Image inspection device and image inspection method

Abstract

PURPOSE: To provide an image inspection device and an image inspection method, in which an original restored image is generated from an image obtained from an SEM device or the like or from a geometrical feature amount generated from design data or the like, on the basis of the geometrical feature amount, the restored image is measured, and a pattern having a complicated shape is measured with high accuracy, rapidly, and with less damage.CONSTITUTION: An image inspection device is provided with: means that pairs an image generated from design data on a pattern formed on a sample with an SEM image generated by detecting a discharged secondary electron or a reflection electron reflected, by scanning while narrowing and emitting an electron beam on a pattern formed on the sample on the basis of the design data, so as to deeply learn a geometrical feature; and restoring means that generates an image having a desired geometrical feature from the geometrical feature amount obtained through the deep learning, which is configured to extract the desired geometrical feature amount from the SEM image to generate an image, on the basis of the geometrical feature amount.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an image inspection apparatus and an image inspection method for measuring dimensions based on images of secondary electrons or reflected electrons emitted by irradiating an electron beam onto a sample. [Background technology]

[0002] As microfabrication technology for semiconductor devices advances, recently EUV exposure technology has been used to form small feature sizes of less than 10 nm on silicon wafers. These semiconductor microfabrication processes are mainly carried out using lithography technology. Lithography is achieved by coating a photosensitive resist on the wafer and transferring a fine pattern on a photomask onto the wafer surface using DUV or EUV light. The limit of microfabrication is roughly determined by the wavelength used, so wavelengths are becoming shorter, from DUV light (193 nm) to EUV light (13.5 nm).

[0003] The exposed photosensitive resist is developed, washed, and annealed to leave a fixed pattern on the wafer. For example, if there is an L&S pattern consisting of straight lines on the photomask, an L&S pattern consisting of straight lines will be formed on the wafer. By further etching, wiring, gates, holes, etc. made of the underlying material can be formed. As described above, semiconductor devices are manufactured through a process in which the design data for the semiconductor circuit is transferred successively onto silicon.

[0004] Meanwhile, in the semiconductor industry, the development of semiconductor exposure equipment using short wavelengths such as EUV has been delayed, so a technology called OPC (optical proximity correction) has been introduced to perform fine processing that exceeds the exposure wavelength. This technology allows the targeted shape to be formed on the wafer by incorporating shape changes caused by the process into the pattern in advance. With the introduction of this technology, the patterns on photomasks, which were previously simple, have become extremely complex. For example, the line width of L&S, which should be constant, now becomes wider or narrower in a local range of several tens of nanometers. CD measurement, which guarantees the final dimensions of the photomask, requires accurate measurement of the distances in these narrow ranges.

[0005] On the other hand, resist is mainly made of polymers and has a molecular weight of over 1000. Therefore, the resist pattern formed on the wafer reflects the molecular size or molecular shape, so if the pattern size formed on the wafer is sufficiently large compared to the molecular size, it appears to be transferred in a straight line, but when the transferred pattern size becomes small enough to be comparable to the molecular weight, the molecular shape is reflected on the pattern edge, and the effect of uneven components appears rather than a pure straight line (edge ​​roughness). The unevenness of the resist edge can reach several nm, which is a major obstacle to pattern inspection, which requires CD measurement with sub-nm accuracy and defect detection sensitivity on the nm order.

[0006] Furthermore, electron beams inherently have a noise called shot noise. This noise appears as a salt and pepper pattern when the number of electrons per pixel is small. Therefore, if the electron beam is not irradiated sufficiently, the image will have a rough texture due to the shot noise in addition to the variation in the local shape of the material. These are also major obstacles to local CD measurement. DISCLOSURE OF THEINVENTION [Problem to be solved by the invention]

[0007] In conventional CD measurements, even when the pattern edge contains uneven components, it is possible to measure CD with a high reproducibility accuracy of the order of 0.1 nm by performing averaging processing for patterns with a linear region that is sufficiently wide relative to the electron beam spot size and scanning range.

[0008] However, because the patterns formed on photomasks and wafers used in cutting-edge device manufacturing have complex shapes, the width of the measurement area where the line width can be considered constant may be only 10 nm or less, making it impossible to obtain a sufficient number of measurement points for averaging. As a result, even a slight deviation in the landing point of the electron beam results in a different measured length, making it difficult to measure dimensions with high precision.

[0009] Furthermore, in order to measure or inspect a pattern having uneven pattern edges on the order of nm, it is necessary to scan the electron beam with high resolution, which poses the problem of long inspection times.

[0010] In addition, images inherently contain shot noise due to the scanning electron beam, so in order to reduce the noise, it is necessary to increase the current, which can damage the sample, lengthen the image acquisition time, and reduce the resolution. [Means for solving the problem]

[0011] In order to solve the above-mentioned problems, the present invention generates an original restored image from an image acquired from an SEM device or the like or from its geometric features based on ideal geometric features generated from design data, etc., and measures the restored image, thereby achieving high-precision, rapid measurement with minimal damage of patterns having complex shapes.

[0012] Therefore, the present invention provides an image inspection device that performs length measurements based on images of secondary electrons or reflected electrons, etc. emitted by irradiating an electron beam onto a sample, and the device is provided with design data for a pattern to be formed on the sample, an image generated by irradiating and scanning a finely focused electron beam onto a pattern formed on the sample based on the design data, detecting the emitted secondary electrons or reflected electrons, etc., and a restoration means for generating a restored image from geometric features generated from the image based on ideal geometric features generated from the design data, and the restored image is generated from the image based on the ideal geometric features.

[0013] At this time, the design data is CAD data of the pattern to be formed on the sample.

[0014] Moreover, the geometric feature amount is set to be a geometric feature amount for each type of pattern to be measured.

[0015] The type of pattern is set to be either L&S, holes, or dots.

[0016] In addition, the ideal geometric feature values ​​generated from the design data are ideal geometric feature values ​​generated based on the design data, or pseudo ideal geometric feature values ​​generated based on the design data and noise.

[0017] In addition, the pseudo-ideal geometric feature is calculated so that the similarity between the pseudo-ideal geometric feature and the ideal geometric feature is equal to or less than a predetermined threshold, or the average of a plurality of pseudo-ideal geometric feature values ​​calculated based on different noises is equal to or less than a predetermined threshold. Effect of the Invention

[0018] The present invention generates an original restored image from an image acquired by an SEM device or the like based on ideal geometric feature values ​​generated from design data, etc., and measures the length of the restored image, enabling highly accurate, rapid, and less-damage measurement of patterns with complex shapes on a sample. It also enables faster pattern inspection. Example 1

[0019] FIG. 1 is a diagram illustrating the configuration concept of the present invention.

[0020] In FIG. 1, SEM image 1 is a secondary electron image of a pattern formed on a sample obtained from a scanning electron microscope (SEM).

[0021] CAD data 2 is one type of design data, and is publicly known data (GDSII, OASIS, etc.) of a pattern to be formed on a sample (e.g., a mask, a wafer).The CAD data 2 is used to generate ideal geometric features of the pattern (described later using Figures 2 to 7, etc.).

[0022] The pseudo data generation unit 3 generates pseudo-ideal geometric features in place of the CAD data 2, for example, by using known GAN technology (generative adversarial network, CGAN, pix2pix, cycleGAN, SAGAN, bigGAN, self-supervised GAN, etc.) to generate pseudo-ideal geometric features (described later).

[0023] The learning unit 4 creates learning data (see FIG. 4, for example) that associates the CAD data 2 or data from the pseudo data generating unit 3 with the SEM image 1, and registers it as learning data 5. In registration, the SEM image and the image generated from the CAD data 2 or the image generated by the pseudo data generating unit 3 are registered in association with the type of pattern (described later).

[0024] The learning data 5 is a registered data in which, for each type of pattern, an SEM image is associated with an image generated from the CAD data 2 or an image generated by the pseudo data generating unit 3 (see FIGS. 4 and 5, etc.).

[0025] The estimation unit 6 restores a restored image 8 and a restored profile 9 that are free of or have reduced noise, vibration, blur, and noise due to insufficient signals, based on the input SEM image and with reference to the learning data 5 (described later). The aforementioned GAN technology is also used for this series of image conversions.

[0026] The statistical processing unit 7 reduces variations and the like by performing statistical processing (for example, averaging processing) on ​​the restored image 8 and restored profile 9 restored by the estimation unit 6. This averaging processing can improve the reproducibility of measurements such as CD measurements.

[0027] The restored image 8 and restoration profile 9 are the patterns in the SEM image that have been restored to an image 8 and restoration profile 9 that are free of or have reduced noise, vibration, blur, and noise due to insufficient signals, based on the learning data 5. Based on these restored images 8 and restoration profiles 9, the dimensions of the patterns (L&S, holes, dots, etc.) are precisely measured and compared with the design data to determine whether they pass or fail (outside the acceptable range).

[0028] The operation of the configuration in FIG. 1 will now be described.

[0029] (1) Based on the CAD data 2, an image with ideal geometric features (an image without or with reduced noise due to noise, vibration, blur, or insufficient signal, for example, a GDSII image, see Figure 4) is generated in association with the type of pattern to be formed on a mask or wafer (L&S, holes, dots, etc.), and then an SEM image is acquired from a mask or wafer on which the same pattern is formed, and the two are associated and registered as learning data 4 (see Figures 4 and 5). Similarly, the pseudo data generator 3 uses GAN technology or the like to generate an image with pseudo ideal geometric features, and associates it with the SEM image and registers it as learning data 5 (described later with reference to Figures 13 to 18).

[0030] The above registration process is carried out in advance to accumulate learning data 5.

[0031] (2) Next, an SEM image of the pattern to be subjected to CD measurement is obtained using a scanning electron microscope or the like.

[0032] (3) The estimation unit 6 restores the SEM images acquired in (2) to an image (restored image) without or with reduced noise, vibration, blur, and noise due to insufficient signals, for each type of pattern, based on an image having ideal geometric features for the corresponding type (or ideal geometric features) by referring to the learning data 5. Then, it outputs it as a restored image 8 or restoration profile 9. Since there is fluctuation in the image obtained by deep learning, the image is created multiple times under the same conditions, and the images are statistically processed so that the fluctuation is less than a desired value. Based on the output restored image 8 and restoration profile 9, precise length measurement is performed for each type of pattern, and the pattern is compared with the design data to determine whether it passes or fails (outside a predetermined tolerance range).

[0033] The following will explain in detail.

[0034] FIG. 2 shows a flowchart (part 1) of the registration of learning data according to the present invention.

[0035] 2, S1 acquires a learning image file, which is design data such as the CAD data 2 in FIG.

[0036] S2 converts from GDSII to BMP etc. This converts, for example, GDSII (vector data), which is one type of design data (CAD data 1) acquired in S1, into BMP format data for image display.

[0037] S3 acquires an SEM image using a scanning electron microscope from the mask and wafer on which the pattern acquired in S2 is formed.

[0038] S4 generates a line image by extracting the contour lines of the SEM image 1 acquired in S3 (GAN technology using deep learning can be used for this image conversion).

[0039] In step S5, the type of pattern is identified. The type of pattern (L&S, hole, or dot) is identified from the line image (image of contour lines) generated in step S4.

[0040] S6 registers the SEM image and the line image according to the specified type. For example, as shown in Fig. 4 and Fig. 5 described later, the SEM image is registered in one-to-one correspondence with the image converted by S2 (GDSII image) for each pattern type (L&S, hole, dot type) and further for each material 1, 2, N, as shown in Fig. 4 and Fig. 5.

[0041] As a result of the above, images (e.g., GDSII images) generated from CAD data 2 and SEM images can be registered by type (L&S, holes, dots) and by material, as shown in Figures 4 and 5, making it possible to generate learning data 5.

[0042] FIG. 3 shows a flowchart (part 2) for registering learning data according to the present invention. This is a flowchart for performing new learning when a new pattern is discovered. For example, OPC is also called a subfeature. In order to accurately create a square pattern on a wafer, it is necessary to deform the shape of the four corners of the square so that the corners do not become rounded during the process. For example, small squares are added to the four corners. The shape and size of these subfeatures change depending on the type of process used. Therefore, when a new process is changed, a pattern of a new shape is learned in response to the change and is additionally registered as learning data 5.

[0043] 3, S11 judges whether it is time to learn. As described above, this judges whether a new OPC has been added or changed, and if YES, it is judged that it is time to learn, and executes S12 and subsequent steps. If NO in S11, the process ends.

[0044] In step S12, learning is performed for each pattern using the learning data stored for each pattern, and learning data is generated (learning data in which the SEM images and GDS2 images described in FIG. 2 are associated with each other by type and by material).

[0045] In step S13, learning data is registered for each pattern. As a result, SEM images and GDSII images of newly modified or added patterns (e.g., OPC) are additionally registered for each pattern and material in Fig. 4 and Fig. 5.

[0046] In step S14, it is determined whether the process is to end. If the answer is YES, the process ends. If the answer is NO, the process from step S11 onward is repeated.

[0047] As a result of the above, when additions or changes such as OPC are made, new SEM images and GDSII images are added and registered for each pattern and material in the learning data 5 in Figures 4 and 5.

[0048] FIG. 4 shows an example of learning data (image file) of the present invention.

[0049] In Figure 4, the SEM images are SEM images acquired for each corresponding pattern and material using a scanning electron microscope, etc. GDSII images are images (images with ideal geometric features) generated from CAD data2, etc., and are images generated for each pattern and material. The patterns are patterns formed on a mask or wafer, and here, there are three types: L&S, holes, and dots. In addition, other types (for example, OPC) may be added.

[0050] Materials 1, 2, and N are types of materials formed on the mask and wafer, and N types of materials 1, 2, and N are assumed here.

[0051] The pattern material images are images for each pattern and each material.

[0052] Here, the learning data (image files) consist of a pair of images: an SEM image and an image generated from the GDSII file that is the source of the pattern appearing in the SEM image. A hierarchy is set for each of these images (image files), and here, for example, a hierarchy is created by dividing them by pattern type. A hierarchy is then created by dividing them by material. A hierarchy can also be created by OPC type.

[0053] FIG. 5 shows a schematic example of learning data (image file) of the present invention.

[0054] In Fig. 5, (a) is a schematic representation of a GDSII image and an SEM image of L&S (line and space). The GDSII image is a line image of a dot image generated from design data. The SEM image is a grayscale image, as it is a secondary electron image of the same pattern obtained by a scanning electron microscope or the like.

[0055] Similarly, in FIG. 5(b) a hole and in FIG. 5(c) a dot are also displayed diagrammatically as shown.

[0056] As described above, the contents of the GDSII file, which is the design data that formed the pattern of the SEM image on the right, can be converted into a BMP image and represented diagrammatically as the GDSII image on the left, and the pair of these two images is registered as learning data 5.

[0057] Fig. 6 shows a flowchart (part 3) of the learning data registration process of the present invention. Fig. 6 shows a flowchart of the size calibration of SEM images and pseudo SEM images.

[0058] 6, S21 acquires an SEM image, which is a secondary electron image from a scanning electron microscope or the like.

[0059] In step S22, a pseudo SEM image is acquired. This is a pseudo SEM image generated from the CAD data 2 in FIG.

[0060] S23 finds the same shape.

[0061] S24 normalizes the pseudo SEM image. These steps 23 and S24 find corresponding pattern shapes for each pattern type (L&S, hole, dot type) in the pseudo SEM image to be normalized, and normalize the found patterns of the same type and shape so that the length, width, area, etc. of the SEM image and the pseudo SEM image match.

[0062] As a result, by standardizing and matching a pseudo-SEM image generated from CAD data 2 with an SEM image for the same pattern shape, it is possible to eliminate or minimize the error between the two and improve the dimensional accuracy of the restored image 8.

[0063] Fig. 7 shows a flowchart (part 4) of the learning data registration according to the present invention. Fig. 7 shows a flowchart of the size calibration of the SEM image and the line image.

[0064] 7, S31 acquires an SEM image, which is a secondary electron image from a scanning electron microscope or the like.

[0065] In step S32, a line image is acquired, which is generated from the CAD data 2 in FIG.

[0066] S33 finds the same shape.

[0067] S34 normalizes the line image. These S33 and S34 find corresponding pattern shapes that are the same shape for each pattern type (L&S, hole, dot type) for the line image to be normalized, and normalize the found patterns of the same shape and type so that the length, width, area, etc. of the SEM image and the line image match.

[0068] As a result, by standardizing and matching the line image generated from CAD data 2 with the SEM image for the same pattern shape, it is possible to eliminate or minimize the error between the two and improve the dimensional accuracy of the restored image 8.

[0069] Fig. 8 shows a restoration flowchart based on learning data according to the present invention. Fig. 8 shows a flowchart for generating a pseudo SEM image from CAD data (GDSII image file) and generating a restored image free of noise, etc., for the SEM image based on this pseudo SEM image.

[0070] 8, S41 acquires CAD data (design data). This is data such as GDSII and OASIS as the CAD data 2 in FIG. 1 described above.

[0071] In step S42, a pseudo SEM image is generated. For example, a pseudo SEM image shown in FIG. 9(b) is generated based on the CAD data shown in FIG. 9(a) (described later) acquired in step S41. The pseudo SEM image is generated by: Line brightness Brightness of the space Brightness of each edge The width of each edge - Corner rounding Based on such attribute data, a pseudo-SEM image with contrast is generated, as shown in Figure 9(b).

[0072] In step S43, geometric features are extracted from the pseudo SEM image generated in step S42 (feature values ​​such as geometric shapes and dimensions for each pattern type).

[0073] Similarly, in S44 and S45, geometric features are extracted from an SEM image acquired from a scanning electron microscope or the like.

[0074] In step S46, a geometric feature included in the pseudo SEM image is extracted from the feature of the SEM image. For example, the width (the width indicated by two short thick vertical lines) of the L (line) of the measurement target in the type (L&S) is extracted as the feature of the SEM image in FIG. 10(a) described later.

[0075] In step S47, the SEM image is restored from the extracted feature amount. This is done by setting (replacing) the feature amount (the line width (the width indicated by the two short thick vertical lines) in FIG. 10(a)) extracted in step S46 to the corresponding part of the pseudo SEM image, and generating a restored SEM image.

[0076] The details are explained below. (1) CAD data, for example a GDSII file, is data that represents polygons, so it is used to first generate a pseudo-SEM image (see Figure 9 (b)). SEM images have parameters such as the brightness of lines, the brightness of spaces, the brightness of each edge, the width of each edge, and the amount of corner rounding, which indicates how rounded the corners are, so a pseudo-SEM image is generated by specifying these. It is also possible to directly generate this pseudo-SEM image itself using the GAN technology described below.

[0077] (2) Extract the features of the generated pseudo-SEM image.

[0078] (3) Next, an SEM image having a corresponding pattern obtained using a scanning electron microscope or the like is obtained, and its geometric features are extracted.

[0079] (4) Then, the geometric features contained in the pseudo-SEM image are extracted from the geometric features contained in the SEM image.

[0080] (5) The features extracted in (4) are set (replaced) in the pseudo-SEM image to restore a very clear restored image free of noise. This is equivalent to replacing the local texture of the actual SEM image with the texture of the ideal SEM image. This results in an ideal SEM image (restored image) with no unevenness in the edge areas.

[0081] (6) To give a specific example, as shown in Figure 10, the original SEM image to which OPC has been added is represented by a complex curve (Figure 10(a)), but this can be replaced with a simple straight line as shown in Figure 10(b). Although it is difficult to accurately measure distance with a complex curve, the simple figure obtained by image transformation (restoration) has a wider measurement area and at the same time retains the properties of the original image as a whole, making it easy to achieve stable line width measurement.

[0082] (7) By evaluating the obtained image using a conventional CD measurement method, it is possible to measure and evaluate distances, etc. on a noise-free restored image.

[0083] FIG. 9 is an explanatory diagram of an example of conversion from CAD data to an SEM image according to the present invention.

[0084] FIG. 9(a) shows an example of CAD data, where white indicates lines and black indicates spaces.

[0085] Figure 9(b) shows an example of an SEM image (pseudo SEM image). This is a pseudo SEM image with shading that was generated based on the CAD data in Figure 9(a) and reflects the attribute data (brightness of lines, brightness of spaces, brightness of each edge, width of each edge, and corner rounding).

[0086] FIG. 10 is an explanatory diagram of the restoration of an SEM image from feature quantities according to the present invention.

[0087] Figure 10(a) shows an example of an SEM image (actual image). This is a schematic example of an SEM image acquired with a scanning electron microscope, in which a line has become distorted due to OPC. Here, the width of two thick and short vertical lines are the measurement points specified as the line width. It is extremely difficult to directly measure the width of a line that has been bent due to OPC, as in Figure 10(a).

[0088] Figure 10(b) shows a schematic example of a restored SEM image (restored image). This is achieved by replacing (restoring) the curved parts of the line in the SEM image in Figure 10(a) with the straight lines in Figure 10(b) (straight lines in a pseudo SEM image generated from CAD data), and as a result, a restored image is generated in which the measurement points of the line (the width of the two short, thick vertical lines) are set to the corresponding parts of the line in the pseudo SEM image. By making measurements on this restored image, it is possible to perform highly accurate measurements on the straight, unbent parts of the line.

[0089] FIG. 11 shows an explanatory diagram (part 2) of the restoration of an SEM image from the feature amount of the present invention.

[0090] Figure 11(a) shows an example of roughness. The left side is an SEM image obtained by a scanning electron microscope. The two on the right side are a pseudo SEM image and a line drawing (line image) generated from CAD data 2.

[0091] Similarly, (b) of FIG. 11 shows an example of OPC, (c) of FIG. 11 shows an example of noise, (d) of FIG. 11 shows an example of vibration, (e) of FIG. 11 shows an example of defocus, and (f) of FIG. 11 shows an example of increasing the resolution.

[0092] Fig. 12 shows a restoration flowchart (part 2) based on learning data according to the present invention. Fig. 12 shows the pseudo SEM image of S42 in Fig. 8 already described, which has been changed to the line image of S52. The line image is a BMP image generated based on CAD data (e.g., GDSII), so for other explanations, refer to Fig. 8.

[0093] In Fig. 12, the line image of S52 is a BMP image generated based on CAD data (for example, GDSII). Here, when contour extraction is performed from a GDSII file in particular, the contour is extracted according to the geometric shape expressed on the GDS. Since there is no edge roughness in a GDS file, the converted image does not have the roughness caused by edge roughness or shot noise. In other words, it is converted into an image with a very high SNR. Therefore, it has the characteristic that the length and area of ​​any local location can be easily and accurately measured.

[0094] Deep learning requires a large amount of data, but that data can be generated using GAN technology. Fig. 13 is an explanatory diagram of the generation of learning data using the GAN technology of the present invention. Fig. 13 shows an example in which the same initial conditions are given to one GAN generator to generate multiple images, which are then statistically processed.

[0095] 13, S61 inputs a noise+SEM image, which is one SEM image and noise for this image are sequentially generated and input by the number of images to be generated.

[0096] In S62, the GAN generator sequentially generates (generates images 1, 2, 3, ... N) pseudo SEM images (SEM images that are mistaken for real SEM images) based on one SEM image input in S61 and sequentially input noise. In this case, the same SEM image is used as the input image, and images 1, 2, 3 ... B are generated with only the noise being different.

[0097] S63 stores images 1, 2, 3, . . . N generated in S62 in a temporary storage device.

[0098] S64 performs statistical processing. This applies statistical processing (e.g. averaging) to images 1, 2, 3, ... N temporarily stored in S63 to reduce variation and create a single image. For example, when averaging 10 images, the image fluctuation can be reduced to roughly one-third of the original image. Unlike actual SEM images, the amount of images generated by the GAN generator depends on the computing power of the computer, so the number of N can be increased as much as necessary until it falls within the required statistical error, which effectively greatly reduces the fluctuation of images generated by the GAN generator (technology).

[0099] As described above, it is now possible to generate multiple pseudo SEM images (pseudo SEM images with a similarity to a real SEM image that can be mistaken for a real SEM image) based on one SEM image and multiple noises, and further reduce the variability (fluctuation) caused by the GAN generator by performing statistical processing (e.g., averaging processing) to generate pseudo SEM images that are more similar to the real SEM image.Then, it becomes possible to input the generated pseudo SEM images to the learning unit 4 in Figure 1 to automatically generate learning data (learning data 5 consisting of pairs of SEM images and pseudo SEM images generated by GAN technology).

[0100] Fig. 14 is an explanatory diagram (part 2) of generating training data using the GAN technology of the present invention. Fig. 14 shows an example in which the same initial conditions are given to multiple GAN generators to generate multiple images, which are then statistically processed.

[0101] 14, S71 inputs a noise+SEM image. This sequentially generates one SEM image and corresponding noise for the number of images to be generated, and outputs them to GAN generators 1, 2, 3, ...N, respectively.

[0102] S72 outputs pseudo SEM images and images 1, 2, 3, ...N based on the SEM images input to GAN generators 1, 2, 3, ...N and different noises 1, 2, 3, ...N, respectively. The pseudo SEM images are SEM images that are similar enough to be distinguished from real SEM images.

[0103] A step S73 performs statistical processing.

[0104] S74 outputs the image. S73 and S74 perform statistical processing (e.g., averaging) on ​​images 1, 2, 3, ... N output by S72 to reduce variation and output a single image (restored image). For example, when averaging 10 images, the image fluctuation can be reduced to about one-third of the original image. Unlike actual SEM images, the amount of images generated by the GAN generator depends on the computing power of the computer, so the number of N can be increased as much as necessary until it falls within the required statistical error, so the fluctuation of the images generated by the GAN generator (technology) can be effectively reduced significantly.

[0105] As described above, it is possible to generate multiple pseudo SEM images (pseudo SEM images having a similarity to a real SEM image) in parallel using multiple GAN generators based on one SEM image and multiple noises, and further reduce the variation (fluctuation) caused by the GAN generator by performing statistical processing (for example, averaging processing) to generate a pseudo SEM image that is more similar to the real SEM image. Then, it is possible to input this generated pseudo SEM image to the learning unit 4 in FIG. 1 to automatically generate learning data (learning data 5 consisting of a pair of an SEM image and a pseudo SEM image generated by GAN technology). As shown in FIG. 1, it is also effective to perform the statistical processing shown in this embodiment on an estimated image. The accuracy of the final output image can be increased, which helps improve the measurement reproducibility in CD measurement, etc.

[0106] Fig. 15 shows an explanatory diagram (part 3) of generating learning data using the GAN technology of the present invention. Fig. 15 shows an explanatory diagram in which a low-resolution SEM image and a high-resolution SEM image are prepared, a pseudo (ideal) high-resolution SEM image that is similar enough to be mistaken for a real high-resolution SEM image is generated by a GAN generator, and these low-resolution SEM images and the pseudo (ideal) high-resolution SEM image are paired and registered in learning data 5 (see Fig. 4).

[0107] In FIG. 15, S81 inputs a low-resolution SEM image.

[0108] S82 inputs a high-resolution SEM image. S91. S92 inputs a low-resolution SEM image and a high-resolution SEM image, which is training data (ideal data), into the GAN generator.

[0109] In S83, the GAN generator generates and outputs a pseudo (ideal) high-resolution SEM image. This is because the GAN generator generates a pseudo high-resolution SEM image (a pseudo high-resolution SEM image having a similarity to the real high-resolution SEM image) for the low-resolution SEM image input in S81, using the high-resolution SEM image input in S82 as a teacher (ideal) high-resolution SEM image.

[0110] S84 outputs a pseudo high-resolution SEM image.

[0111] As a result of the above, it is possible to input a low-resolution SEM image and output a pseudo high-resolution SEM image using GAN technology with that high-resolution SEM image as a teacher (ideal high-resolution SEM image).By registering a pair of the low-resolution SEM image and the pseudo high-resolution SEM image as learning data 5 (Fig. 4), as described above, it becomes possible to generate a high-resolution SEM image by restoring it from the input low-resolution SEM image (called super-resolution).

[0112] Fig. 16 is an explanatory diagram (part 4) of generating learning data using the GAN technology of the present invention. Fig. 16 is an explanatory diagram of preparing a low-resolution SEM image and a GDSII or high-resolution pseudo SEM image, generating a pseudo high-resolution SEM image that is similar enough to be mistaken for a real high-resolution SEM image using a GAN generator, and registering the low-resolution SEM image and the pseudo high-resolution SEM image as a pair in learning data 5 (see Fig. 4). .

[0113] In FIG. 16, S91 inputs a low-resolution SEM image.

[0114] S92 inputs a GDSII or high-resolution pseudo-SEM image. These S91. S92 inputs a low-resolution SEM image and a GDS2 file or a high-resolution pseudo-SEM image as training data (ideal data) into the GAN generator.

[0115] In S93, the GAN generator generates and outputs a pseudo high-resolution SEM image. This is because the GAN generator generates a pseudo high-resolution SEM image (a pseudo high-resolution SEM image that is similar enough to be mistaken for a real high-resolution SEM image) for the low-resolution SEM image input in S91, using the GDSII file or high-resolution pseudo SEM image input in S92 as a teacher (ideal) high-resolution SEM image.

[0116] S94 outputs a pseudo high-resolution SEM image.

[0117] As a result of the above, it is possible to input a low-resolution SEM image and output a pseudo high-resolution SEM image using GAN technology, using the GDSII file or a high-resolution pseudo SEM image as a teacher (ideal high-resolution SEM image).By registering a pair of the low-resolution SEM image and the pseudo high-resolution SEM image as learning data (Figure 4), it becomes possible to generate a high-resolution SEM image by restoring it from the input low-resolution SEM image (called super-resolution), as mentioned above.

[0118] FIG. 17 shows an example image created by the GAN technique of the present invention.

[0119] Figure 17(a) shows an example of an original image. This shows an example of a low-resolution SEM image.

[0120] Figure 17(b) shows an example of super-resolution using GAN technology. This super-resolution is an example in which the low-resolution SEM image in Figure 17(a) is restored to the high-resolution image in Figure 17(b). This technology makes it possible to perform inspection using an electron beam with a large spot size, thereby speeding up pattern inspection.

[0121] Figure 18 shows an example of defect learning in the GAN technology of the present invention. This shows a pair of schematic examples of SEM images having defects that appear in SEM images (protrusions, recesses, bridges, cuts, pinholes, residuals, etc. in the figure) and schematic examples of their teacher images (ideal SEM images). Learning from these images enables the pattern inspection device to extract clear images from blurred images, thereby realizing more precise and faster pattern inspection. The various image transformations described above as the embodiments may be performed only once, or the necessary image transformations may be performed in series.

[0122] FIG. 18(a) shows a pair of a schematic example of an SEM image in the case where a “protrusion” is present and a schematic example of its teacher image (ideal SEM image).

[0123] FIG. 18(b) shows a pair of a schematic example of an SEM image in the presence of a "setback" and a schematic example of its teacher image (ideal SEM image).

[0124] FIG. 18(c) shows a pair of a schematic example of an SEM image in the presence of a "bridge" and a schematic example of its teacher image (ideal SEM image).

[0125] FIG. 18(d) shows a pair of a schematic example of an SEM image when a “cut” is present and a schematic example of its teacher image (ideal SEM image).

[0126] FIG. 18(e) shows a pair of a schematic example of an SEM image in the presence of a "pinhole" and a schematic example of its teacher image (ideal SEM image).

[0127] FIG. 18(f) shows a pair of a schematic example of an SEM image in the presence of a "residual" and a schematic example of its teacher image (ideal SEM image). [Brief description of the drawings]

[0128] [Figure 1] FIG. 1 is a conceptual diagram illustrating the configuration of the present invention. [Diagram 2] 1 is a flowchart (part 1) for registering learning data according to the present invention. [Diagram 3] 13 is a flowchart (part 2) of the learning data registration process according to the present invention. [Figure 4] 1 is an example of learning data (image filter) of the present invention. [Diagram 5] 2 is a schematic example of learning data (image file) of the present invention. [Figure 6] 11 is a flowchart (part 3) for registering learning data according to the present invention. [Figure 7] 11 is a flowchart (part 4) for registering learning data according to the present invention. [Figure 8]1 is a flowchart of restoration based on learning data according to the present invention. [Figure 9] FIG. 2 is an explanatory diagram of an example of conversion from CAD data to a SEM image according to the present invention. [Figure 10] FIG. 13 is an explanatory diagram of restoration of an SEM image from feature quantities according to the present invention. [Figure 11] FIG. 2 is an explanatory diagram (part 2) of the restoration of an SEM image from the feature quantities of the present invention. [Figure 12] 13 is a second flowchart of restoration based on learning data according to the present invention. [Figure 13] FIG. 1 is an explanatory diagram of the generation of learning data using the GAN technology of the present invention. [Figure 14] FIG. 2 is an explanatory diagram (part 2) of the generation of training data using the GAN technology of the present invention. [Figure 15] FIG. 3 is an explanatory diagram (part 3) of the generation of training data using the GAN technology of the present invention. [Figure 16] FIG. 4 is an explanatory diagram (part 4) of generating training data using the GAN technology of the present invention. [Figure 17] 1 is an example of an image created by the GAN technology of the present invention. [Figure 18] 1 is an example of defect learning in the GAN technology of the present invention. [Explanation of symbols]

[0129] 1: SEM image 2:CAD data 3: Pseudo data generation section 4: Learning Department 5: Training data 6: Estimation part 7: Statistical processing section 8:Restored image 9: Recovery Profile

Claims

1. In an image inspection device that uses an image of secondary electrons emitted by irradiating an electron beam on a sample or an image of reflected electrons, A means for deep learning geometric features by pairing an image generated from design data of a pattern to be formed on a sample with a SEM image generated by scanning a pattern formed on the sample based on the design data with a narrow electron beam and detecting emitted secondary electrons or reflected electrons, and a restoration means for generating an image having desired geometric features from the geometric feature amount obtained by the deep learning. Equipped with an image inspection device that generates an image by extracting desired geometric features from the SEM image based on the geometric features.

2. 2. The image inspection device according to claim 1, wherein the image is generated using a GAN based on design data of a pattern to be formed on the sample.

3. 3. An image inspection apparatus, characterized in that the throughput of the inspection apparatus is improved by utilizing the image obtained according to claim 1 or 2.

4. 4. The image inspection device according to claim 1, further comprising a learning device configured to output a profile to be used for length measurement based on the extracted and generated image.

5. 5. An image inspection apparatus according to claim 1, wherein the sizes of the pairs of images used for learning are standardized.

6. An image inspection method utilizing an image of secondary electrons emitted by irradiating an electron beam on a sample or an image of reflected electrons, A means for deep learning geometric features by pairing an image generated from design data of a pattern to be formed on a sample with a SEM image generated by scanning a pattern formed on the sample based on the design data with a narrow electron beam and detecting emitted secondary electrons or reflected electrons, and a restoration means for generating an image having desired geometric features from the geometric feature amount obtained by the deep learning. having An image inspection method, comprising: extracting desired geometric features from the SEM image based on the geometric features; and generating an image.

Images