Method and information processing system for roughness analysis

Abstract

To enable highly accurate roughness analysis about one specific line pattern.SOLUTION: A method and information processing system generates one profile to be an object of roughness analysis from a measurement result of one specific line pattern, calculates a value of an HHCF (Height Height Correlation Function) about each of a plurality of lags r from the profile, calculates a roughness index obtained by correcting a component due to noise, a positive integer A and a noise portion included in the standard deviation of the profile by fitting each lag r and the value of the HHCF to the lag r to a model formula of an HHCF including noise defined by the component due to the noise, the positive integer A and the noise portion included in the standard deviation of the profile in an area with the lag r being a threshold or less, and calculates the value of the HHCF obtained by correcting the noise form the component due to the noise and the value of the HHCF included in the standard variation of the profile.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a roughness analysis technique with noise compensation. [Background technology]

[0002] The shape of line and space patterns in semiconductor device manufacturing is measured using a scanning electron microscope (SEM) or similar. From the SEM image obtained by observing the line pattern, a line edge profile is formed by edge detection, and based on this, a roughness evaluation called line edge roughness (LER) or line width roughness (LWR) is performed. Generally, SEM images contain noise, which causes noise in the line edge profile as well, so the roughness evaluation results also contain errors caused by the noise.

[0003] In one prior art technique (e.g., Non-Patent Document 1), in order to analyze the roughness of a line pattern, a large number of line patterns formed on a semiconductor wafer are measured, and each is subjected to PSD (Power Spectral Density) analysis and averaged to find features in the spectrum that are presumed to be caused by errors, and noise correction is performed by subtracting these from the PSD.

[0004] This PSD analysis involves measuring many line patterns and averaging their PSDs. This process is thought to have two purposes. One is to converge the variation in random noise contained in the line-edge profile by averaging, thereby estimating the amount of noise with relatively high accuracy. The other is to converge or suppress the variation in PSD caused by the unique shape of each line-edge profile, so that the averaged PSD matches the theoretical formula, making it possible to estimate the roughness parameters from the degree of agreement.

[0005] In such conventional techniques, since it is usually necessary to measure several hundred or more line patterns, problems in measurement throughput may occur. In addition, the set of roughness parameters obtained indicates the average roughness parameters of the many line patterns measured, and it is not possible to know the distribution of the roughness parameters among the many different line patterns. In addition, the reliability of the obtained roughness parameters is not evaluated.

[0006] As a roughness analysis method different from PSD, there is an analysis method called HHCF (Height-Height Correlation Function). When HHCF is used, unlike PSD, there is no processing equivalent to converging or suppressing the variation of PSD caused by the shape specific to the line edge profile, and roughness parameters can be calculated for one line edge profile. On the other hand, even when HHCF is used, if the line edge profile contains noise, the shape of the HHCF is deformed, which becomes a large error factor when calculating the roughness parameters. Therefore, similar to the above-mentioned method using PSD, a method has been proposed (for example, Non-Patent Document 2) that estimates and corrects the amount of noise with a relatively high accuracy by measuring a large number of different line patterns and averaging the respective HHCFs. However, this method requires a large number of measurements, similar to the analysis using PSD, and the obtained roughness parameters are the average roughness parameters of a large number of line patterns, similar to the analysis using PSD.

[0007] The conventional techniques described above do not allow for highly accurate acquisition of HHCFs or roughness parameters that have been corrected for noise for a specific line pattern or for one line edge profile of a line pattern. Furthermore, measuring multiple line patterns requires many measurements. Furthermore, the reliability of the roughness parameters obtained by these conventional techniques is not evaluated.

[0008] There are also conventional techniques that analyze roughness parameters for a specific line pattern (e.g., Non-Patent Documents 3 and 4). However, Non-Patent Document 3 has an error in a part of the formula, and noise correction is inaccurate. In addition, Non-Patent Document 4 reduces noise by measuring the same line pattern multiple times, and then superimposing and averaging the measurement results of each line edge profile, but it is expected that a very large number of measurements will be required to sufficiently reduce noise. In addition, the reliability of the roughness parameters obtained by these conventional techniques is not evaluated. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] U.S. Patent Application Publication US2020 / 0211813A1 [Non-patent literature]

[0010] [Non-Patent Document 1] GF Lorusso, et al, "Unbiased roughness measurements: Subtracting out SEM effects", Microelectron. Eng. 190, 33-37, (2018) [Non-Patent Document 2] V. Constantoudis, et al, "Line edge roughness metrology: recent challenges and advances toward more complete and accurate measurements", J. Micro / Nanolith. MEMS MOEMS 17, 041014, (2018) [Non-Patent Document 3] V. Constantoudis, et al, "Noise-free estimation of spatial Line Edge / Width Roughness parameters", Proc. of SPIE Vol.7272, 72724B, (2009) [Non-Patent Document 4] L. Azarnouche, et al, "Unbiased line width roughness measurements with critical dimension scanning electron microscopy and critical dimension atomic force microscopy", Journal of Applied Physics 111, 084318 (2012) Summary of the Invention [Problem to be solved by the invention]

[0011] Therefore, according to one aspect, it is an object of the present invention to provide a novel technique for enabling highly accurate roughness analysis of a specific line pattern. [Means for solving the problem]

[0012] The method according to the present invention includes the steps of: (A) generating one profile to be subjected to roughness analysis from the measurement results of one specific line pattern; (B) calculating a height correlation function (HHCF) value for each of a plurality of lags r from the profile; and (C) calculating, in a region where the lag r is equal to or smaller than a threshold value, each lag r and the HHCF value for the lag r by subtracting a noise-induced component σ included in the standard deviation of the profile. noise , a positive constant A and a noise-corrected roughness index α unbiased By fitting to the model equation of noise-incorporated HHCF defined as noise , a positive integer A and a noise-corrected roughness index αunbiased (D) calculating the noise-induced component σ noise and calculating a noise-corrected HHCF value from the HHCF value. Effect of the Invention

[0013] According to one aspect, it is possible to perform highly accurate roughness analysis for a specific line pattern. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram illustrating an example of a system configuration according to an embodiment. [Diagram 2] FIG. 2 is a diagram showing a processing flow according to the embodiment. [Diagram 3] FIG. 3 is a diagram for explaining the line edge profile. [Figure 4] FIG. 4 is a diagram for explaining averaging of the line edge profile. [Diagram 5] FIG. 5 is a diagram showing an example of calculation of HHCF. [Figure 6] FIG. 6 is a diagram showing an example of setting the low r range. [Figure 7] FIG. 7 is a diagram for explaining fitting to f(r) in the low r range. [Figure 8] FIG. 8 is a diagram showing an example of the HHCF value after noise correction. [Figure 9] FIG. 9 is a diagram illustrating a process flow according to the embodiment. [Figure 10] 10A to 10C are diagrams showing examples of determination of noise correction accuracy. [Figure 11] FIG. 11 is a diagram for explaining calculation of the correlation length. [Figure 12] FIG. 12 is a diagram for explaining the effect of the embodiment. [Figure 13] FIG. 13 is a diagram illustrating an example of a functional configuration of a computer that is an information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] FIG. 1 shows an example of a system according to an embodiment of the present invention. An information processing device 1000 according to this embodiment is connected to the SEM 100. Measurement results of the SEM 100 are output to the information processing device 1000. The information processing device 1000 has a measurement result storage unit 1010, a profile generation unit 1030, a first data storage unit 1040, an HHCF calculation unit 1050, a roughness parameter calculation unit 1060, a correction accuracy evaluation unit 1070, a second data storage unit 1080, an SEM control unit 1020, an input unit 1090, and an output unit 1100. Note that the SEM 100 is an example of a measurement device, and may be another measurement device.

[0016] The measurement result storage unit 1010 of the information processing device 1000 stores the measurement results of the SEM 100. The profile generation unit 1030 generates data such as a line edge profile with reduced noise from the measurement results stored in the measurement result storage unit 1010, and stores the data in the first data storage unit 1040. The HHCF calculation unit 1050 calculates the HHCF value from the data such as the line edge stored in the first data storage unit, and stores the calculated data in the second data storage unit 1080. The roughness parameter calculation unit 1060 stores each roughness parameter in the second data storage unit 1080 using the data stored in the first data storage unit 1040 and the second data storage unit 1080. The correction accuracy evaluation unit 1070 executes a process of evaluating whether or not the noise-corrected HHCF value has been corrected with sufficient accuracy. The output unit 1100 outputs the processing results in the information processing device 1000 to a display device or other output device (which may be another information processing device connected to a network). The input unit 1090 receives user inputs of instructions, threshold values, etc. The SEM control unit 1020 executes control processing such as causing the SEM 100 to carry out a predetermined measurement in response to the instruction input from the input unit 1090.

[0017] Next, the processing contents of the system in FIG. 1 will be described with reference to FIGS. First, the SEM control unit 1020 of the information processing device 1000 controls the SEM 100 to measure, for example, a line pattern on a semiconductor wafer, and stores the measurement results in the measurement result storage unit 1010 (FIG. 2: step S1). At this time, the measurement conditions are adjusted to perform a measurement with lower noise. If it is difficult to reduce noise by adjusting the measurement conditions alone, multiple measurements are performed at the same location. Note that, since whether or not the noise reduction is sufficient is evaluated in a later process, it is not necessary to excessively increase the number of measurements or the acceleration voltage of the SEM. For example, as shown typically on the left side of FIG. 3, a line pattern formed on a semiconductor wafer is photographed by the SEM 100, and an SEM image is obtained as shown typically in the center of FIG. 3.

[0018] The profile generating unit 1030 generates a line edge profile from the measurement result and stores it in the first data storage unit 1040 (step S2). More specifically, the line edge profile is extracted from the SEM image, which is the measurement result, by edge detection. For example, as shown in the right side of FIG. 3, the line edge profile is extracted. When measurements are performed multiple times, the extracted multiple line edge profiles are averaged. In addition, since there is a slight shift in the horizontal direction due to drift during the SEM measurement, a process is performed to make them coincide as much as possible (for example, a process using an alignment method such as the Iterative Closest Point algorithm), and then the line edge profiles are averaged. The result of such a process is stored in the first data storage unit 1040.

[0019] In this embodiment, if the SEM image is as shown in the center of FIG. 3, two line edge profiles are extracted as shown in the right side of FIG. 3, and processing is performed by focusing on one of them. However, the following processing may be performed separately for each of the two line edge profiles. Furthermore, if the SEM image contains multiple line patterns, each line pattern and each of the two line edge profiles may be processed separately. Furthermore, instead of the line edge profile, a line width roughness profile calculated in the longitudinal direction of the line pattern between the two line edge profiles may be extracted in this step. That is, the profile according to this embodiment may be a line edge profile or a line width roughness profile.

[0020] A noise-reduced line edge profile is generated by steps S1 and S2, which corresponds to reducing the random noise ε in the following equation. p i =p t +b+ε i (i=1,2,...,n) Here, p i represents the edge point position of the noisy line edge profile, and p t represents the edge point position of the true line edge profile, b represents the offset, and ε i represents random noise.

[0021] In addition, the square of the HHCF (G 2 When noise is included in (r), the square of the HHCF is expressed by the following formula:

number

[0022] In this way, the square of the noise-incorporated HHCF (G 2 (r)) includes the second and third error terms, and reducing the random noise ε by noise reduction corresponds to reducing the second and third error terms. In particular, the second error term is an error whose magnitude and sign are random and cannot be corrected, so it is reduced by reducing the noise in the SEM image. The third error term is estimated and corrected in step S5 and onwards.

[0023] As an example of the measurement conditions, SEM images were taken under the following conditions: acceleration voltage 5 kV, measurement magnification 50,000 times, image size 1920 × 2560 pixels, and the number of frames per measurement was 32. The line pattern of interest was measured 112 times. The 112 SEM images thus obtained were subjected to edge detection to extract line edge profiles, and then averaged to obtain the curve a in Fig. 4. In Fig. 4, the horizontal axis represents the position x (nm) in the longitudinal direction of the line pattern, and the vertical axis represents the position y (nm) in the direction perpendicular to the longitudinal direction of the line pattern. Comparing with the line edge profile without averaging shown in Fig. 4, it can be seen that noise is reduced by averaging. However, noise is still included in the curve a.

[0024] Next, the HHCF calculation unit 1050 calculates the HHCF value G of the line edge profile using the line edge profile data stored in the first data storage unit 1040. meas (r) is calculated and stored in the second data storage unit 1080 (step S3). In this step, G meas Calculate (r).

number

[0025] For example, the HHCF value G for the averaged line edge profile of 112 measurements is shown in Fig. 5. measand the HHCF value without averaging. Thus, the HHCF value of the averaged line edge profile, G meas is smaller than the HHCF value without averaging over the entire range of lag r. The difference is particularly noticeable in the low r range.

[0026] Then, the output section 1100 is meas By outputting the threshold value r t The threshold r is set by the user. t is preferably smaller than the correlation length ξ, which will be described later. t The roughness parameter calculation unit 1060 receives the input of the threshold value r t is set (step S5).

[0027] In the example shown in FIG. 5, for example, the threshold value r t = 10, and set the low r range to the threshold r t The following region is assumed. Note that the threshold r t Instead of having the user set , a fixed value smaller than the correlation length ξ may be set and used. Alternatively, the correlation length ξ may be calculated from the noise-inclusive HHCF value, and a value smaller than the correlation length ξ by a certain value or a certain percentage may be set as the threshold value r t You can also set it to .

[0028] Then, the roughness parameter calculation unit 1060 calculates the G meas For point (r), fitting is performed using the model equation of the noise-incorporated HHCF, and the noise-induced component σ noise , a positive constant A, and a noise-corrected roughness index α unbiased is calculated and stored in the second data storage unit 1080 (step S7). The model formula of the HHCF including noise is expressed as follows.

number

[0029] Figure 7 shows the relationship between equation (3) (f(r)) and G meas In the low r region, the HHCF value is theoretically 2αunbiased ) 1 / 2 However, due to noise, the value is √2σ noise Therefore, the observed value is larger by G meas If we fit equation (3) to (r), we get σ noise , A and α unbiased You will be able to obtain the following.

[0030] Then, the HHCF calculation unit 1050 calculates the G meas (r) and σ noise Then, the noise-corrected HHCF value G unbiased (r) is calculated and stored in the second data storage unit 1080 (step S9). In this step, G unbiased Calculate (r).

number

[0031] In the example of FIG. noise =0.19 nm, α unbiased =0.90 and A=0.0053.

[0032] In this way, one result of the roughness analysis is the noise-corrected Gunbaised (r) is obtained. This result may be output by the output unit 1100. FIG. 8 shows G with noise meas (r) and G with the noise component corrected unbiased (r). Overall, G with the noise component corrected unbiased (r) has a lower value, and the difference between the two is larger in the lower r range.

[0033] Also, the roughness index α, which is one of the roughness parameters unbiased is also obtained, and this result may also be output by the output unit 1100.

[0034] The process proceeds to the process of FIG. 9 via terminal A, and the correction accuracy evaluation unit 1070 evaluates each point of G in the lower r range unbiased (r) using an evaluation formula for evaluating the noise correction accuracy (step S11). The evaluation formula is expressed as follows, for example.

Equation

[0035] In step S11, E(r) is calculated for each r by substituting the r of the data points that satisfy r ≤ r t .

[0036] The correction accuracy evaluation unit 1070 determines whether the noise correction accuracy is within the allowable range based on the evaluation value E(r) (step S13). For example, a threshold value E for the evaluation value E(r) t (for example, 10%) is set, and it is determined whether the number of data points where E t < E(r) is less than or equal to the threshold value n t (for example, 0). E t and n tThe numerical values are for example, and other appropriate values may be set based on experimental results or the like.

[0037] Fig. 10 shows a calculation example of E(r). Fig. 10(c) shows the values of E(r) (r = 3 to 10) calculated when the line edge profiles measured 5 times are averaged. E t When E = 10, at r = 3, r = 7, and r = 9, E(r)>10. On the other hand, Fig. 10(b) shows the values of E(r) calculated when the line edge profiles measured 25 times are averaged. In this case, there is no r such that E t = 10 <E(r). Similarly, Fig. 10(a) shows the values of E(r) calculated when the line edge profiles measured 112 times are averaged. Comparing the case of measuring 112 times with the case of measuring 25 times, the value of E(r) is generally smaller. In this example, just measuring 5 times, the noise correction accuracy is outside the allowable range, and it can be judged that the noise correction accuracy is within the allowable range if measured 25 times or more.

[0038] Based on such a judgment criterion, when it is judged that the noise correction accuracy is outside the allowable range, the correction accuracy evaluation unit 1070 outputs, via the output unit 1100, a prompt for the user to re-measure because the noise correction accuracy is outside the allowable range (step S15). The user inputs the measurement conditions via the input unit 1090, and the SEM control unit 1020 causes the SEM 100 to re-measure the line pattern under the input measurement conditions for low noise (step S16). Then the process returns to S2 in Fig. 2 via terminal B. In step S16, there may be a case where re-measurement is performed by increasing the number of times under the same measurement conditions as in step S1, or for example, re-measurement may be performed by increasing the acceleration voltage included in the measurement conditions of step S1. In the former case, low noise is achieved by averaging the measurement results of step S1 and the re-measurement results together. In the latter case, since a low-noise measurement result is obtained separately from the measurement result of step S1, a line edge profile is generated using the measurement result of step S16.

[0039] In addition, instead of outputting the instruction to the user to perform remeasurement via the output unit 1100 in step S15, an output instructing the user to perform remeasurement may be immediately sent to the SEM control unit 1020 in accordance with preset measurement conditions (including the number of measurements, etc.).

[0040] On the other hand, if it is determined that the noise correction accuracy is within the allowable range, the roughness parameter calculation unit 1060 calculates the standard deviation value σ of the line-edge profile from the line-edge profile data stored in the first data storage unit 1040. meas is calculated and stored in second data storage unit 1080 (step S17). The method of calculating the standard deviation is well known, and therefore will not be described here.

[0041] Furthermore, the roughness parameter calculation unit 1060 calculates the standard deviation value σ meas and the noise-induced component σ noise From this, the noise-corrected standard deviation value σ unbiased is calculated and stored in the second data storage unit 1080 (step S19). meas 2 = σ true 2 +σ noise 2 Therefore, according to the following formula, σ unbiased Calculate.

number

[0042] The roughness parameter calculation unit 1060 also calculates the noise-corrected standard deviation value σ unbiased and the noise-corrected HHCF value G unbiased From (r), the correlation length ξ unbiased is calculated and stored in the second data storage unit 1080 (step S21). As shown in FIG. 11, the noise-corrected HHCF value G unbiased The saturation value of (r) in the high r range is √2×σ unbiasedFrom there, (2(1-1 / e)) 1 / 2 ×σ unbiased The r value at the time when the correlation length ξ unbiased In addition, G unbiased Since (r) is discrete data, G unbiased Linear interpolation equation for (r) and (2(1-1 / e)) 1 / 2 From the intersection with ξ unbiased In the example of FIG. unbiased = 42.0 nm.

[0043] Then, the output unit 1100 outputs the noise-corrected HHCF value G unbiased (r) and the roughness parameters are output to the user (step S23).

[0044] σ unbiased is used as an evaluation value for various semiconductor processes, for example. unbiased Since is related to the line width variation between different line patterns, it is used, for example, when developing a process to reduce line width variation or when investigating the relationship between line width variation and device performance. unbiased It is also expected that this will be used as a development indicator for semiconductor processes.

[0045] As described above, according to the present embodiment, a highly accurate roughness analysis can be performed for a specific line edge profile. That is, if multiple line edge profiles are obtained, a highly accurate roughness analysis can be performed for each of them. It is also possible to evaluate the noise correction accuracy. Furthermore, a method can be adopted in which the number of measurements is gradually increased so that the noise correction accuracy falls within an acceptable range, so that the measurement efficiency can be improved. Note that a line width roughness profile can be obtained from the line edge profiles at both ends of a specific line pattern, so it can also be said that a highly accurate roughness analysis can be performed for a specific line pattern.

[0046] By enabling roughness analysis of a single line edge profile in this way, it becomes possible to obtain new information that was not available before, such as the distribution (average, variance) of roughness parameters among multiple line patterns. Until now, only the average roughness parameter for the multiple line patterns that were the subject of measurement was obtained, and the estimated accuracy of this value was unknown. Understanding the variance is expected to lead to efficient and reliable measurements based on statistics.

[0047] Furthermore, as shown diagrammatically in Fig. 12, depending on the selection of multiple line patterns, it may be possible to evaluate the distribution of roughness parameters between adjacent line patterns or between line patterns spaced a certain distance apart on the semiconductor wafer surface, which is expected to lead to advanced semiconductor process management. In addition, if roughness analysis with highly accurate noise correction becomes possible, it may be possible to capture even slight changes in LER when the processing conditions for the line patterns are changed.

[0048] In this embodiment, it is assumed that the line edge profile has the characteristics of self-affine fractals, and its autocorrelation function (or autocovariance) R(r) is expressed by the following equation using the lag r and three roughness parameters (standard deviation (σ), correlation length (ξ), and roughness index (α)).

number

[0049] Although the embodiments of the present invention have been described above, the present invention is not limited to these. Elements of each embodiment may be combined in any way. In addition, each embodiment may be implemented by removing any element. As for the processing flow, the order may be changed or multiple steps may be executed in parallel as long as the processing result does not change. In addition, the functional block configuration of the information processing device shown in FIG. 1 is merely an example, and may differ from the program module configuration, file configuration, etc.

[0050] 2 and 9, steps S17 to S23 are performed after step S13, but they may be performed before step S13. In addition, since step S13 determines only the low r range, calculations may also be performed in steps S3 and S9 only in the low r range, and calculations in the high r range may be performed after it is determined that the noise correction accuracy is within an acceptable range.

[0051] In the above, in step S13, the evaluation value E(r) and the threshold value E t and n t We have given an example of automatic judgment based on G unbiased In some cases, E(r) or E(r) is presented to the user, and the user himself / herself determines whether the noise correction accuracy is within the allowable range. Furthermore, in the process flows of Figs. 2 and 9, an example is shown in which remeasurement is performed when the noise correction accuracy is outside the allowable range, but the fitting in step S7 may be optimized by changing the setting in the low r range. For example, the threshold r t If you set a value that is too large or too small, fitting may not be performed properly. If the noise correction accuracy is judged to be outside the acceptable range, set the threshold r t In some cases, the above information may be reviewed and revised.

[0052] The formula (5) for calculating the evaluation value is just an example, and various modifications are possible. unbiased (r)-(Ar 2αunbiased ) 1 / 2 The threshold value is set accordingly.

[0053] 1 shows an example in which the information processing device 1000 is connected to the SEM 100, but there are also cases where the information processing device is not connected to the SEM 100 and is independent of the SEM 100. In this case, the SEM control unit 1020 is not included. Also, instead of realizing the functions of the information processing device 1000 with one information processing device, there are also cases in which multiple information processing devices work together to realize the functions of the information processing device 1000. In either case, the information processing device 1000 may be referred to as an information processing system.

[0054] The information processing device 1000 described above is, for example, a computer device, and as shown in FIG. 13, a memory 2501, a CPU (Central Processing Unit) 2503, a hard disk drive (HDD: Hard Disk Drive) 2505, a display control unit 2507 connected to a display device 2509, a drive device 2513 for a removable disk 2511, an input device 2515, a communication control unit 2517 for connecting to a network, and a peripheral device connection unit 2521 for connecting to peripheral devices (including the image sensor 100, the damping mechanism 300, the cooling mechanism 400, etc.) are connected by a bus 2519. The HDD may be a storage device such as a solid state drive (SSD: Solid State Drive). An operating system (OS) and an application program for implementing the processing in the embodiment of the present invention are stored in the HDD 2505, and are read from the HDD 2505 to the memory 2501 when executed by the CPU 2503. The CPU 2503 controls the display control unit 2507, the communication control unit 2517, and the drive device 2513 according to the processing contents of the application program to perform a predetermined operation. Data during processing is mainly stored in the memory 2501, but may be stored in the HDD 2505. For example, an application program for performing the above-mentioned processing is stored in a computer-readable removable disk 2511 and distributed, and is installed from the drive device 2513 to the HDD 2505. It may also be installed in the HDD 2505 via a network such as the Internet and the communication control unit 2517. Such a computer device realizes the above-mentioned various functions by organic cooperation between the above-mentioned hardware such as the CPU 2503 and the memory 2501 and programs such as the OS and application programs.

[0055] The above-described embodiment can be summarized as follows.

[0056] The method according to the present embodiment includes the steps of: (A) generating one profile to be subjected to roughness analysis from the measurement results of one specific line pattern; (B) calculating a height correlation function (HHCF) value for each of a plurality of lags r from the profile; and (C) calculating, in a region where the lag r is equal to or smaller than a threshold value, each lag r and the HHCF value for the lag r by subtracting a noise-induced component σ included in the standard deviation of the profile. noise , a positive constant A and a noise-corrected roughness index α unbiased By fitting to the model equation of noise-incorporated HHCF defined as noise , a positive integer A and a noise-corrected roughness index α unbiased (D) calculating the noise-induced component σ noise and calculating a noise-corrected HHCF value from the HHCF value.

[0057] By carrying out such processing, it becomes possible to perform roughness analysis with noise corrected with high accuracy for a specific line pattern. The measurement result may be the result of multiple measurements, or may be the result of a single measurement under measurement conditions that allow for low noise. The profile may be generated, for example, by averaging multiple measurement results to generate a line edge profile or a line width roughness profile.

[0058] The above method also includes (E) calculating the noise-corrected HHCF value, the positive constant A, and the noise-corrected roughness index α unbiased The method may further include a step of judging whether the accuracy of the noise correction is within an allowable range based on the above. In this way, it becomes possible to evaluate the noise correction accuracy. The above judgment may be made for an area where the lag r is equal to or less than a threshold value.

[0059] Furthermore, the above method may further include a step of (F) outputting an output to prompt or instruct to remeasure a specific line pattern when the accuracy of the noise correction is not within an allowable range. The remeasurement may be performed under different measurement conditions to reduce noise, or may be performed under the same measurement conditions to increase the number of measurements.

[0060] The method further includes the steps of: (G) calculating a standard deviation of the profile from the profile; and (H) calculating the standard deviation of the profile and a noise-induced component σ noise The method may further include a step of calculating a standard deviation of the noise-corrected profile from the above. This allows the standard deviation (σ unbiased ) can be obtained. Note that these steps may be performed before step (E).

[0061] The method further comprises: (I) calculating a noise-corrected correlation length ξ from the standard deviation of the noise-corrected profile and the noise-corrected HHCF value; unbiased The method may further include a step of calculating the correlation length ξ unbiased is also important as an index of roughness.

[0062] A program for causing a computer to execute the above-described method can be created, and the program is stored in various storage media. [Explanation of symbols]

[0063] 1000 Information processing device 1010 Measurement result storage unit 1020 SEM control unit 1030 Profile generation unit 1040 first data storage unit 1050 HHCF calculation unit 1060 Roughness parameter calculation unit 1070 Correction accuracy evaluation unit 1080 second data storage unit 1090 input unit 1100 Output section

Claims

1. generating one profile to be subjected to roughness analysis from the measurement results of one specific line pattern; calculating a height correlation function (HHCF) value for each of a plurality of lags r from the profile; In the region where the lag r is equal to or smaller than a threshold, the noise-induced component σ noise , a positive constant A and a noise-corrected roughness index α unbiased By fitting to the model equation of the noise-containing HHCF defined as noise , the positive integer A and the noise-corrected roughness index α unbiased and calculating The noise-induced component σ included in the standard deviation of the profile noise and calculating a noise-corrected HHCF value from the HHCF value; The method includes:

2. The noise-corrected HHCF value, the positive constant A, and the noise-corrected roughness index α unbiased and determining whether the accuracy of the noise correction is within an allowable range based on the result. The method of claim 1 further comprising:

3. a step of outputting a prompt to perform remeasurement of the specific one line pattern or an instruction to perform remeasurement when the accuracy of the noise correction is not within an allowable range; The method of claim 2 further comprising:

4. calculating a standard deviation of the profile from the profile; The standard deviation of the profile and the noise-induced component σ included in the standard deviation of the profile noise calculating a standard deviation of the noise-corrected profile from the above; The method of claim 1 or 2, further comprising:

5. From the standard deviation of the noise-corrected profile and the noise-corrected value of the HHCF, a noise-corrected correlation length ξ unbiased Step of calculating The method of claim 4 further comprising:

6. A program for causing a computer to execute the method according to any one of claims 1 to 5.

7. A means for generating one profile to be subjected to roughness analysis from the measurement result of one specific line pattern; means for calculating a height correlation function (HHCF) value for each of a plurality of lags r from the profile; In the region where the lag r is equal to or smaller than a threshold, the noise-induced component σ noise , a positive constant A and a noise-corrected roughness index α unbiased By fitting to the model equation of the noise-containing HHCF defined as noise , the positive integer A and the noise-corrected roughness index α unbiased and a means for calculating The noise-induced component σ included in the standard deviation of the profile noise and a means for calculating a noise-corrected HHCF value from the HHCF value; An information processing system having the above configuration.

Images