JP2021047677A - Reference image generation device, inspection device, alignment device, and reference image generation method - Google Patents

Abstract

To reduce a lead time before reference image generation.SOLUTION: A reference image generation device comprises: an input unit 401; a learning unit 402; and a reference image generation unit 403. The input unit 401 inputs learning data as a combination of design data for a pattern on a substrate 9 and a learning image obtained by imaging the pattern. The learning unit 402 creates a learned model through learning of a relation between the design data of the pattern and the learning image by deep learning using a learning data set as a collection of learning data. The reference image generation unit 403 generates a reference image from the design data of the pattern by using the learned model. Thus, a lead time before reference image generation can be significantly reduced.SELECTED DRAWING: Figure 5

Description

The present invention relates to a technique for generating a reference image used for comparison with an image of the pattern in a process of manufacturing a substrate having a pattern on the surface.

Conventionally, in the manufacturing process of a semiconductor substrate, a printed wiring board, or the like, an appearance inspection for determining the quality of the substrate is performed. In the visual inspection of a semiconductor substrate, for example, defects are detected by comparing a plurality of dies provided on one substrate. Specifically, the image obtained by imaging one die is used as a reference image, the image obtained by imaging another die is used as an image to be inspected, and the image to be inspected is compared with the reference image to make a difference. To detect. However, the inspection cannot detect the same defect present on multiple dies. Further, the same pattern to be compared may not exist on the substrate such as a photomask or a reticle.

Therefore, in Patent Document 1, with respect to the circuit pattern formed on the substrate of the semiconductor wafer, CAD created from the SEM image obtained by capturing the circuit pattern with a scanning electron microscope (SEM) and the design data (CAD data) of the circuit pattern. A technique for detecting defects on an SEM image by comparing with an image has been proposed.

Japanese Unexamined Patent Publication No. 2015-7563

By the way, when the reference image is generated from the design data as described above, it is important to easily generate the reference image in a short time. However, in Patent Document 1, regarding the creation of a CAD image which is a reference image for visual inspection, it is described that CADPloy data is created from CAD data to create a CAD image, but a specific creation method is described. Absent.

In addition, when a technician tries to create a reference image from CAD data, a skilled technician needs to create a creation logic based on experience for each of the various patterns shown by the CAD data, and until the start of creating the reference image. It takes a lot of time (for example, months to years). Furthermore, the quality of the created reference image depends on the proficiency of the technician.

The present invention has been made in view of the above problems, and an object of the present invention is to shorten the lead time until the reference image is generated.

The invention according to claim 1 is a reference image generator that generates a reference image used for comparison with an image of the pattern in a process of manufacturing a substrate having a pattern on the surface, wherein the pattern on the substrate is used. By deep learning using an input unit for inputting learning data, which is a combination of design data and a learning image obtained by imaging the pattern, and a learning data set, which is a set of the learning data. It includes a learning unit that learns the relationship between the pattern design data and the training image to create a trained model, and a reference image generation unit that generates a reference image from the pattern design data using the trained model. ..

The invention according to claim 2 is the reference image generator according to claim 1, wherein the learning data set includes one learning data, the design data of the one learning data, and the one. The training data includes other training data that is a combination with the training image in which the brightness or contrast of the training image is changed.

The invention according to claim 3 is the reference image generator according to claim 1 or 2, wherein the learning data set rotates one learning data and design data of the one learning data. Alternatively, it includes other learning data that is a combination of the inverted design data and the learning image of the one learning data that is rotated or inverted in the same manner as the design data.

The invention according to claim 4 is the reference image generator according to any one of claims 1 to 3, wherein each learning data of the learning data set is provided with the pattern on the substrate. It is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region, and the position of the partial region of each of the learning data in the pattern region is randomly determined.

The invention according to claim 5 is the reference image generator according to any one of claims 1 to 4, wherein each learning data of the learning data set is provided with the pattern on the substrate. It is a combination of the design data of the pattern and the training image in the partial region which is a part of the pattern region, and each position of the pattern region corresponds to the training data of any of the training data sets. It is included in the said partial region.

The invention according to claim 6 is the reference image generator according to any one of claims 1 to 5, wherein a plurality of pattern regions in which the patterns are provided are arranged on the substrate. The learning data set includes a plurality of learning data that are a combination of the design data of the pattern and the training image in each of the plurality of pattern regions.

The invention according to claim 7 is the reference image generator according to any one of claims 1 to 6, wherein the learning data set is on each of a plurality of substrates having the pattern on the surface. It includes a plurality of training data that are a combination of the design data of the pattern and the training image.

The invention according to claim 8 is an inspection apparatus for inspecting a substrate having a pattern on the surface, wherein an imaging unit that images a pattern on the substrate to acquire an image to be inspected and the image to be inspected are inspected. It is provided with an inspection unit for inspecting the presence or absence of defects in the substrate by comparing with the reference image of the pattern generated by the reference image generator according to any one of 1 to 7.

The invention according to claim 9 is an alignment device that adjusts the position of a substrate having a pattern on the surface, and claims an imaging unit that captures a pattern on the substrate to acquire a target image, and the target image. An alignment unit that acquires the position of the substrate as compared with a reference image of the pattern generated by the reference image generator according to any one of 1 to 7, and a position of the substrate acquired by the alignment unit. Based on the above, the substrate moving mechanism for relatively moving the substrate to adjust the position is provided.

The invention according to claim 10 is a reference image generation method for generating a reference image used for comparison with an image of the pattern in a process of manufacturing a substrate having a pattern on the surface, and a) on the substrate. A step of inputting learning data which is a combination of a pattern design data and a learning image obtained by imaging the pattern, and b) deep learning using a learning data set which is a set of the learning data. To create a trained model by learning the relationship between the pattern design data and the training image, and c) to generate a reference image from the pattern design data using the trained model. Be prepared.

The invention according to claim 11 is the reference image generation method according to claim 10, wherein the learning data set includes one learning data, the design data of the one learning data, and the one. The training data includes other training data that is a combination with the training image in which the brightness or contrast of the training image is changed.

The invention according to claim 12 is the reference image generation method according to claim 10 or 11, wherein the learning data set rotates one learning data and design data of the one learning data. Alternatively, it includes other learning data that is a combination of the inverted design data and the learning image of the one learning data that is rotated or inverted in the same manner as the design data.

The invention according to claim 13 is the reference image generation method according to any one of claims 10 to 12, wherein each learning data of the learning data set is provided with the pattern on the substrate. It is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region, and the position of the partial region of each of the learning data in the pattern region is randomly determined.

The invention according to claim 14 is the reference image generation method according to any one of claims 10 to 13, wherein each learning data of the learning data set is provided with the pattern on the substrate. It is a combination of the design data of the pattern and the training image in the partial region which is a part of the pattern region, and each position of the pattern region corresponds to the training data of any of the training data sets. It is included in the said partial region.

The invention according to claim 15 is the reference image generation method according to any one of claims 10 to 14, wherein a plurality of pattern regions in which the patterns are provided are arranged on the substrate. The learning data set includes a plurality of learning data that are a combination of the design data of the pattern and the training image in each of the plurality of pattern regions.

The invention according to claim 16 is the reference image generation method according to any one of claims 10 to 15, wherein the learning data set is on each of a plurality of substrates having the pattern on the surface. It includes a plurality of training data that are a combination of the design data of the pattern and the training image.

In the present invention, the lead time until the reference image is generated can be shortened.

It is a figure which shows the structure of the inspection apparatus. It is a figure which shows the structure of the 1st computer. It is a figure which shows the functional structure realized by the 1st computer. It is a figure which shows the structure of the 2nd computer. It is a figure which shows the functional structure realized by the 2nd computer. It is a figure which shows the flow of the generation of a reference image. It is a figure which shows the flow of the inspection of a substrate. It is a figure which shows the structure of the alignment apparatus. It is a figure which shows the functional structure realized by the 1st computer.

FIG. 1 is a diagram showing a configuration of an inspection device 1 according to a first embodiment of the present invention. The inspection device 1 is, for example, a device that inspects the appearance of the semiconductor substrate 9 (hereinafter, also simply referred to as “board 9”). A pattern is formed on the surface of the substrate 9.

The inspection device 1 includes a device main body 2 that captures an image of the substrate 9, a first computer 3, and a second computer 4. The first computer 3 and the second computer 4 are processing devices including a calculation unit, respectively. The first computer 3 also controls the overall operation of the inspection device 1. The second computer 4 is a reference image generation device that generates a reference image described later. The apparatus main body 2 has an imaging unit 21, a stage 22 for holding the substrate 9, and a stage moving mechanism 23. The image pickup unit 21 takes an image of the pattern on the substrate 9 and acquires an image. The image is, for example, a multi-gradation color image. The stage moving mechanism 23 moves the stage 22 relative to the imaging unit 21.

The image pickup unit 21 includes an illumination unit 211 that emits illumination light, an optical system 212, and an image pickup device 213. The optical system 212 guides the illumination light to the substrate 9 and guides the light from the substrate 9 to the image pickup device 213. The image pickup device 213 converts the image of the substrate 9 imaged by the optical system 212 into an electric signal. The lighting unit 211 includes a lamp such as an LED or a light bulb, and an optical element such as a lens or a reflecting member that regulates the light from the lamp. The optical system 212 includes optical elements such as a plurality of lenses and half mirrors. The image pickup device 213 is, for example, a two-dimensional imaging sensor. The imaging device 213 may be a one-dimensional imaging sensor, in which case imaging is performed while moving the stage 22.

The stage moving mechanism 23 is composed of a ball screw, a guide rail, a motor, and the like. Of course, various mechanisms can be adopted as the stage moving mechanism 23, and for example, a linear motor can be used. When the first computer 3 controls the stage moving mechanism 23 and the imaging unit 21, the stage 22 moves in the horizontal direction and a desired region on the substrate 9 is imaged. The stage 22 is a holding portion for holding the substrate 9. The substrate 9 may be held in various ways. For example, a groove is formed in the stage 22, and suction is performed from the suction port formed in the groove, so that the substrate 9 is attracted to the stage 22. A large number of suction ports may be formed on the stage 22 to suck the substrate 9. The stage 22 may be formed of a porous material, and suction may be performed from the porous material. The stage 22 may hold the substrate 9 by a mechanical mechanism.

FIG. 2 is a diagram showing the configuration of the first computer 3. The first computer 3 is a general computer 3 including a CPU 31, a ROM 32, a RAM 33, a fixed disk 34, a display 35, an input unit 36, a reading device 37, a communication unit 38, a GPU 39, and a bus 30. It has a computer system configuration. The CPU 31 performs various arithmetic processes. The GPU 39 performs various arithmetic processes related to image processing. The ROM 32 stores the basic program. The RAM 33 stores various information. The fixed disk 34 stores information. The display 35 displays various information such as images. The input unit 36 includes a keyboard 36a and a mouse 36b that receive input from the operator. The reading device 37 reads information from a computer-readable recording medium 81 such as an optical disk, a magnetic disk, a magneto-optical disk, or a memory card. The display 35, the keyboard 36a, the mouse 36b, and the reader 37 are connected to the bus 30 via the interface I / F. The communication unit 38 transmits / receives signals to / from other configurations of the inspection device 1 and an external device. The bus 30 is a signal circuit that connects the CPU 31, GPU 39, ROM 32, RAM 33, fixed disk 34, display 35, input unit 36, reader 37, and communication unit 38.

In the first computer 3, the program 811 is read in advance from the recording medium 81 via the reading device 37 and stored in the fixed disk 34. Program 811 may be stored on a fixed disk 34 via a network. The CPU 31 and the GPU 39 execute arithmetic processing while using the RAM 33 and the fixed disk 34 according to the program 811. The CPU 31 and the GPU 39 function as arithmetic units in the first computer 3. In addition to the CPU 31 and the GPU 39, other configurations that function as arithmetic units may be adopted.

FIG. 3 is a diagram showing a functional configuration realized by the first computer 3 executing arithmetic processing or the like according to the program 811. These functional configurations include a storage unit 301, an inspection unit 302, and a control unit 303. All or part of these functions may be realized by a dedicated electric circuit. Further, these functions may be realized by a plurality of computers. Among the functional configurations shown in FIG. 3, the inspection unit 302 and the control unit 303 are realized by the CPU 31, GPU 39, ROM 32, RAM 33, fixed disk 34, and peripheral configurations thereof. Further, the storage unit 301 is mainly realized by the RAM 33 and the fixed disk 34.

The storage unit 301 stores an image of the pattern on the substrate 9 acquired by the imaging unit 21 (hereinafter, also referred to as an “image to be inspected”), a reference image used for inspection of the substrate 9, and the like. The reference image is an image showing a pattern on the substrate 9 having substantially no defects, and is generated by the second computer 4 as described later. The inspection unit 302 compares the image to be inspected with the reference image and inspects the presence or absence of defects in the substrate 9. The control unit 303 controls the operation of the imaging unit 21, the stage moving mechanism 23, and each functional configuration.

FIG. 4 is a diagram showing the configuration of the second computer 4. Similar to the first computer 3, the second computer 4 includes a CPU 41, a ROM 42, a RAM 43, a fixed disk 44, a display 45, an input unit 46, a reading device 47, a communication unit 48, and a GPU 49. , A general computer system configuration including a bus 40. The CPU 41 performs various arithmetic processes. The GPU 49 performs various arithmetic processes related to image processing. The ROM 42 stores the basic program. The RAM 43 stores various information. The fixed disk 44 stores information. The display 45 displays various information such as images. The input unit 46 includes a keyboard 46a and a mouse 46b that receive input from the operator. The reading device 47 reads information from a computer-readable recording medium 82 such as an optical disk, a magnetic disk, a magneto-optical disk, or a memory card. The display 45, keyboard 46a, mouse 46b and reader 47 are connected to the bus 40 via the interface I / F. The communication unit 48 transmits / receives signals to / from other configurations of the inspection device 1 and an external device. The bus 40 is a signal circuit that connects a CPU 41, a GPU 49, a ROM 42, a RAM 43, a fixed disk 44, a display 45, an input unit 46, a reading device 47, and a communication unit 48.

In the second computer 4, the program 821 is read in advance from the recording medium 82 via the reading device 47 and stored in the fixed disk 44. The program 821 may be stored on the fixed disk 44 via the network. The CPU 41 and the GPU 49 execute arithmetic processing while using the RAM 43 and the fixed disk 44 according to the program 821. The CPU 41 and the GPU 49 function as arithmetic units in the second computer 4. In addition to the CPU 41 and the GPU 49, another configuration that functions as a calculation unit may be adopted.

FIG. 5 is a diagram showing a functional configuration realized by the second computer 4 executing arithmetic processing or the like according to the program 821. These functional configurations include an input unit 401, a learning unit 402, a reference image generation unit 403, and a storage unit 404. All or part of these functions may be realized by a dedicated electric circuit. Further, these functions may be realized by a plurality of computers. Among the functional configurations shown in FIG. 5, the input unit 401, the learning unit 402, and the reference image generation unit 403 are realized by the CPU 41, the GPU 49, the ROM 42, the RAM 43, the fixed disk 44, and their peripheral configurations. Further, the storage unit 404 is mainly realized by the RAM 43 and the fixed disk 44.

The input unit 401 inputs the learning data to the storage unit 404. The storage unit 404 stores a learning data set which is a set of a large number of learning data. The learning data is a combination of the design data of the pattern on the substrate 9 and the learning image obtained by imaging the pattern formed on the substrate 9 which has substantially no defects. .. The learning image included in the learning data is, for example, a part of one die among a plurality of dies arranged on the substrate 9 (that is, a plurality of pattern regions in which the above pattern is provided). It is an image of a partial area of a certain predetermined size. In this case, the design data included in the learning data is the design data of the portion of the above pattern included in the partial region corresponding to the learning image. The design data is, for example, CAD data which is vector data, vertex image data generated from the CAD data, or run length data generated by rasterizing the vertex image data. The design data may be data in other formats.

The learning unit 402 causes the initial model to learn the relationship between the pattern design data on the substrate 9 and the learning image by deep learning (for example, GAN (hostile generation network)) using the above-mentioned learning data set. Create a trained model. The trained model is stored in the storage unit 404. The details of the training data set will be described later. The reference image generation unit 403 uses the trained model stored in the storage unit 404 to generate a reference image (that is, an image showing a good product pattern having substantially no defects) from the pattern design data. The reference image is stored in the storage unit 301 of the first computer 3.

FIG. 6 is a diagram showing a flow of generating a reference image in the second computer 4. In the generation of the reference image, first, a learning data set, which is a set of a large number of learning data, is prepared (step S11). As described above, each learning data is a combination of the design data of the pattern on the substrate 9 and the learning image obtained by imaging the pattern. The number of training data included in the training data set is not limited, but is, for example, about 10,000.

A plurality of learning data included in the learning data set are prepared by, for example, the following method. First, in the inspection device 1, a non-defective substrate 9 (that is, a learning substrate 9) having substantially no defects is held by the stage 22. Then, one die out of the plurality of dies arranged on the substrate 9 is selected, and the entire pattern on the selected die is imaged by the imaging unit 21. The imaging by the imaging unit 21 is performed by, for example, scanning or step-and-repeat. The image of the die captured by the imaging unit 21 is stored in the storage unit 301 of the first computer 3.

Subsequently, in the first computer 3, the image of the die is equally divided into a plurality of partial images of a predetermined size (for example, 512 pixels × 512 pixels) arranged vertically and horizontally in a matrix, and each partial image is used for learning. It is regarded as an image. Then, the learning data is created by combining the learning image with the data of the region (that is, the partial region) corresponding to the learning image in the above-mentioned die design data. The plurality of learning images described above are arranged in a matrix without overlapping with adjacent learning images, and constitute the entire image of one die. In other words, each position on the die is included in a partial region corresponding to any of the plurality of learning images. Therefore, the entire pattern on the die can be included in the training data set. It should be noted that the plurality of learning images may partially overlap with other learning images that are vertically and horizontally adjacent to each other. Moreover, the size of the above-mentioned partial image may be changed as appropriate.

The learning data set (that is, a set of a plurality of learning data) prepared in the first computer 3 is sent to the second computer 4 and input to the storage unit 404 by the input unit 401 (step S12). The image of the die captured by the imaging unit 21 may be sent to the second computer 4, and the second computer 4 may create the learning data (that is, prepare the learning data set).

Subsequently, the learning unit 402 of the second computer 4 learns the initial model for generating the reference image by deep learning using the learning data set. The learning unit 402 creates a trained model by training the initial model in the relationship between the pattern design data and the learning image of the pattern (step S13). The trained model is stored in the storage unit 404. The storage unit 404 also stores in advance the design data of the pattern on the substrate 9 scheduled to be inspected by the inspection device 1.

When the creation of the trained model is completed, the reference image generation unit 403 generates a reference image using the trained model from the design data stored in the storage unit 404 (step S14). The reference image generated in step S14 is sent from the second computer 4 to the first computer 3 and stored in the storage unit 301. The reference image is used for inspection of the substrate 9 by the inspection unit 302 of the first computer 3.

In the second computer 4, learning by deep learning in the learning unit 402 (step S13) and generation of the reference image in the reference image generation unit 403 (step S14) are performed by, for example, Pix2Pix, PointNet, or a combination thereof. It is said. In step S13, for example, Adam (Adaptive moment estimation) and backpropagation method are used as methods for updating internal parameters in deep learning. Further, as a reference of the error between the output image and the original image, for example, an L1 error (that is, the sum of the absolute values of the differences of each pixel value), an L2 error (that is, the sum of the squares of the differences of each pixel value), or an intersection. Entropy error is used. The internal parameters are updated, for example, by using the mini-batch learning method. The unit of the mini-batch and the number of epochs are not limited, but are, for example, 4 and 50. The update of the internal parameter does not necessarily have to be repeated a predetermined number of times, and the end of the update may be determined from the error value or transition.

FIG. 7 is a diagram showing an inspection flow of the substrate 9 in the inspection apparatus 1. In the inspection device 1, first, the substrate 9 to be inspected is held by the stage 22. Then, one die out of the plurality of dies arranged on the substrate 9 is selected, and the entire pattern on the selected die is imaged by the imaging unit 21 (step S21). The imaging by the imaging unit 21 is performed by, for example, scanning or step-and-repeat. The image of the die (that is, the image to be inspected) captured by the imaging unit 21 is stored in the storage unit 301 of the first computer 3.

Then, the inspection unit 302 compares the image to be inspected (that is, the image of the pattern) stored in the storage unit 301 with the above-mentioned reference image, and the image to be inspected having a significant difference is classified as "defective". The other images to be inspected are classified as "no defects" (step S22). In the substrate 9 classified as “with defects”, the position, size, defect image data, and the like of the defects are acquired and displayed on the display 35 (see FIG. 2) and the like. The comparison between the image to be inspected and the reference image may be performed by various known methods (for example, a difference comparison method or a shaking comparison method). In the inspection device 1, the above inspections are sequentially performed on a plurality of dies on the substrate 9. When the inspection of one substrate 9 is completed, the next inspection of the substrate 9 is performed.

In the inspection device 1, in the above-mentioned generation of the reference image (steps S11 to S14), various learning data may be included in the learning data set in addition to the above-exemplified learning data. For example, in step S11, the plurality of learning images do not necessarily have to be an equal division of the die image into a matrix, and are appropriately of a predetermined size (for example, 512 pixels × 512 pixels) from the die image. The cut out partial image may be used as the above-mentioned learning image. In this case, the cutout position of the plurality of learning images (that is, the position of the partial region corresponding to each learning image on the die) may be randomly determined. As a result, for various pattern elements included in the pattern on the die, a plurality of learning data in which the positions of the pattern elements on the learning image and the combination of the pattern elements included in the learning image are different are converted into the learning data set. Can be included. The cropping positions of the plurality of learning images may be regularly determined.

When the cutout position of the training image is randomly determined, for example, the x-coordinate and the y-coordinate (that is, the positions in the left-right direction and the up-down direction) are randomly determined using a random number or the like with one vertex on the upper left of the die as the origin. Is decided on. Subsequently, with the determined x-coordinate and y-coordinate as the upper left origin, a partial region having the predetermined size is set on the image of the die. When the entire partial region is included in the die (that is, the pattern region), the partial image corresponding to the partial region is cut out as a learning image. Then, the cut-out learning image and the design data corresponding to the partial region are combined to create the learning data.

On the other hand, when the partial area extends beyond the die, the partial image corresponding to the partial area is not cut out as a learning image, and learning data is not created. Alternatively, after the position of the partial region is shifted in the vertical direction and / or the horizontal direction so that the entire partial region is included in the die, the partial image corresponding to the partial region may be cut out as a learning image. In this case, it is preferable that the shift amount in the vertical direction and the horizontal direction of the partial region is as small as possible. Then, the cut-out learning image and the design data corresponding to the partial region are combined to create the learning data.

Further, in step S11, when one learning data which is a combination of the design data and the learning image is created, new learning data may be created by using the one learning data. ..

For example, a new learning image in which at least one of the brightness and the contrast of the learning image of the one learning data is changed is created, and the new learning image and the design data of the one learning data are combined. In combination, new learning data is created. The brightness of the learning image and the contrast may be changed by various known methods. The new learning data is included in the learning data set together with the above-mentioned one learning data. As a result, the influence of imaging conditions (for example, individual differences in the imaging unit and external factors such as temperature and atmospheric pressure) on the image, and individual differences in the die or substrate (for example, differences caused by subtle changes in process conditions) It is possible to generate a highly robust reference image in consideration of the difference in images due to). As a result, even when the brightness and contrast of the image to be inspected obtained in step S21 fluctuate, the image to be inspected and the reference image in step S22 can be compared with high accuracy.

Further, for example, a new learning image is created by rotating or reversing the learning image of the above-mentioned one learning data, and the design data of the one learning data is used in the same manner as the above-mentioned learning image. New rotated or inverted design data is created. Then, new learning data is created by combining the new design data and the new learning image. The rotation and inversion of the design data and the training image may be performed by various known methods. The new learning data is included in the learning data set together with the above-mentioned one learning data. As a result, it is possible to create a trained model capable of generating a reference image with high accuracy corresponding to a pattern facing various directions.

In step S11, the imaging unit 21 may perform imaging on a plurality of dies on the substrate 9, and learning images may be acquired from each of the images of the plurality of dies to create a plurality of learning data. In other words, the training data set includes a plurality of training data that are a combination of pattern design data and training images in each of a plurality of dies (that is, a plurality of pattern regions). As a result, it is possible to acquire a plurality of learning images corresponding to the same design data from a plurality of dies having different positions on the substrate. As a result, it is possible to create a plurality of learning data in which one design data and a plurality of learning images having different brightness, contrast, line width of pattern elements, and the like are combined.

Further, in step S11, the imaging unit 21 may perform imaging on the plurality of substrates 9, and learning images may be acquired from each of the images on the plurality of substrates to create a plurality of learning data. In other words, the training data set includes a plurality of training data that are a combination of pattern design data and training images on each of the plurality of substrates 9. As a result, it is possible to acquire a plurality of learning images corresponding to the same design data from the plurality of substrates 9 in which the process conditions and the like at the time of manufacturing may be slightly changed. As a result, it is possible to create a plurality of learning data in which one design data and a plurality of learning images having different brightness, contrast, line width of pattern elements, and the like are combined.

As described above, the second computer 4 is a reference image generation device that generates a reference image. The reference image is used for comparison with the image of the pattern in the manufacturing process (inspection step in the above example) of the substrate 9 having the pattern on the surface. The reference image generation device includes an input unit 401, a learning unit 402, and a reference image generation unit 403. The input unit 401 inputs learning data which is a combination of the design data of the pattern on the substrate 9 and the learning image obtained by imaging the pattern. The learning unit 402 creates a trained model by learning the relationship between the pattern design data and the learning image by deep learning using the learning data set which is a set of learning data. The reference image generation unit 403 generates a reference image from the pattern design data using the trained model.

As a result, preparations for reference image generation can be made in a very short period of time (for example, one day) compared to the case where a technician builds logic for generating a reference image from design data (work period: several months to several years). Can be completed. That is, the lead time until the reference image is generated can be significantly shortened. As a result, the reference image can be easily generated in a short time.

The reference image generation method described above includes a step of inputting learning data (step S12), which is a combination of the design data of the pattern on the substrate 9 and the learning image obtained by imaging the pattern, and the learning data. The process of creating a trained model by learning the relationship between the pattern design data and the training image by deep learning using the training data set, which is a set of the above (step S13), and using the trained model. A step (step S14) of generating a reference image from the pattern design data is provided. As a result, the lead time until the reference image is generated can be significantly shortened as described above.

As described above, the learning data set includes one learning data, the design data of the one learning data, and the learning image in which the brightness or contrast of the learning image of the one learning data is changed. It is preferable to include other learning data that is a combination. By changing the brightness, it is possible to perform learning in consideration of the brightness fluctuation of the learning image due to the uneven illuminance of the illumination unit 211 of the imaging unit 21, the difference in the sensitivity characteristics of the imaging device 213, and the fluctuation of the temperature and atmospheric pressure. Become. Further, by changing the contrast, it is possible to perform learning in consideration of the contrast fluctuation between the pattern elements and the contrast fluctuation between the pattern element and the background in the learning image due to the fluctuation of the film thickness of the pattern. Therefore, as described above, it is possible to generate a reference image having high robustness and capable of improving the uniformity of the comparison result. Further, since a plurality of learning data can be created from one learning image, the number of learning data included in the learning data set can be easily increased. As a result, a highly accurate reference image can be generated.

As described above, the training data set includes one training data, design data obtained by rotating or reversing the design data of the one training data, and a training image of the one training data as design data. Similarly, it is preferable to include other training data that is a combination with the rotated or inverted training image. As a result, it is possible to create a trained model capable of generating a reference image with high accuracy corresponding to a pattern facing various directions. Further, since a plurality of learning data can be created from one learning image, the number of learning data included in the learning data set can be easily increased. As a result, a highly accurate reference image can be generated.

As described above, each learning data of the learning data set is for pattern design data and learning in a partial area which is a part of a pattern area (die in the above example) in which the pattern is provided on the substrate 9. It is a combination with an image, and it is preferable that the position of a partial area of each learning data in the pattern area is randomly determined. Thereby, for various pattern elements included in the pattern on the pattern area, a plurality of learning data having different positions and combinations of the pattern elements on the learning image can be included in the learning data set. As a result, it is possible to generate a reference image having high robustness and capable of improving the uniformity of the comparison result.

As described above, each training data of the training data set is for pattern design data and training in a partial region that is a part of a pattern region (die in the above example) in which the pattern is provided on the substrate 9. It is a combination with an image, and it is preferable that each position of the pattern region is included in a partial region corresponding to any training data of the training data set. This makes it possible to generate a reference image that can be preferably compared with any part of the pattern region.

As described above, when a plurality of pattern regions (die in the above example) each of which is provided with a pattern are arranged on the substrate 9, the training data set is a pattern in each of the plurality of pattern regions. It is preferable to include a plurality of learning data which is a combination of the design data of the above and the learning image. As a result, it is possible to acquire a plurality of learning images corresponding to the same design data from a plurality of pattern regions having different positions on the substrate. Therefore, it is possible to create a plurality of learning data in which one design data and a plurality of learning images having different brightness, contrast, line width of pattern elements, and the like are combined. As a result, it is possible to generate a reference image having high robustness and capable of improving the uniformity of the comparison result.

As described above, the learning data set preferably includes a plurality of learning data which is a combination of the pattern design data and the learning image on each of the plurality of substrates 9 having the pattern on the surface. As a result, it is possible to acquire a plurality of learning images corresponding to the same design data from the plurality of substrates 9 in which the process conditions and the like at the time of manufacturing may be slightly changed. Therefore, it is possible to create a plurality of learning data in which one design data and a plurality of learning images having different brightness, contrast, line width of pattern elements, and the like are combined. As a result, it is possible to generate a reference image having high robustness and capable of improving the uniformity of the comparison result.

As described above, the inspection device 1 includes an imaging unit 21 and an inspection unit 302. The image pickup unit 21 takes an image of the pattern on the substrate 9 to acquire an image to be inspected. The inspection unit 302 compares the image to be inspected with the reference image of the pattern generated by the reference image generator (that is, the second computer 4) described above, and inspects the substrate 9 for defects. As described above, the reference image generation device can shorten the lead time until the reference image is generated. Therefore, in the inspection device 1, the defect inspection of the substrate 9 can be performed quickly. In addition, the reference image generator can generate a reference image that has high robustness and can improve the uniformity of comparison results. Therefore, in the inspection device 1, the defect inspection of the substrate 9 can be performed with high accuracy.

In the inspection device 1, when the lot of the substrate 9 to be inspected is changed, the trained model may be retrained. In the re-learning of the trained model, for example, a non-defective substrate 9 is extracted from the lot to be newly inspected, and the substrate 9 is imaged to prepare a new learning data set (FIG. 6: Step S11). ). Then, the trained model is retrained by a procedure substantially the same as in steps S12 to S14 described above. As a result, the defect inspection of the substrate 9 can be performed with higher accuracy. The same applies to the alignment device 1a described below.

FIG. 8 is a diagram showing a configuration of an alignment device 1a according to a second embodiment of the present invention. The alignment device 1a is a device that adjusts the position of the substrate 9 having a pattern on the surface. The alignment device 1a is included in, for example, a processing device that performs various processing on the substrate 9. The processing device is, for example, an inspection device for detecting defects by comparing a plurality of dies provided on one substrate 9, or a drawing apparatus for irradiating the substrate 9 with light to perform drawing. The processing apparatus includes, for example, a processing head that performs various processing on the substrate 9.

The alignment device 1a has substantially the same configuration as the inspection device 1 except that the functional configuration realized by the first computer 3 is different. In the following description, the same reference numerals are given to the configurations of the alignment apparatus 1a corresponding to the respective configurations of the inspection apparatus 1. The alignment device 1a includes a device main body 2, a first computer 3, and a second computer 4. The configuration of the apparatus main body 2, the first computer 3 and the second computer 4, and the functional configuration realized by the second computer 4 are substantially the same as those shown in FIGS. 1, 2, 4, and 5.

FIG. 9 is a diagram showing a functional configuration realized by the first computer 3 of the alignment device 1a executing arithmetic processing or the like according to a program 811 (see FIG. 2). These functional configurations include a storage unit 301, an alignment unit 304, and a control unit 303. All or part of these functions may be realized by a dedicated electric circuit. Further, these functions may be realized by a plurality of computers. Among the functional configurations shown in FIG. 9, the alignment unit 304 and the control unit 303 are realized by the CPU 31, GPU 39, ROM 32, RAM 33, fixed disk 34 (see FIG. 2), and peripheral configurations thereof. Further, the storage unit 301 is mainly realized by the RAM 33 and the fixed disk 34.

The storage unit 301 stores an image of the pattern on the substrate 9 (hereinafter, also referred to as “target image”) acquired by the imaging unit 21 (see FIG. 8), a reference image used for alignment of the substrate 9, and the like. Remember. The reference image is an image of a pattern (hereinafter, also referred to as “alignment pattern”) on the substrate 9 used for alignment, and is generated by the second computer 4 as described later. The alignment pattern may be a part of a circuit pattern or the like on the substrate 9, or may be a mark dedicated to alignment. The alignment unit 304 compares the target image with the reference image and acquires the position of the substrate 9. The control unit 303 controls the operation of the imaging unit 21, the stage moving mechanism 23, and each functional configuration.

In the second computer 4, a trained model is created by deep learning using a training data set which is a set of training data (that is, a combination of alignment pattern design data and a training image) in the same manner as described above. To. Then, using the trained model, a reference image is generated from the design data of the alignment pattern (steps S11 to S14). In steps S11 to S14, since the reference image of the alignment pattern (that is, the alignment template image) can be generated from the design data, compared with the case where the image of the alignment pattern is selected and extracted from the image obtained by imaging the pattern. Therefore, the reference image can be easily obtained. The reference image is sent from the second computer 4 to the first computer 3 and stored in the storage unit 301.

In the first computer 3, the alignment unit 304 compares the target image stored in the storage unit 301 with the reference image, and extracts the alignment pattern corresponding to the alignment pattern of the reference image from the target image. Subsequently, the position of the alignment pattern in the target image is acquired. Then, the deviation between the position of the alignment pattern in the target image and the position of the desired alignment pattern (that is, the design position) is required. The control unit 303 controls the stage moving mechanism 23 based on the deviation to move the substrate 9 so that the alignment pattern on the substrate 9 is located at a desired position. The stage moving mechanism 23 does not necessarily have to move the substrate 9, and may move the above-mentioned processing head that performs processing on the substrate 9. That is, the stage moving mechanism 23 is a substrate moving mechanism that moves the substrate 9 relative to the processing head to adjust the relative position of the substrate 9.

As described above, the alignment device 1a includes an imaging unit 21, an alignment unit 304, and a substrate moving mechanism (in the above example, a stage moving mechanism 23). The imaging unit 21 captures a pattern on the substrate 9 to acquire a target image. The alignment unit 304 compares the target image with the reference image of the pattern generated by the reference image generation device (that is, the second computer 4) described above, and acquires the position of the substrate 9. The substrate moving mechanism adjusts the position by relatively moving the substrate 9 based on the position of the substrate 9 acquired by the alignment unit 304. As described above, the reference image generation device can shorten the lead time until the reference image is generated. Therefore, the alignment device 1a can quickly align the substrate 9. In addition, the reference image generator can generate a reference image that has high robustness and can improve the uniformity of comparison results. Therefore, the alignment device 1a can accurately align the substrate 9.

Various changes can be made in the reference image generation device, the reference image generation method, the inspection device 1 and the alignment device 1a described above.

For example, the learning data set prepared in step S11 may be created outside the inspection device 1 including the acquisition of the learning image, and may be input to the inspection device 1 via a recording disk or a network. The same applies to the alignment device 1a.

The above-mentioned learning data set does not necessarily have to include the learning data on each of the plurality of substrates 9. Further, the above-mentioned learning data set does not necessarily have to include learning data in each of a plurality of pattern regions (die in the above example) on the substrate 9. The plurality of pattern areas may be other than the die. Further, it is not always necessary that a plurality of pattern regions are arranged on the substrate 9.

In the above-mentioned training data set, each position of the pattern region does not necessarily have to be included in a partial region (that is, a region that is a part of the pattern region) corresponding to any training data. Further, each learning data does not necessarily have to be a combination of the pattern design data in the partial area of the pattern area and the learning image. For example, the combination of the pattern design data and the learning image in the entire pattern area may be included in the learning data set as one learning data.

The training data set described above does not necessarily have to include training data in which the brightness or contrast of the training image is changed. Further, the learning data set does not necessarily have to include the learning data obtained by rotating or inverting the learning image.

The substrate 9 handled by the reference image generator, the inspection apparatus 1 and the alignment apparatus 1a described above is not limited to the semiconductor substrate, and is a flat panel display such as a liquid crystal display device or an organic EL (Electro Luminescence) display device. ), Or the glass substrate used for other display devices. Further, the substrate 9 may be an optical disk substrate, a magnetic disk substrate, a magneto-optical disk substrate, a photomask substrate, a ceramic substrate, a solar cell substrate, or the like.

In the inspection device 1 and the alignment device 1a, the first computer 3 and the second computer 4 may be housed in the housing of the device main body 2.

The second computer 4 may be included in a device other than the inspection device 1 and the alignment device 1a. Further, the second computer 4 may be used alone as a reference image generator.

The above-described embodiment and the configurations in each modification may be appropriately combined as long as they do not conflict with each other.

1 Inspection device 1a Alignment device 4 Second computer 9 Board 21 Imaging unit 23 Stage movement mechanism 302 Inspection unit 304 Alignment unit 401 Input unit 402 Learning unit 403 Reference image generation unit S11 to S14, S21 to S22 Steps

Claims (16)

A reference image generator that generates a reference image used for comparison with an image of the pattern in the manufacturing process of a substrate having a pattern on the surface.
An input unit for inputting learning data, which is a combination of the design data of the pattern on the substrate and the learning image obtained by imaging the pattern, and
A learning unit that creates a trained model by learning the relationship between the design data of the pattern and the training image by deep learning using the learning data set which is a set of the training data.
A reference image generator that generates a reference image from pattern design data using the trained model,
A reference image generator comprising. The reference image generator according to claim 1.
The training data set is
One learning data and
Other learning data that is a combination of the design data of the one learning data and the learning image in which the brightness or contrast of the learning image of the one learning data is changed, and
A reference image generator, characterized in that it comprises. The reference image generator according to claim 1 or 2.
The training data set is
One learning data and
Another combination of the design data obtained by rotating or inverting the design data of the one learning data and the learning image obtained by rotating or inverting the learning image of the one learning data in the same manner as the design data. Training data and
A reference image generator, characterized in that it comprises. The reference image generator according to any one of claims 1 to 3.
Each learning data of the learning data set is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region where the pattern is provided on the substrate.
A reference image generator, characterized in that the position of the partial region of the learning data in the pattern region is randomly determined. The reference image generator according to any one of claims 1 to 4.
Each learning data of the learning data set is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region where the pattern is provided on the substrate.
A reference image generation device, wherein each position of the pattern region is included in the partial region corresponding to any training data of the training data set. The reference image generator according to any one of claims 1 to 5.
A plurality of pattern regions in which the patterns are provided are arranged on the substrate.
The learning data set is a reference image generation device including a plurality of learning data which is a combination of the design data of the pattern and the learning image in each of the plurality of pattern regions. The reference image generator according to any one of claims 1 to 6.
The learning data set includes a plurality of learning data that is a combination of the design data of the pattern and the learning image on each of the plurality of substrates having the pattern on the surface. .. An inspection device that inspects a substrate having a pattern on its surface.
An imaging unit that captures a pattern on a substrate and acquires an image to be inspected,
An inspection unit that inspects the presence or absence of defects in the substrate by comparing the image to be inspected with the reference image of the pattern generated by the reference image generator according to any one of claims 1 to 7.
An inspection device comprising. An alignment device that adjusts the position of a substrate that has a pattern on its surface.
An imaging unit that captures a pattern on a substrate and acquires a target image,
An alignment unit for acquiring the position of the substrate by comparing the target image with the reference image of the pattern generated by the reference image generator according to any one of claims 1 to 7.
A substrate moving mechanism that moves the substrate relative to each other to adjust the position based on the position of the substrate acquired by the alignment unit.
An alignment device comprising. A reference image generation method for generating a reference image used for comparison with an image of the pattern in the manufacturing process of a substrate having a pattern on the surface.
a) A process of inputting learning data which is a combination of the design data of the pattern on the substrate and the learning image obtained by imaging the pattern, and
b) A step of creating a trained model by learning the relationship between the design data of the pattern and the learning image by deep learning using the learning data set which is a set of the training data.
c) A process of generating a reference image from pattern design data using the trained model, and
A reference image generation method comprising. The reference image generation method according to claim 10.
The training data set is
One learning data and
Other learning data that is a combination of the design data of the one learning data and the learning image in which the brightness or contrast of the learning image of the one learning data is changed, and
A reference image generation method comprising. The reference image generation method according to claim 10 or 11.
The training data set is
One learning data and
Another combination of the design data obtained by rotating or inverting the design data of the one learning data and the learning image obtained by rotating or inverting the learning image of the one learning data in the same manner as the design data. Training data and
A reference image generation method comprising. The reference image generation method according to any one of claims 10 to 12.
Each learning data of the learning data set is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region where the pattern is provided on the substrate.
A reference image generation method, characterized in that the position of the partial region of each learning data in the pattern region is randomly determined. The reference image generation method according to any one of claims 10 to 13.
Each learning data of the learning data set is a combination of the design data of the pattern and the learning image in the partial region which is a part of the pattern region where the pattern is provided on the substrate.
A reference image generation method, wherein each position of the pattern region is included in the partial region corresponding to any training data of the training data set. The reference image generation method according to any one of claims 10 to 14.
A plurality of pattern regions in which the patterns are provided are arranged on the substrate.
A reference image generation method, wherein the learning data set includes a plurality of learning data that are a combination of the design data of the pattern and the learning image in each of the plurality of pattern regions. The reference image generation method according to any one of claims 10 to 15.
The learning data set includes a plurality of learning data that are a combination of the design data of the pattern and the learning image on each of the plurality of substrates having the pattern on the surface. ..

Images