Evaluation method, search method, and search system

Abstract

To provide a technique for evaluating a shape that appears in a cross-sectional image.SOLUTION: One of the typical evaluation methods of the present invention evaluates a difference between a target shape and a cross-sectional shape of an electron microscopic image, and includes: a first step of measuring a characteristic dimension of the cross-sectional shape; after the first step, a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimension; and a third step of comparing a difference between the template and the cross-sectional shape using a normalized index.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to an evaluation method, a search method, and a search system. [Background technology]

[0002] In general, in manufacturing systems, data about production is very important and valuable, and with the advancement of data analysis tools and algorithms, the constant search for more sophisticated and simpler ways to reduce our workload is crucial for positive growth in a competitive business environment.

[0003] As more information is collected, the amount of data can expand rapidly, requiring more and more resources and time for analysis, which tends to stall process development.

[0004] This situation is also true in the semiconductor industry, where extensive analysis of microscopic images is required to evaluate the success of, for example, etching tests.

[0005] Conventionally, methods for analyzing microscope images have been proposed. For example, Patent Document 1 discloses a dimension measuring device that shortens the time required for dimension measurement and eliminates errors caused by an operator. The dimension measuring device in Patent Document 1 uses a first image recognition model that extracts the boundary between a processed structure and the background and / or the boundary of the interface between different materials over the entire cross-sectional image, and a second image recognition model that outputs information for dividing the boundary over the entire cross-sectional image obtained from the first image recognition model into unit patterns that constitute a repeating pattern, to obtain the coordinates of multiple feature points defined in advance for each unit pattern, and measure a dimension defined as the distance between two predetermined points among the multiple feature points. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] International Publication No. 2020 / 121564 Summary of the Invention [Problem to be solved by the invention]

[0007] In Patent Document 1, a plurality of feature points are extracted from a microscope image in order to evaluate the dimensions of a repeating pattern, but there is still room for improvement in the method of evaluating the shape of the pattern itself. Therefore, an object of the present invention is to provide a technique for evaluating a shape appearing in a cross-sectional image. [Means for solving the problem]

[0008] In order to solve the above problems, one representative evaluation method of the present invention is an evaluation method for evaluating the difference between a target shape and a cross-sectional shape of an electron microscope image, characterized by having a first step of measuring a characteristic dimension of the cross-sectional shape, a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimension after the first step, and a third step of comparing the difference between the template and the cross-sectional shape using a normalized index. Effect of the Invention

[0009] According to the present invention, the shape appearing in the cross-sectional image can be evaluated. Problems, configurations and effects other than those described above will become apparent from the description of the following embodiments. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing an example of a cross section of a semiconductor processing layer imaged by an SEM. [Diagram 2] FIG. 2 is a diagram showing a schematic diagram of an ideal rectangular profile. [Diagram 3] FIG. 3 is a schematic diagram of a more realistic profile found in an etch process. [Figure 4] FIG. 4 is a diagram showing an example of the case where the area method is applied to the trench 4b shown in FIG. [Diagram 5] FIG. 5 is a diagram showing an example of the case where the distance method is applied to the trench 4b shown in FIG. [Figure 6] FIG. 6 is a diagram showing the process of preparing data. [Figure 7] FIG. 7 is a diagram showing an example of a profile diagram. [Figure 8] FIG. 8 is a diagram showing an example of data acquired through a data preparation process. [Figure 9] FIG. 9 is a diagram showing a case where a block region showing an ideal shape is set so as to include a shoulder region above a point having an x-coordinate value xmax. [Figure 10] FIG. 10 is a diagram showing the steps of calculating the area number in method 1. [Figure 11] FIG. 11 is a diagram showing a case where a block region showing an ideal shape is set in the range of a point having an x-coordinate value xmax. [Figure 12] FIG. 12 is a diagram showing a process for calculating the area number in method 2. [Figure 13] FIG. 13 is a diagram showing an example of a scaled profile diagram. [Figure 14] FIG. 14 is a diagram showing the steps of calculating the distance number in the distance method. [Figure 15] FIG. 15 is a diagram showing an example of the configuration of an evaluation system in which the evaluation method shown in the present disclosure is performed. [Figure 16] FIG. 16 is a diagram showing the results of evaluating the semiconductor structure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the present invention is not limited to the embodiment. In addition, in the description of the drawings, the same parts are denoted by the same reference numerals. When there are multiple components having the same or similar functions, they may be described by using the same reference numerals with different subscripts, or when there is no need to distinguish between these multiple components, the subscripts may be omitted. In order to facilitate understanding of the invention, the position, size, shape, range, etc. of each component shown in the drawings may not represent the actual position, size, shape, range, etc. Therefore, the present invention is not necessarily limited to the position, size, shape, range, etc. disclosed in the drawings.

[0012] In this disclosure, when indicating coordinates, conditions may be written in the subscripts. For example, "x y=ymax " means "the value of y is y max "The x-coordinate value when . A profile refers to the outline that appears on a cross section of a semiconductor. For example, a "profile diagram" of a semiconductor structure refers to a diagram showing the outline formed by the boundary between a semiconductor structure portion and a space portion that appears when a cross section of a semiconductor substrate is observed. Furthermore, the width of the profile refers to the length in a direction parallel to the main surface of the semiconductor substrate, and the height of the profile refers to the length in a direction perpendicular to the main surface of the semiconductor substrate.

[0013] [Problems with the prior art] Generally, collecting a large amount of data enables accurate evaluation, but there are challenges when analyzing large amounts of data.

[0014] To offset increases in data collection while maximizing the information in expensive SEM (Scanning Electron Microscope), STEM (Scanning Transmission Electron Microscope), and TEM (Transmission Electron Microscopy) images, it is often useful to represent important structures with as few data points as possible to rapidly analyze and report the outcome of a given test. However, some shapes or behaviors revealed in microscopic images of semiconductor deposits in a test are often difficult to quantify, despite their relevance.

[0015] Attempts to quantify such shapes are often done qualitatively using a single value by manual grading after SEM imaging or other inspection, but this approach leads to subjective measurements depending on who is grading the profile.

[0016] This ambiguity makes this approach less than ideal as a machine learning method and makes it less accurate.

[0017] For example, it is useful to describe the squareness of the profile by using edge detection software and capturing the edges as x and y coordinates, but this suffers from the drawback that the more accurate the shape description, the more sample points are used, resulting in a large collection of asymmetric coordinates that are difficult to manage manually and are not ideal for machine learning analysis.

[0018] Therefore, in this disclosure, we present two methods for quantifying the overall shape of an ambiguous shape using a single dimensionless metric by comparing it to an ideal shape.

[0019] [Embodiment] (Example of microscope image and shape of evaluation target) With reference to Figures 1 to 3, examples of microscope images of semiconductor process substrates are shown. Figures 1 to 3 are perspective views including a portion of a cross section of a semiconductor process substrate. Figure 1 shows an example of a cross section of a semiconductor process layer imaged by SEM. Figure 2 is a schematic diagram of an ideal rectangular profile. Figure 3 is a schematic diagram of a more realistic profile found in an etching process.

[0020] Fig. 1 is a cross-sectional view taken in a direction perpendicular to the main surface of a semiconductor processing substrate. The figure shows a state in which a processing layer 2 including a TiN / Ti layer, an Al-Si-Cu layer, and a TiN layer, and a photoresist layer 3 having a predetermined pattern are formed on a semiconductor substrate 1, and then etching is performed to form trenches 4 according to the predetermined pattern.

[0021] FIG. 2 shows an ideal etching. Under ideal conditions, etching proceeds in a direction perpendicular to the main surface of the semiconductor substrate 1a, and does not proceed in a direction parallel to the main surface of the semiconductor substrate 1a. Since the etching amount is common to the processing layer 2a and the photoresist layer 3a, the processing layer 2a and the photoresist layer 3a have a uniform surface with respect to the trench 4a. As shown in FIG. 2, the ideal trench 4a has a rectangular shape extending in a direction perpendicular to the main surface of the semiconductor substrate 1a. However, in reality, it is difficult to realize the ideal shape because the etching amount differs for each composition contained in the photoresist layer 3a and the processing layer 2a.

[0022] FIG. 3 shows an ideal etching close to a real state. Etching proceeds in a direction perpendicular to the main surface of the semiconductor substrate 1b, and does not proceed in a direction parallel to the main surface of the semiconductor substrate 1b. The trench 4b has an arc-like shape in the semiconductor substrate 1b. The shape shown in FIG. 3 is worthy of evaluation when considering the degree of etching processing. If an index for evaluating the degree of the shape shown in FIG. 3 can be obtained, it can also be applied to the evaluation of the etching processing.

[0023] SUMMARY OF THE PRESENTINVENTION With reference to Figures 4 and 5, this disclosure defines two methods (area method and distance method) for measuring how closely a shape resembles a rectangular shape. The evaluation method in this disclosure is an evaluation method that evaluates the difference between a target shape and a cross-sectional shape of an electron microscope image. Figure 4 is a diagram showing an example of a case where the area method is applied to the trench 4b shown in Figure 3. Figure 5 is a diagram showing an example of a case where the distance method is applied to the trench 4b shown in Figure 3.

[0024] After numerically evaluating the shape using boundary detection software, a template of the ideal shape, specified by unique extrema, is defined.

[0025] Scaling is then performed to allow different shapes to be meaningfully compared to one another.

[0026] In the area method, the area of ​​the profile is subtracted from the ideal template and divided by the area of ​​the ideal template to derive a dimensionless measure.

[0027] More specifically, Fig. 4 is a diagram showing an outline of the area method. In Fig. 4(a), the maximum width w of the trench 4b, which is the length in a direction parallel to the main surface of the semiconductor substrate 1b, is max A rectangular region formed by a side having the same length as the square root of the square root of the trench 4b, a side including the bottom 10 of the trench 4b, and a side of the trench 4b perpendicular to the main surface of the semiconductor substrate 1b is defined as a box region (region indicated by a dashed line) 11. In addition, the part of the box region 11 excluding the part of the trench 4b (the part with a checkered pattern in FIG. 4(a)) is defined as a shoulder region 12. The area number, which is an index calculated by the area method, is derived by formula (1) as shown in FIG. 4(b).

number

[0028] Distance methods also involve scaling the data by local extrema and then evaluating the points of the profile closest to the corners of the ideal template to derive an estimate of a dimensionless quantity.

[0029] More specifically, FIG. 5 is a diagram showing an outline of the distance method. In FIG. 5(b), the box region 11 is set in the same manner as in FIG. 4(a). Here, the y coordinate of the bottom 10 is y0. Also, the maximum width w max The value of the x-coordinate when max The corners (x max The point (x, y) closest to the point (x, y) is extracted. The distance number, which is an index derived by the distance method, is calculated by equation (2) as shown in Figure 5(b). Note that α and β are constants for normalization.

number

[0030] By normalizing these values ​​with the unique extreme value of each shape as described above, a single dimensionless value is obtained and used to compare similar shapes in other images and to assess the rectangularity of the sample.

[0031] These normalized values ​​can be used to better judge the outcome of a test and to simply quantify ambiguous but important shape information for use in statistical analysis.

[0032] (Implementation method) Here we define two quantitative methods that describe the conformance of a feature with an ideal shape by a single dimensionless quantitative measure.

[0033] These methods utilize boundary detection software to quickly obtain a numerical picture of the profile.

[0034] (Data preparation) The preparation of data used in calculating the index will be described with reference to Fig. 6. Fig. 6 is a diagram showing the process of preparing data.

[0035] The data preparation process is the first step in measuring the characteristic dimensions of a cross-sectional feature. The feature dimensions are collected via the profile outline. This is best implemented using boundary extraction software. In the following description, the dimensions are sometimes referred to as CD values ​​(critical dimensions).

[0036] Specifically, first, an SEM image of a cross section of a semiconductor processing substrate is obtained (step S1). Next, the image of the target shape is converted into a profile image in which the boundary line between the image and the background is extracted (step S2). Here, an SEM is used as an example, but the present disclosure is not limited to this. Images may be obtained by other so-called microscopes such as STEM and TEM.

[0037] Next, consider the change in profile width as a function of profile height. The y component of the data is transformed so that the bottom of the profile is at y = 0. Specifically, the y component of the data is transformed so that the bottom of the profile is at y = 0 (step S3).

[0038] Next, the x component of the data is transformed so that the outermost point of the profile is located at x = 0 (step S4). Specifically, the x component of the profile data is transformed using the maximum width value of the profile, and one point of the width when the profile width is at its widest is positioned at x = 0 (step S4).

[0039] The above process provides a set of unique extreme values ​​and parameters necessary to perform the area and distance calculations.

[0040] A specific example of the data in Fig. 6 will be described with reference to Fig. 7 and Fig. 8. Fig. 7 is a diagram showing an example of a profile diagram. Fig. 8 is a diagram showing an example of data acquired through a data preparation process.

[0041] Figure 7 is a profile diagram in which the boundary line is obtained from the SEM image using boundary line detection software. Figure 7 is a diagram obtained after step S2 in Figure 6 is performed. In Figure 7(a), the contours of the processed layer 2c and the photoresist layer 3c are extracted from a microscope image including the processed layer 2c and the photoresist layer 3c. Figure 7(b) shows a case in which a target range 17 to be evaluated is extracted from the profile diagram.

[0042] FIG. 8(a) is a diagram obtained after step S3 in FIG. 6 is performed. In FIG. 8(a), a contour line 20 is shown, which is a boundary line (contour) between the trench and the semiconductor processing layer. The contour line 20 is a collection of data in which N coordinates (N is a positive integer) are defined, and is derived by performing a smoothing process on the boundary line. In the profile diagram, the lowest point 21, which is the lowest point of the contour line 20, indicates the bottom and lowest part of the profile. The profile diagram is transformed so that the lowest point 21 is at the position of y=0. Note that the left side of the contour line 20 with respect to the paper surface indicates a part corresponding to the trench, and the right side of the contour line 20 indicates a part corresponding to the cross section of the semiconductor processing layer. In addition, the top point 23, which is the top point of the contour line 20, is the end of the semiconductor processing layer and corresponds to the boundary part between the semiconductor processing layer and the photoresist layer.

[0043] 8(b) is a diagram obtained after step S4 in FIG. 6 is performed. The outermost point 22 is the maximum width w of the trench on the contour line 20. max In FIG. 8(b), the contour 20 is inverted so that the outermost point 22 is at the position of x=0. In the following description, the y coordinate value of the apex point 23 is referred to as y max Also, the position of x=0 is the maximum width w max Since this is also the point where x=0 is realized, (x max ) is also shown. Although an example of transformation in which the contour 20 is inverted is shown here, the method of transformation is not limited to this. For example, the present disclosure can also be applied to a case in which the contour 20 is translated.

[0044] The prepared profile data is used for area and distance method calculations.

[0045] (Area method calculation) After preparing the profile data and defining the unique extreme values ​​according to the above description, the area number can be calculated. When calculating the area number, a block area showing an ideal rectangular shape is set as shown in FIG. 4. The process of setting the block area is the second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the process of preparing the data (first step). The process of calculating the index is the third step of comparing the difference between the template and the cross-sectional shape by a normalized index. Here, there are several possible methods for setting the block area and calculating the index. Methods of calculating the area number according to the method of setting the block area will be described as Method 1 and Method 2.

[0046] (Method 1) The ideal shape is determined by the characteristic extremes. In the area method, the dimensions are the maximum width and maximum depth of the cross-sectional shape, the template is a rectangle formed based on the maximum width and maximum depth, and the index is the area difference between the rectangle (template) and the cross-sectional shape.

[0047] Generally, a box is created to assume an ideal shape (template) and is defined by the maximum width (maximum width) and maximum height (maximum depth) observed in the profile being evaluated.

[0048] x coordinate of the widest part of the profile max is the y-coordinate value of the top of the profile max If not located at x max Any upper shoulder regions are treated the same as shoulder regions at the bottom side of the profile.

[0049] Therefore, Method 1 can be used to describe and optimize the overall profile match to an ideal shape that has the same inherent extreme value to determine the rectangularity of the profile, and the ideal value would be 1 which is the same as there being no shoulder region to measure.

[0050] FIG. 9 shows a block region 24 showing an ideal shape, with the x-coordinate value x max As shown in FIG. 9(a), in the area method 1, a block region 24 showing an ideal shape is set to include a shoulder region above a point having a width w max The end of the line is x = x max The outermost point 22 is a point located at the position of the maximum height of the trench (y=y max The area surrounded by the contour 20 and the coordinate axes x=0 and y=0 corresponds to the shoulder area.

[0051] 9(b), the area of ​​the shoulder region can be calculated by dividing the shoulder region into rectangles with their long sides parallel to the y-axis direction and taking the sum of the areas of the rectangles. The rectangle has sides with a length Δy in the y-axis direction and a side with a length up to a point set on the contour line 20.

[0052] 10 is a diagram showing the process of calculating the area number in method 1. First, the length Δy of the side in the y-axis direction of the rectangle dividing the shoulder region is calculated by the following formula (3) (step S11). Here, N is the number of coordinates on the contour line 20 set by the boundary detection software. Therefore, Δy is determined based on the resolution of the boundary detection software.

number

[0053] Next, the area of ​​the ideal shape is calculated (step S12). Ideal is calculated as follows:

number

[0054] Next, the area of ​​the shoulder region is calculated (step S13). Shoulder is approximated by the area of ​​N rectangles and calculated as the following formula (5). Here, x i (i is an integer between 1 and N) indicates the x-coordinate value of N coordinates set on the contour 20.

number

[0055] Finally, the area number is calculated using equations (3) to (5) as shown in the following equation (6) (step S14).

number

[0056] (Method 2) To inspect and verify the profile, the height of the ideal shape is calculated as the y-coordinate value y max It is useful to define the height instead as the y-coordinate value of the widest point of the profile.

[0057] This results in an area value of x y=ymax <x max In the case of a profile disadvantage of y x=xmax <y max Such a condition occurs, for example, in a situation encountered in semiconductor process engineering called "side etch".

[0058] Like a perfectly rectangular profile, the x-coordinate value is x maxIf there are many points where the profile is wide, such that the values ​​are close to , then defining the widest point as the top of the ideal shape will result in a less sensitive rectangularity value, whereas defining the narrowest point as the top of the ideal shape will result in a very sensitive rectangularity value if it describes just the bottom of the profile, but may mislead the user into determining the rectangularity value.

[0059] x max y max In other words, if the widest points of the profile are the sidemost points and the apex points, then the ideal shape is the same as the definition in Method 1.

[0060] FIG. 11 shows a block region 34 showing an ideal shape, with the x-coordinate value x max As shown in FIG. 11(a), in the area method 2, a block region 34 showing an ideal shape has a lowest point 21, which is the bottom (y=0), a maximum width w max The end of the line is x = x max The height to the outermost point 22 (the y coordinate of the outermost point 22) is equal to the height of the block region 34.

[0061] As shown in FIG. 11(b), the area of ​​the shoulder region can be calculated, as in Method 1, by dividing the shoulder region into rectangles whose long sides are parallel to the y-axis direction and taking the sum of the areas of the rectangles.

[0062] 12 is a diagram showing a process for calculating the area number in method 2. First, the number P (P is an integer equal to or smaller than N) of rectangles dividing the shoulder region is calculated by the following formula (7) (step S21). Δy is derived from formula (3).

number

[0063] Next, the area of ​​the ideal shape is calculated (step S22). Idealis calculated as follows:

number

[0064] Next, the area of ​​the shoulder region is calculated (step S23). Shoulder is approximated by the area of ​​P rectangles and is calculated as follows:

number

[0065] Finally, the area number is calculated using equations (7) to (9) as shown in the following equation (10) (step S24).

number

[0066] (Distance method calculation) The calculation of the distance method is explained below. In the distance method, the dimensions and template can be set in the same way as in the area method, and the index is the shortest distance from the corner of the rectangle (template) to the cross-sectional shape. After creating the profile data according to Figure 6 and the related explanation and defining the unique extreme value, the distance value is calculated.

[0067] (scaling) To arrive at a dimensionless measure and ensure that different shapes can be sensibly compared, the profile data are scaled by the unique extremum previously calculated. The scaling method can be chosen from the following methods:

[0068] To preserve the aspect ratio, we scale both x and y by the same constant α.

[0069] Furthermore, if we want to ensure that the distance measure is always less than 1, then we can ensure that α is x for all samples considered. maxand y max This often means that max and y max In practice this is difficult since θ is a relative magnitude and both extremes can vary significantly from sample to sample.

[0070] Thus, when a single parameter is used to scale both the x and y coordinate data, it cannot be guaranteed that the distance value will always be less than 1 or that it can be meaningfully compared to other profiles outside the data collection, but it may still be useful for analysis.

[0071] Another scaling parameter β=y max Introducing α=x max By choosing x and y to be scaled by α and β, respectively, it is possible to ensure that the value of the distance law is less than 0. At the cost of some abstraction from the original shape of the profiles due to aspect ratio distortion, we can make universally meaningful comparisons.

[0072] As with the area method, β=y max β = y x=xmax This would penalize any "side-etched" samples as well, but could lead to erratic values ​​observed in the distance law.

[0073] After calculating the region of interest and selecting a scaling method, the distance between all data points in the data collection can be measured to find the point that has the smallest length and is therefore closest.

[0074] This is the definition of the distance law value, where the ideal value for a perfectly rectangular shaped profile is 0.

[0075] FIG. 13 is a diagram showing an example of a scaled profile diagram. FIG. 13(a) shows the scaled profile diagram, and FIG. 13(b) shows an enlarged view of a portion of FIG. 13(a). x=xmaxThe y coordinate data is divided by the y coordinate value of (the outermost point 22), and the x coordinate data is divided by the x coordinate value of the nadir point 21. The scaling method is not limited to that shown in Fig. 13, and can be appropriately set depending on the type of etching shape and the method of evaluation.

[0076] As shown in Fig. 13(b), an ideal shape 44 is set. The ideal shape 44 is a rectangle with vertices (0, 0), (0, 1) which is the outermost point 22, (1, 0) which is the lowermost point 21, and (1, 1). The closest point 45 (x c , y c ) is calculated as a distance number.

[0077] FIG. 14 is a diagram showing a process for calculating the distance number in the distance method. First, an ideal shape is set in a profile diagram (step S31). The ideal shape is (0, 0), (0, y x=0 ), (x y=0 , y x=0 ), (x y=0 , 0) as vertices. Here, y x=0 indicates the y-coordinate value of the point corresponding to the end of the trench when the trench has the maximum width. Also, x y=0 indicates the x-coordinate value of the point corresponding to the bottom of the trench.

[0078] Next, the point closest to (0, 0) in the profile diagram is extracted, and the distance number is calculated (step S32). The distance number is derived by the following formula (11).

number

[0079] Although the distance between the vertex (0, 0) of the ideal shape 44 and the closest point was calculated, depending on how the ideal shape is set, the vertex of the ideal shape may not overlap with the coordinates (0, 0). In that case, the distance between an appropriate vertex of the ideal shape and the closest point is calculated as the distance number.

[0080] (System Configuration) 15 is a diagram showing an example of the configuration of an evaluation system in which the evaluation method described in the present disclosure is performed. Evaluation system 40 includes an input unit 41, a control unit 42, a storage unit 43, and a display unit 44. Evaluation system 40 is a search system that uses machine learning to search for etching conditions that result in a target shape from a process performed by a plasma etching apparatus. In addition, evaluation system 40 performs a search method that uses machine learning to search for etching conditions that result in a target shape from a process performed by a plasma etching apparatus.

[0081] The input unit 41 inputs data between devices and the like to the evaluation system 40, and also functions as an interface for inputting requests from a user of the evaluation system 40. The input unit 41 acquires a microscope image generated by a microscope device such as an SEM. The input unit 41 can also acquire data on processing conditions for a semiconductor processing substrate that is linked to the microscope image.

[0082] The control unit 42 performs processing for image evaluation. For example, the control unit 42 executes boundary detection software, prepares images as shown in Fig. 6, and generates profile diagrams from the microscope images. The control unit 42 also performs area method processing as shown in Figs. 10 and 12, and distance method processing as shown in Fig. 14.

[0083] The storage unit 43 stores boundary detection software. The storage unit 43 also stores a program for performing the area method and a program for performing the distance method. The storage unit 43 also stores the area numbers and distance numbers calculated by the control unit 42 in association with the processing conditions of the semiconductor processing substrate.

[0084] The display unit 44 presents the image evaluation results, including the area number and distance number calculated by the control unit 42, to the user.

[0085] The control unit 42 can also perform machine learning using the image evaluation results as one of the objective variables and perform processing condition optimization to search for optimized processing conditions.

[0086] (Evaluation example) Referring to Fig. 16, an example of applying the area method and distance method to the evaluation of a semiconductor structure is shown. Fig. 16 is a diagram showing the results of evaluating a semiconductor structure. As shown in Fig. 16(a) and Fig. 16(b), two types of semiconductor structures are evaluated. In Fig. 16(a) and Fig. 16(b), Fig. (1) is an SEM image. Fig. (2) is a profile diagram obtained by performing boundary detection processing. Fig. (3) is a diagram in which the profile is highlighted by changing the color.

[0087] In the structure shown in FIG. 16(a), the area number is 0.846351 and the distance number is 0.385486. On the other hand, in the structure shown in FIG. 16(b), the area number is 0.980214 and the distance number is 0.065626. The ideal shape of the trench that appears after the etching process is rectangular, but the area number of the structure in FIG. 16(b) is larger than the area number of the structure in FIG. 16(a). Therefore, it can be inferred that the shape of the trench shown in FIG. 16(b) has a space that is closer to a rectangle.

[0088] Furthermore, the distances show differences from the ideal shape, particularly for the bottom shape of the trench, with the distances for the structure in Fig. 16(b) being smaller than the distances for the structure in Fig. 16(a). Therefore, it can be inferred that the shape of the trench shown in Fig. 16(b) ensures flatness, particularly at the bottom.

[0089] (Action and effect) In this way, according to the present disclosure, an index for evaluating the shape appearing in a cross-sectional image can be obtained. Since the actual shape can be quantitatively evaluated with respect to the ideal cross-sectional shape with one index, absolute quantitative evaluation becomes possible in contrast to the evaluation that was conventionally performed using multiple measured values. Since the evaluation result can be associated with one index, the load on condition optimization by machine learning or the like can be reduced.

[0090] Although the embodiment of the present invention has been described above, the present invention is not limited to the above-described embodiment, and various modifications are possible without departing from the gist of the present invention.

[0091] The following are examples of possible embodiments of the present invention, but the present invention is not limited to these. (Aspect 1) 1. An evaluation method for evaluating a difference between a target shape and a cross-sectional shape of an electron microscope image, comprising: a first step of measuring a characteristic dimension of the cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; and a third step of comparing a difference between the template and the cross-sectional shape using a normalized index. (Aspect 2) In the evaluation method according to aspect 1, the dimensions being a maximum width of the cross-sectional shape and a maximum depth of the cross-sectional shape; The evaluation method, wherein the template is a rectangle formed based on the maximum width and the maximum depth. (Aspect 3) In the evaluation method according to the first or second aspect, An evaluation method, characterized in that the index is an area difference between the rectangle and the cross-sectional shape. (Aspect 4) In the evaluation method according to any one of aspects 1 or 3, An evaluation method, characterized in that the shortest distance from a corner of the rectangle to the cross-sectional shape is used as the index. (Aspect 5) A method for searching for etching conditions using machine learning for obtaining a target shape as a result of processing by a plasma etching apparatus, comprising: A first step of measuring a characteristic dimension of a cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; A searching method characterized in that the evaluation result of the electron microscope image obtained by executing a third step of comparing the difference between the template and the cross-sectional shape using a normalized index is used as one of the objective variables. (Aspect 6) A search system that uses machine learning to search for etching conditions that result in a target shape from a process performed by a plasma etching apparatus, comprising: A first step of measuring a characteristic dimension of a cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; A search system characterized in that the evaluation result of the electron microscope image obtained by executing a third step of comparing the difference between the template and the cross-sectional shape using a normalized index is used as one of the objective variables. [Explanation of symbols]

[0092] 1, 1a, 1b: semiconductor substrate, 2, 2a, 2b, 2c: processing layer, 3, 3a, 3c: photoresist layer, 4, 4a, 4b: trench, 10: bottom, 11: box region, 12 shoulder region, 17: target area, 20: contour line, 21: lowest point, 22: sidemost point, 23: top point, 24, 34: block region, 40: evaluation system, 41: input section, 42: control section, 43: memory section, 44: display section, 44: ideal shape, 45: closest point

Claims

1. 1. An evaluation method for evaluating a difference between a target shape and a cross-sectional shape of an electron microscope image, comprising: a first step of measuring a characteristic dimension of the cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; and a third step of comparing a difference between the template and the cross-sectional shape using a normalized index.

2. The evaluation method according to claim 1, the dimensions being a maximum width of the cross-sectional shape and a maximum depth of the cross-sectional shape; The evaluation method, wherein the template is a rectangle formed based on the maximum width and the maximum depth.

3. The evaluation method according to claim 2, An evaluation method, characterized in that the index is an area difference between the rectangle and the cross-sectional shape.

4. The evaluation method according to claim 2, An evaluation method, characterized in that the shortest distance from a corner of the rectangle to the cross-sectional shape is used as the index.

5. A method for searching for etching conditions using machine learning for obtaining a target shape as a result of processing by a plasma etching apparatus, comprising: A first step of measuring a characteristic dimension of a cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; A searching method characterized in that the evaluation result of the electron microscope image obtained by executing a third step of comparing the difference between the template and the cross-sectional shape using a normalized index is used as one of the objective variables.

6. A search system that uses machine learning to search for etching conditions that result in a target shape from a process performed by a plasma etching apparatus, comprising: A first step of measuring a characteristic dimension of a cross-sectional shape; a second step of creating a template of the target shape in the cross-sectional shape obtained from the dimensions after the first step; A search system characterized in that the evaluation result of the electron microscope image obtained by executing a third step of comparing the difference between the template and the cross-sectional shape using a normalized index is used as one of the objective variables.

Images