JP2012021832A - Image data analysis device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image data analysis device with which image processing conditions (threshold value or the like) enabling desired pattern edges with good reproductivity can be easily obtained by using image data obtained by using a scanning type microscope (CD-SEM) even if the patterns change structures in a height direction.SOLUTION: In an image data analysis device having a recording device 504, a processing part 503 and a display part 502, the processing part 503 displays a graph showing a relationship between noise and an edge roughness after removal of the noise, and a threshold value by using image data recorded in the recording device 504, and a threshold value for determining an edge obtained from the graph on the display part 502.

Description

  The present invention relates to an image data analysis apparatus that analyzes data acquired using a scanning microscope.

  In industries such as semiconductors and recording media, a scanning electron microscope having a length measuring function, that is, a CD-SEM (Critical Dimension Scanning Electron Microscope) is mainly used for pattern shape measurement. In the CD-SEM, a pattern on a substrate can be observed from the upper surface, and a two-dimensional image can be obtained (for example, Patent Documents 1 and 2 and Non-Patent Documents 1 and 2).

  With the miniaturization of pattern processing dimensions, the need to accurately grasp the pattern shape from the CD-SEM image is increasing. For pattern shape evaluation, measurement of typical dimensions of the pattern, such as line width for line patterns, radius or diameter of circles for perfect circle patterns, major axis and minor axis for ellipse patterns, etc. has been the mainstream, but in recent years The required accuracy (reproducibility) for these measurements has become very high. Furthermore, pattern contours, that is, a set of local edge points themselves are being directly analyzed, but high reproducibility is also required when these are extracted from a CD-SEM image. This is because if the reproducibility is low in these pattern inspections, the reliability of the inspection is low, and as a result, the yield of a semiconductor (or recording medium) production line deteriorates.

  On the other hand, the pattern edge position and the pattern size obtained from the set depend largely on how the edge is defined. For example, there is an edge definition method called a threshold method in which a point at which the CD-SEM signal intensity is equal to a certain level is defined as a pattern edge, but the measured value varies greatly depending on how the level (threshold) is set. Is known empirically. In the past, the threshold was often set to 50%. This means that the value at the center of the minimum value and the maximum value of the signal intensity in a certain area is set as the threshold intensity, and the point at which the SEM signal intensity becomes equal to this value is defined as the edge of the pattern. In addition, there is a need for an edge corresponding to the upper part of the pattern, and the threshold value may be set to 100%. These values are used as a result of displaying the position of the edge extracted by using these threshold values superimposed on the CD-SEM image, and judging that the operator is appropriate visually. Value.

JP 2006-2105020 A JP 2009-257937 A

J.S. Villarrubia, et al., “A simulation study of repeatability and bias in the CD-SEM”, Proceedings of SPIE vol. 5038, pp 138-149 (2003). B.D.Bunday, et al., “Determination of optimal parameters for CD-SEM measurement of line edge roughness”, Proceedings of SPIE vol.5375, pp 515-533 (2004).

  In recent years, miniaturization has progressed and the pattern side walls have become non-straight. These include a step on the side wall, a complicated slope, and a pattern with a rounded top. Examples of cross-sectional shapes of these line patterns are shown in FIGS. If a pattern edge is to be extracted from a CD-SEM image having such a pattern, the reproducibility will be very poor if the conventional methods such as the threshold values of 50% and 100% are carried out as they are.

  In order to determine an edge definition with good reproducibility, it is necessary to repeat the experiment of checking the measurement reproducibility by changing the threshold value, which takes a lot of labor. In particular, in the pattern as shown in FIG. 11, since the reproducibility changes greatly only by slightly changing the threshold value, it is necessary to perform an experiment by changing the threshold value in fine steps. Further, when the same spot is observed many times by CD-SEM in order to measure reproducibility, the pattern shape may change. When different spots are observed, variations in the original pattern shape affect the results. Due to these two reasons, the correct result may not be obtained. Nevertheless, such an experiment must be performed every time the sidewall shape of the pattern changes.

  Therefore, it is considered that a method for easily obtaining appropriate image processing conditions (threshold value, etc.) is required even for a pattern whose structure changes in the height direction in the future. Note that “appropriate” as used herein includes not only high reproducibility in measuring a pattern whose structure changes in the height direction, but also that it recognizes where the measurement is in the height direction. It is out. For example, in the case of a pattern having a two-stage structure (as shown in FIG. 11A), the lower edge may be detected or the upper edge may be detected. This is because, for example, depending on the layer or pattern, the portion of the pattern that affects the performance of the device is the top or the bottom.

  In Patent Document 2, line edge roughness and true line edge roughness are measured by changing representative parameters of image processing conditions (hereinafter referred to as image processing parameters), and the height specified by the operator is measured from the graph. A method for obtaining the value of an image processing parameter for producing an edge is disclosed. However, when this method is simply applied to a pattern whose structure changes in the height direction, it is expected that correct results will not be obtained if the noise is large. There is a risk that correct results may not be obtained even with level images. In addition, when true line edge roughness is used, height can be specified by collating with height data obtained separately by AFM (Atomic Force Microscope), but reproducibility of edge position and length measurement is obtained. There is no fear. Furthermore, there is a risk that the detection of the level difference is insufficient in sensitivity (that is, it cannot be sufficiently recognized that there is a change in the structure on the side wall).

  The influence of noise is disclosed in Patent Document 1 and Non-Patent Documents 1 and 2. In Non-Patent Document 1, when calculating the edge position of a pattern pattern from a CD-SEM, if the signal intensity is not averaged on the image, the position of the edge point is the true position due to the influence of image noise. It is described that there is a probability variation of 1 nm or more around. Patent Document 1 and Non-Patent Document 2 disclose a method for calculating a true value of line edge roughness excluding the influence of this variation. However, in these documents, the true CD value excluding the influence of noise and the true edge position itself cannot be determined.

  Furthermore, in order to shorten development time and improve throughput, additional experiments such as AFM and optical measurement, which are always required when acquiring 3D shape information, and collating these data with CD-SEM data, etc. It is not appropriate to introduce time-consuming analysis. That is, as described above, even if the designation of the approximate position (height) in the height direction and high reproducibility are realized, further simplicity is required.

  In other words, when the structure changes in the height direction, it is difficult to determine an image processing condition (threshold value or the like) for determining a pattern edge by a simple method when the structure changes in the height direction.

  An object of the present invention is to determine a desired pattern edge with good reproducibility using image data obtained using a scanning microscope (CD-SEM) even in a pattern whose structure changes in the height direction. An object of the present invention is to provide an image data analysis apparatus capable of easily obtaining image processing conditions (threshold value, etc.).

  In order to achieve the above object, first, the dependency of the pattern edge position or pattern dimension on the image processing parameters (part of the image processing conditions) is grasped, and the dependency of these measured values is reduced. It is effective to select a value of a proper image processing parameter.

  FIG. 12 shows the result of detecting the position of one edge of the line pattern having the cross-sectional shape shown in FIG. 11A by changing the threshold value of the threshold method. The magnitude of the threshold has a strong correlation with the height of the extracted edge point, and the closer to the left in the figure, the closer to the bottom the edge is detected. Since it can be seen that there is a step at a threshold value of about 60%, it can be seen that the side wall is divided into a bottom region and a top region, and that the step portion corresponds to a threshold value of about 60%. In this case, if it is desired to measure the line width and line edge roughness of the bottom region in the future, the threshold value may be made smaller than 60%. Conversely, if it is desired to measure the top region, the threshold value may be made larger than 60%.

  FIG. 13 shows an example in which the line width of the line pattern having the cross-sectional shape shown in FIG. It can be seen that when the threshold value is 80% or more, the line width changes rapidly and becomes unstable. From this, it can be seen that the state of the side wall is divided into a bottom region and a top region, and the boundary corresponds to the threshold value of about 80%, and the line width of the top region cannot be measured stably. In this case, it will be understood that the measurement of the top region should not be performed in the future, or should be limited to a reference level, and the line width and line edge roughness should be measured using a threshold value of less than 80%. It can also be seen that the measurement value obtained in this way is data in the bottom area of the pattern.

  Second, it is necessary to grasp the influence of image noise, which is the cause of reproducibility degradation in the image itself, and to select an image processing condition so as to minimize it. There are various causes of deterioration of reproducibility, but in recent observations of complicated fine patterns, this influence is the greatest, and this is improved by only optimization of image processing parameters. In this case, as in the first case, the dependency of the amount of image noise impact on the image processing parameter is grasped, and a value that minimizes the image noise impact may be selected. At this time, the dependence of the value of the pattern edge roughness on the image processing parameter is also plotted, and it is more preferable to select the image processing parameter value so that the influence of the image noise is minimized in an area where the dependence is small.

  The merit of comparing the second method with the first method is that the sensitivity to the step is high. Since the position of the edge fluctuates greatly with slight noise in the stepped portion, the noise impact increases even if the noise of the image itself is small. For this reason, a level | step difference is emphasized. When the second method is performed on the line pattern whose cross-sectional shape is shown in FIG. 11A, the graph shown in FIG. 14 is obtained. Note that the maximum value of the noise impact is 1.

  When both the first method and the second method are performed, the calculation time is somewhat longer, but the reliability of the inspection is further improved.

  In the conventional method, when determining the image processing conditions for obtaining the edge position of the pattern from the CD-SEM image, the approximate position in the height direction such as the upper part and the lower part of the pattern is specified, and further the image noise is determined. In order to obtain an image processing condition with a small variation in the cause, a complicated and time-consuming analysis is required. In addition, observation with an apparatus other than CD-SEM was also necessary. According to the present invention, it is possible to obtain an image processing condition with a small variation due to noise after designating a position in the approximate height direction of a pattern from only one CD-SEM observation. As a result, accurate pattern shape inspection and dimension measurement can be easily performed, so that inspection accuracy can be improved and inspection costs can be reduced.

It is a processing flow figure when performing image data analysis using the image data analysis apparatus which concerns on a 1st Example. It is a schematic bird's-eye view of the observation sample used in the 1st example. It is a plane schematic diagram of the CD-SEM observation image of the observation sample shown in FIG. (A) is a signal profile of the CD-SEM observation image of the observation sample shown in FIG. 2, and (b) is a conceptual diagram for explaining a method of defining an edge point on the signal profile. It is a conceptual diagram showing the structure of the image data analyzer which concerns on a 1st Example. It is a graph showing the relationship between the output value (roughness value and its noise) of the roughness measurement obtained in 1st Example, and a threshold value. It is a schematic diagram of a GUI screen when calculating an optimum threshold value in the first embodiment. It is a plane schematic diagram of the CD-SEM observation image of the complicated pattern formed in the observation sample used in the 2nd example. It is a plane schematic diagram of the CD-SEM observation image of the line pattern formed in the other observation sample used in the 2nd example. It is a graph showing the relationship between the output value (roughness value and its noise) of the roughness measurement obtained in 2nd Example, and a threshold value. It is an example of a cross-sectional view of a pattern having a structure on a side wall, (a) shows a case where the width is different between the upper part and the lower part, and (b) shows a case where the upper surface is rough. It is a graph which shows the relationship between the pattern edge position extracted from the CD-SEM observation image of the pattern which has a structure in a side wall, and the value of the threshold parameter used when defining an edge. It is a graph which shows the relationship between the dimension of the pattern extracted from the CD-SEM observation image of the pattern which has a structure in a side wall, and the value of the threshold parameter used when defining an edge. It is a graph which shows the relationship between the noise impact and edge roughness at the time of extracting a pattern edge from the CD-SEM observation image of the pattern which has a structure in a side wall, and the value of the threshold value parameter used when defining an edge.

  Hereinafter, an example explains.

  Hereinafter, a first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a flow diagram sequentially showing processing performed by an operator and apparatus for image data analysis, such as optimization of image processing conditions using the image data analysis apparatus according to the present embodiment, and FIG. FIG. 3 is a schematic plan view of an image obtained by observing the line pattern of the observation sample shown in FIG. 2 with a CD-SEM. This is a target pattern analyzed in this example. 4 is a schematic diagram for explaining a part of the secondary electron signal and edge point extraction method of the CD-SEM constituting the image data shown in FIG. 3, and FIG. 5 is an image data analysis used in this embodiment. FIG. 6 is a conceptual diagram showing the configuration of the apparatus, FIG. 6 is a graph showing the relationship between the roughness measurement output value (roughness value and its noise) and the threshold value obtained in this embodiment, and FIG. 7 appears on the display device when this embodiment is implemented. It is a schematic diagram of an input screen.

  First, the operator performed step 101 and specified a CD-SEM image file to be analyzed using the interface 505 for image data analysis apparatus analysis. Then, the file data is loaded from the recording device 504 to the processing unit 503 (step 102), temporarily recorded in the storage area 501, and the image is displayed on the display unit 502. This image is shown in FIG. This is a CD-SEM observation of the line pattern shown in FIG. 2 from vertically above. The pattern was a resist pattern on a silicon substrate. The gray level of each pixel of the image represents the intensity of the secondary electron signal emitted from the pattern and detected by the CD-SEM detector in gray scale. Reference numeral 301 is an area corresponding to the underlying layer of the line pattern, reference numeral 302 is an area having a high signal intensity corresponding to the vicinity of the left edge of the line pattern, reference numeral 303 is an area having a low signal intensity corresponding to a flat portion above the line pattern, Reference numeral 304 denotes an area having a high signal intensity corresponding to the vicinity of the right edge of the line pattern, reference numeral 305 denotes a frame indicating the inspection area, and reference numeral 306 denotes a straight line that divides the image vertically.

  Next, proceeding to step 103, the operator uses the interface 505 to input a region in which an edge is to be extracted from the image, that is, an inspection region. This is a frame 305 in FIG. The edge of the line pattern is defined within the inspection area surrounded by the frame. This region was 400 pixels in the y direction and 50 pixels in the x direction, and when converted into a length, it was 2000 nm in the y direction and 66 nm in the x direction.

  Next, proceeding to step 104, the operator inputs an image processing condition for extracting an edge. At this time, the pattern edge is defined by the threshold method. This threshold method will be described in detail below. FIG. 4A shows the signal intensity of the secondary electrons along the line 306 in FIG. 3, that is, the x dependence of the secondary electron signal intensity. Such x-dependence of the secondary electron signal intensity is called a secondary electron signal profile. There are two peaks, the left peak is the signal from the pattern left edge, and the right peak is the signal from the pattern right edge. FIG. 4B shows an enlarged view of the signal intensity near the peak corresponding to the left edge. On this graph, the region of x value that falls within the frame 305 is the region from a to b. In the threshold method, first, the maximum value and the minimum value of the signal intensity I (x) are searched in a region where a <x <b. The obtained values are I_max and I_min, and the values of x at that time are x_max and x_min. In the threshold method, for the threshold T expressed as a percentage

Let x (T) be the edge position. However, the value of x (T) must be between x_min and x_max.

  Since the inspection area is now 400 pixels along line 302, there are also 400 secondary electron signal profiles included. The edge positions obtained for T on these profiles are denoted by x_1 (T), x_2 (T),. . . . Let x_400 (T).

  Next, in step 105, an index to be calculated from the extracted edge point position is designated. In this example, the noise (3sigma_error) that increases the output value in normal roughness measurement and the roughness value (3sigma_0) from which the influence of noise has been removed are selected.

  Here, these two quantities will be described. These are functions of T. The standard deviation of the distribution of the edge positions (400) is set as sigma (T). This value is so-called line edge roughness, and this value includes random image noise as described above. That is,

  The method of removing noise and obtaining 3sigma_0 (T) is well known, and is omitted here.

  Next, proceeding to step 106, the operator performed the above-described calculation by the analyzer and displayed on the monitor. As a result, the line edge roughness, 3 sigma_0 (T) and the bias amount 3 sigma_error (T) due to image noise are obtained with respect to the value of T, and the graph shown in FIG. 6 is displayed on the monitor. It was done. The state of the display unit 502 at this time is shown in FIG. An image file name is displayed in the band-shaped display area 701, and a graph is displayed in the display area 702.

  Next, from this graph, the operator sees that the roughness is different around T = 50%, and determines that T is divided into a lower part of 50 or less and an upper part of 50 or more. First, T is 5 From 50 to 50, that is, at the bottom, it is determined to calculate T that is least affected by noise, and a condition for searching for the optimum T is input (step 107). In this step 107, a three-dimensional measurement position is determined. In this step, first, as shown in FIG. 7, the cursor is moved to the display area 703, and the area where the optimum T is to be searched for is input from 5 to 50 (input by the operator is possible). In addition, a check mark is placed in front of these items in the display area 704 to determine T based on the graphs of 3sigma_0 and 3sigma_error (input by the operator is possible). In addition, in order to obtain a place where 3 sigma — 0 is stable and a place where 3 sigma — error is minimized, the display portion is pulled down to display these items. As a pull-down item of the display area 704, “maximum” or the like can be registered in addition to “stable” and “minimum”. Thereafter, the process proceeds to step 108, and the execution button of the display area 705 displayed as “Calculate” is clicked, and the threshold value is optimized by the analysis device.

  The processing unit 503 determines a region where 3 sigma — 0 is stable. As a result, it was determined that 3sigma_0 was stable in the region of T = 20% to 40%. Next, the processing unit 503 finds a value of T that minimizes 3 sigma_error in each of the above-described regions. In this example, T = 30%. Therefore, as shown in FIG. 7, a line is displayed at the position of T = 30, and a value of 30 is displayed on the graph of the display unit 502. Basically, this is the end of the operation.

  If it is desired to determine T with reference to the line width, a check mark is put in the check box on the left side of the display of CD in the display area 704 in FIG. However, in that case, it is necessary to select to plot the CD at the same time in Step 105. By selecting CD, the line width of the line 302 is output in this embodiment. In addition, when determining based on stability, stability is displayed, but when selecting the minimum value, minimum is displayed.

  Here, Step 107 and Step 108 were repeated in order to perform the same thing on the upper part of the pattern. First, in order to obtain the optimum T at the top of the pattern, 50 and 100 were input in the area input part (repeating step 107), and the execution button displayed as Calculate in the display area 705 was clicked (repeating step 108). The processing unit 503 determines that 3sigma_0 is stable from T = 75 to 95, further determines that 3sigma_error is minimum at T = 85, and displays a value of 85 on the graph.

  As described above, when 30% is used as a threshold when measuring the lower part of the pattern of the observation sample (wafer) and 85% is used when measuring the upper part, edge detection with a small influence of noise can be performed. I understood it.

  Therefore, an inspection in the mass production process was performed using the above results. In addition to the conventional line width measurement, the size of the line edge roughness obtained with a threshold of 30% (hereinafter referred to as 3 sigma_bottom) is calculated for a wafer whose pattern is formed by the same lithography process as the observation sample (wafer). did. The value of this index represents the roughness of the pattern bottom. In the wafer inspection, these indexes were measured at one location per chip, and chips satisfying 3 sigma_bottom <3 nm with a line width of the target value ± 2 nm were determined as non-defective chips. When the number of chips described above exceeded 70% of the number of chips inspected per wafer, the wafer was sent to the next process, and when it was 70% or less, lithography was repeated.

  Previously, pattern inspection was performed using simple line width measurement and line edge roughness measured without optimizing the threshold, but when the inspection was changed to the above, the inspection time was the first optimum. Only by increasing the number of steps for obtaining T, the yield rate of the final product was improved from 88% to 91%.

  Further, in addition to the measurement of the line width and 3 sigma_bottom described above, the size of the line edge roughness (hereinafter, 3 sigma_top) was also measured at a threshold of 85%. At this time, 3 sigma_top <5 nm was added to the condition of the good chip. Then, the inspection time was increased by 1 minute and 30 seconds per wafer, and the yield rate was 92%.

  Furthermore, the difference (hereinafter referred to as ΔCD) between the line width values measured between the threshold values of 30% and 85% was calculated. This index is considered to represent the inclination of the pattern sidewall. In the wafer inspection, in addition to the above conditions, a chip satisfying ΔCD <3 nm was regarded as a good product. As a result, the inspection time per wafer further increased by 1 minute, but the yield rate improved to 94%.

  In the above example, both 3 sigma_0 and 3 sigma_error are plotted against the threshold T. However, in order to shorten the time, it is also possible to optimize the threshold only from 3 sigma_error. However, in this case, it is impossible to accurately grasp how the roughness of the pattern side wall changes in the height direction, and an error (edge detection failure) may occur when detecting an edge with the obtained threshold value.

  In order to detect a change in the roughness height direction of the pattern side wall, a roughness index such as the skewness γ of the edge position distribution obtained from the roughness data and the autocorrelation length ξ can be used instead of 3 sigma_0. The skewness is an index representing the asymmetry of the distribution, and is applied to edge roughness measurement to detect the roughness that the edge bites into the line or the roughness that protrudes outside the line. The autocorrelation length is a distance until the local fluctuation is attenuated, and the characteristic size of the roughness measured along the edge can be evaluated by applying to the edge roughness measurement. In any case, this is not a degree of roughness but an amount representing a characteristic in a direction along the edge. In order to use these indexes for analysis, a check box on the left side of a display such as Skewness or ACL (abbreviation of Auto-correlation length) may be clicked on the screen shown in FIG. Γ is calculated as follows.

  Here, N is the number of edge points, x_i (T) is the i-th pattern edge position (x coordinate) obtained using the threshold T, and μ (T) is the arithmetic average thereof.

  The autocorrelation length ξ is obtained by solving the following equation.

  Here, x_i (T) is periodic, and the value of x_ (i + N) (T) is the same as x_i (T). Δy is the interval between the edge points in the y direction.

  In general, the threshold T can be optimized with high accuracy by plotting 3 sigma — 0 (3 sigma if image noise is suppressed), but other indices may be desirable. For example, when the formation process of line edge roughness differs greatly between the upper part and the lower part of the pattern, the T dependence of γ and ξ is clear. This corresponds to a case where only the top of the pattern is roughened by dry etching. In particular, γ should be used when roughness that bites into the line appears. Further, when processing polycrystalline silicon, a polymer that easily aggregates, etc., the particle size changes due to process variations, and as a result, the autocorrelation length ξ changes. Therefore, ξ is preferably used.

  Here, the parameter of the image processing condition is the threshold value, but the parameter can be optimized by the above-described procedure even when the edge is defined by another known method. To do this, first determine the main parameters for image processing. Next, an edge is extracted by changing its value within a certain range, and a graph similar to that shown in FIG. 6 (the horizontal axis is not a threshold value but a parameter selected by the operator, the vertical axis is 3 sigma_0, 3 sigma_error Or the roughness index and line width described above). Thereafter, as described in the present embodiment, a parameter value in which 3 sigma_0 is stable and 3 sigma_error is small may be selected.

  As described above, by using the image data analysis apparatus according to the present embodiment, the inspection accuracy of the semiconductor pattern shape is increased without affecting the throughput of the mass production process, and the yield is improved.

  According to the present embodiment, a desired pattern edge can be determined with good reproducibility using image data obtained using a scanning microscope (CD-SEM) even in a pattern whose structure changes in the height direction. It was possible to provide an image data analysis apparatus that can easily obtain possible image processing conditions (threshold value, etc.). In addition, the production yield of the product could be improved.

  Hereinafter, a second embodiment of the present invention will be described with reference to FIGS. Note that the matters described in the first embodiment but not described in the present embodiment can be applied to the present embodiment as long as there is no particular circumstance. FIG. 8 shows an observation image of the pattern to be analyzed by CD-SEM, and FIG. 9 shows an observation image of the line pattern by CD-SEM. FIG. 10 is a graph showing the relationship between the output value of roughness measurement (roughness value and its noise) obtained by analysis and the threshold value. In the second embodiment, an example in which a contour line of a complicated two-dimensional pattern is obtained will be described.

  First, the operator observed a two-dimensional pattern to be analyzed with a CD-SEM, and obtained an image shown in FIG. Next, a sample of a line pattern created under the same conditions using the same material as the pattern to be observed was observed with a CD-SEM. The observation conditions were all the same as when the image shown in FIG. 8 was obtained. As a result, an observation image shown in FIG. 9 was obtained.

  Next, the operator performed the same analysis as in Example 1 using the image shown in FIG. As a result, the graph shown in FIG. 10 was obtained, and it was found that there was no structure on the sidewall and that the error due to noise could be minimized by setting the threshold value to 80%. Therefore, the edge of the pattern was extracted from the image of FIG. 8 with a threshold of 80%.

  Since the pattern having the same layout as that shown in FIG. 8 is one of the most important parts for the device being manufactured, whether or not this part is patterned as designed is an important inspection item. In this manufacturing process, a pattern contour having the same shape as the image of FIG. 8 was previously extracted with a threshold of 50%, and whether the shape was good or not was determined. The defective rate decreased from 3% to 1.5%.

  Image processing conditions that allow a desired pattern edge to be determined with good reproducibility using image data of a pattern obtained using a scanning microscope (CD-SEM) even if the pattern has a structure that changes in the height direction (CD-SEM) It was possible to provide an image data analysis apparatus that can easily obtain a threshold value and the like.

  According to the present embodiment, it is possible to easily obtain image processing conditions (threshold value, etc.) by which a desired pattern edge can be determined with good reproducibility using image data acquired using a scanning microscope (CD-SEM). An image data analysis device could be provided. In addition, a pattern contour having a complicated two-dimensional pattern can be easily obtained, and the defect rate of the product can be reduced.

Although the present invention has been described in detail above, the main invention modes are listed below.
(1) A recording unit that records image data of a pattern formed on a substrate obtained by a scanning electron microscope, a processing unit that processes the image data recorded in the recording unit, and the processing unit An image data analysis apparatus having a display unit for displaying the processed results,
The processor is
From the image data, a roughness parameter value representing the roughness characteristics of the edge of the pattern using a plurality of image processing conditions, and a noise parameter representing the degree of noise of the image or the influence of noise on the position of the edge Find the value and
An image data analyzing apparatus, wherein a graph having one value of the roughness parameter and the noise parameter and the other value representing the image processing condition is displayed on a pre-description display unit.
(2) The image data analysis device described in (1) above,
The processor is
An image characterized in that the noise parameter value is expressed as a function of an image processing parameter in an image processing condition, and an image processing parameter is obtained when the noise parameter value takes a minimum value with respect to the image processing parameter. Data analysis device.
(3) The image data analysis device described in (1) above,
The value of the roughness parameter is expressed as a function of the image processing parameter in the image processing condition, and the value of the image processing parameter in which the dependency of the value of the roughness parameter on the image processing parameter is smaller than a predetermined value. An image data analysis apparatus characterized by being obtained.
(4) The image data analysis device described in (1) above,
As the roughness parameter,
A value representing the standard deviation of the distribution of the x coordinate of the edge point of the pattern or the skewness γ of the distribution of the x coordinate of the edge point of the pattern when the x coordinate is determined in a direction perpendicular to the edge of the pattern Or
If the pattern is a line pattern, the y coordinate is set perpendicular to the x coordinate,
If the pattern is a closed pattern, measure the distance from one point on the pattern edge in the direction along the edge and define it as the y coordinate;
A sequence of Δx (y) expressing a deviation from an average value or a design value of the x coordinate of the edge point of the line pattern as a function of y is calculated, and an autocorrelation length ξ obtained from the sequence is used. Image data analysis device.
(5) The image data analysis device according to (2) or (3) above,
The processor is
The image data obtained by the scanning electron microscope, the plurality of image processing condition processes used for image processing, the value of the roughness parameter in the plurality of image processing conditions, and the value of the obtained image processing parameter, An image data analyzing apparatus comprising a function of recording at least two of the data as a set of data in the recording unit.

101 ... Specifying the image file to be analyzed,
102... The process in which the analysis device loads the designated file from the recording area to the processing unit;
103 ... setting the inspection area on the image,
104: a step of setting image processing conditions at the time of edge extraction;
105 ... a step of designating an index calculated from the edge position data;
106 ... calculating and displaying the threshold dependence of the index,
107 ... a step of setting analysis conditions for threshold optimization,
108... Performing threshold optimization,
301... Region corresponding to the underlying layer of the line pattern,
302 ... a region having a high signal intensity corresponding to the vicinity of the left edge of the line pattern,
303 ... a region having a low signal intensity corresponding to a flat portion on the line pattern,
304... A region having a large signal intensity corresponding to the vicinity of the right edge of the line pattern;
305 ... A frame indicating the inspection area,
306 ... a straight line that bisects the image up and down,
501 ... CD-SEM system or recording area for storing CD-SEM data,
502 ... display section,
503: Processing unit,
504 ... Recording device,
505 ... an interface for controlling the processing device,
701 ... Display area of the name of the image file being analyzed,
702 ... graph display area,
703 ... A display area of T range for searching for the optimum T (operator input is possible),
704 ... Display area for setting conditions for threshold optimization (operator input allowed).

Claims (6)

A recording unit that records image data of a pattern formed on a substrate obtained by a scanning electron microscope, a processing unit that processes the image data recorded in the recording unit, and a processing unit that processes the image data. An image data analysis apparatus having a display unit for displaying results,
The processor is
From the image data, a roughness parameter value representing the roughness characteristics of the edge of the pattern using a plurality of image processing conditions, and a noise parameter representing the degree of noise of the image or the influence of noise on the position of the edge Value and
An image data analyzing apparatus, wherein a graph having one value of the roughness parameter and the noise parameter and the other value representing the image processing condition is displayed on a pre-description display unit. The image data analysis device according to claim 1,
The processor is
An image characterized in that the noise parameter value is expressed as a function of an image processing parameter in an image processing condition, and an image processing parameter is obtained when the noise parameter value takes a minimum value with respect to the image processing parameter. Data analysis device. The image data analysis device according to claim 1,
The value of the roughness parameter is expressed as a function of the image processing parameter in the image processing condition, and the value of the image processing parameter in which the dependency of the value of the roughness parameter on the image processing parameter is smaller than a predetermined value. An image data analysis apparatus characterized by being obtained. The image data analysis device according to claim 1,
As the roughness parameter,
A value representing the standard deviation of the distribution of the x coordinate of the edge point of the pattern or the skewness γ of the distribution of the x coordinate of the edge point of the pattern when the x coordinate is determined in a direction perpendicular to the edge of the pattern Or
If the pattern is a line pattern, the y coordinate is set perpendicular to the x coordinate,
If the pattern is a closed pattern, measure the distance from one point on the pattern edge in the direction along the edge and define it as the y coordinate;
A sequence of Δx (y) expressing a deviation from an average value or a design value of the x coordinate of the edge point of the line pattern as a function of y is calculated, and an autocorrelation length ξ obtained from the sequence is used. Image data analysis device. The image data analysis device according to claim 2,
The processor is
The image data obtained by the scanning electron microscope, the plurality of image processing condition processes used for image processing, the value of the roughness parameter in the plurality of image processing conditions, and the value of the obtained image processing parameter, An image data analyzing apparatus comprising a function of recording at least two of the data as a set of data in the recording unit. The image data analysis device according to claim 3,
The processor is
The image data obtained by the scanning electron microscope, the plurality of image processing condition processes used for image processing, the value of the roughness parameter in the plurality of image processing conditions, and the value of the obtained image processing parameter, An image data analyzing apparatus comprising a function of recording at least two of the data as a set of data in the recording unit.

Images