501 and 2601 can be detected without erroneous detection.

That is, with respect to the image 2601 of Fig. 26, the template matching is performed, and the correlation coefficient image as described with reference to Fig. 22 is created. Next, an image of the arrangement pattern detection result as described with reference to Fig. 23 is created from the correlation coefficient image. Correlation coefficients are defined in each field of the image in which the pattern detection result is matched, and the correlation coefficients of the same coordinates of the correlation coefficient image are multiplied to obtain an integrated value, and the threshold value processing of the integrated value can be used to correctly detect the pattern. .

The template matching 604 of FIG. 6 can detect a pattern by matching the feature points with the characteristic portions of the pattern contained in the template image. The extraction of the characteristic portion can be utilized in the feature amount extraction method called SIFT described in Non-Patent Document 2.

The correlation coefficient is calculated by calculating the feature amount of the coordinate corresponding to the feature point on the template image from the matching target image, and comparing it with the feature amount of the feature point of the template image. In the case of SIFT, the feature quantity is expressed by a histogram of the direction of the luminance gradient, and the matching between the feature quantities is the Kullback-Leibler distance D (n) represented by (Formula 5). ) to calculate.

However, i is a histogram bin number, P(n, i) is a histogram of the feature quantity of the nth feature point, and Q(n, i) is a feature point corresponding to the nth number. A histogram of the feature quantities of the coordinates of the matching end.

When the distance D(n) is larger than the threshold T, the matching coefficient G(n) is calculated by G(n)=0 and G(n)=1, and the correlation coefficient is calculated by (Expression 6).

However, N is the total number of feature points extracted.

Feature point matching is not processing all the pixels on the template image when the correlation coefficient is obtained, and only the feature points can be processed, which is higher than the normal template matching which is related to normalization. Process quickly.

The arrangement pattern detection 603 is a method in which an autocorrelation function can be used in addition to the Fourier transform. Waveform 2702 of Fig. 27 shows an autocorrelation function in the x-axis direction of the pattern image 2701 in which the holes are arranged in a lattice pattern. The center 2703 of the graph indicates when the position offset amount is 0. The autocorrelation function R(t) is relative to the original image 2801 of FIG. 28, such that the offset 2802 in the x-axis direction is t, and the banner 2805 of the field 2804 in which the offset image 2803 overlaps with the original image 2801 is W When the vertical width 2806 is H and the pixel value (x, y) of the original image 2801 is f (x, y), it is obtained by the following (Formula 7).

Accordingly, the threshold value 2704 is applied to the obtained autocorrelation function 2702, and the peak point 2705 other than the center 2703 is detected. The distance 2706 between the center 2703 and the peak point 2705 is the arrangement interval of the pattern in the x-axis direction. The autocorrelation function 2707 is also calculated in the y-axis direction, and the arrangement interval of the patterns is obtained in the same manner. Further, the pattern image in which the lattice-like arrangement is inclined with respect to the image 2901 of FIG. 29 can be obtained by tilting the direction of the shift to the direction of the arrow 2902 and the direction of the arrow 2904 perpendicular thereto to obtain the autocorrelation function. 2903, 2905.

The method of discrete Fourier transform is used to analyze the pattern image as the base of the sin wave. On the one hand, the autocorrelation function uses the pattern shape itself as the base, using signals of more images than Fourier analysis. Component analysis is periodic. For this reason, in an image with many noises or distortions, the arrangement pattern detection by the autocorrelation function can correctly perform the arrangement pattern detection than the method using the discrete Fourier transform.

In FIG. 30, the flow of the process of the measurement cursor setting operation (recipe setting) using the pattern measurement of the measurement algorithm of the measurement cursor of FIG. 6 is demonstrated. Further, the setting target is the pattern image 301 of FIG.

First, in step S3001, the pattern area of the measurement target on the wafer is photographed, and the image 301 of FIG. 3 is obtained. In step S3002, the acquired image 301 is indicated on the GUI. In step S3003, the field 303 used as the template image from the image 301 is designated by the GUI. In step S3004, the image 301 and the template image 303 are used, and the measurement algorithm of the measurement cursor shown in FIG. 6 is performed to detect the measurement cursor. In step S3005, the detection result of the measurement cursor by S3004 and the result of the arrangement pattern detection 603 matched with the template of the intermediate process 604 are output to the GUI.

In step S3006, the user can change the parameters of the arrangement pattern detection 603 and the template match 604 and the integration process 607 by using the GUI. In step S3007, the measured parameters are detected again using the adjusted parameters, and the display of the cursor detection result and the intermediate processing result is updated. If the user judges that it is necessary to correct the parameter, the process returns to step S3006, and the parameter is corrected again.

If there is almost no problem to detect the measurement cursor at the place to be measured, the process proceeds to step S3008. In step S3008, the user can add and delete the measurement cursor according to the detection result of the measurement cursor. In addition, location correction. In step S3009, the photographing position and the relative coordinates of the measurement cursor opposite thereto are saved as recipe data. Using the recipe data, the image 301 on a different wafer or on a different wafer is measured in the same pattern area. When the saving of the recipe data is completed in step S3009, the series of processing ends.

Fig. 31 is a flow chart showing the processing at the time of measurement using the recipe data for carrying out the processing flow of Fig. 30 and storing it. At S3101, the recipe data stored in step S3009 is read from the memory device 206. Next, in S3102, the wafer 107 is moved by the control of the platform 106 in the same manner as the image of the creation end of the recipe data until the field of photography. In S3103, the measurement area is photographed and positioned at a low magnification, and then the measurement is performed at a magnification of the measurement to obtain a pattern image. Next, in S3104, using the coordinates of the cursor in the recipe data, a measurement cursor is set for the acquired image, and measurement processing is performed for each cursor region. If there is a measurement field in which the measurement is not completed, the process returns to S3102, and if the measurement in all the measurement fields is completed, the process proceeds to S3106. The statistical value of the measurement result is calculated in S3106, and the result data is output to the memory device 206 to end the measurement.

Further, the measurement processing performed in S3104 is performed by the calculation unit 202. In the case of the hole pattern, the elliptical fit described in Non-Patent Document 3 is performed in each measurement cursor, and the long diameter, the short diameter, and the area are calculated. In addition, when the measurement target is the distance between the pattern of the end and the end of the line, the calculation is performed by the threshold method described in Patent Document 3.

In Figure 32, the description of the processing flow is illustrated in Figure 30. The GUI used when setting the recipe. 3201 is an example of a window of a GUI used in steps S3003 and S3006. 3202 indicates the image acquired in step S3001. The user can set the template image to be used in the template matching by setting the rectangular field 3203 to the image 3202. The sliders 3204 to 3208 are parameters that can be adjusted in step S3006.

The parameter is: a frame 3205 for the integrated value threshold 3204 or the pattern detection image 2304, a maximum value 3206 of the frequency at which the peak is detected from the Fourier spectrum image 1201, and a plurality of peaks are detected. The threshold value 3207, the result of the integration template matching, the weighting 3208 when the result of the pattern detection is performed, and the like.

Let the value of the correlation coefficient image obtained by matching the template of a certain value on the pattern image be C(0≦C≦1), and the value of the pattern detection image is A(0≦A≦1), by sliding When the weighting factor determined by the lever 3208 is t (0≦t≦1), the value S of the result of the integration processing is multiplied by the weight expressed by the following (Formula 8) [Math 8] S = C ^t . A ^{1- t} ‧‧‧(8)

Or the weight average expressed by the following (Formula 9) [Math 9] S = tC + (1- t ) A‧ ‧ (9)

And can be calculated. Here, the user uses it as a measurement The choice of the hole of the object.

For example, when the pattern image 3202 has a defective pattern like the hole 3209, the hole 3209 is a field in which the correlation coefficient is low in the template matching and is not present as a hole, and it is determined that the hole exists in the pattern detection. When the measurement cursor is not set to the hole 3209 by the slider 3208, the weighting factor t is set to approximately 1, and when the cursor is to be set, t is set to approximately 0, and the user can select the object to be set by the cursor.

When the setting parameter is pressed to execute the button 3219 (click), the pattern detection processing is performed to indicate that the result image 3210 of the measurement cursor is drawn. Further, the grid 3211 indicating the arrangement of the patterns is drawn, and the detection result of the arrangement pattern can be confirmed. Referring to Fig. 30, in step S3008, the respective cursors 3212 are selected by the mouse operation and the cursor is moved, and the position of the cursor can be corrected. In the state where the cursor is selected, the delete button 3214 is pressed to delete the cursor, and the additional button 3213 is pressed to create one cursor at the center, and the cursor can be moved to an arbitrary location.

The image 3215 can display and confirm the Fourier spectrum image of 1201 described in FIG. 12 or the correlation coefficient image of 2201 illustrated in FIG. 22 and the arrangement pattern of 2301 illustrated in FIG. 23 by means of the switching label 3216. The result image of the detection, etc., and the intermediate image of the cursor detection process. In Fig. 32, the state indicating the Fourier spectrum image is displayed on the display field 3215. In the Fourier spectrum image 3215, the frame line represented by the maximum value of the frequency set by the slider 3206 is used. 3217 can confirm the field of detectable peaks.

Further, when the threshold value is changed by the slider 3207, the peak to be detected can be confirmed by the cursor 3218 surrounding the peak point of the threshold or more. When the parameter adjustment and the correction of the cursor are completed, the coordinate information of the cursor can be stored in the memory device 206 so that the save button 3220 on the GUI screen 3201 in step S3009 of the processing flow of FIG. 30 is pushed down.

In Fig. 33, the operational flow of the apparatus for the step of formula setting of Fig. 3 will be described. In order to obtain an image of the wafer 107 (corresponding to step S3001), in S3301, the control platform 106 moves the wafer 107 until the field of the pattern is acquired. In S3302, the electron beam harness 101 is irradiated on the wafer 107, and the detector 108 detects the electrons twice. In S3303, the image data obtained from the detector is transferred to the memory 203 and the control terminal 114, and the image is acquired on the GUI of the control terminal 114 (corresponding to step S3002).

In S3304, the setting data of the template image is received by the user's input through the GUI, and the setting data is transferred to the memory 203 (corresponding to step S3003). In S3305, the calculation unit 202 performs the processing shown in FIG. 6 from the setting data of the image data and the template image located in the memory 203, and detects the measurement cursor (corresponding to step S3004). In S3306, the detection result is transferred to the control terminal unit 114, and is represented by a GUI (corresponding to step S3005). In S3307, when the user has corrected the detection parameter, the process proceeds to S3310, and the correction data of the parameter is received on the GUI, and the process returns to S3305.

If there is no correction, the process proceeds to S3308 (corresponding to steps S3006 and S3007). In S3308, the correction data of the detection result of the measurement cursor is received through the GUI on the control terminal 114, and the correction data is transferred to the memory 203 (corresponding to step S3008). In S3309, the coordinate data of the measurement cursor to which the correction data is applied is stored in the memory device 206 (corresponding to step S3009), and the processing is terminated.

According to the first embodiment, the burden of the user operation for setting a plurality of measurement cursors can be reduced, and the operation rate of the measurement device can be improved by reducing the time set by the recipe.

[Embodiment 2]

In the second embodiment, the procedure and measurement of the recipe setting different from the first embodiment are performed. In the second embodiment, in the pattern images 301 and 501 having the hole patterns having the same size and shape, the pattern images different from the pattern image 301 are also arranged in such a manner that the pattern image 301 is set only by the pattern image 301. 501 is measured. In this case, the coordinates of the measurement cursor set at the time of recipe setting are not used, and the measurement cursor is set and measured every time the pattern is photographed during measurement. Here, in the recipe setting, the adjusted parameters of 3204 to 3208 on the BUI of FIG. 32 and the template image are used for measurement.

In Fig. 34, the processing steps regarding the recipe setting are explained. First, in the same manner as in the first embodiment, the process shown in FIG. 30 is continued from step S3001 to step S3007, and the process proceeds to step S3401. In step S3401, the model image represented by the field 303 set in step S3003 is used. It is saved to the memory medium 114 as a recipe material. Further, the parameters of the template matching, the arrangement pattern detection, and the integration processing set in step S3007 are also saved as the recipe data to the memory medium 114, and the recipe setting is ended.

In Fig. 35, the flow at the time of measurement in the second embodiment will be described. At S3501, the recipe data stored in step S3401 is read from the memory device 206. Next, at S3502, the wafer 107 is moved by the control of the stage 106 until the field of photography of the image to be measured. In S3503, the measurement field is photographed to obtain a pattern image. Next, in S3504, the template image and the parameters of the recipe data are used for the acquired pattern image, and the measurement cursor detection processing shown in FIG. 6 is performed, and the measurement cursor is set. Next, in S3505, measurement processing is performed for each cursor area. If there is a field in which the measurement has not been completed, the process returns to S3502, and if the measurement of all the measurement target areas is completed, the process proceeds to S3507. The statistical value of the measurement result is calculated in S3507, and the result data is output to the memory device 206 to end the measurement.

According to the second embodiment, it is possible to set the recipe in the measurement area only for one area, and it is possible to set the measurement pattern having the same mask data but different from the same single measurement pattern without formula setting. Take measurements. In addition, in the wafers of the mask materials different in the measurement field of the recipe setting, the same material and process are used, and the measurement patterns are the same, and the wafer measurement area can be set without formulating the wafer. Measurement.

Moreover, the reduction in the time set by these recipes reduces the burden on the user and increases the operating rate of the measuring device.

[Example 3]

In the third embodiment, similarly to the second embodiment, the pattern images 301 and 501 having the same hole pattern and the same hole pattern are designed. For example, the pattern image 301 is used in the same processing flow as in the second embodiment. The recipe data created; the pattern image 501 is set in the recipe as described in the first embodiment. Through this, the processing steps of FIG. 30 can be simplified.

The flow of the processing of this embodiment is shown in Fig. 36. First, the measurement area (for example, the pattern image 301) of one place is subjected to the same recipe setting as that described in the second embodiment using the processing flow described in FIG. 34, and the template image and the parameter data are saved. Next, for the fields of the different patterns having the same single measurement pattern (for example, corresponding to the field of the pattern image 501), the recipe setting is performed in the processing flow as shown in FIG.

That is, in step S3601 (corresponding to step S3001 of FIG. 30), a wafer image (for example, pattern image 501) is acquired. Next, in step S3602, the reading of the recipe data (the template image and the parameter data) which was created in the same processing flow as in the case of the second embodiment and which was previously saved in the measurement area (for example, the pattern image 301) of the previous one is performed. . Next, the detection of the measurement cursor is performed in step S3603 (corresponding to step S3004), and in step S3604, the detection result of the measurement cursor by S3603, and the result of the template matching 604 and the arrangement pattern detection 603 of the intermediate processing are output to On the GUI (corresponding to step S3005).

In step S3605, the user can change the parameters of the arrangement pattern detection 603 and the template match 604 and the integration process 607 by using the GUI (corresponding to step S3006). In step S3606, the measured parameter is used again, and the detection of the measurement cursor is performed again, and the display of the cursor detection result and the intermediate processing result is updated (corresponding to step S3007). If the user judges that it is necessary to correct the parameter, the process returns to step S3605, and the parameter correction is performed again.

If there is almost no problem to detect the measurement cursor at the place to be measured, the process proceeds to step S3607. In step S3607, the user can perform the addition, deletion, and correction of the measurement cursor to the detection result of the measurement cursor (corresponding to step S3008). In step S3608, the photographing position and the relative coordinates of the measurement cursor opposite thereto are saved as recipe data (corresponding to step S3009). Using the recipe data, the image 301 on a different wafer or on a different wafer is measured in the same pattern area. When the saving of the recipe data is ended in step S3608, the series of processing ends.

According to the present embodiment, in the sample formed of the same shape but arranged in a different pattern, the pattern having the same shape as the first field after the first field is processed but formed in a different arrangement from the first field is formed. When the second field is measured, the steps S3002 and S3003 are omitted, and the recipe data of the long cursor coordinate data can be created so as to perform the operations in the step S300 and subsequent steps. Using the recipe data, as in the case of the first embodiment, measurement as shown in Fig. 31 can be performed. Through this, the steps of the domain of the step S3003 of the first embodiment can be omitted, and the steps can be utilized. For the adjustment of the parameters, the parameter adjustment of step S3605 is also solved with a minimum correction.

According to the present embodiment, by using the recipe data of the second embodiment, the time set by the recipe of the first embodiment can be reduced, the burden on the user can be reduced, and the operation rate of the measuring device can be improved.

601, 602‧‧ images

603‧‧‧Detection pattern

604‧‧‧Template match

605, 606‧‧‧ Results

607‧‧‧Integrated processing 609

610, 611‧‧‧ fields

Claims (12)

A method for measuring a pattern size, which is a method for measuring a size of a pattern; wherein: a plurality of patterns having the same shape formed on a sample are photographed to obtain an image of the plurality of patterns; The image of the pattern is integrated with the information of the pattern extracted by the template matching method and the information of the pattern extracted by the periodic information of the plurality of patterns using the image of the plurality of patterns, and the measurement result is set. The cursor sets a size measurement area using the set measurement cursor on the image of the plurality of patterns obtained as described above, and processes the image of the pattern of the size measurement area set by the measurement cursor in the image of the plurality of patterns, and measures the image. The size of the pattern. The method for measuring a pattern size according to claim 1, wherein the image of the plurality of patterns is an image obtained by photographing by SEM. The method for measuring a pattern size of claim 1, wherein the information of the pattern extracted by the template matching method for the image of the plurality of patterns is related to the template image extracted from the images of the plurality of patterns and the plurality of images The normalization of the pattern of each image of the pattern is used to obtain information on the correlation coefficient obtained. The method for measuring a pattern size of claim 1, wherein the plurality of patterns of the plurality of patterns are arranged in a circumference of the plurality of patterns The information of the period is information obtained by using the Fourier spectrum image obtained by performing discrete Fourier transform on the images of the plurality of patterns. The method of measuring the pattern size of claim 1, wherein the periodic information of the arrangement of the plurality of patterns of the image of the plurality of patterns is obtained by extracting an autocorrelation function from the images of the plurality of patterns, and Information obtained from the autocorrelation function. The method for measuring a pattern size of claim 1, wherein the integration result is obtained by associating a template image extracted from the image of the plurality of patterns with a normalization pattern of each of the images of the plurality of patterns The obtained correlation coefficient and the pattern of the pattern corresponding to the Fourier spectrum image obtained by discrete Fourier transform from the image of the plurality of patterns have a pattern corresponding to each position. The weighting factor of the probability, the information of the result of the multiplication. A pattern size measuring device which is a device for measuring a size of a pattern, and is characterized in that: the image obtaining means is configured to image a plurality of patterns having the same shape formed on a sample to obtain an image of the plurality of patterns The measurement cursor setting means extracts the information of the pattern by the template matching method for the image of the plurality of patterns acquired by the image capturing means, and uses the periodic information of the plurality of patterns of the plurality of patterns of the image. Extracting the information of the pattern, the information of the pattern extracted by the template matching method, and the period of the arrangement using the plurality of patterns The information of the pattern extracted by the sexual information is integrated to obtain the integration result, and the measurement cursor is set using the integration result obtained as described above; the size measurement field setting means uses the aforementioned measurement using the image of the plurality of patterns obtained by the image acquisition means. The measurement cursor set by the cursor setting means sets the size measurement area; and the size measurement means stores the measurement cursor set by the measurement cursor setting means in the image of the plurality of patterns acquired by the image acquisition means. The image of the pattern of the set size measurement field is processed, and the size of the pattern is measured. The image size measuring device according to claim 7, wherein the image capturing means includes an SEM, and the image of the plurality of patterns obtained by the image capturing means is an image obtained by the SEM imaging. The pattern size measuring device according to claim 7, wherein the information of the pattern extracted by the template matching method for the image of the plurality of patterns by the cursor setting means is a template for extracting images from the plurality of patterns The image is correlated with the normalization of the pattern of each of the plurality of patterns to obtain information on the correlation coefficient obtained. The pattern size measuring device according to claim 7, wherein the information of the pattern extracted by using the periodic information of the plurality of patterns of the image of the plurality of patterns by the cursor setting means is used for the plural The image of the pattern is obtained by the Fourier transform image obtained by the discrete Fourier transform. The pattern size measuring device according to claim 7, wherein the information of the pattern extracted by using the periodic information of the plurality of patterns of the image of the plurality of patterns by the cursor setting means is the plurality of The image of the pattern is obtained from the autocorrelation function and the information obtained from the obtained autocorrelation function. The pattern size measuring device according to claim 7, wherein the integration result integrated by the measurement cursor setting means is: each of the image of the template image extracted from the image of the plurality of patterns and the image of the plurality of patterns Corresponding to the normalization of the pattern, the correlation coefficient obtained and the period information of the pattern calculated based on the information of the Fourier spectrum image obtained by discrete Fourier transform from the image of the plurality of patterns are used. The pattern of each position has a weighting factor of the probability of existence, and the information of the result of the multiplication.