Pattern measurement system, pattern measurement method, and recording medium

Abstract

The present application provides a kind of pattern measurement system, pattern measurement method and recording medium, without the restriction of measurement pattern, can detect random noise component with high precision, and can measure edge roughness with higher precision.According to the present disclosure, in the line pattern image obtained, the left edge and right edge of the line pattern are respectively subjected to pattern matching and edge position correction, and an image without roughness is generated.According to the image, the PSD value is measured, and the average PSD value of the full frequency is taken as the random noise component, so that the random noise component can be detected with high precision.Furthermore, by subtracting the PSD value, i.e.random noise component, from the PSD value of the original image, the edge roughness can be measured with high precision.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a pattern measurement system, a pattern measurement method, and a recording medium. BACKGROUND

[0002] In a fine processing process of a semiconductor manufacturing process, edge roughness of a pattern (unevenness of a pattern end portion) has a large influence on a yield of a device. It is known that the degree of generation of edge roughness greatly varies depending on a material constituting a semiconductor device, an exposure device, or properties, characteristics, and the like of a substrate board. In recent years, with miniaturization of a pattern, higher-precision edge roughness measurement is required. In edge roughness measurement, if there are many random noises in a measured image, an edge of a pattern cannot be accurately captured, and an edge roughness larger than a true edge roughness can be measured. Therefore, in order to perform high-precision edge roughness measurement, it is effective to detect and remove a random noise component of an image.

[0003] For example, in Patent Literature 1, a measurement method is disclosed in which edge roughness of each of left and right of a line pattern is found as a 3σ value indicating an amplitude of unevenness fluctuation from an ideal straight line. In addition, in Patent Literature 1, a method is described in which Fourier analysis is performed on a set of fluctuation data, an intensity with respect to a spatial frequency is found, and thus an edge shape is analyzed. Further, in Patent Literature 2, a method is described in which a random noise component is detected from a PSD (Power: rum Density: Power Spectrum Density) value of a line pattern.

[0004] As described above, in order to perform high-precision edge roughness measurement, it is effective to detect and remove a random noise component of an image.

[0005] However, in the method of Patent Literature 2, since a random noise amount is estimated only from a high-frequency component, the PSD value in the high-frequency portion is not sufficiently converged at times. In this case, an error of an approximate straight line of the PSD value is also large, and detection accuracy of a noise component is reduced. In particular, in a case where line patterns are few in a screen or lines are short, detection accuracy of a noise is reduced. Therefore, in order to make detection accuracy good, the number of patterns (a constraint of a measurement pattern) needs to be increased, but if the number of patterns is increased, processing becomes complicated, and efficiency is deteriorated.

[0006] PRIOR ART DOCUMENTS

[0007] Patent Literature 1: Japanese Patent No. 3870044

[0008] Patent Literature 2: Japanese Patent Application Laid-Open No. 2019-39884 SUMMARY

[0009] In view of the above, this disclosure proposes a technique that can detect random noise components with high precision without the limitation of measurement patterns and can perform edge roughness measurement with even higher precision.

[0010] To address the aforementioned issues, the present invention provides a pattern measurement system that uses signals obtained by scanning a charged particle beam or irradiation light onto a sample to measure patterns formed on the sample. The pattern measurement system comprises: a storage device storing a program for processing pattern information; and a computer that reads the program from the storage device and processes the pattern information. The computer performs the following processing: generating multiple independent line profiles, a summed line profile of the multiple line profiles, and an original image of the pattern based on the signals obtained from the pattern; dividing the multiple independent line profiles into two regions; performing pattern matching between the individual line profiles and the summed line profile, respectively, for the multiple individual line profiles and the summed line profile divided into two regions; performing position correction on the multiple individual line profiles based on the pattern matching results; generating a corrected image of the pattern based on the position-corrected multiple line profiles; calculating the power spectral density at the edge of the pattern in the corrected image, and averaging the power spectral density to obtain a random noise component; and calculating the edge roughness by obtaining the difference between the power spectral density at the edge of the pattern in the original image and the random noise component.

[0011] Further features relating to this disclosure will become clear from the description and accompanying drawings. Furthermore, this disclosure is achieved through elements and combinations of elements, as well as the subsequent detailed description and the scope of protection outlined in the appended claims.

[0012] It should be understood that the description in this specification is merely a typical example and is not intended to limit the scope of protection or application of the claims of this disclosure in any sense.

[0013] According to the technology disclosed herein, random noise components can be detected with high precision without the limitation of the measurement pattern, and edge roughness can be measured with even higher precision. Attached Figure Description

[0014] Figure 1 This is a summary structure example of the CD-SEM in this embodiment.

[0015] Figure 2 The system structure is illustrated in the example, which includes an image transformation unit 109 and a pattern measurement unit 202 for measuring the image pattern obtained therein.

[0016] Figure 3 This is a flowchart illustrating the image processing and actual edge roughness measurement processing in the pattern measurement unit 202.

[0017] Figure 4This represents the contour obtained from each line scan.

[0018] Figure 5 This represents an example of a SUM line profile calculated from the full range of line profiles (e.g., the average value obtained by summing the full range of line profiles and dividing by the number of lines).

[0019] Figure 6 An example representing the positional offset of the SUM line profile from each other line profile.

[0020] Figure 7 This represents an example of an image with the roughness component removed.

[0021] Figure 8 This diagram shows a comparison between the PSD value of the original image (PSD value for spatial frequency) and the PSD value of the processed image obtained in step 305 (PSD value for spatial frequency).

[0022] Figure 9 This is a flowchart used to illustrate the optimal frame rate calculation process.

[0023] Figure 10 This represents the cumulative image generated by accumulating from 2 frames to 32 frames (example).

[0024] Figure 11 This indicates the relationship between the number of frames and the amount of random noise.

[0025] Figure 12 This is a flowchart illustrating the detailed process of determining the optimal smoothing coefficient.

[0026] Figure 13 It is a graph obtained by adding a smoothing coefficient in the Y direction while plotting the calculated amount of random noise.

[0027] Figure 14 It is a graph obtained by adding a smoothing coefficient in the X direction while plotting the calculated amount of random noise.

[0028] Figure 15 This is an example of an SEM image that has been mixed with noise synchronized with the power supply frequency.

[0029] Figure 16 Indicates to Figure 15 The SEM image shown is an example of the change in measured values ​​relative to frequency during PSD measurement.

[0030] Figure 17 It is a flowchart used to illustrate the process of removing only noise components that are synchronized with the power supply frequency.

[0031] Figure 18 The outline of a line is represented by a pattern of at least two lines.

[0032] Figure 19 It is a simulation of the SEM image after the position offset is fed back to the SEM image. Detailed Implementation

[0033] This embodiment proposes a pattern measurement method and a pattern measurement system (apparatus) for detecting noise components with high precision from an image that is the object of measurement, thereby measuring the true edge roughness that appears at the edge of the pattern with higher precision.

[0034] In recent years, with the miniaturization of patterns, it has become essential to measure the edge roughness (the unevenness at the ends of the pattern) of the pattern with high precision. This is because the degree of edge roughness varies greatly depending on the materials constituting the semiconductor device, the exposure equipment, or the properties and characteristics of the substrate. The measurement and control of edge roughness have a significant impact on the yield of the device.

[0035] However, for example, when imaging a pattern in a semiconductor measurement device such as a scanning electron microscope (Critical Dimension SEM: CD-SEM), noise generated from the detector (randomly generated over time) or from the circuitry during the conversion of the detector signal into an electrical signal (random noise) overlaps. Therefore, the edges of the pattern cannot be accurately captured due to this random noise. Consequently, when measuring edge roughness, the output roughness value is larger than the true edge roughness inherent in the pattern itself.

[0036] Therefore, high-precision detection of random noise in the image is required, and this noise must be removed to measure edge roughness. Methods for measuring semiconductor patterns include illumination with light and illumination with charged particle beams. The difference between the two lies in resolution, but both involve digital signal processing during pattern imaging. Therefore, the same image processing techniques can be applied.

[0037] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings, using images obtained via CD-SEM as an example. In the drawings, functionally identical elements are sometimes indicated by the same reference numerals. The drawings illustrate specific embodiments and installation examples conforming to the principles of the present disclosure, but they are intended for understanding the present disclosure and are not intended to limit its interpretation.

[0038] In this embodiment, the present disclosure has been described in detail in a manner that enables those skilled in the art to fully implement it. However, it should be understood that other installations or methods may be used, and structural or constructional changes or substitutions of various elements may be made without departing from the technical concept and spirit of the present disclosure. Therefore, the following description is not intended to be limited thereto.

[0039] In addition, as described later, embodiments of the present disclosure can be implemented by software running on a general-purpose computer, or by dedicated hardware, or a combination of software and hardware.

[0040] <Configuration Example of CD-SEM>

[0041] Figure 1 Shows a schematic structural example of the CD-SEM in this embodiment. The CD-SEM 10 includes an electron gun 101, a condenser lens 103, an aperture 104, an objective lens 105, a detector 108, and an image conversion unit 109.

[0042] The electron gun 101 is used to extract a primary beam 102 from an electron source and accelerate it to an acceleration voltage set by the user. After that, the primary beam 102 is irradiated onto the specimen 106 via the condenser lens 103, the aperture 104, and the objective lens 105. Depending on the energy, specimen potential, unevenness, etc. during irradiation, secondary electrons 107 are emitted from the specimen 106. The secondary electrons 107 are detected by the detector 108. The detected signal is converted into an electrical signal and input to the image conversion unit 109 to be converted into an image.

[0043] <Structural Example of Image Conversion Unit>

[0044] Figure 2 Shows a structural example of a system including the image conversion unit 109 and the pattern measurement unit 202, where the pattern measurement unit 202 measures the pattern in the image obtained here. Figure 2 The structure in which the image conversion unit 109 and the pattern measurement unit 202 are connected via a network is shown, but it is not limited thereto, and the pattern measurement described later can also be performed within the image conversion unit 109.

[0045] Figure 2The illustrated pattern measurement unit 202 includes: an image processing unit 203 that receives an image obtained from the image transformation unit 109 and performs line contour calculation and image processing as described later; an FFT processing unit 204 that detects edges from the obtained image and performs Fourier analysis to obtain the power spectral density (PSD); a waveform processing unit 205 that measures the amount of random noise based on the PSD curve generated by the FFT processing unit 204 using the method described later; and a measurement processing unit 206 that uses the amount of random noise obtained from the waveform processing unit 205 to calculate the true edge roughness measure and outputs the result to the display of the image transformation unit 109. The pattern measurement unit 202 can be configured, for example, by a computer that includes a storage device and a processor as structural elements. In this case, the processor can read various programs for implementing the image processing unit 203, the FFT processing unit 204, the waveform processing unit 205, and the measurement processing unit 206 from the storage device and expand them in the processor's internal memory (not shown) to construct each unit and each processing unit.

[0046] <Detailed content on image processing and true edge roughness measurement>

[0047] Figure 3 This is a flowchart used to illustrate the image processing and actual edge roughness measurement processing in the pattern measurement unit 202. Hereinafter, the action subject of each step is set as each processing unit (image processing unit 203, etc.), but each processing unit is implemented by the processor of the pattern measurement unit 202, so the action subject can also be set as the processor or the pattern measurement unit 202.

[0048] (i) Step 301

[0049] First, the image processing unit 203 acquires an image captured by the SEM, containing at least one line pattern, from the image transformation unit 109. Then, as... Figure 4 As shown, the image processing unit 203 obtains the line contour of each line scan from the SEM image. Here, Figure 4 The diagram shows the contours obtained for each line scan. For example, in... Figure 4 In this case, if the range in the Y direction is 256 pixels, then 256 line profiles are obtained.

[0050] (ii) Step 302

[0051] like Figure 5 In this way, the image processing unit 203 obtains the SUM line profile from the SEM image by summing the entire range specified in the Y direction. Figure 5This represents an example of a SUM line profile calculated based on the line profiles of the entire range (e.g., the average value obtained by summing the line profiles of the entire range and dividing by the number of lines).

[0052] (iii) Step 303

[0053] The image processing unit 203 divides the line contours obtained in step 301 into left and right (L and R) sections, and performs pattern matching at each line edge with the SUM line contour obtained in step 302, such as... Figure 6 The position offset is obtained as shown. Figure 6 An example of the positional offset of the SUM line profile from each other is shown.

[0054] (iv) Step 304

[0055] The image processing unit 203 performs position offset correction (image processing) on ​​the SEM image based on the position offset obtained in step 303, such as... Figure 7 The image shown is generated after removing the roughness component. Figure 7 This represents an example of an image with the roughness component removed.

[0056] (v) Step 305

[0057] The FFT processing unit 204 performs power spectral density (PSD) measurement on the SEM image obtained in step 304, and calculates the average value of the full frequency range of the obtained PSD value (PSD Ave.) as the random noise component.

[0058] (vi) Step 306

[0059] The waveform processing unit 205 measures the true edge roughness by subtracting the random noise component obtained in step 305 from the PSD value of the original image. The measurement processing unit 206 receives the measured edge roughness from the waveform processing unit 205 and displays it on the display of the image conversion unit 109.

[0060] Figure 8 This indicates a comparison between the PSD value of the original image (PSD value for spatial frequency) and the PSD value of the processed image obtained in step 305 (PSD value for spatial frequency).

[0061] like Figure 8As shown, the PSD value after image processing only represents random noise components because the edge position of the pattern has been corrected. Furthermore, random noise is generally known to exist uniformly across the entire frequency band; therefore, by using the average value across all frequencies as the random noise component, deviations in the random noise component can be suppressed. Moreover, by subtracting the random noise component obtained using this method from the PSD value of the original image, the true edge roughness can be calculated with high accuracy. Additionally, when multiple line patterns exist within the image, the above processing can be performed on each line pattern to calculate and average the random noise component, and then the averaged value is subtracted from the PSD value of the original image.

[0062] As described above, an image with roughness components removed is generated, and the average PSD value across all frequencies is used as the random noise component. This enables high-precision detection of noise components and high-precision measurement of roughness. Furthermore, this method places no restrictions on the patterns used in the measurement.

[0063] <Optimal shooting frame rate determination processing>

[0064] This section explains how to determine the optimal number of frames for shooting using high-precision noise detection. In CD-SEM 10, when generating images, a predetermined number of images (frames) are accumulated from multiple scans of the same location, and then measurements are taken (the same applies to the roughness measurement process described above). In this case, the more frames are added, the lower the random noise becomes, and the more stable the measurement becomes.

[0065] However, increasing the frame rate increases the image acquisition time, leading to damage from electron beams irradiating the sample, and pattern shrinkage in resist samples. Furthermore, when the frame rates are summed, the amount of random noise saturates around a certain frame rate.

[0066] Therefore, this paper proposes a calculation process for the minimum number of frames (optimal number of frames) that saturates the random noise level.

[0067] Figure 9 This is a flowchart used to illustrate the optimal frame rate calculation process.

[0068] (i) Step 901

[0069] The image processing unit 203 acquires SEM images at a frame rate that is sufficient for a certain degree of saturation of random noise (a preset frame rate: for example, 32 frames).

[0070] (ii) Step 902

[0071] like Figure 10As shown, the image processing unit 203 generates cumulative images up to 2 frames, 4 frames, 8 frames, 16 frames, and 32 frames. At this time, the number of frames accumulated is uncertain and can be any number of cumulative frames.

[0072] (iii) Step 903

[0073] Image processing unit 203 performs the following steps for each accumulated image: Figure 3 Steps 301 to 304 of the processing generate an image without roughness. Furthermore, the FFT processing unit 204 performs... Figure 3 Step 305 involves measuring the PSD to calculate the amount of random noise.

[0074] (iv) Step 904

[0075] The waveform processing unit 205 generates and displays a graph showing the relationship between the number of frames and the amount of random noise, and sets the lowest number of frames where the noise level is saturated as the optimal number of frames. Figure 11 This is a graph showing the relationship between frame rate and random noise level. Whether the noise level has reached saturation can be determined based on... Figure 11 The judgment is made based on whether the slope in the chart shown is below a predetermined value (if the slope is below the predetermined value, it is judged that the noise level has saturated). Figure 11 In the analysis, since the amount of random noise was saturated in 16 frames (which could be determined that the slope of the graph was below a predetermined value), 16 frames were determined to be the optimal frames.

[0076] As described above, by varying the frame rate while plotting the amount of random noise, the optimal frame rate can be determined.

[0077] <Determination of Optimal Smoothing Coefficient>

[0078] This section explains the process of using the amount of noise detected with high precision to determine the optimal smoothing coefficient during measurement. As mentioned above, in order to measure edge roughness, it is necessary to detect the edge position of the pattern based on the line profile obtained from the image. At this time, by smoothing the line profile in the X or Y direction with an arbitrary number of pixels, it is possible to reduce the random noise component from the line profile and stably detect and determine the edge.

[0079] However, when the number of points (smoothing coefficient) used in the smoothing process is large (more points), it deviates from the actual shape of the pattern edge (when the smoothing coefficient is too large, the shape of the pattern edge becomes blunt), and the roughness of the pattern edge also disappears.

[0080] Therefore, a calculation process for the optimal smoothing coefficient that reduces random noise components while preserving the edge information of the pattern is proposed.

[0081] Figure 12This is a flowchart illustrating the detailed process of determining the optimal smoothing coefficient.

[0082] (i) Step 1201

[0083] The image processing unit 203 acquires SEM images at any number of frames.

[0084] (ii) Step 1202

[0085] The image processing unit 203 performs the following steps on the image acquired in step 1201: Figure 3 Steps 301 to 304 generate an image without edge roughness (refer to...). Figure 7 ).

[0086] (iii) Step 1203

[0087] In the edge-roughness-free image obtained in step 1202, the FFT processing unit 204 measures the PSD while increasing the smoothing coefficient and calculates the amount of random noise.

[0088] (iv) Step 1204

[0089] The waveform processing unit 205 generates a graph showing the relationship between the smoothing coefficient and the amount of random noise, and displays it in the image conversion unit. Then, the waveform processing unit 205 sets the lowest smoothing coefficient that causes noise saturation as the optimal smoothing coefficient. Here, Figure 13 It involves plotting the calculated random noise level while simultaneously increasing the smoothing coefficient in the Y direction. Additionally, Figure 14 It involves plotting the calculated random noise level while simultaneously increasing the smoothing coefficient in the X direction. (See reference...) Figure 13 and Figure 14 It can be seen that the random noise level is approximately saturated at 20 pixels in the Y direction and 19 pixels in the X direction (saturation can be determined based on the slope of the graph, similar to the noise saturation mentioned above). That is, smoothing beyond these values ​​is meaningless. Based on the above results, the optimal smoothing coefficient is 20 pixels in the Y direction and 19 pixels in the X direction.

[0090] As described above, by varying the smoothing coefficient while plotting the amount of random noise in an image without roughness, the optimal smoothing coefficient can be determined.

[0091] <Detection and Removal of Noise Components Synchronized with Power Supply Frequency>

[0092] This section describes the method for detecting and removing noise components synchronized with the power supply frequency. In CD-SEM10, in addition to random noise, noise synchronized with the power supply frequency is sometimes mixed in. In CD-SEM10, as mentioned above, the same area is typically scanned multiple times and frame accumulation is performed to acquire an image. When scanning begins synchronized with the power supply frequency to acquire an image and accumulation is performed, noise appears in the image as distortion.

[0093] Figure 15 This shows an example of an SEM image mixed with noise synchronized with the power supply frequency. (By...) Figure 15 It can be seen that a common distortion exists in the multiple line patterns within the image. Furthermore, Figure 16 It is a chart (example) that represents the... Figure 15 The SEM image shown illustrates the variation of PSD measurements relative to frequency. Figure 16 It can be seen that the noise synchronized with the power supply frequency appears in the PSD.

[0094] Therefore, this paper proposes a process that removes only the noise components that are synchronized with the power supply frequency.

[0095] Figure 17 This is a flowchart illustrating the process of removing only noise components synchronized with the power supply frequency.

[0096] (i) Step 1701

[0097] The image processing unit 203 acquires SEM images in a manner that allows for the measurement of at least two line patterns.

[0098] (ii) Step 1702

[0099] The image processing unit 203 acquires the line profile for each line scan. However, in this case (during the process of detecting and removing noise components synchronized with the power supply frequency), the profile is not acquired in a single pattern as in the actual edge roughness measurement process described above, but rather... Figure 18 As shown, obtain the line outline from at least two line patterns.

[0100] (iii) Step 1703

[0101] The image processing unit 203 obtains the SUM line profile obtained by summing the entire range specified in the Y direction.

[0102] (iv) Step 1704

[0103] The image processing unit 203 performs pattern matching between the line contour of each line scan and the SUM line contour, and feeds back the positional offset to the SEM image. At this time, the same amount of positional offset is fed back for both lines. Figure 19This is a simulation image, representing the SEM image after the positional offset has been fed back into the SEM image. Observe. Figure 19 It can be seen that the distortion caused by noise synchronized with the power supply frequency disappears, and an image that retains only the edge roughness of the pattern and random noise can be obtained.

[0104] As described above, by performing a detection / removal process on noise components synchronized with the power supply frequency, it is possible to detect and remove noise components synchronized with the power supply frequency.

[0105] Summary

[0106] (i) Specific matters of this disclosure

[0107] Here, we list the specific aspects of the optimal shooting frame rate determination process, the optimal smoothing coefficient determination process, and the detection and removal of noise components synchronized with the power supply frequency.

[0108] (i-1) Specific Item 1

[0109] A pattern measuring system uses a signal obtained by scanning a sample with a charged particle beam or irradiation light to measure a pattern formed on the sample, the pattern measuring system comprising:

[0110] A storage device that stores a program for processing information about the pattern;

[0111] A computer reads the program from the storage device to process the information of the pattern.

[0112] The computer performs the following processing:

[0113] Generate original images of the various frame accumulation patterns;

[0114] Calculate the amount of random noise in the original image for each cumulative number;

[0115] Based on the relationship between the cumulative number of frames and the amount of random noise, the optimal cumulative number of frames is determined, which is neither excessive nor insufficient.

[0116] (i-2) Specific Item 2

[0117] In the pattern measurement system for specific item 1, during the processing of calculating the random noise amount for the original image of each cumulative number, the computer generates multiple individual line contours and a total line contour for the original image of each cumulative number. The multiple individual line contours are divided into two regions. For the multiple individual line contours and the total line contour divided into the two regions, pattern matching is performed between the individual line contours and the total line contour. Based on the result of the pattern matching, position correction is performed on the multiple individual line contours. A corrected image of the pattern is generated based on the multiple line contours after position correction. The power spectral density at the edge of the pattern in the corrected image is calculated. The random noise component is calculated by averaging the power spectral density.

[0118] (i-3) Specific Items 3

[0119] A pattern measuring system uses a signal obtained by scanning a sample with a charged particle beam or irradiation light to measure a pattern formed on the sample, the pattern measuring system comprising:

[0120] A storage device that stores a program for processing information about the pattern;

[0121] A computer reads the program from the storage device to process the information of the pattern.

[0122] The computer performs the following processing:

[0123] Obtain the original image containing the line pattern;

[0124] The line outline is obtained from the original image containing the line pattern, and an image with the edge roughness subtracted is generated;

[0125] In the image after removing edge roughness, the power spectral density is measured while the smoothing coefficient is changed, and the amount of random noise is calculated.

[0126] The minimum smoothing coefficient for noise saturation is determined based on the relationship between the smoothing coefficient and the amount of random noise.

[0127] (i-4) Specific Items 4

[0128] In the pattern measurement system for specific item 3, during the processing of generating an image with reduced edge roughness, the computer generates multiple individual line contours and a summed line contour for an original image containing at least two line patterns. The multiple individual line contours are divided into two regions. For the multiple individual line contours divided into the two regions and the summed line contour, pattern matching is performed between the individual line contours and the summed line contour. Based on the pattern matching results, position correction is performed on the multiple individual line contours. A corrected image of the at least two line patterns is generated based on the multiple line contours with the position correction. The corrected image is then used to generate an image with reduced edge roughness.

[0129] (i-5) Specific Items 5

[0130] A pattern measuring system uses a signal obtained by scanning a sample with a charged particle beam or irradiation light to measure a pattern formed on the sample, the pattern measuring system comprising:

[0131] A storage device that stores a program for processing information about the pattern;

[0132] A computer reads the program from the storage device to process the information of the pattern.

[0133] The computer performs the following processing:

[0134] Obtain multiple line contours from the line pattern contained in the original image of the sample;

[0135] A summation line profile in a predetermined direction is generated based on the plurality of line profiles;

[0136] The positional offset is calculated by performing pattern matching between the plurality of line profiles and the sum of line profiles;

[0137] The positional offset is fed back and reflected in the original image.

[0138] (ii) The pattern measurement system (also called a pattern measurement device) of this embodiment performs the following processes: generating multiple individual line contours, a sum of multiple line contours (the above-mentioned SUM line contour), and an original image of the pattern based on the signal obtained by irradiating the pattern with a charged particle beam; dividing the multiple individual line contours into two regions (for example, left and right regions); performing pattern matching between the individual line contours and the sum of line contours, respectively, for the multiple individual line contours and the sum of line contours divided into two regions; performing position correction processing on the multiple individual line contours based on the pattern matching results; and generating a corrected image of the pattern (see reference) based on the multiple line contours that have undergone position correction. Figure 7The process involves: processing the image to obtain the power spectral density (PSD) at the edge of the pattern in the corrected image; averaging this PSD to determine the random noise component; and obtaining the difference between the PSD at the edge of the pattern in the original image and the random noise component to calculate the edge roughness. Therefore, since the average PSD value across all frequencies is used as the random noise component, the random noise component can be detected with high precision. Furthermore, by subtracting the average PSD value (random noise component) across all frequencies from the PSD value obtained from the original image, the true edge roughness can be measured with high precision.

[0139] The summation line profile is configured with a first line template image (e.g., an image of the right region) containing a first peak waveform and a second line template image (e.g., an image of the left region) containing a second peak waveform. This enables high-precision pattern matching with the original image.

[0140] (iii) The pattern measurement system of this embodiment performs the following processes: generating original images (original images corresponding to each frame accumulation number) of patterns with various frame accumulation numbers (e.g., accumulation numbers from 2 frames to 32 frames); calculating the amount of random noise for the original images of each accumulation number; and determining the optimal frame accumulation number (the optimal accumulation number that neither over-accumulates nor under-accumulates) based on the relationship between the frame accumulation number and the amount of random noise. As described above, the process of calculating the amount of random noise for the original images of each accumulation number corresponds to... Figure 3 The processing steps 301 to 305 are as follows: In calculating the random noise amount, for each cumulative number of original images, multiple individual line contours and a summed line contour (SUM line contour) are generated. Each individual line contour is divided into two regions (left and right). For each individual line contour and the summed line contour divided into two regions, pattern matching is performed between the individual line contours and the summed line contour. Based on the pattern matching results, position correction is performed on each individual line contour. A corrected image of the pattern is generated based on the corrected line contours. The power spectral density at the edges of the pattern in the corrected image is calculated, and the random noise component is obtained by averaging the power spectral density. This approach balances image acquisition time and pattern measurement stability. Even if the cumulative frame count is excessively increased, it only increases the image acquisition time; accumulating more frames than the optimal frame count renders the accumulation process useless. On the other hand, if the cumulative frame count is too low, the random noise is relatively large, and the measurement results are unstable. Therefore, this embodiment achieves a balance between these two aspects by performing a process to determine the optimal frame accumulation number. When determining the optimal frame accumulation number, a graph representing the relationship between the frame accumulation number and the amount of random noise is generated (see [reference]). Figure 11The optimal cumulative frame count is the number of frames whose slope changes below a predetermined value in the graph.

[0141] (iv) The pattern measurement system of this embodiment performs the following processes: acquiring an original image containing a line pattern; acquiring line contours from the original image containing the line pattern and generating an image with reduced edge roughness; measuring the power spectral density and calculating the amount of random noise while changing the smoothing coefficient in the image with reduced edge roughness (preferably an image with no edge roughness at all, but it may also be an image with reduced edge roughness compared to the original image); and determining the minimum smoothing coefficient for noise saturation based on the relationship between the smoothing coefficient and the amount of random noise. Thus, the optimal smoothing coefficient (the number of pixels used in the smoothing process) suitable for the actual pattern edge shape can be determined. To be more specific, in the process of generating an image with reduced edge roughness, for the original image containing the line pattern, multiple individual line contours and a summed line contour (SUM line contour) are generated. Each of the individual line contours is divided into two regions (e.g., left and right regions). For both the individual and summed line contours divided into two regions, pattern matching is performed between the individual and summed line contours. Based on the pattern matching results, position correction is performed on each of the individual line contours. A corrected image of the line pattern is generated from the corrected line contours. This corrected image is then used to generate the image with reduced edge roughness. This is equivalent to... Figure 3 Steps 301 to 304.

[0142] (v) The pattern measurement system of this embodiment performs the following processes: acquiring multiple line contours from at least two line patterns contained in the original image; generating a summed line contour (SUM line contour) in a predetermined direction (e.g., the Y direction) based on the multiple line contours; performing pattern matching between the multiple line contours and the summed line contour, and calculating the positional offset; and feeding back the positional offset and reflecting it in the original image. This enables the detection and removal of noise components synchronized with the power supply frequency. Here, the detection and removal of this power supply frequency noise is performed on an image containing at least two line patterns because when noise synchronized with the power supply frequency is present, passivation of the patterns in the two line patterns in the same direction is detected.

[0143] (vi) The functions of the embodiments of this disclosure can also be implemented by software program code. In this case, a storage medium storing program code is provided to a system or device, and the computer (or CPU or MPU) of the system or device reads the program code stored in the storage medium. In this case, the program code read from the storage medium itself implements the functions of the above-described embodiments, and the program code itself and the storage medium storing the program code constitute this disclosure. Examples of storage media for supplying such program code include floppy disks, CD-ROMs, DVD-ROMs, hard disks, optical disks, optical discs, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, etc.

[0144] Alternatively, the actual processing can be performed by an OS (operating system) running on a computer based on instructions from program code, or in some or all of the processing, thereby achieving the functions of the embodiments described above. Furthermore, program code read from a storage medium can be written into the computer's memory, and then the computer's CPU or similar device can perform some or all of the actual processing based on instructions from that program code, thereby achieving the functions of the embodiments described above.

[0145] Furthermore, software program code for implementing the functions of the implementation method can be distributed via a network, thereby storing the program code in a storage unit such as a hard disk or memory of the system or device, or in a storage medium such as a CD-RW or CD-R. When in use, the computer (or CPU or MPU) of the system or device reads and executes the program code stored in the storage unit or storage medium.

[0146] Finally, the processes and techniques described herein are not inherently related to any particular device and can be implemented through any suitable combination of components. Furthermore, various types of general-purpose devices can be used according to the teachings described herein. It is also advantageous to construct dedicated devices to perform the steps described herein. Additionally, various inventions can be formed by appropriately combining the multiple structural elements disclosed in the embodiments. For example, several structural elements can be removed from all the structural elements shown in the embodiments, and different structural elements can also be appropriately combined. In this disclosure, descriptions are given in connection with specific examples, but these descriptions are not intended to be limiting but rather illustrative in all respects. Those skilled in the art will understand that various combinations of hardware, software, and firmware are suitable for implementing this disclosure. For example, the described software can be installed in a wide range of programming or scripting languages ​​such as assembler, C / C++, perl, shell, or Java (registered trademark).

[0147] Furthermore, in the above embodiments, control lines and information lines are considered as lines required for description, and the product may not necessarily represent all control lines and information lines. All structures can be interconnected.

[0148] Explanation of reference numerals in the attached figures

[0149] 10 CD-SEM

[0150] 101 Electron Gun

[0151] 102 single beam

[0152] 103 Condensing Lens

[0153] 104 aperture

[0154] 105 Objective Lens

[0155] 106 samples

[0156] 107 releases electrons

[0157] 108 detectors

[0158] 109 Image Conversion Unit

[0159] 202 Pattern Measurement Unit

[0160] 203 Image Processing Department

[0161] 204 FFT Processing Department

[0162] 205 Waveform Processing Department

[0163] 206 Measurement and Processing Department.

Claims

1. A pattern measuring system that uses a signal obtained by scanning a beam of charged particles or irradiating light onto a sample to measure a pattern formed on said sample, characterized in that, The pattern measuring system includes: Storage device, which stores a program for processing information of the pattern; and The computer reads the program from the storage device and processes the information of the pattern. The computer performs the following processing: Based on the signal obtained from the pattern, an original image of the pattern, multiple individual line profiles based on the original image, and a sum of the multiple individual line profiles are generated. The multiple individual line outlines are each divided into two regions; For the multiple individual line profiles and the total line profile divided into the two regions, pattern matching is performed between the individual line profiles and the total line profile, respectively. Based on the pattern matching results, position corrections are performed on the plurality of individual line contours respectively; A corrected image of the pattern is generated based on multiple line contours that have undergone the position correction. The power spectral density at the edge of the pattern in the corrected image is calculated, and the random noise component is obtained by averaging the power spectral density. Edge roughness is calculated by obtaining the difference between the power spectral density at the edge of the pattern in the original image and the random noise component.

2. The pattern measuring system according to claim 1, characterized in that, The computer divides the multiple individual line contours into two regions, left and right.

3. The pattern measuring system according to claim 1, characterized in that, The summation line profile has a first line template image containing a first peak waveform and a second line template image containing a second peak waveform.

4. The pattern measuring system according to claim 1, characterized in that, The computer performs the following processing: Generate the original image of the pattern with multiple types of cumulative frame counts; Calculate the amount of random noise in the original image for each cumulative number; Based on the relationship between the cumulative number of frames and the amount of random noise, the optimal cumulative number of frames is determined, which is neither excessive nor insufficient.

5. The pattern measuring system according to claim 4, characterized in that, In the process of calculating the random noise amount for the original image of each cumulative number, the computer generates multiple individual line contours and a total line contour for the original image of each cumulative number. The multiple individual line contours are divided into two regions. For the multiple individual line contours and the total line contour divided into the two regions, pattern matching is performed between the individual line contours and the total line contour. Based on the result of the pattern matching, position correction is performed on the multiple individual line contours. A corrected image of the pattern is generated based on the multiple line contours with position correction. The power spectral density at the edge of the pattern in the corrected image is calculated. The random noise component is calculated by averaging the power spectral density.

6. The pattern measuring system according to claim 4, characterized in that, The computer generates a graph showing the relationship between the cumulative number of frames and the amount of random noise, and determines the cumulative number of frames when the slope of the graph changes to a predetermined value or below as the optimal cumulative number of frames.

7. The pattern measuring system according to claim 1, characterized in that, The computer performs the following processing: Obtain the line outline of each line and generate an image with edge roughness subtracted; In the image after removing edge roughness, the power spectral density is measured while the smoothing coefficient is changed, and the amount of random noise is calculated; The minimum smoothing coefficient for noise saturation is determined based on the relationship between the smoothing coefficient and the amount of random noise.

8. The pattern measuring system according to claim 7, characterized in that, In the process of generating an image with reduced edge roughness, the computer generates multiple individual line contours and a summed line contour for the original image containing the line pattern. Each of the multiple individual line contours is divided into two regions. For both the individual line contours and the summed line contour, pattern matching is performed between them. Based on the pattern matching results, position correction is performed on each of the individual line contours. A corrected image of the line pattern is generated based on the corrected line contours. The image with reduced edge roughness is then used to generate the image.

9. The pattern measuring system according to claim 1, characterized in that, The computer performs the following processing: Obtain the multiple line outlines from at least two line patterns contained in the original image; The summation line profile in a predetermined direction is generated based on the plurality of line profiles; Perform pattern matching between the plurality of line contours and the sum of line contours, and calculate the position offset; The position offset is fed back and reflected in the original image.

10. A pattern determination method, which uses a signal obtained by scanning a charged particle beam or irradiating light onto a sample to determine a pattern formed on said sample, characterized in that, The pattern determination method includes: The computer reads the program from a storage device storing the program for processing information about the pattern, and processes the information about the pattern, and generates an original image of the pattern, multiple individual line contours based on the signal obtained from the pattern, and a sum of the multiple individual line contours based on the original image. The computer divides the multiple individual line contours into two regions respectively; The computer performs pattern matching between the individual line contours and the total line contour, which are divided into the two regions, respectively. Based on the pattern matching result, the computer performs position correction on each of the multiple individual line contours; The computer generates a corrected image of the pattern based on multiple line contours that have undergone the position correction. The computer calculates the power spectral density at the edge of the pattern in the corrected image, and averages the power spectral density to determine the random noise component. The computer calculates edge roughness by obtaining the difference between the power spectral density at the edge of the pattern in the original image and the random noise component.

11. The pattern measurement method according to claim 10, characterized in that, The pattern determination method includes: The computer generates the original image of the pattern, which is a cumulative number of frames of multiple types. The computer calculates the amount of random noise for the original image of each cumulative number; The computer determines the optimal number of frames to accumulate, which is neither excessive nor insufficient, based on the relationship between the cumulative number of frames and the amount of random noise.

12. The pattern measurement method according to claim 10, characterized in that, The pattern determination method includes: The computer acquires the original image; The computer obtains the line outline of each line from the original image and generates an image with the edge roughness subtracted. The computer measures the power spectral density and calculates the amount of random noise in the image after the edge roughness has been reduced, while changing the smoothing coefficient. The computer determines the minimum smoothing coefficient for noise saturation based on the relationship between the smoothing coefficient and the amount of random noise.

13. The pattern measurement method according to claim 10, characterized in that, The pattern determination method includes: The computer obtains the multiple line outlines from at least two line patterns contained in the original image; The computer generates the sum of line profiles in a predetermined direction based on the plurality of line profiles; The computer performs pattern matching between the plurality of line contours and the sum of line contours, and calculates the position offset; The computer feeds back the position offset and reflects it in the original image.

14. A recording medium storing a program for a computer to perform a pattern measurement process, the pattern measurement process using signals obtained by scanning a charged particle beam or irradiation light onto a sample to measure a pattern formed on said sample, characterized in that, The program causes the computer to perform the following processes: A computer that processes information about the pattern generates, based on the signal obtained from the pattern, an original image of the pattern, multiple individual line contours based on the original image, and a sum of the multiple individual line contours. The computer divides the multiple individual line contours into two regions respectively; The computer performs pattern matching between the individual line contours and the total line contour, which are divided into the two regions, respectively. Based on the pattern matching result, the computer performs position correction on each of the multiple individual line contours; The computer generates a corrected image of the pattern based on multiple line contours that have undergone the position correction. The computer calculates the power spectral density at the edge of the pattern in the corrected image, and averages the power spectral density to determine the random noise component. The computer calculates edge roughness by obtaining the difference between the power spectral density at the edge of the pattern in the original image and the random noise component.

Images