Lithographic device, information processing apparatus, estimation method, program, and method for manufacturing article

JP2024165899APending Publication Date: 2024-11-28CANON KK
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2023082483
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-18
Publication Date
2024-11-28

AI Technical Summary

Benefits of technology

【0010】 本発明によれば、例えば、計測器から出力されたアライメント画像から得られる情報であるアライメントマーク像情報のデータを効率的に利用して、基板上に形成されるパターンの位置ずれ量を推定するのに有利な技術を提供することができる。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024165899000001_ABST
    Figure 2024165899000001_ABST
Patent Text Reader

Abstract

To provide a technique in a lithographic device that performs processing of transferring a pattern of an original plate to a substrate, the technique advantageous for estimating the amount of positional deviation of the pattern formed on the substrate.SOLUTION: A lithographic device has: a plurality of measuring instruments that each measure marks provided on an original plate and a substrate and output the amount positional deviation between the original plate and the substrate; and an estimation unit that performs alignment between the original plate and the substrate on the basis of the amount of positional deviation by using a first learning model and a second learning model different from the first learning model, and estimates an alignment error that is an error in the measurement of the amount of positional deviation from mark image information obtained from mark images obtained from observation made by the measuring instruments.SELECTED DRAWING: Figure 8
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a lithography apparatus, an information processing apparatus, an estimation method, a program, and a method for manufacturing an article. [Background technology]

[0002] With the increasing demand for miniaturization of semiconductor devices and Micro Electro Mechanical Systems (MEMS), imprinting technology has been attracting attention in addition to conventional photolithography technology. Imprinting technology is a microfabrication technology that forms a pattern of imprinting material on a substrate that corresponds to the pattern of the mold by molding an imprinting material placed (supplied) on a substrate with a mold (master or mold). With imprinting technology, it is possible to form fine structures on the order of several nanometers on a substrate.

[0003] One of the imprinting technologies is the photocuring method. In an imprinting device that employs the photocuring method, first, an imprinting material is placed on a substrate, and the mold (pattern) and the substrate (shot area) are aligned while the mold and the imprinting material on the substrate are brought into contact (imprinted), filling the mold (pattern). Then, while the imprinting material on the substrate and the mold are in contact, light is irradiated to harden the imprinting material, and the mold is detached from the hardened imprinting material on the substrate to form a pattern of the imprinting material on the substrate.

[0004] In such an imprint apparatus, it is important to form a pattern by aligning the mold and the substrate with high accuracy, and therefore, alignment-related technologies have been proposed in the past (see Patent Document 1). Patent Document 1 discloses a technology for inferring a correction value (alignment correction value) when aligning the mold and the substrate using a neural network (multilayer perceptron). In the technology disclosed in Patent Document 1, a neural network is trained using device information when the pattern is formed and the amount of misalignment of the pattern obtained using an overlay inspection device or the like, and the amount of misalignment of the pattern is estimated from only the device information to calculate an alignment correction value. Such a technology is effective as a method for estimating an alignment correction value, and can realize highly accurate alignment between the mold and the substrate. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2019-102495 A Summary of the Invention [Problem to be solved by the invention]

[0006] In estimation using statistical methods including machine learning, the accuracy of the estimation generally depends on the number of data used to train the learning model. Even when estimating the amount of misalignment of a pattern formed on a substrate, a considerable amount of data must be prepared to achieve the accuracy expected to be effective. Furthermore, if the tendency of measurement errors differs for each measuring instrument used for alignment, a learning model must be generated (created) for each measuring instrument. In such a case, the learning model for each measuring instrument is generated using only the data acquired by each measuring instrument. Therefore, the data that can be collected from one substrate is limited, and collecting more data requires more costs (such as increasing the number of substrates from which data is collected).

[0007] An exemplary object of the present invention is to provide a technology that is advantageous for estimating the amount of misalignment of a pattern formed on a substrate by efficiently utilizing alignment mark image information data, which is information obtained from an alignment image output from a measuring instrument. [Means for solving the problem]

[0008] In order to achieve the above object, a lithography apparatus according to one aspect of the present invention is a lithography apparatus that performs a process of transferring a pattern of an original onto a substrate, and includes a plurality of measuring instruments that measure marks provided on the original and the substrate and each output an amount of misalignment between the original and the substrate, and an estimation unit that uses a first learning model and a second learning model different from the first learning model to align the original and the substrate based on the amount of misalignment, while estimating an alignment error, which is a measurement error of the amount of misalignment, from mark image information obtained from a mark image observed by the measuring instruments, and the first learning model uses a first data group including all of the mark image information output from each of the plurality of measuring instruments as a learning data set for an input side of the first learning model. the second learning model is a learning model generated using as learning data on the output side of the first learning model a fourth data group including the alignment errors output by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model for each of the plurality of measuring instruments, and a fifth data group including the alignment errors corresponding to each of the mark image information included in the third data group as learning data on the output side of the second learning model.

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

[0010] According to the present invention, for example, it is possible to provide a technology that is advantageous for estimating the amount of positional deviation of a pattern formed on a substrate by efficiently utilizing alignment mark image information data, which is information obtained from an alignment image output from a measuring instrument. [Brief description of the drawings]

[0011] [Figure 1] 1 is a diagram for explaining a configuration of an imprint apparatus according to one aspect of the present invention. [Diagram 2] FIG. 1 is a schematic diagram showing a configuration of a lithography system. [Diagram 3] 1 is a flowchart for explaining the operation of an imprint apparatus or a lithography system. [Figure 4] 10A to 10C are diagrams for explaining an example of a process for calculating the position of an alignment mark. [Diagram 5] 11 is a flowchart illustrating an example of a process for calculating the position of an alignment mark. [Figure 6] FIG. 1 is a diagram for explaining an overview of learning processing and estimation processing in the prior art. [Figure 7] FIG. 2 is a diagram for explaining an overview of a learning process in the present embodiment. [Figure 8] FIG. 2 is a diagram for explaining an outline of an estimation process in the present embodiment. [Figure 9] FIG. 2 is a diagram for explaining an overview of a learning process in the present embodiment. [Figure 10] FIG. 2 is a diagram for explaining an outline of an estimation process in the present embodiment. [Figure 11] 8 is a flowchart for explaining the learning process shown in FIG. 7. [Figure 12] 9 is a flowchart for explaining the estimation process shown in FIG. 8. [Figure 13] 10 is a flowchart for explaining the learning process shown in FIG. [Figure 14] 11 is a flowchart for explaining the estimation process shown in FIG. 10. [Figure 15] 1A to 1C are diagrams 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] 1(a) and 1(b) are diagrams for explaining the configuration of an imprint apparatus IMP according to one aspect of the present invention. The imprint apparatus IMP is a lithography apparatus that is employed in a lithography process, which is a manufacturing process for devices such as semiconductor elements, liquid crystal display elements, and magnetic storage media as articles, and forms a pattern on a substrate. The imprint apparatus IMP performs an imprint process in which a pattern of the imprint material is formed on the substrate by molding the imprint material on the substrate using a mold (original). In this embodiment, the imprint apparatus IMP brings uncured imprint material arranged (supplied) on the substrate into contact with the mold, and applies energy for curing to the imprint material to form a pattern of a cured material to which the pattern of the mold has been transferred.

[0014] The imprint material is a material (curable composition) that is cured by applying energy for curing. The energy for curing may be electromagnetic waves, heat, or the like. The electromagnetic waves include, for example, light having a wavelength selected from the range of 10 nm to 1 mm, specifically infrared rays, visible rays, ultraviolet rays, and the like.

[0015] The curable composition is a composition that is cured by irradiation with light or by heating. 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, a polymer component, and the like.

[0016] The imprint material may be applied onto the substrate by a liquid jet head in the form of droplets, or in the form of islands or a film formed by a plurality of droplets connected together. The viscosity of the imprint material (at 25° C.) is, for example, 1 mPa·s or more and 100 mPa·s or less.

[0017] The substrate may be made of glass, ceramics, metal, semiconductor, resin, etc., and may have a member made of a material different from that of the substrate formed on its surface as required. Specifically, the substrate may be made of a silicon wafer, a compound semiconductor wafer, quartz glass, etc.

[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 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, the Y direction, and the 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 specified based on the coordinates of the X-axis, Y-axis, and Z-axis, and the orientation is information specified by the values ​​of the θX-axis, θY-axis, and θZ-axis. Positioning means controlling the position and / or orientation. Alignment involves controlling the position and / or orientation of at least one of the substrate and the mold.

[0019] The imprint apparatus IMP has 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 includes a motor such as a linear motor.

[0020] The imprint apparatus IMP has a sensor 151 that detects a substrate driving force (alignment load) required for the substrate driving mechanism 105 to drive the substrate S (substrate holder 102) in alignment. The substrate driving force in alignment, which is performed in a state where the imprint material IM on the substrate and the pattern region MP of the mold M are in contact with each other, corresponds to, for example, a shear force acting between the substrate S and the mold M. The shear force is a force that mainly acts in the planar direction of the substrate S and the mold M. The substrate driving force in alignment is correlated with, for example, the magnitude of a current supplied to the substrate driving mechanism 105 (the motor thereof) to perform alignment. Therefore, the sensor 151 can detect the substrate driving force by detecting (the magnitude of) a current supplied to the substrate driving mechanism 105. The sensor 151 also functions as a measuring instrument that measures the influence (for example, shear force) on the mold M in the imprint process of forming a pattern on the substrate.

[0021] The imprint apparatus IMP has 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 includes a motor such as a voice coil motor.

[0022] The imprint apparatus IMP has a sensor 152 that detects the release force (separation load) and / or the pressing force. The release force is the force required to separate (separate) the hardened product of the imprint material IM on the substrate from the mold M. The pressing force is the force with which the mold M is pressed to bring the mold M into contact with the imprint material IM on the substrate S. The release force and the pressing force are mainly forces that act in a direction perpendicular to the planar direction of the substrate S and the mold M. The release force and the pressing force are correlated with, for example, the magnitude of the current supplied to the mold driving mechanism 122 (the motor of the mold driving mechanism 122). Therefore, the sensor 152 can detect the release force and the pressing force by detecting (the magnitude of) the current supplied to the mold driving mechanism 122. The sensor 152 also functions as a measuring instrument that measures the influence (for example, the release force and the pressing force) on the mold M in the imprint process that forms a pattern on the substrate.

[0023] The substrate driving mechanism 105 and the mold driving mechanism 122 constitute a driving mechanism for adjusting the relative position and the relative posture of the substrate S and the mold M. The adjustment of the relative position of the substrate S and the mold M by such a driving mechanism includes driving for contacting the mold M with the imprint material IM on the substrate and for separating the mold M from the cured imprint material IM (pattern of the cured material) on the substrate. The substrate driving mechanism 105 is configured to drive the substrate S (the substrate holding unit 102 holding the substrate S) about a plurality of axes, for example, three axes of the X axis, the Y axis, and the θZ axis, and 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 is configured to drive the mold M (the mold holding unit 121 holding the mold M) about a plurality of axes, for example, three axes of the Z axis, the θX axis, and the θY axis, and preferably six axes of the X axis, the Y axis, the Z axis, the θX axis, the θY axis, and the θZ axis.

[0024] The imprint apparatus IMP has a mold transport mechanism 140 that transports the mold M, and a mold cleaner 150. The mold transport mechanism 140, for example, transports the mold M to the mold holding unit 121, and transports the mold M from the mold holding unit 121 to a stocker (not shown) or the mold cleaner 150. The mold cleaner 150 cleans the mold M using ultraviolet light, a chemical solution, or the like.

[0025] The mold holding part 121 includes a window member 125 that forms a pressure-controlled space CS on the back surface (the surface opposite to the pattern area MP in which the pattern to be transferred to the substrate S is formed) of the mold M. The imprint apparatus IMP has a deformation mechanism 123 that controls the pressure in the pressure-controlled space CS (hereinafter referred to as "cavity pressure") to deform the pattern area MP of the mold M into a convex shape toward the substrate S, as shown typically in FIG. 2(b).

[0026] The imprint apparatus IMP has an alignment measurement instrument 106. The alignment measurement instrument 106 illuminates an alignment mark provided on the mold M and an alignment mark provided on the substrate S, and captures an image of them to measure the relative position between the marks (the relative position between the mold M and the substrate S). The relative position measured by the alignment measurement instrument 106 is used for alignment when the mold M is brought into contact with the imprint material IM on the substrate. Specifically, under the control of the control unit 110, the position and / or attitude of at least one of the mold M and the substrate S is adjusted based on the relative position measured by the alignment measurement instrument 106, thereby aligning the mold M and the substrate S. The alignment measurement instrument 106 is positioned via a drive mechanism (not shown) according to the positions of the alignment marks provided on the mold M and the substrate S.

[0027] The imprint apparatus IMP is provided with a plurality of alignment measuring instruments 106 that simultaneously detect (observe) a plurality of alignment marks provided in each of the shot areas of the mold M and the substrate S. In this embodiment, four alignment measuring instruments 106 that detect alignment marks provided at the four corners of each of the shot areas of the mold M and the substrate S are provided.

[0028] Here, an image output (acquired) by capturing an image of an alignment mark with the alignment measuring instrument 106 is referred to as an alignment image. Also, the position of the alignment mark measured by the alignment measuring instrument 106 (obtained from the alignment image) is referred to as an alignment measurement value. Alignment images include, for example, images capturing reflected light from alignment marks provided on the mold M and the substrate S, and images capturing images generated by moire of alignment marks provided on the mold M and the substrate S, respectively.

[0029] The imprint apparatus IMP has a curing unit 107, an imaging unit 112, and an optical member 111. The curing unit 107 irradiates the imprint material IM on the substrate 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 mold M, the imprint material IM on the substrate, and the substrate S via the optical member 111 and the window member 125 to obtain images. The image obtained by the imaging unit 112 is also referred to as a spread image.

[0030] The imprint apparatus IMP has a dispenser 108 that places the imprint material IM on the substrate. The dispenser 108 ejects (droplets of) the imprint material IM onto the substrate S, for example, in accordance with a drop recipe that indicates the placement (of droplets) of the imprint material IM on the substrate.

[0031] The imprint apparatus IMP has a control unit 110 that performs imprint processing by comprehensively controlling each unit of the imprint apparatus IMP. The control unit 110 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 is configured, for example, by a PLD (abbreviation of Programmable Logic Device) such as an FPGA (abbreviation of Field Programmable Gate Array), or an ASIC (abbreviation of Application Specific Integrated Circuit), or a general-purpose computer (information processing device) with a program built in, or a combination of all or part of these.

[0032] 2 is a schematic diagram showing a configuration of a lithography system 1001 for manufacturing an article such as a semiconductor device. The lithography system 1001 includes one or more imprint apparatuses IMP, one or more inspection apparatuses 1005, one or more processing apparatuses 1006, one or more generation apparatuses 1007, and a control apparatus 1003. Note that the lithography system 1001 may further include one or more exposure apparatuses.

[0033] The inspection apparatus 1005 includes, for example, an overlay inspection apparatus, a foreign substance inspection apparatus, a CD (Critical Dimension) inspection apparatus, and a defect inspection apparatus. The processing apparatus 1006 includes, for example, an etching apparatus and a film forming apparatus. The generating apparatus 1007 includes a learning model generating apparatus that generates (creates) a learning model described below. The imprint apparatus IMP, the inspection apparatus 1005, the processing apparatus 1006, the generating apparatus 1007, and the control apparatus 1003 are connected to each other via a network 1002. In the lithography system 1001, the imprint apparatus IMP, the inspection apparatus 1005, the processing apparatus 1006, and the generating apparatus 1007 are controlled by the control apparatus 1003, which is an external apparatus (information processing apparatus) different from the imprint apparatus IMP.

[0034] The generating device 1007, like the control unit 110, is configured by, for example, a PLD (abbreviation of Programmable Logic Device) such as an FPGA (abbreviation of Field Programmable Gate Array), or an ASIC (abbreviation of Application Specific Integrated Circuit), or a general-purpose computer (information processing device) with a program installed, or a combination of all or part of these. In addition, the generating device 1007 can be realized by the control unit 110 or the control device 1003, or may be realized by using the control unit 110 and the control device 1003 in combination.

[0035] The operation of the imprint apparatus IMP or the lithography system 1001 including the imprint apparatus IMP will be described with reference to Fig. 3. This operation is controlled by at least one of the control unit 110 and the control device 1003. It is assumed that a mold M has been transported to the imprint apparatus IMP in advance and held by the mold holding unit 121.

[0036] In S101, the substrate S is carried into the imprint apparatus IMP. Specifically, using a substrate transport mechanism (not shown), the substrate S is transported from the transport source (an intermediary portion with the pre-processing apparatus) to the substrate holding portion 102, and the substrate S is held by the substrate holding portion 102.

[0037] In S102 to S106, each process (each step of the imprint process) for forming a pattern on the substrate is performed on a shot area (target shot area) that is to be subjected to the imprint process among multiple shot areas (areas where a pattern is to be formed) of the substrate S.

[0038] In S102, the imprint material IM is placed (supplied) in a target shot area among the multiple shot areas of the substrate S. Specifically, while the substrate S is driven by the substrate driving mechanism 105, the imprint material IM is discharged from the dispenser 108 to place the imprint material IM in the target shot area.

[0039] In S103, the imprint material IM on the target shot area of ​​the substrate S is brought into contact with the mold M (pattern area MP). Specifically, the substrate S and the mold M are driven relatively by at least one of the substrate driving mechanism 105 and the mold driving mechanism 122, thereby bringing the imprint material IM on the target shot area of ​​the substrate S into contact with the mold M. For example, the mold M is driven by the mold driving mechanism 122 so that the mold M comes into contact with the imprint material IM on the target shot area of ​​the substrate S. When bringing the imprint material IM on the target shot area of ​​the substrate S into contact with the mold M, it is preferable to deform the pattern area MP of the mold M into a convex shape toward the substrate S by the deformation mechanism 123.

[0040] In S104, the alignment between the mold M and the substrate S, specifically, the alignment between the pattern area MP of the mold M and the target shot area of ​​the substrate S, is performed. Specifically, while measuring the relative position between the alignment mark of the mold M and the alignment mark of the target shot area of ​​the substrate S by a plurality of alignment measuring instruments 106, the alignment is performed so that the relative position falls within the allowable range of the target relative position. For example, the substrate S and the mold M are driven relatively by at least one of the substrate driving mechanism 105 and the mold driving mechanism 122, so that their relative positions fall within the allowable range of the target relative position. At this time, in the case of the estimation process described later, a first learning model and a second learning model described later are used to estimate the alignment error, which is an error when measuring the above-mentioned relative position, from the alignment mark image information obtained from the alignment image, and the alignment error is reflected in the alignment. This makes it possible to realize a highly accurate alignment between the mold M and the substrate S. The alignment mark image information and the alignment error will be described in detail later.

[0041] In the case of the learning process described later, in S104, alignment images and alignment mark image information obtained by the multiple alignment measuring instruments 106 are accumulated and transmitted to the generation device 1007 in order to generate a first learning model and a second learning model. Note that the alignment images and alignment mark image information used to generate the first learning model and the second learning model are, for example, those obtained after the alignment is completed or immediately before the imprint material IM is hardened.

[0042] Here, an example of a process for calculating the position of an alignment mark (alignment measurement value) will be described with reference to Fig. 4(a), Fig. 4(b) and Fig. 5. Fig. 4(a) shows an alignment mark image 7001 included in an alignment image output (acquired) by imaging an alignment mark for measuring the position in the X direction with the alignment measuring device 106. Fig. 4(b) shows an alignment waveform 7006 obtained from the alignment mark image 7001, which is obtained by integrating the value of the alignment mark image 7001 in the Y direction. It is assumed that each of the substrate S and the mold M is provided with an alignment mark corresponding to the alignment mark image 7001 shown in Fig. 4(a) and an alignment mark for measuring the position in the Y direction, which is obtained by rotating the alignment mark by 90 degrees. In addition, when the alignment marks provided on the substrate S and the mold M are diffraction gratings, a moire pattern, which is an interference image generated by overlapping them, is obtained as an alignment image (instead of the alignment mark image 7001) by the alignment measurement instrument 106. In this case, the alignment measurement instrument 106 can be configured with a simple optical system. In S104 (alignment), the alignment mark position 7002, which is the center position of the alignment mark in the measurement direction (X direction), is calculated from the alignment mark image 7001 shown in FIG. 4(a). In this way, the alignment mark positions in the X direction and Y direction on the substrate S and the mold M are calculated, and the relative positions (X direction and Y direction) between the mold M and the substrate S are measured.

[0043] FIG. 5 is a flow chart for specifically explaining an example of a process for calculating the position 7002 of the alignment mark (alignment measurement value) from the alignment mark image 7001 shown in FIG. 4(a). In S201, the alignment mark is imaged by the alignment measurement instrument 106 to obtain the alignment mark image 7001 shown in FIG. 4(a). In S202, an alignment waveform 7006 shown in FIG. 4(b) is calculated from the alignment mark image 7001 obtained in S201. For example, the alignment waveform 7006 is calculated by accumulating the measurement values ​​in a direction 7005 (Y direction) perpendicular to the measurement direction (X direction) for a measurement region 7003 including the alignment mark image 7001 in the image obtained by the alignment measurement instrument 106. In S203, the position 7002 of the alignment mark shown in FIG. 4(a) and FIG. 4(b) is calculated from the alignment waveform 7006 calculated in S202. For example, the center of gravity of the alignment waveform 7006 may be set as the alignment mark position 7002, or the alignment mark position 7002 may be calculated by determining the phase of the alignment waveform 7006 using a Fourier transform or the like. Furthermore, the alignment mark position 7002 may be calculated from the alignment waveform 7006 using a pattern mashing method.

[0044] 3, in S105, the imprint material IM is cured in a state in which the imprint material IM on the target shot area of ​​the substrate S is in contact with the mold M. Specifically, energy for curing the imprint material IM is irradiated from the curing unit 107 to the imprint material IM between the substrate S and the pattern area MP of the mold M. As a result, the imprint material IM between the substrate S and the pattern area MP of the mold M is cured, and a cured product (pattern) of the imprint material IM is formed in the target shot area of ​​the substrate S.

[0045] In S106, the mold M is separated from the cured imprint material IM on the target shot area of ​​the substrate S (mold release). Specifically, the substrate S and the mold M are driven relatively by at least one of the substrate driving mechanism 105 and the mold driving mechanism 122 so that the cured product of the imprint material IM on the target shot area of ​​the substrate S is separated from the pattern area MP of the mold M. For example, the mold M is driven by the mold driving mechanism 122 so that the mold M is separated from the cured product of the imprint material IM on the target shot area of ​​the substrate S. When the mold M is separated from the cured imprint IM on the target shot area of ​​the substrate S, it is preferable to deform the pattern area MP of the mold M into a convex shape toward the substrate S by the deformation mechanism 123. In addition, the image capturing unit 112 captures images of the mold M, the imprint material IM on the substrate, and the substrate S to obtain a spread image, and the state of separation between the imprint material IM on the substrate and the mold M is observed based on the spread image.

[0046] In S107, it is determined whether or not the imprint process (S102 to S106) has been performed on all the shot areas of the substrate S. If the imprint process has not been performed on all the shot areas, the process proceeds to S102 to perform the imprint process on the next shot area (target shot area) of the substrate S. On the other hand, if the imprint process has been performed on all the shot areas, the process proceeds to S108.

[0047] In S108, the substrate S is unloaded from the imprint apparatus IMP. Specifically, the substrate S is transported from the substrate holding unit 102 to a destination (for example, an intermediary unit of a post-processing apparatus) using a substrate transport mechanism (not shown).

[0048] When a lot consisting of a plurality of substrates is processed, the steps shown in FIG. 3 are carried out for each of the plurality of substrates.

[0049] In this embodiment, in the alignment between the mold M and the substrate S (S104), the alignment error is estimated from the alignment mark image information obtained from the alignment image output from the alignment measurement instrument 106. The alignment mark image information is a feature value obtained from the alignment image, and the above-mentioned alignment waveform is an example of the alignment mark image information. The alignment waveform may be processed by thinning out the waveform data. Another example of the alignment mark image information may include information on the brightness of the alignment image. Here, the alignment error refers to the amount of misalignment between the substrate S and the mold M, and corresponds to the measurement error of the relative position measured by the measurement instrument during the alignment. In addition, the measurement value of the amount of misalignment between the patterns of the substrate S and the mold M measured by the overlay inspection device also corresponds to the alignment error.

[0050] In this embodiment, in addition to the measurement direction information (feature amount) included in the alignment mark image information, the non-measurement direction information included in the alignment mark image information is also used to estimate the alignment error. Here, the non-measurement direction information is information on a direction different from the measurement direction, for example, information on a direction perpendicular to the measurement direction. By using the non-measurement direction information, it is possible to obtain the position of the alignment mark (alignment measurement value) with high accuracy.

[0051] In this embodiment, machine learning is used to estimate the alignment error. Specifically, a model (learning model) for estimating the alignment error is generated (learned) in advance from alignment mark image information and a measurement value of the alignment error obtained by an overlay inspection device. Then, in the alignment between the mold M and the substrate S (S104), the previously generated learning model is used to estimate the alignment error from the alignment mark image information, and the estimated value of the alignment error is reflected in the alignment operation.

[0052] In estimation using statistical methods including machine learning, the estimation accuracy generally depends heavily on the number of data (amount of data) used to train the learning model. For example, to estimate the alignment error with useful accuracy, it is necessary to collect a large amount of data on alignment mark image information used for training and the measurement values ​​of the alignment error obtained by an overlay inspection device.

[0053] In this embodiment, the accuracy of estimating the alignment error can be improved by generating (learning) a learning model for each of the multiple alignment measuring instruments 106 used for alignment. There are not a few manufacturing errors (variations) and installation errors between the multiple alignment measuring instruments 106. Therefore, the correlation between the alignment mark image information and the measurement value of the alignment error obtained by the overlay inspection device differs for each alignment measuring instrument 106. Therefore, generating each correlation as a separate learning model for the multiple alignment measuring instruments 106 is advantageous from the viewpoint of improving the accuracy of estimating the alignment error. However, when generating a learning model for each alignment measuring instrument 106, the data available for learning is limited to only the alignment mark image information acquired for each alignment measuring instrument 106 and the data group of the corresponding measurement value of the alignment error. Therefore, in order to collect enough data to be used for learning, more costs (such as an increase in the number of substrates for which data is collected) are required.

[0054] With reference to FIG. 6, an overview of a generation process (learning process) for generating a learning model for each alignment measurement instrument 106 and an estimation process for estimating an alignment error using such a learning model in the conventional technology will be described.

[0055] First, a description will be given of the generation process (step 1-1) for generating a learning model 2001. Reference numeral 2002 denotes a data group of alignment mark image information acquired by one measuring instrument N among the multiple alignment measuring instruments 106. Reference numeral 2003 denotes a data group of measurement values ​​of alignment errors acquired by an overlay inspection apparatus corresponding to each piece of alignment mark image information included in the data group 2002.

[0056] When generating (learning) the model, the data group 2002 of the alignment mark image information is used as learning data on the input side of the learning model, and the data group 2003 of the measurement values ​​of the alignment error is used as learning data on the output side of the learning model. By performing such learning, a learning model 2001 of the alignment measuring instrument N is generated. The learning data on the input side of the learning model is sometimes called training data, and the learning data on the output side of the learning model is sometimes called teacher data. As for the other alignment measuring instruments, a learning model of each alignment measuring instrument is generated using a data group of the alignment mark image and a data group of the measurement values ​​of the alignment error collected by each alignment measuring instrument. In this way, a learning model for estimating the alignment error is generated for each of the multiple alignment measuring instruments 106. Although an example in which the alignment mark image information is used as the input side learning data of the learning model will be described, the input side learning data of the learning model may include information other than the alignment mark image information. For example, the input side learning data may include the above-mentioned information on the shear force and the imprint force. In this case, there is a possibility that the estimation accuracy is further improved. This also applies to the input of the learning model in the estimation method. This is also true not only for the conventional technology but also for this embodiment.

[0057] Next, an estimation method (step 2-1) for estimating the alignment error using the learning model 2001 generated in step 1-1 will be described. 2004 is alignment mark image information obtained from an alignment image acquired by the alignment measuring instrument N in the alignment (S104) between the mold M and the substrate S. By inputting the alignment mark image information 2004 to (the input side of) the learning model 2001 of the alignment measuring instrument N, an alignment error 2005 corresponding to the alignment mark image information 2004 is output from (the output side of) the learning model 2001. The alignment error 2005 corresponds to the alignment error estimated from the alignment mark image information 2004 using the learning model 2001, that is, an estimate of the measurement value of the alignment error obtained by the overlay inspection device. For other alignment measuring instruments, an estimate of the alignment error is obtained from the alignment mark image information obtained by each alignment measuring instrument using the learning model of each alignment measuring instrument.

[0058] Here, the correlation between the alignment mark image information and the alignment error also includes elements common to all of the alignment measurement instruments 106, such as alignment errors caused by the shape of the alignment mark on the substrate. Since the alignment mark on the substrate is formed through a lithography process, an etching process, and the like, its shape is not necessarily ideal, and distortion of the shape of the alignment mark becomes an alignment error. Furthermore, since the above-mentioned process is performed in the same manner on the entire substrate, the distortion of the shape of the alignment mark is generally similar within the same substrate. These become alignment errors common to all of the alignment measurement instruments 106, and are also expressed as a correlation between the alignment mark image information and the alignment error.

[0059] Therefore, in this embodiment, a learning model for each alignment measuring instrument 106 is generated using a data group of alignment mark image information obtained by all alignment measuring instruments 106. Specifically, a first learning model common to all alignment measuring instruments 106 is generated using a data group of all alignment mark image information output from each of the multiple alignment measuring instruments 106. Furthermore, a second learning model different from the first learning model is generated for each of the multiple alignment measuring instruments 106 using a data group of alignment mark image information output from one of the multiple alignment measuring instruments 106. Then, in the alignment between the mold M and the substrate S (S104), the alignment error is estimated using the first learning model and the second learning model.

[0060] With reference to Figures 7 and 8, in this embodiment, an overview of a generation process (learning process) that generates a first learning model and a second learning model, and an estimation process that estimates an alignment error using such first learning model and second learning model will be described.

[0061] First, as shown in FIG. 7, a generation process for generating a first learning model 3001 and a second learning model 3001 will be described. In step 3-1, the first learning model 3001 is generated. 3002 is a data group of alignment mark image information acquired by all of the multiple alignment measuring instruments 106, i.e., all of the alignment measuring instruments 106. In this way, the data group 3002 corresponds to a data group (first data group) including all of the alignment mark image information output from each of the multiple alignment measuring instruments 106. 3003 is a data group (second data group) of measurement values ​​of alignment errors obtained by the overlay inspection device corresponding to each of all the alignment mark image information included in the data group 3002. In other words, the data group 3003 is a data group that collects measurement values ​​of alignment errors corresponding to all the alignment measuring instruments 106. When generating (learning) the model, a data group 3002 of alignment mark image information is used as learning data on the input side of the learning model, and a data group 3003 of the measurement values ​​of the alignment error is used as learning data on the output side of the learning model. By performing such learning, a first learning model 3001 is generated. In this embodiment, a first learning model 3001 for estimating alignment errors for all alignment measuring instruments 106, that is, a first learning model 3001 common to all alignment measuring instruments 106, is generated. Note that the number of substrates required to collect data used for learning the first learning model 3001 may be smaller than the number of substrates required to collect data used for learning in the conventional technology. As described later, in this embodiment, it is possible to estimate alignment errors with the same accuracy even with a number of substrates smaller than the number of substrates required to collect data used for learning in the conventional technology.

[0062] Next, in steps 3-2 and 3-3, a second learning model 3006 of each alignment measurement instrument 106 is generated. In step 3-2, a data group 3004 (third data group) of alignment mark image information obtained by one alignment measurement instrument N among data group 3002 of alignment mark image information is input to (the input side of) first learning model 3001. As a result, a data group 3005 (fourth data group) of alignment errors corresponding to each piece of alignment mark image information included in data group 3004 is output from (the output side of) first learning model 3001. Data group 3005 corresponds to a data group of estimated values ​​(first estimated values) of alignment errors obtained by an overlay inspection device corresponding to each piece of alignment mark image information included in data group 3004. Such processing is performed on data groups of alignment mark image information obtained by all alignment measurement instruments 106, i.e., all alignment mark image information included in such data groups.

[0063] In step 3-3, the data group 3005 of the first estimated value of the alignment error of the alignment measuring instrument N is used as learning data on the input side of the learning model, and the data group 3007 of the measurement value of the alignment error is used as learning data on the output side of the learning model. By performing such learning, a second learning model 3006 of the alignment measuring instrument N is generated. Note that the data group 3007 is a data group (fifth data group) of the measurement value of the alignment error obtained by the overlay inspection device corresponding to each of the alignment mark image information included in the data group 3004. In this way, in step 3-3, the difference between the first estimated value of the alignment error estimated using the first learning model 3001 and the corresponding measurement value of the alignment error is learned. For the other alignment measuring instruments, a second learning model of each alignment measuring instrument is generated using the data group of the first estimated value of the alignment error estimated using the first learning model and the data group of the measurement value of the alignment error. In this way, a second learning model is generated for each of the multiple alignment measuring instruments 106. In this embodiment, the case where the data group 3005 of the first estimated value of the alignment error is used as learning data on the input side of the learning model to generate the second learning model 3006 has been described, but the present invention is not limited to this. For example, the second learning model 3006 may be generated by using the data group 3004 of the alignment mark image information as learning data on the input side of the learning model in addition to the data group 3005 of the first estimated value of the alignment error.

[0064] Next, a method for estimating an alignment error using the first learning model 3001 and the second learning model 3001 generated in steps 3-1, 3-2, and 3-3 will be described as shown in Fig. 8. Reference numeral 4001 denotes alignment mark image information (first mark image information) obtained from an alignment image acquired by an alignment measuring instrument N (first measuring instrument) in the alignment (S104) between the mold M and the substrate S.

[0065] In step 4-1, alignment mark image information 4001 is input to (the input side of) the first learning model 3001. As a result, a first estimate value 4002 of the alignment error (first alignment error) corresponding to the alignment mark image information 4001 is output from (the output side of) the first learning model 3001.

[0066] In step 4-2, the first alignment error estimate 4002 is input to (the input side of) the second learning model 3006 of the alignment measurement instrument N. This causes an alignment error 4003 (second alignment error) to be output from (the output side of) the second learning model 3006. The alignment error 4003 corresponds to the alignment error estimated from the alignment mark image information 4001 using the first learning model 3001 and the second learning model 3006, i.e., an estimate of the measurement value of the alignment error obtained by the overlay inspection device.

[0067] When second learning model 3006 is generated using data group 3004 of alignment mark image information as learning data on the input side of the learning model in addition to data group 3005 of the first alignment error estimate, the alignment error is estimated as follows: By inputting first alignment error estimate 4002 and alignment mark image information 4001 to second learning model 3006 of alignment measurement instrument N, alignment error 4003 (third alignment error) is estimated.

[0068] According to this embodiment, when generating a learning model (second learning model) for each alignment measuring instrument, data from all alignment measuring instruments can be used while reflecting the characteristics unique to each alignment measuring instrument. Therefore, compared to the conventional technology in which a learning model is generated using only data from each alignment measuring instrument, the same number of data can be used even if data is collected from a small number of substrates. This makes it possible to generate a learning model capable of estimating alignment errors with the same accuracy as the conventional technology. For example, in the conventional technology, a data group containing 10,000 pieces of data (data collected from 10,000 substrates) is required to estimate alignment errors with a certain accuracy. In contrast, in this embodiment, for example, when four alignment measuring instruments are provided, a learning model is generated using a data group collected from half the number of substrates (50,000 substrates). In this case, the second learning model is generated with half the amount of data compared to the conventional technology, but the first learning model is generated with twice the amount of data compared to the conventional technology (the product of the number of substrates and the number of alignment measuring instruments). This is expected to make it possible to estimate alignment errors with the same accuracy as conventional techniques.

[0069] Referring to Figures 9 and 10, as another embodiment, an overview of a generation process (learning process) for generating a first learning model and a second learning model, and an estimation process for estimating an alignment error using such first learning model and second learning model will be described.

[0070] First, as shown in FIG. 9, a generation process for generating a first learning model 5001 and a second learning model 5004 will be described. In step 5-1, the first learning model 5001 is generated in the same manner as in step 3-1. 5002 is a data group of alignment mark image information acquired by all of the multiple alignment measuring instruments 106, that is, all of the alignment measuring instruments 106. In this manner, the data group 5002 corresponds to a data group (first data group) including all of the alignment mark image information output from each of the multiple alignment measuring instruments 106. 5003 is a data group (second data group) of measurement values ​​of alignment errors obtained by the overlay inspection device corresponding to each of all the alignment mark image information included in the data group 5002. The data group 5003 is a data group that collects measurement values ​​of alignment errors corresponding to all the alignment measuring instruments 106. When generating (learning) the model, a data group 5002 of alignment mark image information is used as learning data on the input side of the learning model, and a data group 5003 of the measurement values ​​of the alignment error is used as learning data on the output side of the learning model. By performing learning in this manner, a first learning model 5001 is generated.

[0071] In step 5-2, a second learning model 5004 is generated in the same manner as in the conventional technique (step 1-1). 5005 is a data group (third data group) of alignment mark image information acquired by one measuring instrument N of the multiple alignment measuring instruments 106. 5006 is a data group (fourth data group) of measurement values ​​of alignment errors acquired by an overlay inspection device corresponding to each of the alignment mark image information included in the data group 5005. When generating (learning) a model, the data group 5005 of the alignment mark image information is used as learning data on the input side of the learning model, and the data group 5006 of the measurement values ​​of the alignment errors is used as learning data on the output side of the learning model. By performing such learning, a second learning model 5004 of the alignment measuring instrument N is generated. For the other alignment measuring instruments, a second learning model of each alignment measuring instrument is generated using the data group of alignment mark image information and the data group of the measurement values ​​of the alignment errors collected by each alignment measuring instrument. In this manner, a second learning model for estimating the alignment error is generated for each of the multiple alignment measurement instruments 106.

[0072] Next, a method for estimating an alignment error using the first learning model 5001 and the second learning model 5004 generated in steps 5-1 and 5-2 will be described as shown in Fig. 10. Reference numeral 6001 denotes alignment mark image information (first mark image information) obtained from an alignment image acquired by an alignment measuring instrument N (first measuring instrument) in the alignment (S104) between the mold M and the substrate S.

[0073] In step 6-1, alignment mark image information 6001 is input to (the input side of) the first learning model 5001. As a result, a first estimate 6002 of the alignment error (first alignment error) corresponding to the alignment mark image information 6001 is output from (the output side of) the first learning model 5001.

[0074] In step 6-2, alignment mark image information 6001 is input to (the input side of) second learning model 5004. As a result, a second estimate 6003 of the alignment error (second alignment error) corresponding to the alignment mark image information 6001 is output from (the output side of) second learning model 5004.

[0075] In step 6-3, a registration error 6004 is calculated based on the first estimated value 6002 obtained in step 6-1 and the second estimated value 6003 obtained in step 6-2. The registration error 6004 corresponds to a registration error estimated from the alignment mark image information 6001 using the first learning model 5001 and the second learning model 5004, that is, an estimate of the measurement value of the registration error obtained by the overlay inspection device. In this embodiment, the statistical value of the first estimated value 6002 obtained in step 6-1 and the second estimated value 6003 obtained in step 6-2 is set as the registration error 6004. For example, the average value of the first estimated value 6002 and the second estimated value 6003 may be set as the registration error 6004, or the weighted average value of the first estimated value 6002 and the second estimated value 6003 may be set as the registration error 6004 using a preset weighting coefficient.

[0076] 9 and 10, it is possible to estimate the alignment error by using the data of all the alignment measurement instruments and by reflecting the characteristics specific to each alignment measurement instrument. In other words, compared to the conventional technology in which a learning model is generated using only the data of each alignment measurement instrument, it is possible to generate a learning model that can estimate the alignment error with the same accuracy as the conventional technology even with a smaller amount of data.

[0077] In this embodiment, there is also an advantage in updating (relearning) the first learning model and the second learning model. The process of updating the first learning model and the second learning model may be performed by the generation device 1007, or may be performed by the control unit 110 of the imprint apparatus IMP. In other words, the generation device 1007 and the control unit 110 function as an update unit that updates the first learning model and the second learning model. As described above, the first learning model is a model that represents the relationship between alignment image information and alignment error that is common to all alignment measurement instruments, and the second learning model is a model that represents the relationship unique to each alignment measurement instrument.

[0078] For example, the process conditions for transferring the pattern of the mold M to the substrate (including the lithography process and the etching process performed on the substrate) may be changed in order to improve the quality of the final product. In such a case, the shape of the alignment mark may change, and the correlation between the alignment mark image information and the alignment error may also change, so it is necessary to update (re-learn) the learning model. However, only the first learning model, which expresses the correlation common to all alignment measuring instruments, is affected by changing the process conditions for transferring the pattern of the mold M to the substrate (such as the lithography process and the etching process). Therefore, when the process conditions for the lithography process and the etching process are changed, it is sufficient to update only the first learning model without updating the second learning model (only step 3-1 and step 5-1 may be performed).

[0079] Furthermore, the shape of the alignment mark may gradually change due to changes over time in the equipment used in the lithography and etching processes. In such cases, only the first learning model, which expresses the correlation common to all alignment measurement instruments, is affected. Therefore, in response to changes over time in the equipment used in the lithography and etching processes, it is sufficient to periodically update only the first learning model without updating the second learning model (only step 3-1 or step 5-1 needs to be performed).

[0080] On the other hand, when at least some of the alignment measuring instruments 106 are replaced and adjusted, the learning model needs to be updated (re-learned), but only the second learning model that expresses the relationship unique to each alignment measuring instrument is affected. Therefore, when the alignment measuring instruments 106 are replaced and adjusted, it is sufficient to update only the second learning model without updating the first learning model (only steps 3-2 and 3-3 and step 5-2 need to be performed).

[0081] In this manner, in this embodiment, when relearning of the learning model is required, it is possible to deal with the situation by updating only one of the first learning model and the second learning model depending on the cause. In other words, in this embodiment, the first learning model and the second learning model can be updated individually depending on the cause of the need for relearning of the learning model. Note that the relationship between the cause of the need for relearning of the learning model and the learning model that is affected by it (whether it is the first learning model or the second learning model) can be obtained in advance as a table from simulations, empirical rules, etc.

[0082] Next, the generation process (the learning process shown in FIG. 7) for generating the first learning model and the second learning model will be specifically described with reference to FIG. 11. Such generation process may be performed by the generation device 1007 or by the control unit 110 of the imprint apparatus IMP. In other words, the generation device 1007 and the control unit 110 function as a generation unit that generates the first learning model and the second learning model.

[0083] In S301, the measurement values ​​of the alignment errors corresponding to all the alignment measuring instruments 106, i.e., the measurement values ​​of the alignment errors obtained by the overlay inspection device, are obtained. In the shot area of ​​the substrate S, the overlay inspection device measures the overlay inspection marks in the vicinity of the alignment marks detected by each alignment measuring instrument 106 during alignment (S104), and the measurement values ​​of the alignment errors are obtained. The overlay inspection device measures the amount of misalignment between, for example, the base layer of the substrate (its overlay inspection mark) and the layer formed by the imprint device IMP (its overlay inspection mark) to obtain the alignment error.

[0084] In S302, alignment mark image information corresponding to the measurement values ​​(of all alignment measuring instruments 106) of the alignment errors acquired in S301 is acquired. The alignment mark image information includes feature quantities such as the shape of the alignment waveform extracted from the alignment image, and in this embodiment, includes information regarding the measurement direction and non-measurement direction. Note that the alignment mark image information may be the phase or amplitude obtained by Fourier transforming the alignment waveform. The alignment image is an image acquired by the alignment measuring instruments 106 immediately after alignment (S104) or immediately after curing (S105).

[0085] In S303, a first learning model is generated. Specifically, the alignment mark image information acquired in S302 is used as learning data on the input side of the learning model, and the measurement value of the alignment error acquired in S301 is used as learning data on the output side of the learning model, thereby generating the first learning model.

[0086] Here, a specific method of learning and estimation in this embodiment will be described. Specific examples of the method include Gaussian process regression and Bayesian estimation, which perform estimation taking uncertainty into account by treating variables as probabilities. When Gaussian process regression and Bayesian estimation are used, the learning model is a function that inputs alignment mark image information and outputs a probability distribution of alignment errors, and the internal variables are optimized by learning. The expected value of the probability distribution of alignment errors obtained in this way can be used as an estimate of the alignment error (measurement value).

[0087] Furthermore, if it is necessary to reduce the amount of calculations involved in the estimation, it is advisable to use a statistical model with a low amount of calculations, such as multiple regression analysis. If the pixel values ​​of the alignment image are used as features (high-dimensional features), a method of optimizing internal variables using a neural network composed of a multi-layer perceptron is suitable. If there are many abnormal values ​​in the alignment error or alignment image information, a method based on a decision tree (analysis) that is robust to abnormal values ​​may be applied. When multiple regression analysis or a neural network is used, the learning model inputs the alignment image information and outputs the alignment error, and optimizes the internal variables through learning. When a decision tree is used, the learning model inputs the alignment image information and outputs the alignment error, and constructs a decision tree through learning.

[0088] In S304, the alignment error corresponding to the target alignment measurement instrument 106 (target alignment measurement instrument) among the multiple alignment measurement instruments 106 is estimated to calculate a first estimated value. Specifically, the alignment mark image information obtained by the target alignment measurement instrument from the alignment mark image information obtained in S302 is input to (the input side of) the first learning model generated in S303 to calculate the first estimated value of the alignment error. This process is repeated until all alignment measurement instruments 106 are determined to be target alignment measurement instruments, and a data group of the first estimated values ​​of the alignment errors corresponding to all alignment measurement instruments 106 is obtained.

[0089] In S305, a second learning model of the target alignment measurement instrument 106 (target alignment measurement instrument) is generated. Specifically, learning is performed by using the first estimated value of the alignment error calculated in S304 as learning data on the input side of the learning model, and using the measured value of the alignment error acquired in S301 as learning data on the output side of the learning model. In this way, a second learning model of the target alignment measurement instrument is generated. This process is repeated until all alignment measurement instruments 106 are determined to be target alignment measurement instruments, and a second learning model of each alignment measurement instrument 106 is generated. Note that the average value of the difference between the first estimated value of the alignment error and the corresponding measured value of the alignment error may be obtained as the offset. Also, the coefficients may be calculated by polynomial approximation using a learning model in which the relationship with the first estimated value of the alignment error is expressed as a polynomial.

[0090] In S306, it is determined whether or not second learning models have been generated for all alignment measurement instruments 106. If second learning models have not been generated for all alignment measurement instruments 106, the process proceeds to S304 to generate a second learning model for the next alignment measurement instrument 106. On the other hand, if second learning models have been generated for all alignment measurement instruments 106, the process proceeds to S307.

[0091] In S307, the first learning model generated in S303 and the second learning model generated in S305 are stored (saved). In this embodiment, the first learning model and the second learning model are stored in the control unit 110 (such as a memory) of the imprint apparatus IMP.

[0092] Next, the estimation process (the estimation process shown in FIG. 8) for estimating the alignment error using the first learning model and the second learning model will be specifically described with reference to FIG. 12. Such estimation process is basically performed by the control unit 110 of the imprint apparatus IMP. In other words, the control unit 110 functions as an estimation unit that estimates the alignment error using the first learning model and the second learning model. However, this does not exclude the estimation process for estimating the alignment error using the first learning model and the second learning model being performed by an external information processing device or the like.

[0093] In S401, the first learning model and the second learning model stored in S305 are acquired (read out). Note that the first learning model and the second learning model are acquired when performing alignment (S104), but they may be acquired prior to alignment, for example, when placing the imprint material IM (S102).

[0094] In S402, alignment mark image information obtained by the alignment measurement instrument 106 in the alignment (S104) is acquired.

[0095] In S403, the alignment mark image information acquired in S402 is input to (the input side of) the first learning model acquired in S401 to calculate a first estimate of the alignment error. For example, when Gaussian process regression is used as the first learning model, the alignment mark image information is input to the first learning model, and the expected value of the probability distribution of the alignment error output from the first learning model is set as the first estimate of the alignment error.

[0096] In S404, the first estimated value of the registration error calculated in S403 is input to (the input side of) the second learning model acquired in S401 to calculate an estimated value (final estimated value) of the registration error. For example, when polynomial approximation is used as the second learning model, the first estimated value of the registration error is input to a polynomial, and the registration error output from the polynomial is treated as the final estimated value.

[0097] In S405, it is determined whether or not the alignment error estimates (final estimates) have been calculated for all alignment measurement instruments 106. If the alignment error estimates have not been calculated for all alignment measurement instruments 106, the process proceeds to S403. On the other hand, if the alignment error estimates have been calculated for all alignment measurement instruments 106, the process ends.

[0098] Next, the generation process (the learning process shown in FIG. 9) for generating the first learning model and the second learning model will be specifically described with reference to FIG. 13. Such generation process may be performed by the generation device 1007 or by the control unit 110 of the imprint apparatus IMP. In other words, the generation device 1007 and the control unit 110 function as a generation unit that generates the first learning model and the second learning model.

[0099] In S501, similarly to S301, the measurement values ​​of the alignment errors corresponding to all the alignment measuring instruments 106, that is, the measurement values ​​of the alignment errors obtained by the overlay inspection apparatus, are obtained.

[0100] In S502, similarly to S302, alignment mark image information corresponding to the measurement values ​​of the alignment errors (for all alignment measurement instruments 106) acquired in S501 is acquired.

[0101] In S503, a first learning model is generated in the same manner as in S303. Specifically, the alignment mark image information acquired in S502 is used as learning data on the input side of the learning model, and the measurement value of the alignment error acquired in S501 is used as learning data on the output side of the learning model, thereby generating the first learning model.

[0102] In S504, a second learning model of the target alignment measuring instrument 106 (target alignment measuring instrument) is generated. Specifically, the alignment mark image information obtained by the target alignment measuring instrument from the alignment mark image information obtained in S502 is used as learning data on the input side of the learning model. In addition, the measurement value of the alignment error corresponding to the target alignment measuring instrument obtained in S501 is used as learning data on the output side of the learning model. In this way, a second learning model of the target alignment measuring instrument is generated. Such processing is repeated until all alignment measuring instruments 106 are determined to be target alignment measuring instruments, and a second learning model of each alignment measuring instrument 106 is generated.

[0103] In S305, it is determined whether or not second learning models have been generated for all alignment measuring instruments 106. If second learning models have not been generated for all alignment measuring instruments 106, the process proceeds to S504 to generate a second learning model for the next alignment measuring instrument 106. On the other hand, if second learning models have been generated for all alignment measuring instruments 106, the process proceeds to S506.

[0104] In S506, the first learning model generated in S503 and the second learning model generated in S504 are stored (saved). In this embodiment, the first learning model and the second learning model are stored in the control unit 110 (such as a memory) of the imprint apparatus IMP.

[0105] Next, with reference to Fig. 14, the estimation process (the estimation process shown in Fig. 10) for estimating the alignment error using the first learning model and the second learning model will be specifically described. Such estimation process is basically performed by the control unit 110 of the imprint apparatus IMP. In other words, the control unit 110 functions as an estimation unit that estimates the alignment error using the first learning model and the second learning model. However, this does not exclude the estimation process for estimating the alignment error using the first learning model and the second learning model being performed by an external information processing device or the like.

[0106] In S601, similar to S401, the first learning model and the second learning model stored in S505 are obtained (read).

[0107] In S602, similarly to S402, alignment mark image information obtained by the alignment measurement instrument 106 in the alignment (S104) is acquired.

[0108] In S603, similarly to S403, the alignment mark image information acquired in S602 is input to (the input side of) the first learning model acquired in S601, and a first estimated value of the alignment error is calculated.

[0109] In S604, the alignment mark image information acquired in S602 is input to (the input side of) the second learning model acquired in S601 to calculate a second estimate of the alignment error.

[0110] In S605, an estimate (final estimate) of the registration error is calculated based on the first estimate of the registration error calculated in S603 and the second estimate of the registration error calculated in S604. For example, the average value of the first estimate of the registration error and the second estimate of the registration error is set as the final estimate of the registration error.

[0111] In S606, it is determined whether or not the alignment error estimates (final estimates) have been calculated for all alignment measurement instruments 106. If the alignment error estimates have not been calculated for all alignment measurement instruments 106, the process proceeds to S603. On the other hand, if the alignment error estimates have been calculated for all alignment measurement instruments 106, the process ends.

[0112] The pattern of the cured product formed using the imprint apparatus IMP or the lithography system 1001 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 the 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 sensor, and FPGA. Examples of the molds include molds for imprinting.

[0113] The pattern of the cured product may be used as it is as at least a part of the component of the above-mentioned article, or may be used temporarily as a resist mask, which is removed after etching or ion implantation is performed in a substrate processing step.

[0114] Next, a specific method for manufacturing the article will be described. As shown in Fig. 15(a), a substrate such as a silicon wafer with a workpiece such as an insulator formed on its surface is prepared, and then an imprint material is applied to the surface of the workpiece by an inkjet method or the like. Here, the state in which the imprint material in the form of multiple droplets is applied to the substrate is shown.

[0115] As shown in Fig. 15(b), the imprinting mold is placed so that the side on which the concave-convex pattern is formed faces the imprinting material on the substrate. As shown in Fig. 15(c), the substrate on which the imprinting material has been applied is brought into contact with the mold, and pressure is applied. The imprinting material fills the gap between the mold and the workpiece. When light is irradiated through the mold in this state as hardening energy, the imprinting material hardens.

[0116] As shown in Fig. 15(d), when the imprint material is cured and then the mold and the substrate are separated, a pattern of the cured product of the imprint material is formed on the substrate. 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 patterns of the mold are transferred to the imprint material.

[0117] As shown in Fig. 15(e), when etching is performed using the pattern of the cured material as an etching-resistant mask, the portions of the surface of the workpiece where there is no cured material or where only a thin portion remains are removed, forming grooves. As shown in Fig. 15(f), when the pattern of the cured material is removed, an article having grooves formed on the surface of the workpiece can be obtained. Here, the pattern of the cured material is removed, but it may be used as an interlayer insulating film included in a semiconductor element, i.e., a component of an article, without being removed after processing.

[0118] 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.

[0119] In the present embodiment, an imprint apparatus has been described as an example of a lithography apparatus for forming a pattern on a substrate, but the present invention is not limited thereto. For example, the present invention can also be applied to an exposure apparatus that performs an exposure process in which a pattern of an original (mask or reticle) is projected onto a substrate via a projection optical system to expose the substrate, and the pattern of the original is transferred onto the substrate. In an exposure apparatus, in order to measure the position of the substrate relative to the position of the original, an alignment measuring instrument that observes an alignment mark on the substrate without using the original is generally provided. In addition, there is also a form having a plurality of alignment measuring instruments for the purpose of improving throughput. In such an exposure apparatus, even when an alignment mark on a substrate is observed without using an original, it is possible to correct the alignment position in the same manner by using the feature amount of the alignment mark image when forming a pattern as alignment mark image information. In an exposure apparatus, the relationship between alignment mark image information and positioning error includes a relationship common to alignment measuring instruments (such as processing conditions of the substrate) and a relationship specific to alignment measuring instruments (such as adjustment state and installation state of the measuring instrument). Therefore, the effects of the present invention can be expected to be obtained from an exposure apparatus as well as an imprint apparatus. The present invention can also be applied to a drawing apparatus that draws a pattern on a substrate using a charged particle beam (such as an electron beam or an ion beam).

[0120] The disclosure of this specification includes the following lithography apparatus, information processing apparatus, estimation method, program, and method for manufacturing an article.

[0121] (Item 1) A lithography apparatus that performs a process of transferring a pattern of an original onto a substrate, comprising: a plurality of measuring instruments each measuring marks provided on the original and the substrate and outputting an amount of misalignment between the original and the substrate; an estimation unit that estimates an alignment error, which is a measurement error of the positional deviation amount, from mark image information obtained from a mark image observed by a measurement device, while aligning the original and the substrate based on the positional deviation amount using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model, and using a fourth data group including the alignment error outputted as input learning data of the second learning model, and using a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group as output learning data of the second learning model. 13. A lithography apparatus comprising:

[0122] (Item 2) The estimation unit is inputting first mark image information output from a first measuring instrument among the plurality of measuring instruments into the first learning model to obtain a first alignment error output from the first learning model; inputting the first alignment error into the second learning model generated for the first measuring instrument to obtain a second alignment error output from the second learning model; the second alignment error being an estimate of the alignment error corresponding to the first mark image information; Item 2. A lithographic apparatus according to item 1, characterized in that

[0123] (Item 3) The second learning model is a learning model generated by using the fourth data group and the third data group as input learning data of the second learning model and the fifth data group as output learning data of the second learning model, The estimation unit is inputting the first mark image information into the second learning model in addition to the first alignment error to obtain a third alignment error output from the second learning model; the third alignment error being an estimate of the alignment error corresponding to the first mark image information; 3. A lithographic apparatus according to item 2, characterized in that

[0124] (Item 4) A lithography apparatus that performs a process of transferring a pattern of an original onto a substrate, comprising: a plurality of measuring instruments each measuring marks provided on the original and the substrate and outputting an amount of misalignment between the original and the substrate; an estimation unit that estimates an alignment error, which is a measurement error of the positional deviation amount, from mark image information obtained from a mark image observed by a measurement device, while aligning the original and the substrate based on the positional deviation amount using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by using, for each of the plurality of measuring instruments, a third data group including the mark image information output from one of the plurality of measuring instruments as learning data on the input side of the second learning model, and a fourth data group including the alignment errors corresponding to each of the mark image information included in the third data group as learning data on the output side of the second learning model. 13. A lithography apparatus comprising:

[0125] (Item 5) The estimation unit is inputting first mark image information output from a first measuring instrument among the plurality of measuring instruments into the first learning model to obtain a first alignment error output from the first learning model; inputting the first mark image information into the second learning model to obtain a second alignment error output from the second learning model; a statistical value of the first alignment error and the second alignment error is used as an estimate of the alignment error corresponding to the first mark image information; 5. A lithographic apparatus according to item 4, characterized in that

[0126] (Item 6) 6. The lithographic apparatus of item 5, wherein the statistical value comprises an average value of the first alignment error and the second alignment error, or a weighted average value of the first alignment error and the second alignment error.

[0127] (Item 7) 7. The lithography apparatus according to any one of items 1 to 6, wherein the estimation unit uses Gaussian process regression, Bayesian estimation, multi-layer perceptron, multiple regression analysis, decision tree, polynomial approximation, or a combination thereof.

[0128] (Item 8) 8. The lithography apparatus according to any one of items 1 to 7, further comprising an update unit that updates the first learning model and the second learning model individually.

[0129] (Item 9) 9. The lithography apparatus according to item 8, wherein the update unit updates only the first learning model when processing conditions for performing the processing are changed.

[0130] (Item 10) 10. The lithography apparatus according to item 8 or 9, wherein the update unit periodically updates only the first learning model.

[0131] (Item 11) The lithography apparatus described in any one of items 8 to 10, characterized in that, when at least a portion of the plurality of measuring instruments is replaced or adjusted, the update unit updates only the second learning model for the at least a portion of the measuring instruments.

[0132] (Item 12) 12. A lithography apparatus according to any one of items 1 to 11, characterized in that the process includes an imprint process in which an imprint material on a substrate is molded using the original to form a pattern of the imprint material on the substrate.

[0133] (Item 13) 12. The lithography apparatus according to any one of items 1 to 11, wherein the processing includes an exposure process of projecting a pattern of the original onto the substrate via a projection optical system.

[0134] (Item 14) An information processing device, The method comprises: measuring marks on an original and a substrate with a plurality of measuring instruments, and performing a process of transferring a pattern of the original to the substrate while aligning the original and the substrate based on the amount of misalignment between the original and the substrate output from each of the plurality of measuring instruments; and generating a first learning model used to estimate, from mark image information obtained from a mark image observed by the measuring instruments, an alignment error which is a measurement error of the amount of misalignment of a pattern formed on the substrate, from mark image information obtained from a mark image observed by the measuring instruments; and generating a second learning model different from the first learning model; The generation unit is generating the first learning model by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model; generating the second learning model by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model, the fourth data group including the alignment error output from the first learning model as learning data on the input side of the second learning model, and a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group as learning data on the output side of the second learning model; 23. An information processing apparatus comprising:

[0135] (Item 15) An information processing device, The method comprises: measuring marks on an original and a substrate with a plurality of measuring instruments, and performing a process of transferring a pattern of the original to the substrate while aligning the original and the substrate based on the amount of misalignment between the original and the substrate output from each of the plurality of measuring instruments; and generating a first learning model used to estimate, from mark image information obtained from a mark image observed by the measuring instruments, an alignment error which is a measurement error of the amount of misalignment of a pattern formed on the substrate, from mark image information obtained from a mark image observed by the measuring instruments; and generating a second learning model different from the first learning model; The generation unit is generating the first learning model by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model; generating the second learning model by using, for each of the plurality of measuring instruments, a third data group including the mark image information output from one of the plurality of measuring instruments as input learning data of the second learning model, and a fourth data group including the alignment errors corresponding to each of the mark image information included in the third data group as output learning data of the second learning model; 23. An information processing apparatus comprising:

[0136] (Item 16) a step of measuring marks provided on the original and the substrate with a plurality of measuring instruments to obtain mark image information obtained from mark images observed by each of the plurality of measuring instruments; and estimating, from the mark image information, an alignment error, which is a measurement error of a positional deviation amount of a pattern formed on the substrate by performing a process of transferring a pattern of the original onto the substrate while aligning the original and the substrate, using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model, and using a fourth data group including the alignment error outputted as input learning data of the second learning model, and using a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group as output learning data of the second learning model. The estimation method according to claim 1,

[0137] (Item 17) Item 17. A program for causing a computer to execute the estimation method according to item 16.

[0138] (Item 18) forming a pattern on a substrate using a lithography apparatus according to any one of items 1 to 13; processing the substrate on which the pattern has been formed in the process; producing an article from the processed substrate; A method for producing an article, comprising the steps of:

[0139] 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]

[0140] IMP: Imprint device M: Mold S: Substrate IM: Imprint material 106: Alignment measuring device 110: Control unit 1007: Generation device

Claims

1. A lithography apparatus that performs a process of transferring a pattern of an original onto a substrate, comprising: a plurality of measuring instruments each measuring marks provided on the original and the substrate and outputting an amount of misalignment between the original and the substrate; an estimation unit that estimates an alignment error, which is a measurement error of the positional deviation amount, from mark image information obtained from a mark image observed by a measuring instrument, while aligning the original and the substrate based on the positional deviation amount using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model, and using a fourth data group including the alignment error outputted as input learning data of the second learning model, and using a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group as output learning data of the second learning model.

13. A lithography apparatus comprising:

2. The estimation unit is inputting first mark image information output from a first measuring instrument among the plurality of measuring instruments into the first learning model to obtain a first alignment error output from the first learning model; inputting the first alignment error into the second learning model generated for the first measuring instrument to obtain a second alignment error output from the second learning model; the second alignment error being an estimate of the alignment error corresponding to the first mark image information; A lithographic apparatus according to claim 1 .

3. The second learning model is a learning model generated using the fourth data group and the third data group as input learning data of the second learning model, and the fifth data group as output learning data of the second learning model, The estimation unit is inputting the first mark image information into the second learning model in addition to the first alignment error to obtain a third alignment error output from the second learning model; the third alignment error being an estimate of the alignment error corresponding to the first mark image information; A lithographic apparatus according to claim 2 .

4. A lithography apparatus that performs a process of transferring a pattern of an original onto a substrate, comprising: a plurality of measuring instruments each measuring marks provided on the original and the substrate and outputting an amount of misalignment between the original and the substrate; an estimation unit that estimates an alignment error, which is a measurement error of the positional deviation amount, from mark image information obtained from a mark image observed by a measuring instrument, while aligning the original and the substrate based on the positional deviation amount using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by using, for each of the plurality of measuring instruments, a third data group including the mark image information output from one of the plurality of measuring instruments as learning data on the input side of the second learning model, and a fourth data group including the alignment errors corresponding to each of the mark image information included in the third data group as learning data on the output side of the second learning model.

13. A lithography apparatus comprising:

5. The estimation unit is inputting first mark image information output from a first measuring instrument among the plurality of measuring instruments into the first learning model to obtain a first alignment error output from the first learning model; inputting the first mark image information into the second learning model to obtain a second alignment error output from the second learning model; a statistical value of the first alignment error and the second alignment error is used as an estimate of the alignment error corresponding to the first mark image information; A lithographic apparatus according to claim 4 .

6. The lithographic apparatus of claim 5, wherein the statistical value comprises an average value of the first alignment error and the second alignment error, or a weighted average value of the first alignment error and the second alignment error.

7. The lithography apparatus according to claim 1 or 4, wherein the estimation unit uses Gaussian process regression, Bayesian estimation, multi-layer perceptron, multiple regression analysis, decision tree, polynomial approximation, or a combination thereof.

8. The lithography apparatus according to claim 1 , further comprising an update unit that updates the first learning model and the second learning model individually.

9. The lithography apparatus according to claim 8 , wherein the update unit updates only the first learning model when a processing condition for performing the processing is changed.

10. The lithography apparatus according to claim 8 , wherein the update unit periodically updates only the first learning model.

11. 9. The lithography apparatus according to claim 8, wherein the update unit updates only the second learning model for at least a portion of the plurality of measuring instruments when the at least a portion of the measuring instruments is replaced or adjusted.

12. 5. The lithographic apparatus according to claim 1, wherein the process includes an imprint process for forming a pattern of the imprint material on the substrate by molding an imprint material on the substrate with the original.

13. 5. The lithography apparatus according to claim 1, wherein the processing includes an exposure process of projecting a pattern of the original onto the substrate via a projection optical system.

14. An information processing device, The apparatus includes a generation unit that generates a first learning model used to estimate, from mark image information obtained from a mark image observed by a measuring instrument, a measurement error of a positional deviation amount of a pattern formed on the substrate by performing a process of transferring a pattern of the original onto the substrate while aligning the original and the substrate based on the positional deviation amount between the original and the substrate output from each of the plurality of measuring instruments, and a second learning model different from the first learning model, and The generation unit is generating the first learning model by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model; a fourth data group including the alignment error output by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model for each of the plurality of measuring instruments, the fourth data group being used as learning data on the input side of the second learning model, and a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group being used as learning data on the output side of the second learning model to generate the second learning model.

23. An information processing apparatus comprising:

15. An information processing device, The apparatus includes a generation unit that generates a first learning model used to estimate, from mark image information obtained from a mark image observed by a measuring instrument, a measurement error of a positional deviation amount of a pattern formed on the substrate by performing a process of transferring a pattern of the original onto the substrate while aligning the original and the substrate based on the positional deviation amount between the original and the substrate output from each of the plurality of measuring instruments, and a second learning model different from the first learning model, and The generation unit is generating the first learning model by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model; generating the second learning model by using, for each of the plurality of measuring instruments, a third data group including the mark image information output from one of the plurality of measuring instruments as learning data on an input side of the second learning model, and a fourth data group including the alignment errors corresponding to each of the mark image information included in the third data group as learning data on an output side of the second learning model; 23. An information processing apparatus comprising:

16. a step of measuring marks provided on the original and the substrate with a plurality of measuring instruments to obtain mark image information obtained from mark images observed by each of the plurality of measuring instruments; and estimating, from the mark image information, an alignment error, which is a measurement error of a positional deviation amount of a pattern formed on the substrate by performing a process of transferring a pattern of the original onto the substrate while aligning the original and the substrate, using a first learning model and a second learning model different from the first learning model, the first learning model is a learning model generated by using a first data group including all of the mark image information output from each of the multiple measuring instruments as learning data on an input side of the first learning model, and using a second data group including the alignment errors corresponding to all of the mark image information included in the first data group as learning data on an output side of the first learning model, The second learning model is a learning model generated by inputting a third data group including the mark image information output from one of the plurality of measuring instruments into the first learning model, and using a fourth data group including the alignment error outputted as input learning data of the second learning model, and using a fifth data group including the alignment error corresponding to each of the mark image information included in the third data group as output learning data of the second learning model. The estimation method according to claim 1,

17. A program causing a computer to execute the estimation method according to claim 16.

18. forming a pattern on a substrate using a lithographic apparatus according to claim 1 or 4; processing the substrate on which the pattern has been formed in the process; producing an article from the processed substrate; A method for producing an article, comprising the steps of:

Citation Information

Patent Citations

  • Information processing apparatus, program, lithographic apparatus, lithographic system, and method of manufacturing article

    JP2019102495A