Generation method, learning method, program, generation device, evaluation device, membrane formation system, and manufacturing method for article

Abstract

To provide a technique advantageous for easily generating training data for a model for evaluating a film of a composition formed on a substrate using a mold.SOLUTION: A generation method is provided for generating training data for a model that evaluates a film of a composition formed on a substrate using a mold, by means of an information processing device, wherein the method includes generating a training image as the training data by processing a design image that shows geometric features of at least a portion of the substrate.SELECTED DRAWING: Figure 12

Description

[Technical field]

[0001] The present invention relates to a generation method, a learning method, a program, a generation device, an evaluation device, a film forming system, and an article manufacturing method. [Background technology]

[0002] Imprinting technology, which is a technology for forming a film having a fine pattern on a substrate, is being put into practical use. One of the imprinting technologies is a photocuring method. An imprinting device that employs the photocuring method hardens an imprinting material (photocurable composition) supplied onto a substrate by irradiating the imprinting material with light while a mold (original plate) is in contact with the imprinting material. Thereafter, a film (hardened film) of the imprinting material having a pattern is formed on the substrate by separating the mold from the hardened imprinting material. For example, an apparatus that applies step-and-flash imprint lithography is effective for manufacturing semiconductor devices and the like (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-80047 A Summary of the Invention [Problem to be solved by the invention]

[0004] When forming a film of imprint material on a substrate using imprint technology, which is one of the film formation technologies, if too much imprint material is supplied, the imprint material may overflow outside the pattern area of ​​the mold (hereinafter, this may be referred to as seepage). On the other hand, if too little imprint material is supplied, the imprint material may not reach the entire pattern area of ​​the mold, and the film of the imprint material may not be formed in parts (hereinafter, this may be referred to as non-filling). If seepage occurs, not only will the seeped portion cause a poor formation of the film of the imprint material, but it may also cause the pattern of the mold that comes into contact with that portion to be destroyed. Furthermore, if non-filling occurs, the film of the imprint material will not be formed in that portion, resulting in a defective semiconductor device.

[0005] Therefore, after the imprinting process, it is necessary to detect the presence or absence of such seepage and non-filling, and adjust the supply amount and position of the imprinting material according to the detection result to prevent defective formation of the imprinting material film. However, since seepage and non-filling occur in a small area, it is necessary to check a huge number of observation images obtained by a high-magnification microscope with a narrow detection range, which is difficult for a human to do. Therefore, there is a demand for a technology that can inspect the seepage and non-filling from the observation images without human intervention, and detect defective formation of the imprinting material film caused by seepage and non-filling (i.e., evaluate the imprinting material film). Hereinafter, defective formation of the imprinting material film caused by seepage and non-filling may be referred to as "abnormal".

[0006] Methods for detecting anomalies from observed images include a method using machine learning. In this method, a model that receives an observed image as an input and outputs an anomaly detection result is generated by machine learning, and the anomaly can be detected by using the model. In order to generate such a model with high accuracy by machine learning, it is necessary to prepare a large number of images with abnormality features as training data in advance. However, in order to prepare a large number of images as training data, it is cumbersome and may take a considerable amount of time to perform imprint processing by changing conditions so that the position, size, etc. of anomalies contained in the images are changed.

[0007] Therefore, an object of the present invention is to provide an advantageous technique for easily generating training data for a model for evaluating a film of a composition formed on a substrate using a mold. [Means for solving the problem]

[0008] In order to achieve the above-mentioned object, one aspect of the present invention is a generation method that generates training data for a model for evaluating a film of a composition formed on a substrate using a mold using an information processing device, and is characterized in that a training image as the training data is generated by processing a design image that shows at least a portion of the geometric characteristics of the substrate.

[0009] Further objects and other aspects of the present invention will become apparent from the following description of preferred embodiments with reference to the accompanying drawings. Effect of the Invention

[0010] According to the present invention, for example, an advantageous technique can be provided for easily generating training data for a model for evaluating a film of a composition formed on a substrate using a mold. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 illustrates an example of the configuration of an imprint apparatus; [Diagram 2]A diagram showing an example of the configuration of a wide-angle alignment measurement instrument [Diagram 3] FIG. 1 is a diagram showing an example of the configuration of an article manufacturing system. [Figure 4] FIG. 1 shows an example of the configuration of an evaluation device. [Diagram 5] Flowchart showing the operation of the imprint apparatus [Figure 6] Diagram illustrating seepage and underfilling [Figure 7] FIG. 1 illustrates an image including seepage and non-filling. [Figure 8] FIG. 13 illustrates an example of a non-filling that occurs in a mark. [Figure 9] (a) Schematic diagram showing a method for generating an evaluation model by machine learning, (b) Schematic diagram showing a method for evaluating a film on a substrate using the evaluation model. [Figure 10] Flowchart showing how machine learning generates a valuation model [Figure 11] 1 is a flowchart showing a method for evaluating a film on a substrate using an evaluation model. [Figure 12] 1 is a flowchart showing how training images are generated. [Figure 13] FIG. 13 is a diagram showing an example of a design image. [Figure 14] FIG. 1 is a schematic diagram for explaining a process performed by a generating device; [Figure 15] FIG. 1 is a schematic diagram for explaining a process performed by a creating unit; [Figure 16] FIG. 1 is a schematic diagram for explaining a process performed by a generation unit; [Figure 17] A diagram for explaining a method for manufacturing an article DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, the embodiments will be described in detail with reference to the attached drawings. Note that the following embodiments do not limit the invention according to the claims. Although the embodiments describe a number of features, not all of these features are essential to the invention, and the features may be combined in any manner. Furthermore, in the attached drawings, the same reference numbers are used for the same or similar configurations, and duplicated descriptions are omitted.

[0013] The following embodiment relates to a film formation system including a film formation apparatus. The film formation apparatus is used in the manufacture of devices such as semiconductor devices as articles, and places an uncured (liquid) composition on a substrate, molds the placed composition in a mold, and forms a film of the composition on the substrate. The film formation apparatus may be called a molding apparatus, and similarly, the film formation process may be called a molding process.

[0014] The film-forming process includes a contacting step of contacting the curable composition (moldable material) supplied onto a substrate with a mold (master plate, template). Through this contacting step, a film of the curable composition molded by the mold is formed on the substrate. The film-forming process may further include a curing step of curing the curable composition while the curable composition is in contact with the mold. Through this curing step, the film of the curable composition formed on the substrate is cured, and a cured film made of a cured product of the curable composition is formed on the substrate. The film-forming process may further include a separation step of separating the cured film made of a cured product of the curable composition from the mold.

[0015] The film forming apparatus can be used as an imprinting apparatus that performs an imprinting process in which an imprinting material, which is a curable composition supplied onto a shot area of ​​a substrate, is brought into contact with a pattern area of ​​a mold, thereby transferring a pattern of the mold to the imprinting material. The pattern of the mold can be, for example, a pattern of a semiconductor device (device pattern). The shot area is an area in which a pattern made of a cured product of the imprinting material is to be formed, and a plurality of shot areas are arranged on the substrate. The imprinting apparatus can perform an imprinting process (film forming process) individually for each of the plurality of shot areas on the substrate. The imprinting apparatus may be configured to perform an imprinting process collectively for at least two shot areas among the plurality of shot areas on the substrate (for the entire substrate, or for one or more rows of shot areas). The film forming apparatus can also be used as a planarizing apparatus that performs a planarizing process to form a planarized film of the curable composition on the substrate by bringing the curable composition on the substrate into contact with a member having a flat surface (flat surface of the mold). In the following, in order to show a specific example, a system including an imprinting apparatus, which is an example of a film forming apparatus, will be described.

[0016] 1(a) shows a schematic configuration example of an imprint apparatus IMP according to an embodiment of the present invention. The imprint apparatus IMP performs an imprint process in which an imprint material IM on a substrate S is brought into contact with a pattern region MP of a mold M, the imprint material IM is cured, and the cured product of the imprint material IM is separated from the mold M. By this imprint process, a pattern made of the cured product of the imprint material IM is formed on the substrate S.

[0017] As the imprint material IM, a curable composition (sometimes called an uncured resin) that is cured by applying energy for curing is used. As the energy for curing, electromagnetic waves, heat, etc. can be used. The electromagnetic waves can be, for example, light having a wavelength selected from the range of 10 nm to 1 mm, such as infrared rays, visible light, and ultraviolet rays. The curable composition can be a composition that is cured by irradiation with light or by heating. Among these, the photocurable composition that is cured by irradiation with light contains at least a polymerizable compound and a photopolymerization initiator, and may further contain a non-polymerizable compound or a solvent as necessary. The non-polymerizable compound is at least one selected from the group consisting of a sensitizer, a hydrogen donor, an internal mold release agent, a surfactant, an antioxidant, and a polymer component. The imprint material can be arranged on the substrate S in the form of droplets, or in the form of islands or a film formed by connecting a plurality of droplets. The viscosity (viscosity at 25° C.) of the imprint material IM can be, for example, 1 mPa·s to 100 mPa·s. Examples of materials that can be used for the substrate S include glass, ceramics, metals, semiconductors, and resins. If necessary, a member made of a material different from that of the substrate S may be provided on the surface of the substrate S. The substrate S is, for example, a silicon wafer, a compound semiconductor wafer, or quartz glass.

[0018] In this specification and the accompanying drawings, directions are shown in an XYZ coordinate system in which the direction parallel to the surface of the substrate S is the XY plane. The directions parallel to the X-axis, Y-axis, and Z-axis in the XYZ coordinate system are the X direction, Y direction, and Z direction, respectively, and the rotation around the X-axis, the Y axis, and the Z axis are θX, θY, and θZ, respectively. Control or drive regarding the X-axis, Y-axis, and Z-axis means control or drive regarding the direction parallel to the X-axis, the direction parallel to the Y axis, and the direction parallel to the Z axis, respectively. Furthermore, control or drive regarding the θX-axis, θY-axis, and θZ-axis means control or drive regarding the rotation around an axis parallel to the X-axis, the rotation around an axis parallel to the Y axis, and the rotation around an axis parallel to the Z axis, respectively. Furthermore, the position is information that can be specified based on the coordinates of the X-axis, Y-axis, and Z-axis, and the orientation is information that can be specified by the values ​​of the θX-axis, θY-axis, and θZ-axis. Positioning means controlling the position and / or orientation. Alignment may include control of the position and / or orientation of at least one of the substrate and mold.

[0019] The imprint apparatus IMP may include a substrate holding unit 102 that holds a substrate S, a substrate driving mechanism 105 that drives the substrate S by driving the substrate holding unit 102, a base 104 that supports the substrate holding unit 102, and a position measurement unit 103 that measures the position of the substrate holding unit 102. The substrate driving mechanism 105 may include, for example, a motor such as a linear motor.

[0020] The imprint apparatus IMP may include a mold holding part 121 that holds a mold M, a mold driving mechanism 122 that drives the mold M by driving the mold holding part 121, and a support structure 130 that supports the mold driving mechanism 122. The mold driving mechanism 122 may include, for example, a motor such as a voice coil motor.

[0021] The substrate driving mechanism 105 and the mold driving mechanism 122 constitute a driving mechanism that adjusts the relative position and the relative attitude of the substrate S and the mold M. The adjustment of the relative position of the substrate S and the mold M by the driving mechanism includes driving for contact of the mold M with the imprint material IM on the substrate S and separation of the mold M from the cured imprint material IM (pattern of the cured material). The substrate driving mechanism 105 can be configured to drive the substrate S about a plurality of axes (e.g., three axes of the X axis, the Y axis, and the θZ axis, preferably six axes of the X axis, the Y axis, the Z axis, the θX axis, the θY axis, and the θZ axis). The mold driving mechanism 122 can be configured to drive the mold M about a plurality of axes (e.g., three axes of the Z axis, the θX axis, and the θY axis, preferably six axes of the X axis, the Y axis, the Z axis, the θX axis, the θY axis, and the θZ axis).

[0022] The imprint apparatus IMP may include a mold transport mechanism 140 that transports the mold M, and a mold cleaner 150. The mold transport mechanism 140 may be configured to, for example, transport the mold M to the mold holding unit 121, or transport the mold M from the mold holding unit 121 to an original plate stocker (not shown) or the mold cleaner 150. The mold cleaner 150 cleans the mold M with ultraviolet light, a chemical solution, or the like.

[0023] 1(b), the imprint apparatus IMP can include a deformation mechanism 123 that deforms the pattern region MP of the mold M into a convex shape toward the substrate S. The mold holding part 121 is provided with a window member 125 for forming a pressure-controlled space CS on the back surface of the mold M (the surface opposite to the pattern region MP in which the pattern to be transferred to the substrate S is formed). The deformation mechanism 123 can deform the pattern region MP of the mold M by controlling the pressure in the pressure-controlled space CS (hereinafter referred to as cavity pressure).

[0024] The imprint apparatus IMP may include an alignment measurement instrument 106, a wide-angle alignment measurement instrument 151, a curing unit 107, an imaging unit 112, and an optical member 111. The alignment measurement instrument 106 illuminates the alignment mark of the substrate S and the alignment mark of the mold M, and measures the relative position between the marks by capturing their images. The alignment measurement instrument 106 can be positioned by a driving mechanism (not shown) according to the position of the alignment mark to be observed. The wide-angle alignment measurement instrument 151 is a measurement instrument having a wider field of view than the alignment measurement instrument 106, and measures the position of the substrate S by illuminating the alignment mark of the substrate S and capturing its image. By measuring the position of the substrate S with the wide-angle alignment measurement instrument, the alignment mark of the substrate S can be moved within the field of view of the alignment measurement instrument 106.

[0025] The curing unit 107 irradiates the imprint material IM with energy for curing the imprint material IM (for example, light such as ultraviolet light) via the optical member 111, thereby curing the imprint material IM. The imaging unit 112 captures images of the substrate S, the mold M, and the imprint material IM via the optical member 111 and the window member 125.

[0026] The wide-angle alignment measuring instrument 151 may have a mechanism for switching the wavelength of the illumination light. For example, the wide-angle alignment measuring instrument 151 has a wavelength filter arranged on the optical path and has a mechanism for switching the wavelength filter. Alternatively, the wide-angle alignment measuring instrument 151 may have a configuration capable of simultaneously capturing images of multiple wavelengths as shown in FIG. 2. The wide-angle alignment measuring instrument 151 shown in FIG. 2 includes a light source 161, multiple half mirrors 162 that branch the optical path, multiple wavelength filters 163 that transmit different wavelengths, and multiple image sensors 164, and can capture images of different wavelengths simultaneously. In addition, the wide-angle alignment measuring instrument 151 may have a mechanism for switching the amount of light of the illumination light. For example, the wide-angle alignment measuring instrument 151 may have a mechanism for switching the ND filter arranged on the optical path. Furthermore, the wide-angle alignment measuring instrument 151 may have multiple optical systems such as a bright-field optical system and a dark-field optical system, and may have a mechanism for switching the optical system through which the image to be captured passes. The wide-angle alignment measurement instrument 151 may have a mechanism for switching the polarization of the illumination light or the received light. For example, the wide-angle alignment measurement instrument 151 may also have a mechanism for switching a polarizing filter disposed on the optical path.

[0027] The imprint apparatus IMP can include a dispenser 108 that disposes (supplies) the imprint material IM onto the substrate S. The dispenser 108 ejects the imprint material IM so that the imprint material IM is disposed on the substrate S according to, for example, a drop recipe that indicates the placement of the imprint material IM.

[0028] The imprint apparatus IMP may include a control unit 110 that controls the substrate driving mechanism 105, the mold driving mechanism 122, the deformation mechanism 123, the mold transport mechanism 140, the mold cleaner 150, the alignment measuring instrument 106, the curing unit 107, the imaging unit 112, the dispenser 108, and the like. The control unit 110 may include an information processing device 113 (computer) having a processor such as a CPU (Central Processing Unit) and a storage unit such as a memory. The control unit 110 may be configured, for example, by a PLD (abbreviation for Programmable Logic Device) such as an FPGA (abbreviation for Field Programmable Gate Array), an ASIC (abbreviation for Application Specific Integrated Circuit), a general-purpose computer with a built-in program, or a combination of all or part of these.

[0029] FIG. 3 illustrates an example of the configuration of an article manufacturing system 401 for manufacturing an article such as a semiconductor device. The article manufacturing system 401 may include, for example, one or more imprint apparatuses IMP (film forming apparatuses) and one or more inspection apparatuses 405 (for example, an overlay inspection apparatus, a CD inspection apparatus, a defect inspection apparatus, an electrical property inspection apparatus). The article manufacturing system 401 may also include one or more substrate processing apparatuses 406 (etching apparatus, film forming apparatus). The article manufacturing system 401 further includes an evaluation apparatus 407 that evaluates a film (for example, a cured film of an imprint material on which a pattern is formed) formed on a substrate S by the imprint apparatus IMP (film forming apparatus) as described later. The article manufacturing system 401 includes a generation apparatus 408 that generates training data for an evaluation model for evaluating a film formed on a substrate S in the evaluation apparatus 407 as described later. These apparatuses are connected to a control apparatus 403, which is an external apparatus different from the imprint apparatus IMP, via a network 402, and may be controlled by the control apparatus 403.

[0030] Each of the evaluation device 407 and the generation device 408 may be configured by an information processing device (computer) having a processor such as a CPU (Central Processing Unit) and a storage unit such as a memory. The information processing device may be configured by, for example, a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), a general-purpose computer with a built-in program, or a combination of all or part of these. In this embodiment, the evaluation device 407 and the generation device 408 are configured by separate information processing devices, but may be configured by a single information processing device.

[0031] FIG. 4 shows a configuration example of the evaluation device 407. The evaluation device 407 may include a control unit 501, a RAM 502 that stores temporary data and provides a work area for the control unit 501, and a ROM 503 that stores fixed data and programs. The evaluation device 407 may further include a storage device 504, a display device 505, and an input device 506. The storage device 504 stores a program (evaluation program) for executing an evaluation method for evaluating a film formed on a substrate S by the imprint apparatus IMP, and a program (learning program) for executing a learning method of an evaluation model for evaluating the film. The network I / F 507 is an interface for connecting to the network 402. In this embodiment, the network I / F 507 may function as an acquisition unit that acquires an image or the like including a composition formed on a substrate by a film formation process. The control unit 501 may also function as an image processing unit that processes the acquired image for evaluation. The control unit 501 may also function as a display control unit that controls the display of the display device 505.

[0032] 4 can also be applied to the generating device 408. That is, like the evaluating device 407, the generating device 408 can include a control unit 501, a RAM 502, a ROM 503, a storage device 504, a display device 505, an input device 506, and a network I / F 507. In this case, the storage device 504 stores a program (generating program) for executing a generating method for generating training data for an evaluation model.

[0033] The functions of the evaluation apparatus 407 and the generation apparatus 408 may be realized by any one of the control unit 110 of the imprint apparatus IMP, the control apparatus 403, or the control unit of the inspection apparatus 405, or a combination of these. In this embodiment, a system including the imprint apparatus IMP and the evaluation apparatus 407 may be understood as a film formation system (lithography system). The film formation system may further include the generation apparatus 408.

[0034] The lithography method according to this embodiment will be described below. In this embodiment, after performing an imprint process, an image of a shot area (film formation area) in which a film of the imprint material IM is formed by the imprint process and an area (evaluation area) including the surrounding area is acquired by imaging. Using this image, an evaluation is performed on the film of the imprint material IM (composition part) formed on the substrate S. That is, an abnormality occurring in the film of the imprint material IM, such as seepage or non-filling, is detected. Machine learning is used to detect an abnormality such as seepage or non-filling. Furthermore, in machine learning, an object detection algorithm can be used to detect an abnormality.

[0035] The operation of the imprint apparatus IMP will be described with reference to the flowchart of Fig. 5. The operation shown in Fig. 5 can be executed by the control unit 110.

[0036] In step S101, the substrate S is transported from a transport source (for example, an intermediary between a pre-processing device and an imprint device IMP) onto the substrate holding part 102 by a substrate transport mechanism (not shown). At this time, the position of the transported substrate S on the substrate holding part 102 is measured by observing a mark on the substrate S using a wide-angle alignment measuring instrument 151. The control part 110 positions the substrate S based on the position obtained by the measurement.

[0037] Steps S102 to S106 are steps for performing imprint processing on a target shot area for performing imprint processing (pattern formation) among a plurality of shot areas on the substrate S. Hereinafter, the target shot area for performing imprint processing may be referred to as a target shot area.

[0038] In step S102, the imprint material IM is placed (supplied) onto the target shot area by the dispenser 108. This process can be performed by discharging the imprint material IM from the dispenser 108 while the substrate S is driven relative to the dispenser 108 by the substrate driving mechanism 105.

[0039] In step S103, the substrate S and the mold M are driven relatively by at least one of the mold driving mechanism 122 and the substrate driving mechanism 105 so that the pattern region MP of the mold M comes into contact with the imprint material IM on the target shot area. In one example, the mold M is driven by the mold driving mechanism 122 so that the pattern region MP of the mold M comes into contact with the imprint material IM on the target shot area. In the process of bringing the pattern region MP of the mold M into contact with the imprint material IM, the deformation mechanism 123 can deform the pattern region MP of the mold M into a convex shape toward the substrate S.

[0040] In step S104, alignment between the target shot area and the pattern area MP of the mold M can be performed. The alignment can be performed by measuring the relative position between the alignment mark of the target shot area and the alignment mark of the mold M by the alignment measurement instrument 106 so that the relative position falls within a tolerance range of the target relative position. In the alignment, the substrate S and the mold M are driven relatively by at least one of the mold driving mechanism 122 and the substrate driving mechanism 105. The target relative position between the alignment mark of the target shot area and the alignment mark of the mold M can be determined by a correction value determined from past results of the overlay inspection device, etc.

[0041] In step S105, the curing unit 107 irradiates the imprint material IM between the substrate S and the pattern region MP of the mold M with energy for curing the imprint material IM. This cures the imprint material IM, and a cured product (cured film) of the imprint material IM is formed.

[0042] In step S106, the substrate S and the mold M are driven relatively by at least one of the mold driving mechanism 122 and the substrate driving mechanism 105 so that the cured product of the imprint material IM and the pattern region MP of the mold M are separated. In one example, the mold M is driven by the mold driving mechanism 122 so that the cured product of the imprint material IM and the pattern region MP of the mold M are separated. When the cured product of the imprint material IM and the pattern region MP of the mold M are separated, the pattern region MP of the mold M can also be deformed into a convex shape toward the substrate S. Furthermore, imaging is performed by the imaging unit 112, and the state of separation between the imprint material IM and the mold M is observed based on the captured image.

[0043] In step S107, the control unit 110 determines whether the imprint processing of steps S102 to S106 has been performed on all shot areas on the substrate S. If the imprint processing of steps S102 to S106 has been performed on all shot areas on the substrate S, the process proceeds to step S108. If an unprocessed shot area exists, the process returns to step S102. In this case, the imprint processing of steps S102 to S106 is performed on a shot area selected from the unprocessed shot areas as a target shot area.

[0044] In step S108, an image of an area (evaluation area) including a shot area (film-formed area) after the imprint process is acquired as an evaluation image for evaluating a film (cured film) formed on the substrate S by the imprint process. For example, the control unit 110 uses the wide-angle alignment measuring instrument 151 to capture an image of the shot area and its periphery. When the field of view of the wide-angle alignment measuring instrument 151 is narrower than the shot area, the substrate driving mechanism 105 may be driven to change the position of the substrate S relative to the field of view of the wide-angle alignment measuring instrument 151, and multiple images may be captured to obtain an image of a desired area. The image obtained in step S108 may be used as input data (evaluation image) for an evaluation model for evaluating (detecting) an abnormality in the cured film on the substrate S, but may also be used as training data (learning image) for performing machine learning of the evaluation model. Here, in this embodiment, an example of capturing an image using the wide-angle alignment measuring instrument 151 will be described, but the present invention is not limited thereto, and for example, an image may be captured using the alignment measuring instrument 106 or the imaging unit 112.

[0045] In the above procedure, the image acquisition (step S108) is performed after the imprint process is performed on all of the multiple shot areas, but the present invention is not limited to this. For example, after a pattern is formed in each shot area (after step S106), an image may be acquired for each shot area by a method similar to step S108. Also, an image may be acquired for the substrate S unloaded from the imprint apparatus by a device other than the imprint apparatus by a method similar to step S108.

[0046] In step S109, the substrate S is transported from the substrate holding unit 102 to a destination (for example, an intermediary unit between the imprint apparatus IMP and the post-processing apparatus) by a substrate transport mechanism (not shown). When a lot consisting of multiple substrates is processed, the operation shown in FIG. 5 is performed for each of the multiple substrates.

[0047] Next, an example of an abnormality that occurs in the imprint process will be described. Fig. 6 is a side view of a state in which the mold M and the imprint material IM on the substrate S are in contact (after completion of step S103, for example, steps S104 and S105). "Seeping out" refers to a state in which the imprint material IM protrudes from the pattern region MP of the mold M, as shown in Fig. 6(a). "Unfilled" refers to a state in which a portion not filled with the imprint material IM occurs between the pattern region MP of the mold M and the substrate S, as shown in Fig. 6(b).

[0048] When such an abnormality such as seepage or non-filling occurs, an image as shown in FIG. 7 is acquired in step S108. FIG. 7 is a diagram showing an example of an image acquired in step S108 when an abnormality such as seepage or non-filling occurs, and shows an image including a shot area in which a cured film (pattern) of the imprint material IM is formed by the imprint process and its surroundings. Normally (i.e., when no abnormality occurs), as shown in FIG. 7(a), the imprint material IM is filled up to the boundary 601 of the shot area, and the imprint material IM does not protrude beyond the boundary 601. On the other hand, in the case of non-filling, as shown in FIG. 7(b), the part where the imprint material IM is not filled up to the boundary 601 of the shot area is imaged as white (or black). Also, in the case of seepage, as shown in FIG. 7(c), the imprint material IM protruding beyond the boundary 601 of the shot area is imaged as black (or white).

[0049] FIG. 8 shows an example of a mark used for alignment or inspection that is not filled. Each of FIGS. 8(a) to 8(b) shows an image obtained by photographing the periphery of a shot area and its surroundings. The area shown in gray in each image is the shot area to be inspected that is filled with the imprint material IM. The shot area includes a first mark 701 and a second mark 702 that are different in shape from each other. In FIG. 8(a), the first mark 701 and the second mark 702 are each shown entirely in black. This shows a normal state in which the imprint material IM is filled inside the mark. On the other hand, in FIG. 8(b), a part of each of the first mark 701 and the second mark 702 is shown in white. This shows a state in which the imprint material IM is not sufficiently filled inside the mark in that part, resulting in an unfilled area N.

[0050] When seepage as described above occurs, there is a possibility that the protruding, hardened imprint material IM will come into contact with the mold M during imprint processing of the shot area beyond which the imprint material IM has protruded, damaging the pattern of the mold M. Furthermore, when an unfilled portion occurs, no pattern is formed there, resulting in a defective semiconductor device. Therefore, it is necessary to prevent the above-mentioned defects by observing the presence or absence of abnormalities such as seepage or unfilled portions after the imprint processing and adjusting the imprint conditions.

[0051] As an example of an adjustment method, a method of changing the amount of imprint material IM supplied onto the substrate according to the position, size, and shape of an abnormality (seepage, non-filling) occurring in the film on the substrate is considered. To perform this adjustment, it is necessary to acquire information on the position, size, and shape of the abnormality occurring in the film on the substrate, that is, information on the position, size, and shape of the area where the amount of imprint material IM is insufficient or excessive (hereinafter, this may be referred to as abnormality information). In this embodiment, an abnormality information is acquired from the evaluation image obtained in step S108 using a trained model generated by machine learning. The trained model is a model for evaluating the film formed on the substrate S, and may be referred to as an evaluation model hereinafter. The evaluation model uses the evaluation image acquired in step S108 as input data, and abnormality information, which is the evaluation result of the film formed on the substrate S, as output data. Examples of the evaluation model include the models exemplified below. In the method using these models, whether or not an object to be detected exists is calculated for each pixel of the image, and the detailed shape of the object can be obtained by labeling them. Convolutional Neural Network structure model · Models with an Encoder / Decoder mechanism such as U-net Model based on Region-Convolutional Neural Network (R-CNN)

[0052] Next, an evaluation model (trained model) for evaluating a film on a substrate in the evaluation device 407 will be described with reference to Figs. 9 to 11. Fig. 9(a) is a schematic diagram showing a method for generating an evaluation model by machine learning, and Fig. 9(b) is a schematic diagram showing a method for evaluating a film on a substrate using the generated evaluation model (i.e., a method for acquiring information on an abnormality occurring in a film on a substrate). Fig. 10 is a flowchart showing a method for generating an evaluation model by machine learning. Fig. 11 is a flowchart showing a method for evaluating a film on a substrate using an evaluation model (evaluation method).

[0053] First, a method for generating an evaluation model by machine learning will be described with reference to Fig. 9(a) and Fig. 10. The method for generating an evaluation model by machine learning may be understood as a learning method for the evaluation model. For example, a program (learning program) corresponding to the flowchart of Fig. 10 is stored in the storage device 504 in the evaluation device 407, and can be executed by the control unit 501 (image processing unit) after being loaded into the RAM 502.

[0054] In step S201, the evaluation device 407 acquires a training image 801 (learning image) as training data for the evaluation model. Conventionally, the training image 801 can be acquired by going through a process similar to step S108 in FIG. 1. Specifically, the training image 801 similar to the evaluation image is acquired based on the conditions of the material and the measurement conditions used when acquiring the evaluation image used for evaluating the film on the substrate S. The training image 801 can be, for example, an image obtained by capturing an image of the periphery of a shot area and its surroundings, similar to the evaluation image. Here, the training image 801 may be acquired in a plurality of shot areas on each substrate, using a plurality of substrates. It is desirable to have a large number of training images, and it is preferable that the training images include a sufficient amount of samples of the abnormality to be detected.

[0055] In step S202, the evaluation device 407 acquires anomaly information 802 indicating information on anomalies (seepage, non-filling) contained in each training image 801 acquired in step S201. Conventionally, an operator visually inspects each training image 801 acquired in step S201, and creates anomaly information 802 including information on the category (type), size, position, etc. of the anomaly for all anomalies present in each training image 801. The anomaly information 802 created and input by the operator in this manner can be acquired by the evaluation device 407.

[0056] In step S203, the evaluation device 407 performs machine learning of the evaluation model using the training image 801 acquired in step S201 as training data. For example, the evaluation device 407 performs optimization by using the training image 801 as input data for a neural network created in advance and the anomaly information 802 as output data (teacher data). This makes it possible to generate an evaluation model 803 as a trained model (neural network). Next, in step S204, the evaluation device 407 stores the evaluation model 803 generated in step S203 in the storage unit 804. The storage unit 804 may be understood as the storage device 504 of the evaluation device 407.

[0057] Next, a method for evaluating a film on a substrate using the evaluation model (i.e., a method for detecting an abnormality occurring in a film on a substrate) will be described with reference to Fig. 9(b) and Fig. 11. A program (evaluation program) corresponding to the flowchart in Fig. 11 is stored in the storage device 504 in the evaluation device 407, for example, and can be executed by the control unit 501 (image processing unit) after being loaded into the RAM 502.

[0058] In step S301, the evaluation device 407 reads an evaluation model that outputs an evaluation result of a film based on an image of the film formed on a substrate by the imprint device IMP. The evaluation model is a model that outputs, when an evaluation image of a film formed on a substrate by the imprint device IMP is input, information on an abnormality occurring in the evaluation image as an evaluation result. The evaluation model is generated according to the above-mentioned flowchart of FIG. 10 based on the conditions of the imprint material to be inspected and an image acquired under conditions similar to the measurement conditions of the evaluation image acquired in step S108, and is stored in the storage unit 804.

[0059] In step S302, the evaluation device 407 reads the evaluation image acquired in step S108. Next, in step S303, the evaluation device 407 detects anomaly information (characteristics of anomaly) occurring in the evaluation image as an evaluation of a portion of the film on the substrate corresponding to the evaluation image based on the evaluation model read in step S301 and the evaluation image read in step S302. Specifically, the evaluation device 407 inputs the evaluation image into the evaluation model, thereby acquiring "anomaly information (characteristics of anomaly) occurring in the evaluation image" output from the evaluation model as an evaluation result. The anomaly information (characteristics of anomaly) is obtained for each anomaly present in the evaluation image. The anomaly information (characteristics of anomaly) may include the category (type), position, size, and shape of the anomaly, as well as the type of anomaly (seepage / unfilled), the coordinates of the vertices of a rectangle surrounding the anomaly area, and the certainty of the detected anomaly. Here, the certainty of a detected anomaly is a value that indicates the reliability of the evaluation result (detection result of anomaly information), and is automatically calculated by the evaluation model for each detected anomaly. As mentioned above, types of anomalies include seepage and non-filling, but other types of anomalies can also be detected by having the evaluation model learn.

[0060] In step S304, the evaluation device 407 performs post-processing on the anomaly information obtained by the evaluation model in step S303. For example, the post-processing may include classifying the anomaly by comparing the certainty of the anomaly with a specific threshold for each detected anomaly. As an example, when the certainty is expressed as a value between 0 and 1, an anomaly with a certainty of 0.5 or less can be classified as an anomaly that is not subject to detection. Classification conditions such as thresholds when classifying based on the certainty can be changed depending on various data such as the imprint material to be inspected, recipe information, dimming conditions at the time of imaging, and mode at the time of imaging.

[0061] In step S305, the evaluation device 407 judges whether evaluation results (inspection results) have been obtained for all of the multiple evaluation images obtained by the imprint device IMP. If evaluation results have been obtained for all evaluation images, the process ends. On the other hand, if there is an evaluation image for which evaluation results have not been obtained, the process proceeds to step S302, and steps S302 to S304 are executed for that evaluation image. In this way, steps S302 to S304 are repeatedly executed so that evaluation results (inspection results) are obtained for each of the multiple evaluation images obtained by the imprint device IMP.

[0062] In the above example, the evaluation device 407 is described as an information processing device that executes both learning and evaluation (inspection), but the information processing device that performs learning and the information processing device that performs evaluation (inspection) may be configured separately. In this case, the first information processing device creates an evaluation model and transfers the evaluation model to the second information processing device that performs evaluation. The second information processing device uses the evaluation model transferred from the first information processing device to evaluate (inspect) the film on the substrate based on the input evaluation image.

[0063] Incidentally, in order to generate the above-mentioned evaluation model with high accuracy by machine learning, it is necessary to prepare (collect) a large number (sufficient amount) of training images with abnormality features as training data in advance. However, in order to prepare a large number of training images as training data, it is cumbersome and may take a considerable amount of time to perform imprint processing by changing conditions so that the positions and sizes of abnormalities (seepage, non-filling) contained in the training images are changed. In addition, if an operator visually inspects each of the large number of prepared training images to obtain abnormality information of each training image as teacher data, the operator's work hours may increase. In particular, since it is difficult to control the frequency of occurrence of abnormalities, it may be necessary to perform a larger number of imprint processes in order to obtain a large number of training images that can be used as training data for the evaluation model.

[0064] Therefore, in this embodiment, a design image showing at least a part of the geometric characteristics of the substrate S is processed to generate a training image as training data for the evaluation model. More specifically, the design image is processed to generate a training image to which abnormalities that may occur in the film on the substrate, such as seepage or non-filling, are pseudo-imparted. This makes it possible to easily generate training data (training image) for the evaluation model, and to reduce the imprint process required to prepare the training image and the number of worker's work steps. Note that the generation of the training image in this embodiment can be executed by the generation device 408 (control unit 501).

[0065] First, a basic idea of ​​generating anomalies (seepage, non-filling) to be added to a training image will be described. In this embodiment, anomalies such as seepage or non-filling of the imprint material IM are generated in a pseudo manner by processing a design image showing at least a part of the geometric characteristics of the substrate. The design image may be an image including the boundary of the shot area as at least a part of the geometric characteristics of the substrate S, or an image including a mark used in alignment or inspection. As an example, the design image including the boundary of the shot area may be an image in which no anomalies (seepage, non-filling) occur, as shown in FIG. 7(a). Here, the design image may be obtained from design information expressed in the form of an image relating to at least a part of the substrate S, but may also be obtained from design information in which at least a part of the substrate is expressed by vertices, lines, or polygons. The design image may also be obtained by capturing an image of a substrate on which imprint processing has been normally performed under similar conditions to those of the substrate to be inspected, or by measuring the substrate using a device for measuring the surface shape.

[0066] As described above with reference to FIG. 7(b), in the case of unfilled, the portion where the imprint material IM is not filled up to the boundary 601 of the shot area is imaged as white (or black). Therefore, by processing the design image (i.e., applying image processing) so that the area inside the boundary 601 becomes white (or black), it is possible to simulate the unfilled state of the imprint material IM. On the other hand, as described above with reference to FIG. 7(c), in the case of seepage, the imprint material IM that protrudes beyond the boundary 601 of the shot area is imaged as black (or white). Therefore, by processing the design image (i.e., applying image processing) so that the area outside the boundary 601 becomes black (or white), it is possible to simulate the seepage of the imprint material IM.

[0067] The marks 701-702 used in alignment and inspection are similar to the boundary 601 of the shot area. As shown in FIG. 8B, in the case of unfilled areas, the portions of the marks 701-702 that are not filled with the imprint material IM are imaged as white (or black). Therefore, by processing the design image (i.e., applying image processing) so that a portion of each of the marks 701-702 becomes white (or black), it is possible to create a pseudo unfilled area in each of the marks 701-702. Here, there are cases where gas bubbles accumulate inside the imprint material IM during filling with the imprint material IM, resulting in an unfilled area of ​​the imprint material IM. In this case, the bubbles tend to be circular in areas where there is no pattern, and to have a shape that follows the pattern in areas where a pattern exists. Therefore, by processing the design image so that the shape of the unfilled area follows this tendency, it is possible to create a pseudo unfilled area that is closer to reality.

[0068] Furthermore, in this embodiment, the category (type), size, position, etc. of the anomaly pseudo-assigned to the training image are known information to the generating device 408 (information processing device) that generated the training image. Therefore, the generating device 408 can automatically generate anomaly information indicating the characteristics (category, size, position, etc.) of the anomaly contained in the training image when creating the training image. That is, the generating device 408 can easily generate anomaly information about the training image as teacher data without visual inspection by an operator.

[0069] In this manner, in this embodiment, by processing the design image, a training image can be easily generated as training data for the evaluation model. In addition, abnormality information can be easily generated as teacher data for the evaluation model. In other words, in this embodiment, since a large amount of training data and teacher data for the evaluation model can be easily generated, machine learning of the evaluation model can be performed with high accuracy using the large amount of training data and teacher data. As a result, the trained evaluation model can accurately evaluate (i.e., detect abnormalities in) the film of the imprint material IM formed on the substrate by the imprint apparatus IMP based on the evaluation image actually captured in step S108.

[0070] Next, a method for generating training images to which anomalies are artificially added as training data for an evaluation model will be described. Fig. 12 is a flowchart showing the method for generating training images. A program (generation program) corresponding to the flowchart of Fig. 12 is stored in the storage device 504 in the generating device 408, for example, and can be executed by the control unit 501 (image processing unit) after being loaded into the RAM 502.

[0071] In step S401, the generating device 408 acquires a reference image representing at least a portion of the substrate S. The reference image is an image actually captured by the imprinting device IMP, and is preferably an image free of abnormalities such as seepage or non-filling. For example, the reference image may be an image (hereinafter, sometimes referred to as a post-processing image) obtained by capturing an image of at least a portion of the substrate S on which a film of the imprinting material IM has been formed by the imprinting process. The reference image may be an evaluation image previously obtained in the above-mentioned step S108.

[0072] In step S402, the generating device 408 acquires a design image for the same region of the substrate S as the reference image acquired in step S401 (i.e., at least a part of the substrate acquired as the reference image). As described above, the design image is an image showing at least a part of the geometric characteristics of the substrate S. The design image may be created from design information expressed in the form of an image relating to at least a part of the substrate S, or may be created from design information in which at least a part of the substrate is represented by vertices, lines, or polygons. FIGS. 13(a)-(b) show an example of a design image. The design image 900a shown in FIG. 13(a) includes a shot region boundary 901 (i.e., a boundary of a region to be filled with the imprint material IM) as at least a part of the geometric characteristics of the substrate S. Moreover, the design image 900b shown in FIG. 13(b) includes the shot region boundary 901 and marks 902-903 as at least a part of the geometric characteristics of the substrate S. Note that the design image may include only the marks 902-903 without including the shot region boundary 901.

[0073] In step S403, the generating device 408 acquires an image (hereinafter, sometimes referred to as a pre-processing image) obtained by imaging the substrate S before a film of the imprint material IM is formed by the imprinting process. The pre-processing image is preferably acquired under the same imaging conditions as those used to acquire the post-processing image (reference image). The pre-processing image is used as an image representing an anomaly (seepage, non-filling) to be pseudo-imparted to the post-processing image (reference image). In other words, the pre-processing image is used as an image representing a pseudo anomaly to be imparted to an anomaly-imparted region defined by a region defining image described below.

[0074] In step S404, the generating device 408 creates an area defining image for defining an anomaly-imparted area in the processed image (reference image) acquired in step S401, based on the design image acquired in step S402. The anomaly-imparted area is an area in the processed image (reference image) to which an anomaly (seepage, non-filling) should be artificially imparted. The area defining image is an image that indicates the position, size, shape, etc. of the anomaly-imparted area in the processed image (reference image). For example, the generating device 408 may create the area defining image by processing (i.e., applying image processing) the design image acquired in step S402. A specific example of creating the area defining image in this step S403 will be described later.

[0075] In step S405, the generating device 408 generates training images by pseudo-adding anomalies to the anomaly-added regions of the processed image (reference image) acquired in step S401 based on the region defining image created in step S404. That is, the generating device 408 generates training images by pseudo-adding anomalies to the anomaly-added regions of the processed image (reference image). The pre-processing image acquired in step S403 is used as an image showing anomalies pseudo-added to the anomaly-added regions of the processed image (reference image). A specific example of generating training images in step S404 will be described later.

[0076] 14 is a schematic diagram for explaining the processing performed by the generating device 408. As described above, the design image 1301 is acquired in step S402, the processed image 1302 as a reference image is acquired in step S401, and the pre-processing image 1303 is acquired in step S403. These images (design image 1301, processed image 1302, pre-processing image 1303) are input to an image processing unit 1304 (control unit 501) of the generating device 408. The image processing unit 1304 may include a creation unit 1307 and a generating unit 1308.

[0077] The creation unit 1307 creates an abnormal part image 1306 and a normal part image 1309 based on the design image 1301. The abnormal part image 1306 is an image for opening an abnormality-imparted region in the processed image 1302 (reference image) and masking a region other than the abnormality-imparted region. That is, the abnormal part image 1306 is an image in which an abnormality is imparted only to the abnormality-imparted region and a region other than the abnormality-imparted region is masked so that an abnormality is not imparted to the region other than the abnormality-imparted region. The abnormal part image 1306 may be understood as a region defining image for defining an abnormality-imparted region in the processed image 1302 (reference image). Moreover, the normal part image 1309 is an image for opening a region other than the abnormality-imparted region in the processed image 1302 (reference image) and masking the abnormality-imparted region. The normal part image 1309 may be understood as an image in which the mask region in the abnormal part image 1306 is inverted. The abnormal area image 1306 and normal area image 1309 created by the creation unit 1307 are input to a generation unit 1308 .

[0078] The generating unit 1308 generates a training image 1305 to which an abnormality is pseudo-added based on the design image 1301, the processed image 1302 (reference image), the pre-processing image 1303, the abnormal part image 1306, and the normal part image 1309. The training image 1305 may be understood as a pseudo defect image having a pseudo defect. The abnormal part image 1306 created by the creating unit 1307 and the training image 1305 generated by the generating unit 1308 are stored in the storage unit 1310 (e.g., the storage device 504).

[0079] Next, the processing contents performed by the creation unit 1307 will be described with reference to Fig. 15. Fig. 15 is a schematic diagram for explaining the processing performed by the creation unit 1307. In each of the abnormal area image 1306 and normal area image 1309 in Fig. 15, white parts indicate opened areas, and black parts indicate masked areas.

[0080] The creation unit 1307 creates (outputs) an abnormal part image 1306 (region defining image) for defining an abnormality-imparting region in the processed image 1302 (reference image), and a normal part image 1309 which is an inverted image of the abnormal part image 1306. The abnormal part image 1306 and the normal part image 1309 are two-dimensional images in which each pixel has a value of "0" or "1". When the design image 1301 is input, the creation unit 1307 creates the abnormal part image 1306 by setting restrictions on the region to which an abnormality (seepage, non-filling) is to be added, based on the design image 1301. The creation unit 1307 also creates the normal part image 1309 by inverting a mask region in the abnormal part image 1306.

[0081] For example, when adding a pseudo defect simulating an unfilled portion, the creation unit 1307 performs geometric transformation of a geometric feature in the design image 1301 as shown in FIG. 15(a) to generate a geometrically transformed image 1311 (first image). In FIG. 15(a), a shot region boundary 1301a is illustrated as an example of the geometric feature in the design image 1301. The creation unit 1307 then calculates the difference between the design image 1301 and the geometrically transformed image 1311, and creates an abnormal portion image 1306 in which the region where the difference occurs is opened as an abnormality addition region and the region other than the calculated region is masked. The creation unit 1307 also creates a normal portion image 1309 by inverting the mask region in the abnormal portion image 1306.

[0082] As shown in FIG. 15(b), the creation unit 1307 can also create the abnormal part image 1306 and the normal part image 1309 by adding noise components to the design image 1301. In this case, the creation unit 1307 generates a first noise image 1312 using a texture creation method such as Perlin noise, and generates a second noise image 1313 by performing threshold processing (image binarization) on the first noise image 1312. The first noise image 1312 may be understood as an image having a random noise pattern. The second noise image 1313 is a noise image obtained by performing threshold processing on each pixel of the first noise image 1312, and may be understood as an image (second image) having noise components to be added to the design image 1301. Then, the creation unit 1307 creates the abnormal part image 1306 by calculating the Hadamard product of the design image 1301 and the second noise image. Moreover, the creating unit 1307 creates a normal area image 1309 by inverting the mask region in the abnormal area image 1306 .

[0083] Next, the processing contents performed by the generating unit 1308 will be described with reference to Fig. 16. Fig. 16 is a schematic diagram for explaining the processing performed by the generating unit 1308. Note that Fig. 16 describes an example of generating a training image 1305 (pseudo defect image) using an abnormal part image 1306 and a normal part image 1309 created by adding a noise component to a design image 1301 as shown in Fig. 15(b). However, even when using an abnormal part image 1306 and a normal part image 1309 created by performing a geometric transformation of the design image 1301 as shown in Fig. 15(a), the training image 1305 can be generated by a similar method.

[0084] The generating unit 1308 calculates the Hadamard product between the abnormal area image 1306 and the pre-processing image 1303 to generate an abnormal area image 1321 in which the pre-processing image 1303 is reflected (inserted) into the opening area (white portion) of the abnormal area image 1306. Here, the generating unit 1308 may use a luminance-converted image 1320 in which the luminance of the post-processing image 1302 (reference image) is converted, instead of the pre-processing image 1303. The luminance-converted image 1320 is generated, for example, by converting the luminance of the post-processing image 1302 so that it is approximately the same as the luminance of the image of the substrate S before the imprint processing is performed, and is used as an image representing an abnormality that is pseudo-added to the post-processing image 1302. In this case, the generating unit 1308 calculates the Hadamard product between the abnormal area image 1306 and the luminance-converted image 1320 to generate the abnormal area image 1321.

[0085] Furthermore, the generating unit 1308 calculates the Hadamard product between the normal portion image 1309 and the processed image 1302 (reference image) to generate a normal area image 1322 in which the processed image 1302 is reflected (inserted) in the opening region (white portion) of the normal portion image 1309. Then, the generating unit 1308 adds the abnormal area image 1321 and the normal area image 1322. In this way, the generating unit 1308 can generate a training image 1305 (pseudo defect image) in which an abnormality is pseudo-added to the abnormality-added region of the processed image 1302 (reference image).

[0086] Here, in this embodiment, the unfilled area is exemplified, but the training image 1305 can be generated for abnormalities such as seepage, mark abnormalities, patterns other than the marks formed on the shot area, and marks and patterns inside the shot area in the same manner as for the unfilled area. In addition, in this embodiment, an example of generating a training image using the unprocessed image 1303 or the brightness conversion image 1320 is described, but the present invention is not limited to this. For example, the training image 1305 can be generated by directly adding or multiplying the abnormal part image 1306 to the processed image 1302 without using the unprocessed image 1303 or the brightness conversion image 1320. In particular, seepage often appears extremely black compared to normal areas, and the training image 1305 to which the abnormality is pseudo-added with high accuracy can be generated without using the unprocessed image 1303. In this case, the step S403 in FIG. 12 can be omitted.

[0087] In this embodiment, post-processing such as smoothing processing and contrast adjustment may be performed on the abnormal part image 1306, normal part image 1309, and / or training image 1305 generated as described above. By performing post-processing, it is possible to generate a training image 1305 to which an abnormality is pseudo-added with higher accuracy. In addition, in this embodiment, an example has been described in which the training image 1305 generated by steps S401 to S405 in FIG. 12 is used as training data to train an evaluation model. However, this is not limited to this, and in addition to the training image 1305, images visually inspected by an operator or abnormality information may be used as training data to train an evaluation model as described in step S202 in FIG. 10.

[0088] As described above, in this embodiment, a training image 1305 to which a pseudo abnormality that may occur in a film on a substrate is imparted is generated as training data for an evaluation model by processing a design image 1301 showing at least a part of the geometric characteristics of a substrate S. According to this embodiment, training data (training image) for an evaluation model can be easily generated, and the implementation of an imprint process for preparing a training image and the number of worker's work steps can be reduced.

[0089] <Embodiment of the article manufacturing method> The article manufacturing method according to the embodiment of the present invention is suitable for manufacturing articles such as microdevices such as semiconductor devices and elements having a fine structure. The article manufacturing method of the present embodiment may include a forming step of forming a film of a composition on a substrate by a film forming apparatus (e.g., an imprinting apparatus) in a film forming system, a processing step of processing the substrate that has undergone the forming step, and a manufacturing step of manufacturing an article from the substrate that has undergone the processing step. Furthermore, such a manufacturing method includes other well-known steps (oxidation, film formation, deposition, doping, planarization, etching, resist stripping, dicing, bonding, packaging, etc.). The article manufacturing method of the present embodiment is advantageous in at least one of the performance, quality, productivity, and production cost of the article compared to conventional methods.

[0090] When an imprinting apparatus is used as the film forming apparatus, the pattern of the cured material formed by using the imprinting apparatus is used permanently on at least a part of various articles, or temporarily when manufacturing various articles. The articles include electric circuit elements, optical elements, MEMS, recording elements, sensors, and molds. Examples of electric circuit elements include volatile or non-volatile semiconductor memories such as DRAM, SRAM, flash memory, and MRAM, and semiconductor elements such as LSI, CCD, image sensors, and FPGA. Examples of molds include molds for imprinting.

[0091] The pattern of the cured product is used as it is as at least a part of a component of the article, or is used temporarily as a resist mask, which is removed after etching or ion implantation in a substrate processing step.

[0092] Next, a specific article manufacturing method will be described. Here, an example of an article manufacturing method using an imprinting apparatus will be described with reference to Fig. 17. In step SA of Fig. 17, a substrate 1z such as a silicon substrate having a workpiece 2z such as an insulator formed on its surface is prepared, and then an imprinting material 3z is applied to the surface of the workpiece 2z by an inkjet method or the like. Here, a state in which multiple droplets of the imprinting material 3z have been applied onto the substrate is shown.

[0093] In step SB of Fig. 17, the mold 4z for imprinting is placed with the side on which the concave-convex pattern is formed facing the imprint material 3z on the substrate. In step SC of Fig. 17, the substrate 1z to which the imprint material 3z has been applied is brought into contact with the mold 4z, and pressure is applied. The imprint material 3z fills the gap between the mold 4z and the workpiece 2z. When light is irradiated through the mold 4z in this state as energy for curing, the imprint material 3z is cured.

[0094] 17, after the imprint material 3z is cured, the mold 4z and the substrate 1z are separated, and a pattern of the cured product of the imprint material 3z is formed on the substrate 1z. In this cured product pattern, the recesses of the mold correspond to the protrusions of the cured product, and the protrusions of the mold correspond to the recesses of the cured product, i.e., the recessed and protruding pattern of the mold 4z is transferred to the imprint material 3z.

[0095] In step SE of Fig. 17, etching is performed using the pattern of the cured material as an etching-resistant mask, and the portions of the surface of the workpiece 2z where there is no cured material or where only a thin layer remains are removed to form grooves 5z. In step SF of Fig. 17, the pattern of the cured material is removed to obtain an article in which grooves 5z are formed on the surface of the workpiece 2z. Here, the pattern of the cured material is removed, but it may be used as an interlayer insulating film included in a semiconductor element or the like, that is, a component of an article, without being removed after processing.

[0096] <Other embodiments> The present invention can also be realized by a process in which a program for implementing one or more of the functions of the above-described embodiments is supplied to a system or device via a network or a storage medium, and one or more processors in a computer of the system or device read and execute the program. The present invention can also be realized by a circuit (e.g., ASIC) for implementing one or more of the functions.

[0097] <Summary of the embodiment> The disclosure of this specification includes at least the following generating method, learning method, program, generating device, evaluation device, film forming system, and article manufacturing method. (Item 1) A method for generating training data for a model for evaluating a film of a composition formed on a substrate using a mold, comprising: A generation method comprising the steps of: generating a training image as the training data by processing a design image showing at least a portion of the geometric characteristics of the substrate. (Item 2) 2. The method of claim 1, further comprising the step of processing the design image to generate the training image having simulated abnormalities that may occur in the membrane. (Item 3) 3. The method for generating the training image according to item 2, further comprising the steps of: creating an area defining image for defining an anomaly-added area in a reference image representing at least a portion of the design image based on the design image; and generating the training image by pseudo-adding the anomaly to the anomaly-added area of ​​the reference image based on the area defining image. (Item 4) 4. The generating method according to item 3, further comprising the steps of: performing a geometric transformation of the geometric features in the design image; and generating the region defining image based on a first image obtained thereby. (Item 5) 5. The generating method according to item 4, characterized in that an image in which an area where a difference occurs between the design image and the first image is set as the abnormality-added area is created as the area defining image. (Item 6) 6. The method according to any one of items 3 to 5, further comprising generating the area defining image based on the design image and a second image having a noise component. (Item 7) The method for generating the training image is characterized in that, based on the region definition image, a portion of an image obtained by converting the luminance of the reference image is pseudo-assigned as the abnormality to the abnormality-assigned region of the reference image. (Item 8) The method for generating the training image is characterized in that, based on the region defining image, a portion of an image of the substrate obtained before the film is formed is pseudo-assigned as the anomaly to the anomaly-assigned region of the reference image. (Item 9) The method of any one of claims 2 to 8, wherein the anomalies include at least one of seepage of the composition from an area on the substrate where the film is to be formed and non-filling of the composition in the area. (Item 10) 10. The generating method according to any one of items 1 to 9, wherein the design image includes, as the geometric features, information on the boundary of an area of ​​the substrate where a pattern is to be formed. (Item 11) 11. The method of claim 1, wherein the design image includes information of a mark to be formed on the substrate as the geometric features. (Item 12) 12. The method according to any one of claims 1 to 11, characterized in that the design image is obtained from design information expressed in the form of an image relating to at least the portion. (Item 13) 13. The method according to any one of items 1 to 12, wherein the design image is obtained from design information, at least a portion of which is represented by vertices, line segments, or polygons. (Item 14) 14. A program for causing a computer to execute the generating method according to any one of items 1 to 13. (Item 15) A model learning method for inputting an image of a film of a composition formed on a substrate using a mold and outputting an evaluation result of the film, the method comprising: A learning method, comprising: learning the model using training images generated by the generation method according to any one of items 1 to 13 as training data. (Item 16) Item 16. A program for causing a computer to execute the learning method according to item 15. (Item 17) A generator for generating training data for a model for evaluating a film of a composition formed on a substrate using a mold, the generator comprising: A generating device that generates a training image as the training data by processing a design image showing at least a portion of the geometric characteristics of the substrate. (Item 18) An evaluation apparatus for evaluating a film of a composition formed on a substrate using a mold, comprising: a model that outputs an evaluation result of the film by inputting an image of the film; 18. An evaluation device, comprising: a generating device according to claim 17, configured to generate training images for the model; (Item 19) a film forming apparatus for forming a film of a composition on a substrate using a mold; Item 19. An evaluation apparatus according to item 18, which evaluates the film based on an image of the film formed on the substrate by the film forming apparatus; A film forming system comprising: (Item 20) A forming step of forming a film of the composition on a substrate by a film forming apparatus in the film forming system according to item 19; a processing step of processing the substrate that has been subjected to the forming step; a manufacturing process for manufacturing an article from the substrate that has been subjected to the processing process; A method for manufacturing an article, comprising:

[0098] The invention is not limited to the above-described embodiments, and various modifications and variations are possible without departing from the spirit and scope of the invention. Accordingly, the following claims are appended to apprise the public of the scope of the invention. [Explanation of symbols]

[0099] M: mold, S: substrate, IM: imprint material (composition), IMP: imprint device (film forming device), 407: evaluation device, 408: generation device, 1301: design image, 1302: processed image (reference image), 1303: pre-processed image, 1304: image processing unit, 1305: training image, 1306: abnormal part image (region defining image), 1307: creation unit, 1308: generation unit, 1309: normal part image

Claims

1. A method for generating training data for a model for evaluating a film of a composition formed on a substrate using a mold, comprising: A generation method comprising the steps of: generating a training image as the training data by processing a design image showing at least a portion of the geometric characteristics of the substrate.

2. The generating method according to claim 1, further comprising the step of processing the design image to generate the training image having simulated abnormalities that may occur in the membrane.

3. The generation method described in claim 2, characterized in that an area defining image is created based on the design image to define an anomaly-added area in a reference image representing at least a portion of the design image, and the training image is generated by simulating the anomaly in the anomaly-added area of ​​the reference image based on the area defining image.

4. 4. The generating method according to claim 3, further comprising the step of: performing a geometric transformation of the geometric features in the design image; and generating the area defining image based on a first image obtained thereby.

5. The generating method according to claim 4 , further comprising the step of generating an image in which a region where a difference occurs between the design image and the first image is set as the abnormality-added region as the region defining image.

6. The generating method according to claim 3 , further comprising generating the area defining image based on the design image and a second image having a noise component.

7. The generation method described in claim 3, characterized in that the training image is generated by pseudo-assigning a portion of an image obtained by converting the luminance of the reference image as the abnormality to the abnormality-assignment area of ​​the reference image based on the area definition image.

8. The generation method described in claim 3, characterized in that the training image is generated by pseudo-adding a portion of an image of the substrate obtained before forming the film as the anomaly to the anomaly-added area of ​​the reference image based on the area definition image.

9. 3. The method of claim 2, wherein the anomalies include at least one of the composition seeping out of an area on the substrate where the film is to be formed and the composition not filling the area.

10. The generating method according to claim 1 , wherein the design image includes, as the geometric features, information on a boundary of an area of ​​the substrate where a pattern is to be formed.

11. The generating method according to claim 1 , wherein the design image includes information of a mark to be formed on the substrate as the geometric features.

12. The method of claim 1 , wherein the design image is derived from design information represented in the form of an image relating to the at least part.

13. The generating method according to claim 1 , wherein the design image is obtained from design information, the at least a portion of which is represented by vertices, line segments, or polygons.

14. A program for causing a computer to execute the generating method according to any one of claims 1 to 13.

15. A model learning method for inputting an image of a film of a composition formed on a substrate using a mold and outputting an evaluation result of the film, the method comprising: A learning method, comprising: learning said model using training images generated by the generation method according to any one of claims 1 to 13 as training data.

16. A program for causing a computer to execute the learning method according to claim 15.

17. A generator for generating training data for a model for evaluating a film of a composition formed on a substrate using a mold, the generator comprising: A generating device that generates a training image as the training data by processing a design image showing at least a portion of the geometric characteristics of the substrate.

18. An evaluation apparatus for evaluating a film of a composition formed on a substrate using a mold, comprising: a model that outputs an evaluation result of the film by inputting an image of the film; 20. An evaluation device, comprising: a generation device according to claim 17; and a training image generated by the generation device according to claim 17, which is used as training data to train the model.

19. a film forming apparatus for forming a film of a composition on a substrate using a mold; The evaluation device according to claim 18 , which evaluates the film based on an image of the film formed on the substrate by the film forming device; A film forming system comprising:

20. A forming step of forming a film of a composition on a substrate by a film forming device in the film forming system according to claim 19; a processing step of processing the substrate that has been subjected to the forming step; a manufacturing process for manufacturing an article from the substrate that has been subjected to the processing process; A method for manufacturing an article, comprising:

Images