Etching shape evaluation system and method
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- HITACHI HIGH TECH CORP
- Filing Date
- 2025-01-24
- Publication Date
- 2026-07-30
Smart Images
Figure JP2025002290_30072026_PF_FP_ABST
Abstract
Description
Etching Shape Evaluation System and Method
[0001] The present invention relates to an etching shape evaluation system and method.
[0002] In a semiconductor wafer processing step called the previous step in the semiconductor manufacturing process, an electronic circuit or integrated circuit including transistors is formed on the surface of a silicon wafer. As part of the basic steps for forming these integrated circuits, there is a step of using a photoresist developed in a mask or the like and processing it into a shape such as a wiring by an etching process using an etching apparatus or the like. These are formed by repeating precise processing based on film formation, pattern transfer, and etching.
[0003] The improvement of the performance of integrated circuits has been achieved by developing miniaturization processing technology and miniaturizing elements and improving the integration density. In recent years, however, the miniaturization of wiring has reached a dead end, and new integration technologies have been required. Therefore, as new integration technologies, 2.5D packaging technology and 3D packaging technology have been attracting attention. These adopt a three-dimensional structure to improve the integration density, different from the method of improving integration in the planar structure by conventional miniaturization.
[0004] The etching apparatus used in the wiring processing step is used to etch (cut) the laminated film formed on the wafer to form a target structure. It is a device required to form a pattern according to the designed dimensions and shape and precisely drill long holes (deep holes) and grooves (deep grooves) for the wiring contained therein, and is indispensable for the manufacturing process. In the etching process, it is necessary to repeatedly perform processing tests while changing and adjusting the processing conditions to form a pattern according to the designed dimensions and shape. Since more precise processing accuracy is required in the processed shape, as an index indicating the processed shape, a method of measuring the actual cross-sectional shape and dimensions from the cross-sectional shape and quantitatively confirming the change amount is often used.
[0005] Patent Document 1 discloses a technique for estimating the cross-sectional shape of a sample to be inspected, comprising a scanning electron microscope (SEM) image of a pattern on a substrate, an array pattern detection process for detecting a regular arrangement of patterns within the image area, and template matching for detecting the position of patterns within the image area. The results of the array pattern detection and template matching are then integrated to detect the position of individual patterns and set a measurement cursor, and a measurement process for the pattern dimensions is performed for each measurement cursor. Patent Document 2 discloses a method for estimating the cross-sectional shape of a sample to be inspected, comprising a fitting step for fitting multiple shape models to the cross-sectional shape data of the sample to be inspected, and a selection step for selecting at least one shape model from the multiple shape models as the optimal model based on an error function value, which is an indicator of the accuracy of the fitting model fitted in the fitting step.
[0006] Japanese Patent Publication No. 2015-21942 Japanese Patent Publication No. 2013-73776
[0007] However, when observing semiconductor circuit patterns with deep holes or grooves on a Si wafer, which have three-dimensional structures such as fine cylinders, cones, and N-shaped prisms, in a planar state from the top view of the Si wafer using a SEM, it is difficult to estimate the cross-sectional shape consisting of the three-dimensional structure of the deep holes or grooves. For this reason, the conventional method for confirming the cross-sectional shape is to cleave the sample to be etched and confirm the cross-section with an SEM. In this case, since it is important to quickly start up the manufacturing process and move to a high-yield mass production system as soon as possible in device manufacturing, it is important to shorten the time required for observing and evaluating the cross-sectional shape when evaluating the performance of etching in repeated processing tests.
[0008] However, if the fine pattern is a pattern with circular (hole) deep holes (also called a hole pattern), in order to confirm its cross-sectional shape, it is necessary to process the sample by cleaving it vertically (making a small scratch along the surface with a diamond cutter, and then applying force to break it open, which forms a smooth cleavage surface). In this case, the finer the circular dimensions of the pattern, the more time is required to measure the roundness of each individual hole, and it is difficult to manually cleave the sample while maintaining the positional relationship of the center points and aiming for the same position as the pattern position determined from the top surface using a CD-SEM (Critical Dimension-SEM). Furthermore, it is even more difficult to cleave the sample while aiming for the center of the circular pattern in a symmetrical state, and to confirm that the cross-section corresponds to the true cross-section (i.e., the position corresponding to the center of the pattern at all positions from the top (surface) to the bottom (surface)).
[0009] On the other hand, observation and measurement are easily affected by the way cracks occur due to cleavage, and in etching tests where processing conditions are repeatedly adjusted to achieve the desired shape, obtaining the cross-sectional shape, which serves as an indicator of the processed shape, easily affects the accuracy of dimensional measurement results and shape evaluation. Furthermore, observation and measurement are easily influenced by human experience (discretion) and skill (how to cleave, identification of structure, etc.). Focused ion beam (FIB) processing can produce high-precision samples for cross-sectional observation, but it requires the acquisition of advanced processing techniques, and sample preparation takes several hours, often even longer depending on the material. Patent documents 1 and 2 do not adequately disclose the issues regarding accuracy and efficiency when performing shape evaluation by cleaving such processed samples and observing the cross-sectional shape.
[0010] Therefore, the present invention aims to provide a technology that enables improvements in accuracy and efficiency when evaluating the etching shape by observing the cross-section obtained by cleaving a sample for processing.
[0011] To solve the above problems, one representative etching shape evaluation system of the present invention includes a computer device, the computer device having a processing unit and a data storage unit, the data storage unit storing a base model image, the processing unit acquiring one or more cross-sectional observation images for creation captured by a scanning electron microscope, applying a grid template to the cross-sectional observation images for creation, identifying the image (material image) within the section of the grid template corresponding to each depth position where the dimension width of the cross-sectional contour line is maximum, extracting the material image, stitching it together to create a montage image, and the data storage unit storing the montage image as the base model image.
[0012] According to the present invention, the accuracy and efficiency of evaluating the etching shape by observing the cross-section obtained by cleaving a sample for processing can be improved. Other problems, configurations, and effects will be shown in the embodiments for carrying out the invention.
[0013] Figure 1 is a flowchart showing the steps for creating a montage image in the first embodiment. Figure 2 is a schematic diagram of the sample to be processed before and after etching in the first embodiment. Figure 3 is a schematic diagram illustrating several mask pattern shapes and deep hole pattern shapes. Figure 4 is a diagram showing an example of obtaining multiple cross-sectional observation images (A) to (C). Figure 5 is a schematic diagram showing how to apply a grid template to a cross-sectional observation image. Figure 6 is a schematic diagram showing how to apply another grid template to a cross-sectional observation image. Figure 7 is a schematic diagram showing how to apply grid templates A and B of different sizes to multiple cross-sectional observation images. Figure 8 is a schematic diagram showing how to apply a grid template to a cross-sectional observation image and how to measure the cross-sectional contour in all sections. Figure 9 is a schematic diagram illustrating the cleavage of a cylindrical deep hole pattern and the cross-sectional shape after cleavage. Figure 10 is a schematic diagram illustrating the cleavage of a conical deep hole pattern and the cross-sectional shape after cleavage. Figure 11 is a schematic diagram showing the shift in the mask pattern shape before and after etching. Figure 12 is a schematic diagram illustrating the three-dimensional structure of an actual deep hole pattern. Figure 13 is a schematic diagram showing how the grid-like section with the maximum value is identified. Figure 14 is a schematic diagram showing how the image within the grid-like section with the maximum value is extracted as the source image. Figure 15 is a schematic diagram showing how the cut-out source image data is joined together to create a montage image. Figure 16 is a flowchart showing the steps for evaluating the performance of etching by image recognition sequence processing using a base model image in the second embodiment. Figure 17 is a schematic diagram showing how similar cross-sectional pattern images similar to the montage image are identified and extracted.
[0014] Embodiments of the present invention will be described below with reference to the drawings. However, the present invention is not limited to these embodiments. Furthermore, in the drawings, identical parts are denoted by the same reference numerals.
[0015] [First Embodiment] (Creation of Montage Image) In the first embodiment of the present invention, the main focus is on creating a montage image to be used as an indicator for evaluating the etching shape from an image of a cross-section obtained by cleaving a processing sample (also simply called a sample) after etching. Figure 1 is a flowchart showing the steps for creating a montage image in the first embodiment. The following will be explained in accordance with steps 101 to 109 in Figure 1.
[0016] (Step 101) First, an etching test is performed on the sample to be processed. Figure 2 is a schematic diagram of the sample to be processed before and after etching in the first embodiment. In this embodiment, a scanning electron microscope (SEM) was used as the imaging device to photograph the sample to be processed. However, an imaging device other than an SEM may be used.
[0017] The sample for processing 201 is composed of at least an etchable layer 203 and a mask layer 202 stacked on a Si wafer. As shown in Figure 2(a), a schematic diagram (pattern top view) of the sample for processing observed from the top surface of the mask layer, and Figure 2(b), a schematic diagram observed from a cross-section, the mask layer 202 is patterned with regularly arranged circular, concave mask patterns on its surface. When etching is performed using an etching apparatus with a concave mask pattern, a deep-hole pattern having a three-dimensional cavity structure such as a cylinder, cone, or N-shaped prism is formed in the etchable layer 203 removed by processing, depending on the shape of the mask pattern. Figure 3 is a schematic diagram illustrating several mask pattern shapes and deep-hole pattern shapes. Figure 3(a) is a pattern top view of a circular mask pattern and a three-dimensional view of a cylindrical deep-hole pattern. Figure 3(b) shows a top view of a circular mask pattern and a three-dimensional view of a conical deep hole pattern, Figure 3(c) shows a top view of a triangular mask pattern and a three-dimensional view of a triangular prism-shaped deep hole pattern, and Figure 3(d) shows a top view of a square mask pattern and a three-dimensional view of a rectangular prism-shaped deep hole pattern.
[0018] On the other hand, if the mask pattern is convex, the shape after etching is such that the etched layer other than the one directly beneath the mask layer is removed, forming a pillar-like three-dimensional structure. If the regularly arranged convex patterns are densely arranged, deep grooves are formed between the patterns. In this embodiment, we consider the case of deep holes created by a concave mask pattern, but it can also be applied to the case of deep grooves.
[0019] (Step 102) Next, the sample to be processed is cleaved open, and an SEM observation image of the cross-section of the sample to be measured (referred to as a "cross-sectional observation image") is obtained (the cross-sectional observation image obtained to create a montage image is also called a "cross-sectional observation image for creation"). The shape of the deep hole pattern after etching is important for determining the etching shape. At that time, the ideal cleavage position A passes through the center of the circular mask pattern. ＊ -B ＊ The ideal method is to observe the cross-section obtained by cleaving the sample for processing, that is, the cross-sectional shape in which the three-dimensional structure of the deep pore pattern is symmetrical (see Figures 2(a) and 2(b)). In this embodiment, the three-dimensional structure of the deep pore pattern is at the ideal cleavage position A. ＊ -B ＊ The method involves making minute scratches on the Si wafer surface with a diamond cutter or similar tool along an arbitrary position, ensuring that the cleavage is contained as much as possible in a plane parallel to the Y direction, and then applying force to break the wafer. This method forms a relatively smooth cleavage surface, allowing for cross-sectional observation. However, the mask pattern is very fine, and most of the cleavage is performed manually, so the ideal cleavage position A is always required for the sample being processed. ＊ -B ＊ Performing cleavage is usually difficult. When more precise cleavage at the ideal cleavage position is required, there are methods using specialized scribing equipment equipped with microscopes, but high-precision processing like that of the focused ion beam (FIB) mentioned above is difficult, and it is not possible to target the center of each circle in the fine circular mask pattern observed from above.
[0020] Therefore, in practice, the sample to be processed is observed using a cross-sectional observation image obtained by cleaving it at the actual cleavage position A-B. Figure 2(c) is a schematic diagram of a cross-sectional observation image obtained by cleaving the sample to be processed after etching. When the sample to be processed is cleaved after etching, the cross-section of the three-dimensional structure (cavity) formed by the three-dimensional structure of the mask pattern of the mask layer 202 and the three-dimensional structure of the deep hole pattern of the layer to be etched 203 is revealed by the cross-sectional contour line 204. As shown in Figure 2(c), the cross-sectional contour line 204 can take on various shapes, and the dimensional width (CD: Critical Dimension) 205 at any depth 206 can also take on various values. Therefore, by measuring the shape and dimensions of the cross-sectional contour line 204, data such as the difference and variation from the shape and dimensions of the target cross-sectional contour line after processing can be obtained. In the following explanation, the direction horizontal to the upper surface of the mask layer along the actual cleavage position A-B is defined as the X direction, and the direction perpendicular to the upper surface of the mask layer is defined as the Y direction.
[0021] The observation magnification used to acquire the cross-sectional observation image was adjusted so that the depth 206 at any position (Y direction) from the upper end of the mask layer 202 to the lower end of the etched layer 203 relative to the cross-sectional contour line 204 could be measured, the amount of change in the dimensional width 205 at each depth position (X direction) could be quantitatively measured using, for example, a measurement tool on the screen of a computer device's display unit, and the observation area was included within the range of a standard image size.
[0022] In this embodiment, a 200K magnification setting, as indicated by the scanning electron microscope (SEM) used for observation, was employed, and multiple cross-sectional observation images were acquired from arbitrary observation positions within the sample. Figure 4 shows an example of acquiring multiple cross-sectional observation images (A) to (C). Note that the size and resolution of the image data may differ depending on the basic performance of the scanning electron microscope (SEM) used and the settings of its parameters. Furthermore, when importing the multiple cross-sectional observation image data after etching shown in Figure 4 into a computer system, it is advisable to confirm that the image data consists of regularly arranged deep-hole patterns of the same three-dimensional structure (e.g., cylindrical), and that the image data is a standardized image with consistent size, resolution, observation range, and observation magnification.
[0023] (Step 103) Next, a grid template is applied to the cross-sectional observation image. Figure 5 is a schematic diagram showing how to apply a grid template to a cross-sectional observation image. The grid template A501 is composed of grid-like sections (referred to as "grid sections" or simply "sections"), and by applying the grid template to the cross-sectional observation image, the shape of the cross-sectional contour line 204 can be measured for each section cut out (referred to as "section images"). The shape of the grid sections is not necessarily limited to a rectangle, as long as the shapes of multiple cross-sectional contour lines 204 can be measured for each section image and a montage image can be created.
[0024] Regarding the horizontally aligned sections of the grid-like template A501, since the mask pattern of the mask layer 202 is regularly arranged at regular intervals (see Figure 2(a)), and the deep hole pattern after etching is also expected to be regularly arranged according to the mask pattern, the sections should be arranged at the same intervals as the mask pattern and the deep hole pattern, and the dimensional width of each cross-sectional contour line 204 should be set to fit within the horizontal section size (section X size). The horizontal size of the entire grid-like template (template X size) is not particularly limited, but it should be set appropriately to accommodate the multiple cross-sectional contour lines to be measured that are included in the cross-sectional observation image when applied.
[0025] On the other hand, the vertically aligned sections of the grid-patterned template A501 can be adjusted as appropriate depending on the accuracy or efficiency of the etching shape evaluation. Making the size of each vertical section (section Y size) finer allows for the acquisition of more detailed measurement data on the cross-sectional contour line, improving the accuracy of the evaluation. However, this also increases the time and cost required for evaluation, so adjustments should be made based on this balance.
[0026] The overall vertical size of the grid-pattern template (template Y size) is not particularly limited, but it should be set appropriately to accommodate the multiple cross-sectional contour lines to be measured that are included in the cross-sectional observation image when applied. However, when measuring depth in the vertical direction, a reference position for the uppermost or lowermost position that serves as the starting point of the depth may be specified. For example, the uppermost position of the starting point of the depth may be placed at position C1 on the upper surface of the mask layer 202, or at position C2 on the interface between the mask layer 202 and the layer to be etched 203. In such cases, the template Y size should be adjusted so that the uppermost position (C2) or lowermost position (D) specified as the reference position is included, as shown in the example in Figure 5.
[0027] Figure 6 is a schematic diagram showing how a different grid template is applied to a cross-sectional observation image. Grid template B601 has an increased number of sections (divisions) in the Y direction compared to grid template A501. In this way, the number of sections in the vertical direction (Y direction) and the Y size of the sections of the grid template can be flexibly changed depending on the observation magnification, the range of depths to be observed, or the accuracy of performance evaluation.
[0028] If the size of the grid-like sections is subdivided, the number of source images for the montage composite image formed by combining the images within each section at each depth position, as described later, will also increase, allowing for the creation of a more detailed montage composite image. If detailed measurements are to be performed within a specific depth range, the size of the grid-like sections can be arbitrarily set, for example, by making the grid-like sections finer only within that range. Figure 7 is a schematic diagram showing how grid-like templates A and B of different sizes are applied to multiple cross-sectional observation images.
[0029] (Step 104) Next, the cross-sectional contour lines 204 are measured in all sections of the applied grid template A501. Figure 8 is a schematic diagram showing how the grid template is applied to a cross-sectional observation image and the cross-sectional contour lines are measured in all sections. As shown in Figure 8, all sections of the grid template A501 are identified by numbering in the X and Y directions. In the example in Figure 8, the X direction is assigned numbers 1 to 10 corresponding to the hole positions of the circular mask pattern, and the Y direction is assigned numbers a to e, so that all sections can be identified by section numbers such as 1-a, 1-b, ~10-d, 10-e. Then, the depth 206 of the cross-sectional contour line 204 and the dimension width 205 at that depth position are measured in all sections and numerical data is obtained. Note that the position where the depth 206 is measured in each section is not particularly limited and may be the centroid position of the section.
[0030] However, if the three-dimensional structure of the deep hole pattern deviates significantly from the ideal in parts due to the influence of the cleavage position and the state of the fractured cross-section (see the dotted line frame in Figure 2), and the cross-sectional contour line 204 disappears or becomes unclear, the measurement of the dimensions width 205 and depth 206 within that section may be omitted (see the × mark in Figure 8).
[0031] If multiple cross-sectional observation image data are acquired (see Figure 7), it is advisable to standardize the image data specifications, including image size and magnification, before importing them into the computer system. Then, apply a grid template of the same size to these images and measure each section to acquire numerical data. Regarding the measurement method, applications with applicable measurement functions on the computer system can be used, but it is not limited to just one method as long as it is possible to measure quantitative changes. However, it is desirable that the measurement conditions are constant and that output values can be obtained stably.
[0032] In one or more cross-sectional observation images, the numerical values of the width 205 and depth 206 measured within all grid sections of the applied grid template are entered into a spreadsheet application on a computer device to create a list sheet. It is desirable to include the section number (e.g., 1-a) of the measured section along with the measurement data of the width and depth in the list sheet. The width and depth can be measured by any means. They may be automatically measured by identifying both ends of the cross-sectional contour line at a predetermined depth using known image recognition technology, or they may be measured using a measurement tool on the display screen of a computer device.
[0033] It is desirable to record the maximum dimension width within each section numbered (a to e) at the same depth position on the list sheet. Furthermore, by calculating and recording the minimum, average, or median values in addition to the maximum value, it becomes possible to verify whether the obtained maximum value deviates from the desired design dimensions, which can be beneficial in improving the accuracy of the montage image described later. These calculations can be easily performed using the calculation functions of a spreadsheet application or similar tool, based on all the obtained measurement results. The list sheet can also be edited as needed, and highlighting the sections that take the maximum or minimum values in the grid sheet by color-coding will make it easier to understand the positional relationship of the sections that take the maximum value in the grid template later.
[0034] (Step 105) Next, select the maximum value for the dimensional width of the cross-sectional contour line within the grid template images corresponding to each depth position. Therefore, we will first explain the case where there is a discrepancy between the ideal cleavage position and the actual cleavage position when cleaving the sample for processing. Figure 9 is a schematic diagram illustrating the cleavage of a cylindrical deep hole pattern and the cross-sectional shape after cleavage. The inside of the deep hole pattern is an etched cavity, and the top and bottom surfaces of the three-dimensional structure of the cylindrical pattern are circular. Figure 9(a) shows the case where cleavage is performed along the direction Pa which is perpendicular to the top surface and passes through the center position A-A' (ideal cleavage position) of the circle (see (a-1)), and Qa in the cross section corresponds to the dimensional width of the cross-sectional contour line (see (a-2)). In this case, Qa indicates the maximum value that the dimensional width of the cross-sectional contour line can take at any depth position in the vertical direction, and this state is the state where cleavage has been performed at the center of the circle, i.e., on the plane of symmetry.
[0035] Figure 9(b) shows the case where cleavage occurs along a direction Pb perpendicular to the top surface and passing through a position B-B' slightly offset from the center position A-A' of the circle (see (b-1)), where Qb in the cross section corresponds to the dimension width of the cross section contour line (see (b-2)). Since the dimension width Qb of the cross section contour line decreases compared to the maximum value Qa, it can be inferred that the cleavage position is a state in which cleavage occurs on an asymmetric plane offset from the center position A-A'.
[0036] Similarly, Figure 9(c) shows the case where cleavage occurs along a direction Pc that is perpendicular to the top surface and passes through a position C-C' significantly offset from the center position A-A' of the circle (see (c-1)), and Qc in the cross section corresponds to the dimension width of the cross section contour line (see (c-2)). Since the dimension width Qc of the cross section contour line is significantly smaller than the maximum value Qa, it can be inferred that the cleavage is performed by an asymmetrical plane that is significantly offset from the center position A-A'.
[0037] Next, we will explain what happens when the cleavage position deviates from the ideal position, specifically when the three-dimensional structure of the etched deep hole pattern is conical (see Figure 3(b)). In this case, the dimensional width of the cross-sectional contour line of the deep hole pattern decreases at each depth position from the top surface to the bottom surface. Figure 10 is a schematic diagram illustrating the cleavage of a conical deep hole pattern and the cross-sectional shape after cleavage. The interior of the deep hole pattern is an etched cavity, and the top surface of the three-dimensional structure of the conical pattern is circular. Figure 10(a) shows the case where cleavage occurs along the direction Pa, which is perpendicular to the top surface and passes through the center position A-A' (ideal cleavage position) of the circle (see (a-1)), and Qa in the cross section corresponds to the dimensional width of the cross-sectional contour line (see (a-2)). In this case, Qa is the maximum value that the dimensional width of the cross-sectional contour line can take at any depth position in the vertical direction, and the vertical depth Ra of the cross-sectional contour line is also at its maximum value at any position in the horizontal direction. This state is the state of cleavage at the center of the circle, i.e., the plane of symmetry.
[0038] Figure 10(b) shows the case where cleavage occurs along a direction Pb perpendicular to the top surface and passing through positions B-B' offset from the center position A-A' of the circle (see (b-1)), where Qb in the cross section corresponds to the dimension width of the cross section contour line (see (b-2)). Since the dimension width Qb of the cross section contour line decreases compared to the maximum value Qa, and the vertical depth Rb also decreases compared to the maximum value Ra, it can be inferred that the cleavage position is a state in which cleavage has occurred on an asymmetric plane offset from the center position A-A'.
[0039] Furthermore, we will discuss cases where the three-dimensional structure of the deep hole pattern deforms or bends in the vertical direction. In the case of the regularly arranged circular mask pattern shown in Figure 2(a) and the N-shaped mask pattern shown in Figure 3, the finer the design dimensions, the greater the discrepancy between the dimensions of the etched pattern and the design dimensions. Figure 11 is a schematic diagram showing the discrepancy in the shape of the mask pattern before and after etching. Ideally, the designed circular mask pattern 1101 and the actual mask patterns 1102 to 1104 observed from above after etching should be identical in shape and dimensions. However, in reality, as shown in Figure 11, there are minute differences, and it is assumed that there are various discrepancies, including parts that exceed and fall below the design dimensions of the mask pattern 1101.
[0040] Furthermore, even in the three-dimensional structure of a deep hole pattern formed by etching a circular mask pattern vertically, there are discrepancies between the ideal shape and the actual shape. Figure 12 is a schematic diagram illustrating the three-dimensional structure of an actual deep hole pattern. Figure 12(a) shows the three-dimensional structure of an ideal deep hole pattern, and also shows the top surface 1201 and bottom surface 1202 (both circular) of the deep hole pattern as observed from above. In contrast, Figure 12(b) shows the three-dimensional structure and bottom surface 1203 when the hole axis of the deep hole pattern is bent, Figure 12(c) shows the three-dimensional structure and bottom surface 1204 when the dimensional width of the deep hole pattern is reduced, Figure 12(d) shows the three-dimensional structure and bottom surface 1205 when the deep hole pattern is tilted and insufficiently etched, and Figure 12(e) shows the three-dimensional structure and bottom surface 1206 when the deep hole pattern is tilted and deviates from the center line. Thus, the three-dimensional structure of an actual deep hole pattern deviates in various ways from the ideal shape.
[0041] When the three-dimensional structure of the deep hole pattern is deformed or bent vertically as described above, a state similar to the splitting of the conical three-dimensional structure shown in Fig. 10 is considered to occur. Even if it could split along the direction Pa passing through the center position A - A' of the circle, it is also expected that the shape of the cross-sectional contour line will partially show the dimension width of Qb and the depth of Rb due to the deviation of the shape state in the depth direction. In this case as well, it is assumed that the dimension width at each depth position will decrease from the maximum value that is plane-symmetric, and the measured values will also decrease with depth.
[0042] As described above, when comparing the dimension widths of the cross-sectional contour lines appearing in the cross-sectional observation image obtained by splitting the three-dimensional structure of the deep hole pattern at each depth position, it can be estimated that the larger the value, the closer it is to the splitting state in the symmetric plane (true cross-section) passing through the center position of the mask pattern. Therefore, when it is desired to estimate the splitting state in the symmetric plane, it is preferable to select the image with the maximum value. Furthermore, it can be estimated that the larger the depth of the cross-sectional contour line in the vertical direction, the less the bending and deformation of the deep hole pattern in the processing state.
[0043] (Step 106) Next, the data of the selected maximum value is compared with the position information of the square grid sections of the square grid template to identify the square grid section that takes the maximum value. Fig. 13 is a schematic diagram showing the state of identifying the square grid section that takes the maximum value. As described in step 104, the square grid template A501 is applied to the cross-sectional observation image (A), and the data of the dimension widths 205 of the cross-sectional contour line 204 in all the square grid sections (1 - a to 10 - e) have been obtained (Fig. 13(a)). However, it is not necessary to obtain the measurement data of the sections where the cross-sectional contour line is not clear (see the × marks in Fig. 13).
[0044] The comparison of the data of the maximum value with the position information of the square grid section is specified by section numbers such as 1 - a, 8 - b, 3 - c, 8 - d, 2 - e in the order of depth as shown in the example of Fig. 1(c). At this time, in order to make it easier to select the image with the maximum value, processing that is easy to select may be performed, such as emphasizing the display within the section by changing the bold line or color, or erasing or hiding the images within the other sections that do not apply.
[0045] In addition, when there is one cross-sectional observation image as shown in FIG. 13, the section with the maximum value is specified from within that image. However, when there are multiple image data as shown in FIG. 7, measurements are taken for all sections in all the image data, the maximum value is selected for each depth position from all the images and all the sections, and the section that takes the maximum value can be specified. However, the accuracy of the measurement data (dimension width and depth) may vary depending on the amount of information shown in the image data (resolution, pattern size, and observation magnification) and the size of the variable section size.
[0046] (Step 107) Next, the in-section image of the square mesh section that takes the maximum value is extracted as the material image. FIG. 14 is a schematic diagram showing the state of extracting the in-section image of the square mesh section that takes the maximum value as the material image. From the position of the square mesh section that takes the maximum value specified in Step 106 (FIG. 14(a)), as shown in FIG. 14(b), a method of extracting is adopted in which the image data is cut out fragmentarily while maintaining the size of the section size (the fragmentarily cut-out image is referred to as the "material image"). When there are the same maximum values, for example, any one can be adopted by comparing factors such as consistency with the design of the film structure or the boundary line of the laminated film, clarity of the cross-sectional contour line, presence or absence of deformation or foreign matter adhesion.
[0047] (Step 108) Next, the cropped material image data are joined together to create a montage image. Figure 15 is a schematic diagram showing how the cropped material image data is joined together to create a montage image. The material images 1-a, 8-b, 3-c, 8-d, and 2-e extracted in step 107 (Figure 15(a)) are joined together to create a montage image 1501 (Figure 15(b)). At this time, it is desirable to join the material images while making fine adjustments to maintain an intermediate position that provides as much continuity as possible so that the cross-sectional contour lines of the montage image do not shift significantly at the joints of adjacent material images. Furthermore, by using a smaller and narrower Y-size section of the grid template to be applied, it is possible to further smooth the joints between the upper and lower images when creating the montage image, and to form a more accurate montage composite image. However, if the number of vertical sections (divisions) in the grid template is increased and the Y-size of the sections is narrowed too much, the amount of measurement data will increase significantly. In addition, it is expected that a lot of time will be spent on selecting and extracting source images and creating montage images by stitching the source images together, so it is important to strike a balance with accuracy.
[0048] (Step 109) The montage image created in this way is an image within a section where the material image has the maximum dimension width at each depth position. Therefore, the montage image as a whole can be estimated to be an ideal cross-sectional contour line on the plane of symmetry (true cross-section) cleaved at the center of the mask pattern shape. Thus, the montage image can be used as an index image for evaluating the etching shape (the image used as an evaluation index is called the "base model image"). The montage image is then stored in the data storage unit as a base model image that serves as an index for evaluating the performance of the etching process. It is then used as an index for evaluating the performance of the etching process.
[0049] (Hardware Configuration) This embodiment can be constructed as a system with a computer device comprising a processing unit, memory, input interface, instruction input unit, data storage unit, communication interface, output interface, and display unit as its hardware configuration. The processing unit is a processor (CPU) that executes instructions according to a program stored in memory. The memory is a medium for storing data and programs, and may include random access semiconductor memory, a storage device, or a storage medium (either volatile or non-volatile). The input interface is connected to an instruction input unit such as a keyboard or mouse and is an interface for inputting user instructions. The data storage unit stores data such as base model images including montage images and grid templates, and is composed of known storage such as a hard disk drive (HDD) or solid-state drive (SSD). The communication interface is connected to a scanning electron microscope (SEM) or etching apparatus and is an interface for sending and receiving data with other devices, such as captured images and etching conditions. The output interface is connected to a display unit such as a display and is an interface for outputting various display information. The computer device may be any electronic device such as a desktop computer or a notebook computer. In this embodiment, for example, steps 103 (application of a grid template), 104 (measurement of cross-sectional contour lines), 105 (selection of the maximum value), 106 (identification of the section that takes the maximum value), 107 (extraction of the section that takes the maximum value), and 108 (creation of a montage image) can be automatically processed by the processing unit of a computer device.
[0050] Thus, montage images can be estimated to represent ideal cross-sectional contours, serving as an indicator that contributes to improving the accuracy of shape evaluation. Furthermore, they have the advantage of being efficiently created using computer equipment based on cross-sectional observation images.
[0051] [Second Embodiment] (Evaluation of Etching Performance) The second embodiment of the present invention evaluates the performance of etching using a base model image including a montage image created in the first embodiment. Figure 16 is a flowchart showing the steps of evaluating the performance of etching by image recognition sequence processing using the base model image in the second embodiment. The following will be explained in accordance with steps 1601 to 1609 of Figure 16. The hardware configuration is the same as in the first embodiment, so the differences will be explained in detail.
[0052] (Step 1601) An etching test is performed on the sample to be processed under etching conditions different from those of the first embodiment. However, the sample to be processed is the same as the one used when creating the montage image in the first embodiment, and has the same mask pattern arrangement. However, the processing test may be performed under the same etching conditions as the first embodiment.
[0053] (Step 1602) Obtain a cross-sectional observation image of the object to be measured, similar to the first embodiment (the cross-sectional observation image obtained to evaluate the etching shape is also called the "evaluation cross-sectional observation image"). The observation environment, such as the observation range and observation magnification, shall be the same as that of the first embodiment.
[0054] (Step 1603) The cross-sectional observation images, consisting of any number of images obtained in Step 1602, are stored in the data storage unit of the computer device.
[0055] (Step 1604) Image recognition sequence processing is started in the processing unit of the computer device.
[0056] (Step 1605) First, the montage image created in the first embodiment and saved as the base model image (Step 109) is read from the data storage unit. Similarly, the cross-sectional observation image saved in Step 1603 is also read.
[0057] (Step 1606) Next, the similarity between the base model image and the cross-sectional observation image is determined. The method for determining similarity is not particularly limited, and it is also possible to employ AI (artificial intelligence) image recognition technology. If it is determined that there is no similarity, that is, that no image of a cross-sectional contour line with equivalent features can be detected, return to step 1602 and repeat the cleavage to obtain a cross-sectional observation image. Alternatively, return to step 1601 and repeat the processing test. The following are possible reasons for determining that there is no similarity: (1) The etching conditions could not reach the desired design dimensions. (2) The cross-sectional observation image was not cleaved near the center position (plane of symmetry) of the mask pattern, but was taken at a large deviation from the center position or did not capture the deep hole pattern. (3) The cross-sectional area was soiled or damaged, making the cross-sectional contour line unclear. (4) The deep hole area was blocked by deposits, etc.
[0058] If it is determined that images with similar cross-sectional contour lines, i.e., images with equivalent features, are detected, images composed of sections containing similar cross-sectional contour lines (referred to as "similar cross-sectional pattern images") are identified and extracted along with their locations. Figure 17 is a schematic diagram showing how similar cross-sectional pattern images are identified and extracted from a montage image. Multiple similar cross-sectional pattern images may be extracted.
[0059] (Step 1608) With respect to the similar cross-sectional pattern images extracted in Step 1607, the width and depth of the cross-sectional contour lines in all grid sections are measured, similar to Step 104 of the first embodiment.
[0060] (Step 1609) Similar cross-sectional pattern images can be used as new base model images because, like montage images, they can be estimated as ideal cross-sectional contour lines in the plane of symmetry (true cross-section) cleaved at the center of the mask pattern shape. Therefore, similar cross-sectional pattern images are saved in the data storage unit along with the measurement data acquired in step 1608. The saved similar cross-sectional pattern images are then read as base model images in the subsequent image recognition sequence processing (Step 1605).
[0061] In this way, by using AI-based image recognition technology to determine the similarity between cross-sectional observation images obtained from etching processing tests and montage images as base model images, the performance of the etching process can be evaluated efficiently in a short time, and the influence of differences in human experience can be reduced. Furthermore, similar cross-sectional pattern images determined to be similar to the montage image or existing base model image can be used as new base model images to update the AI training data, thus improving the accuracy of similarity determination in AI-based image recognition sequence processing.
[0062] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above, and various modifications are possible without departing from the spirit of the present invention.
[0063] The following describes embodiments that may constitute the present invention, but are not limited thereto. (Etchology 1) An etching shape evaluation system comprising a computer device, wherein the computer device comprises a processing unit and a data storage unit, the data storage unit stores a base model image, the processing unit acquires one or more cross-sectional observation images for creation captured by a scanning electron microscope, applies a grid template to the cross-sectional observation images for creation, identifies an image (material image) within a section of the grid template corresponding to each depth position where the width of the cross-sectional contour line is the maximum value, extracts the material image, stitches it together to create a montage image, and the data storage unit stores the montage image as a base model image. (Etchology 2) The etching shape evaluation system according to Embodiment 1, wherein the vertical size of the sections of the grid template can be changed according to the accuracy of the performance evaluation. (Etchology 3) The etching shape evaluation system according to Embodiment 1 or 2, wherein the mask layer of the processing sample to be evaluated for etching shape has a circular mask pattern formed on it. (Aspect 4) An etching shape evaluation system according to any one of aspects 1 to 3, wherein the processing unit acquires one or more evaluation cross-sectional observation images captured by a scanning electron microscope and determines the similarity between a base model image and the evaluation cross-sectional observation images. (Aspect 5) An etching shape evaluation system according to aspect 4, wherein the base model image includes a similar cross-sectional pattern image extracted from the evaluation cross-sectional observation images that were determined to be similar in the determination. (Aspect 6) An etching shape evaluation system according to aspect 4 or 5, wherein the determination is performed by image recognition using artificial intelligence (AI).(Aspect 7) An etching shape evaluation method comprising: in the processing unit of a computer device, acquiring one or more cross-sectional observation images for creation captured by a scanning electron microscope; applying a grid template to the cross-sectional observation images for creation; identifying an image (material image) within a section of the grid template corresponding to each depth position where the width of the cross-sectional contour line is the maximum value; extracting the material image and stitching it together to create a montage image; and saving the montage image as a base model image in the data storage unit of the computer device. (Aspect 8) An etching shape evaluation method according to aspect 7, wherein in the processing unit, acquiring one or more cross-sectional observation images for evaluation captured by a scanning electron microscope; and determining the similarity between a base model image and the cross-sectional observation images for evaluation.
[0064] 201...Sample for processing, 202...Mask layer, 203...Etching layer, 204...Cross-sectional contour line, 205...Dimensional width (CD), 206...Depth, 501...Grid template A, 601...Grid template B, 1101-1104...Mask pattern, 1201...Top surface, 1202-1206...Bottom surface, 1501...Montage image
Claims
1. An etching shape evaluation system comprising a computer device, wherein the computer device comprises a processing unit and a data storage unit, the data storage unit stores a base model image, the processing unit acquires one or more cross-sectional observation images for creation captured by a scanning electron microscope, applies a grid template to the cross-sectional observation images for creation, identifies a material image which is an image within the section of the grid template corresponding to each depth position where the width of the cross-sectional contour line is at its maximum, extracts the material image and stitches it together to create a montage image, and the data storage unit stores the montage image as the base model image.
2. An etching shape evaluation system according to claim 1, wherein the size of the compartments of the grid template in the vertical direction can be changed according to the accuracy of the etching shape evaluation.
3. An etching shape evaluation system according to claim 1, wherein the etching shape to be evaluated is a shape etched using a hole mask pattern.
4. Etching shape evaluation system according to claim 1, wherein the processing unit acquires one or more evaluation cross-sectional observation images captured by a scanning electron microscope and determines the similarity between a base model image and the evaluation cross-sectional observation images.
5. An etching shape evaluation system according to claim 4, wherein the base model image includes a similar cross-sectional pattern image extracted from evaluation cross-sectional observation images that have been determined to be similar in the above determination.
6. An etching shape evaluation system according to claim 4, characterized in that the determination is performed by image recognition using artificial intelligence (AI).
7. An etching shape evaluation method comprising: in the processing unit of a computer device, acquiring one or more cross-sectional observation images for creation captured by a scanning electron microscope; applying a grid template to the cross-sectional observation images for creation; identifying a material image that is an image within the section of the grid template corresponding to each depth position where the dimension width of the cross-sectional contour line is at its maximum; extracting the material image and stitching it together to create a montage image; and saving the montage image as a base model image in the data storage unit of the computer device.
8. An etching shape evaluation method according to claim 7, wherein the processing unit acquires one or more evaluation cross-sectional observation images captured by a scanning electron microscope, and determines the similarity between a base model image and the evaluation cross-sectional observation images.