Mark detecting apparatus, mark learning apparatus, substrate processing apparatus, mark detecting method, and learning model generation method

Abstract

To automatically and quickly adjust parameters related to image capture to obtain good alignment mark images.SOLUTION: A mark detecting apparatus includes: imaging means 131 for generating an alignment mark image by imaging an alignment mark on an object; detection means 110 for detecting the alignment mark in the alignment mark image; and adjustment means 110 for adjusting parameters relating to the imaging. The adjustment means acquires second parameters as a result of inference processing by a learning model generated through a learning process that uses first parameters which are parameters during imaging the alignment mark images where alignment marks could not be detected and alignment mark images where alignment marks were detected. The imaging means performs the imaging after the parameters are adjusted to the second parameters.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technique for detecting alignment marks from image data obtained by capturing an image of the alignment marks. [Background technology]

[0002] The above-described alignment mark detection is performed, for example, in an exposure apparatus that uses photolithography to manufacture semiconductor devices and display devices. The exposure apparatus projects and transfers a pattern from an original, such as a mask or reticle, onto a substrate (such as a wafer or glass plate) via a projection optical system. When transferring multiple layers of different original patterns onto the same substrate, highly accurate alignment is required to prevent deviation of the actual transfer position from the target transfer position. This alignment involves detecting the alignment mark position using image data (alignment mark image) obtained by capturing an image of the alignment mark, and then correcting the position based on the detection result.

[0003] Patent Document 1 discloses a method for detecting the position of an alignment mark in an alignment mark image by matching using a template, in which the template is optimized by automatically deforming it in response to deformation of the alignment mark and changes in brightness. This method improves the alignment mark detection rate by self-learning so that the optimized template can be used in subsequent matching. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2003-338455 A Summary of the Invention [Problem to be solved by the invention]

[0005] However, even when using the method of Patent Document 1, an operator must first manually align the alignment mark with the imaging position and adjust multiple parameters, such as the wavelength and illuminance of the light used for imaging, to ensure good imaging of the alignment mark. It takes time for the operator to determine the combination of parameters to use. Furthermore, because the state of the alignment mark in the alignment image changes depending on the physical properties of the resist applied to the substrate, the operator must have sufficient knowledge to determine the combination.

[0006] On the other hand, there are exposure tools that automate parameter adjustment, but these tools automatically change parameter combinations in sequence and determine whether they are appropriate, so it can take time to determine the parameter combination to use.

[0007] The present invention provides an alignment mark detection device and the like that can automatically and quickly adjust parameters related to imaging to obtain good alignment mark images. [Means for solving the problem]

[0008] A mark detection device as one aspect of the present invention comprises an imaging means for generating an alignment mark image by imaging an alignment mark on an object, a detection means for detecting an alignment mark in the alignment mark image, and an adjustment means for adjusting parameters related to imaging based on the output of a learning model that outputs parameters inferred when an alignment mark image is input, wherein the learning model is generated by learning using an alignment mark image in which an alignment mark could not be detected and a first parameter as a parameter at the time of imaging the alignment mark image in which an alignment mark could be detected, the adjustment means obtains a second parameter output from the learning model by inputting the first alignment mark image into the learning model, and the imaging means performs imaging with the parameter adjusted from the initial parameter to the second parameter.

[0009] A substrate processing apparatus having the above-described mark detection apparatus also constitutes another aspect of the present invention.Furthermore, a method for manufacturing an article using the substrate processing apparatus also constitutes another aspect of the present invention.

[0010] Another aspect of the present invention is a mark detection method comprising the steps of generating an alignment mark image by capturing an image of an alignment mark on an object, detecting the alignment mark in the alignment mark image, and adjusting parameters related to the image capture, wherein the parameter adjustment step adjusts the parameters based on the output of a learning model that outputs parameters inferred by inputting an alignment image captured with initial parameters, the learning model being generated by learning using an alignment mark image in which an alignment mark could not be detected and a first parameter as a parameter at the time of capturing the alignment mark image in which an alignment mark could be detected, and the parameter adjustment step acquires a second parameter output from the learning model, and captures the image with the parameter adjusted from the initial parameter to the second parameter. Note that a program for causing a computer to execute processing in accordance with the above-described mark detection method also constitutes another aspect of the present invention. [Effects of the Invention]

[0011] According to the present invention, by using a learning model that uses alignment mark images as input, it is possible to automatically and quickly adjust parameters related to imaging in order to obtain good alignment mark images. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a block diagram showing the configuration of a mark detection device according to a first embodiment. [Figure 2] FIG. 2 is a diagram showing the relationship between parameters and units to be adjusted in the first embodiment. [Figure 3] 10 is a flowchart showing a parameter inference process in the first embodiment. [Figure 4] 10A to 10C are diagrams showing alignment mark images with different illuminances of observation light. [Figure 5] 10A to 10C are diagrams showing alignment mark images in which the positions of the alignment marks within the observation field of view are different. [Figure 6] 4 is a flowchart showing a parameter learning process in the first embodiment. [Figure 7] FIG. 10 is a diagram showing a network configuration of a learning shared server in the second embodiment. [Figure 8] FIG. 1 is a diagram showing the configuration of an exposure apparatus equipped with a mark detection apparatus according to a first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. [Example]

[0014] 1 shows the configuration (mark detection device) related to alignment mark measurement in the exposure apparatus of Example 1. The exposure apparatus has an operation terminal 100, a control unit 110, a learning server 120, and a drive unit 130. A sequence controller 111 in the control unit 110 performs sequence control by issuing measurement and drive instructions to a measurement controller 112 and a drive controller 113, respectively, in accordance with commands from the operation terminal 100 operated by an operator.

[0015] The measurement controller 112 communicates with the drive unit 130 to control the measurement unit 131, observation light unit 133, and wavelength correction unit 134 within the drive unit 130. The measurement unit 131, which serves as an imaging means, captures an image of an alignment mark provided on an object such as a mask or plate (substrate) within its observation field of view (imaging area) to generate an alignment image. The observation light unit 133 includes multiple light sources such as LEDs with different wavelengths. The wavelength correction unit 134 controls the light reflected from the mask or plate to correct the wavelength of the light captured by the measurement unit 131.

[0016] The drive controller 113 communicates with the drive unit 130 to control the objective lens drive unit 132, mask stage drive unit 135, and plate stage drive unit 136 in the drive unit 130. The objective lens drive unit 132 moves an objective lens for transferring a mask pattern onto a plate by exposure. The mask stage drive unit 135 and plate stage drive unit 136 move stages that support the mask and plate, respectively.

[0017] The operation terminal (operation means) 100 is equipped with an operation unit 101 that allows adjustment of multiple parameters related to imaging for acquiring an alignment mark image, and adjusts the parameters in response to an operator's operation of the operation unit 101. The multiple parameters related to imaging are values that, when adjusted, change the state (brightness, contrast, position, etc.) of the alignment mark in the alignment mark image.

[0018] The learning server (alignment mark learning device) 120 includes a learning model 121 used to acquire (infer) a suitable combination of parameters, and a learning unit 122 that updates the learning model 121.

[0019] Imaging to obtain an alignment mark image in an exposure apparatus is performed by a measurement unit 131 after adjusting the imaging conditions by adjusting parameters such as those shown in FIG. 2. Specifically, the position of the alignment mark within the observation field of view of the measurement unit 131 is adjusted by driving the objective lens driving unit 132, the mask stage driving unit 135, and the plate stage driving unit 136. Furthermore, during imaging, the observation field of view is illuminated with observation light. The wavelength of the observation light is adjusted by switching to a light source of a different wavelength in the observation light unit 133, adjusting an aperture arranged in the observation light path to extract light of a specific wavelength band, or suppressing specular reflection light using a wavelength correction unit 134. Furthermore, the illuminance of the observation light is adjusted by increasing or decreasing the current value input to the light source in the observation light unit 133 to change the light emission amount.

[0020] The flowchart in Fig. 3 shows an inference process for inferring a suitable combination of parameters using a learning model 121 that receives as input an alignment mark image captured by a measurement unit 131. A control unit 110 (sequence controller 111, measurement controller 112, and drive controller 113) configured by a computer executes this process according to a computer program. The control unit 110 corresponds to a detection means and an adjustment means.

[0021] First, the sequence controller 111 issues a command to the measurement controller 112 and the drive controller 113 to measure the position of the alignment mark. In response to this, the drive controller 113 acquires a combination of parameters (hereinafter referred to as initial parameters) already stored in the operation terminal 100 and sets them in each unit in the drive section 130. As a result, the position of the alignment mark moves within (or outside) the observation field of view to a position according to the initial parameters, and the wavelength and illuminance of the observation light from the observation light unit 133 are set according to the initial parameters.

[0022] In step S300, measurement controller 112 causes measurement unit 131 to capture an image of the alignment mark. The alignment mark image obtained by this image capture is defined as a first alignment mark image.

[0023] In step S301, the sequence controller 111 inputs the first alignment mark image to the learning model 121 in the learning server 120. The learning model 121 outputs a combination of suitable parameters (second parameters: hereinafter referred to as inference parameters) as an inference result.

[0024] The learning model 121 is generated by machine learning using various alignment mark images where measurement has failed in the past and parameters (first parameters) at the time of capturing alignment mark images where measurement was successful. However, the learning model 121 does not have to be generated within the exposure apparatus of this embodiment, and may be a learning model generated by another exposure apparatus, or may be a learning model generated using a log of alignment mark images and parameters acquired from outside.

[0025] Machine learning can be performed using, for example, a convolutional neural network. A convolutional neural network is an algorithm that has an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer, and that takes an image as input and outputs a single inference result. In this embodiment, a learning model 121 that uses a convolutional neural network will be described, but the learning model 121 may also be generated using other models or algorithms.

[0026] In step S302, the sequence controller 111 applies the combination of inference parameters output from the learning model 121 to each unit in the drive section 130 via the measurement controller 112 and the drive controller 113. As a result, each parameter related to imaging is adjusted to the inference parameter.

[0027] Then, in step S303 after application of the inference parameters, the sequence controller 111 causes the measurement unit 131 to capture an image of the alignment mark again via the measurement controller 112. The alignment image obtained by this image capture is defined as a second alignment mark image.

[0028] 4 shows examples of a first alignment mark image and a second alignment mark image. If the illuminance of the observation light set by the initial parameters is low, for example, 20%, a first alignment mark image 400 with insufficient brightness or low contrast is obtained. If the illuminance of the observation light is increased, for example, to 70%, by applying the inference parameters obtained by inputting first alignment mark image 400 into learning model 121, a second alignment mark image 401 with high brightness or high contrast is obtained.

[0029] FIG. 5 shows another example of a first alignment mark image and a second alignment mark image. If the position of the alignment mark set by the initial parameters is such that at least a portion of the mark is outside the observation field of view, a first alignment mark image 500 is obtained in which at least a portion of the alignment mark is missing. When the first alignment mark image 500 is input to the learning model 121, the plate stage drive amount is output as an inference parameter. When the plate stage drive unit 136 drives the plate stage on which the plate is mounted according to this plate stage drive amount, the alignment mark is positioned at the center of the observation field of view. This results in a second alignment mark image 501 in which the alignment mark is centered.

[0030] In step S304, the sequence controller 111 measures the position of the alignment mark in the second alignment mark image. Then, in step S305, it is determined whether the measurement was possible (in other words, whether the alignment mark was detected). If the alignment mark position could be measured (success), the inference process ends. If the measurement was impossible (failed), the process proceeds to the learning process described next.

[0031] According to this embodiment, by using a learning model that uses alignment mark images as input, it is possible to automatically and quickly adjust parameters related to imaging in order to obtain good alignment mark images. Furthermore, by using good alignment mark images, it is possible to position a mask or plate with high accuracy.

[0032] The flowchart in Figure 6 shows the learning process (learning model generation method). Here, we will explain the learning process that is executed when measurement of the position of the alignment mark in the second alignment mark image obtained by imaging using the combination of inference parameters obtained by the above-mentioned inference process fails. However, the cases in which the learning process is executed are not limited to this case.

[0033] First, in step S600, the sequence controller 111 automatically adjusts a combination of parameters that enables measurement according to a predetermined adjustment procedure. At this time, the sequence controller 111 searches for a combination of parameters that enables measurement while driving the objective lens drive unit 132, the observation light unit 133, and the wavelength correction unit 134 according to predetermined rules via the measurement and drive controllers 112 and 113. In this step, the sequence controller 111 may also acquire a combination of parameters input by an operator via the operation unit 101.

[0034] Next, in step S601, the sequence controller 111 applies the combination of parameters obtained in step S600 to each unit in the drive section 130 via the measurement and drive controllers 112 and 113. Then, the sequence controller 111 causes the measurement unit 131 to capture an image of the alignment mark.

[0035] Next, in step S602, the sequence controller 111 measures the position of the alignment mark in the alignment mark image obtained by imaging. Then, in step S603, it is determined whether the measurement was possible. If the measurement was possible (successful), in step S604, the sequence controller 111 causes the learning unit 122 to update the learning model 121. The learning unit 122 corresponds to an acquisition means and a learning means.

[0036] In updating the learning model 121, as described above, the learning unit 122 performs learning using as input a combination of the first alignment mark image where measurement failed and the parameters when measurement was successful in step S602 (hereinafter referred to as success parameters). When machine learning using a convolutional neural network is performed as the learning here, for example, backpropagation can be used.

[0037] In machine learning using a convolutional neural network, a combination of successful parameters is used as the correct answer data, and a combination of inference parameters obtained using the learning model 121 is used as the output data. Then, the weights between neurons are adjusted so that the error between this correct answer data and the output data is small. As the error between the correct answer data and the output data, the sum of squares error shown in the following formula (1) or the cross entropy error shown in formula (2) can be used. In each formula, E is the error value, y is the output data value, and t is the correct answer data value.

[0038]

number

[0039]

number

[0040] At this time, the learning model 121 is updated using the first alignment mark image acquired in the inference process.

[0041] If the alignment mark measurement is not possible (failed) in step S602, the sequence controller 111 returns to step S600 and repeats the processing described above.

[0042] According to this embodiment, a learning model for inferring parameters for imaging using an alignment mark image as an input can be optimized automatically or with little burden on the user.

[0043] In this embodiment, the present invention has been described as being used in an exposure apparatus serving as a substrate processing apparatus that forms a pattern on a substrate by exposing the substrate positioned using an alignment mark detected by a mark detection device. However, the present invention can also be used in other substrate processing apparatuses. For example, the present invention can be used in substrate processing apparatuses such as an imprinting apparatus that forms a pattern of an imprint material on a substrate using a mold, or a drawing apparatus that irradiates a substrate with a charged particle beam to form a pattern on the substrate. The present invention can also be used in substrate processing apparatuses such as a coating apparatus that coats a photosensitive medium on the surface of a substrate, or a developing apparatus that develops the photosensitive medium to which a pattern has been transferred. Furthermore, the present invention can be used in substrate processing apparatuses such as a film formation apparatus (CVD apparatus, etc.), a processing apparatus (laser processing apparatus, etc.), an inspection apparatus (overlay inspection apparatus, etc.), and a measurement apparatus (mark measurement apparatus, etc.).

[0044] 8 shows the configuration of an exposure apparatus 10. The exposure apparatus 10 projects an image of a pattern of a mask M onto a plate W serving as a substrate via a projection optical system (objective lens) 14, thereby exposing the plate W. The plate W and the mask M are provided with alignment marks that are detected by a mark detection device.

[0045] The direction parallel to the optical axis of the projection optical system 14 is defined as the Z-axis direction, and the two directions perpendicular to each other in a plane perpendicular to the Z-axis direction are defined as the X-axis direction and the Y-axis direction. Rotation around the X-axis, rotation around the Y-axis, and rotation around the Z-axis are defined as θX, θY, and θZ, respectively.

[0046] The exposure apparatus 10 includes a light source 11, an illumination optical system 12, a mask stage 13, the above-mentioned projection optical system 14, a plate stage 15, and a main controller 16. The exposure apparatus 10 also includes a mask stage drive unit 21 (135 in FIG. 1) that drives the mask stage 13, and an objective lens drive unit 22 (132 in FIG. 1) that drives the lens 14a of the projection optical system 14. The exposure apparatus 10 also includes a plate stage drive unit 51 (136 in FIG. 1) that drives the plate stage 15. The mask stage drive unit 21, the objective lens drive unit 22, and the plate stage drive unit 51 are controlled by a mask stage controller 31, a lens controller 32, and a plate stage controller 41, respectively, which are included in the measurement controller 112 shown in FIG. 1. The light source 11 to the plate stage controller 41 correspond to processing means that perform processing on the plate W (forming a pattern by exposure).

[0047] Main control unit 16 controls the overall operation of exposure apparatus 10, and includes control unit 110 and learning server 120 shown in Fig. 1. In addition, main control unit 16 is connected to operation terminal 100 shown in Fig. 1.

[0048] The light source 11 emits exposure light. The illumination optical system 12 illuminates the mask M using the light emitted from the light source 11. The mask stage 13 holds the mask M and is moved by a mask stage drive unit 21 within an XY plane perpendicular to the optical axis of the projection optical system 14. The projection optical system 14 projects an image of the pattern of the mask M illuminated by the illumination optical system 12 onto the plate W. The projection optical system 14 includes an optical element 14a that can be moved in the X-axis direction by an objective lens drive unit 22. The plate stage 15 holds the plate W and is moved in parallel within the XY plane and rotated in the θZ direction by a plate stage drive unit 51.

[0049] The main control unit 16 controls the mask stage drive unit 21, the objective lens drive unit 22, and the plate stage drive unit 51 via the mask stage control unit 31, the lens control unit 32, and the plate stage control unit 41 in accordance with the measurement results of the positions of the alignment marks detected by the mark detection device. In this way, the mask M and the plate W are positioned within the XY plane.

[0050] The exposure apparatus 10 is suitable for manufacturing articles such as microdevices such as semiconductor devices, elements having fine structures, and flat panel displays. A method for manufacturing an article includes a step of processing a substrate using the exposure apparatus 10 and a step of manufacturing an article from the substrate processed in the step. The manufacturing method may further include well-known processes (exposure, oxidation, film formation, vapor deposition, doping, planarization, etching, resist stripping, dicing, bonding, packaging, etc.). A method for manufacturing an article using the exposure apparatus 10 is advantageous over conventional methods in at least one of the performance, quality, productivity, and production costs of the article. [Example]

[0051] 7 shows a network configuration as Example 2. Here, an exposure tool 701 is connected to a learning shared server 700 via a network together with other exposure tools.

[0052] In the first embodiment, a case was described in which inference parameters were obtained and the learning model was updated using a learning model within one exposure tool (701). In contrast, in this embodiment, multiple exposure tools including the exposure tool 701 are connected to a learning shared server 700, and inference parameters are obtained and the learning model is updated using the learning model within the learning shared server 700. In other words, the learning model is shared among multiple exposure tools.

[0053] By sharing the learning model among multiple exposure tools, it is possible to obtain inference parameters using the learning model even for newly installed exposure tools that have not yet undergone learning. Furthermore, by performing learning using alignment mark images and success parameters obtained from multiple exposure tools, the accuracy of learning can be further improved.

[0054] The learning shared server 700 may be installed either inside or outside the factory where the exposure apparatus is installed. Furthermore, the same learning shared server 700 may be used by exposure apparatuses in multiple factories via a network. (Other Examples) The present invention can also be realized by supplying a program that realizes one or more functions of the above-described embodiments to a system or device via a network or a storage medium, and having one or more processors in the computer of the system or device read and execute the program. It can also be realized by a circuit (e.g., ASIC) that realizes one or more functions. The embodiments described above are merely representative examples, and various modifications and alterations are possible to each embodiment when implementing the present invention. [Explanation of symbols]

[0055] 111 Sequence Controller 112 Measurement Controller 113 Drive Controller 121 Learning Model 131 Measurement Unit

Claims

1. an imaging means for capturing an image of an alignment mark on an object to generate an alignment mark image; a detection means for detecting an alignment mark in the alignment mark image; an adjustment means for adjusting the parameters related to the imaging based on an output of a learning model that outputs parameters inferred by inputting the alignment mark image; the learning model is generated by learning using an alignment mark image in which the alignment mark cannot be detected and a first parameter as the parameter at the time of capturing the alignment mark image in which the alignment mark can be detected, the adjusting means inputs a first alignment mark image into the learning model to obtain a second parameter output from the learning model; The mark detection device is characterized in that the imaging means performs the imaging in a state where the parameters are adjusted from the initial parameters to the second parameters.

2. the learning model outputs, as the second parameters, a position of an alignment mark in an observation field of view, a wavelength of imaging light, and an illuminance of imaging light; 2. The mark detection device according to claim 1, wherein said adjusting means adjusts only the imaging conditions that should be adjusted with respect to said initial parameters.

3. 2. The mark detection device according to claim 1, further comprising a learning means for performing learning to generate the learning model.

4. The mark detection device according to claim 3, characterized in that the learning means performs the learning when it is unable to detect an alignment mark in an alignment mark image generated by the imaging while the parameter is adjusted to the second parameter.

5. generating an alignment mark image by imaging an alignment mark on an object; detecting an alignment mark in the alignment mark image; adjusting parameters related to the imaging; In the step of adjusting the parameters, the parameters are adjusted based on the output of a learning model that outputs the parameters inferred by inputting the alignment image captured with initial parameters; the learning model is generated by learning using an alignment mark image in which the alignment mark cannot be detected and a first parameter as the parameter at the time of capturing the alignment mark image in which the alignment mark can be detected, In the step of adjusting the parameters, a second parameter output from the learning model is acquired; a mark detection method, characterized in that the imaging is performed in a state in which the parameters are adjusted from the initial parameters to the second parameters;

6. the learning model outputs, as the second parameters, a position of an alignment mark in an observation field of view, a wavelength of imaging light, and an illuminance of imaging light; 6. The mark detection method according to claim 5, wherein in the step of adjusting the parameters, only the imaging conditions to be adjusted with respect to the initial parameters are adjusted.

7. 7. A program for causing a computer to execute a process according to the mark detection method of claim 5.

8. The mark detection device according to any one of claims 1 to 4, and processing means for performing processing on a substrate positioned using the alignment mark detected by the mark detection device.

9. Processing a substrate using the substrate processing apparatus according to claim 8; and producing an article from the processed substrate.

Images