Film thickness estimation device, film thickness estimation method, and computer program

Abstract

The present invention provides a film thickness estimation apparatus and method for accurately estimating the film thickness of a coating film formed on a substrate. [Solution] The autoencoder 200 of the base model Md1 has a first encoder unit and a decoder unit. The first encoder unit takes film thickness data x1 as input and outputs a first latent variable z1. The decoder unit takes the first latent variable z1 as input and outputs reconstructed film thickness data x1′. The second encoder unit takes process parameters x2 as input and outputs a second latent variable z2. The first learning process ST1 trains the autoencoder to minimize the error between the film thickness data x1 input to the autoencoder and the output film thickness data x1′. In the second learning process ST2, the second encoder unit is trained to minimize the error between the first latent variable z1 obtained by inputting film thickness data x1 to the first encoder unit and the second latent variable z2 obtained by inputting process parameters x2 to the second encoder unit.

Description

Technical Field

[0001] The subject matter disclosed in this specification relates to a technique for estimating the film thickness of a coating film applied to a substrate.

Background Art

[0002] In the manufacturing process of flat panel displays, a device called a coater is used. A coater is a substrate processing device that discharges a processing liquid from a slit nozzle by a pump and applies the processing liquid to the entire substrate being conveyed. With such a coater, along with the higher quality of recent products, it has been required to apply the processing liquid so that the film thickness of the processing liquid becomes uniform over the entire substrate. For example, in Patent Document 1, the discharge characteristics of the processing liquid are repeatedly measured, and the parameters of pump control are adjusted for optimization.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the case of the prior art, by repeating the control process and the measurement of the discharge characteristics a plurality of times, there was a risk that the time required for optimizing the control parameters would be prolonged. Therefore, it is conceivable to alternately perform optimization using a film thickness estimation model and optimization by time-consuming actual measurement to minimize the number of actual measurements. However, in this system, it is necessary to prepare a film thickness estimation model, and its accuracy greatly affects the accuracy of optimization. [[ID=?]] [[ID=?]]

[0005] [[ID=?]] An object of the present invention is to provide a technique capable of accurately estimating the film thickness of a coating film formed on a substrate.

Means for Solving the Problems

[0006] To solve the above problems, the first embodiment is a film thickness estimation device for estimating the film thickness of a coating film applied to a substrate, comprising: a learning unit for training a base model; and a film thickness estimation unit that estimates film thickness data from process parameters related to the coating process using a submodel based on the trained base model obtained by the learning unit, wherein the base model includes an autoencoder having a first encoder unit that takes the film thickness data as input and outputs a first latent variable compressed to a low dimension, and a decoder unit that takes the first latent variable output from the first encoder unit as input and outputs reconstructed film thickness data, and a second latent variable that takes the process parameters as input and whose number of dimensions matches that of the first latent variable The submodel comprises a second encoder unit that outputs a film thickness data, and the submodel comprises the second encoder unit and the decoder unit connected to the output side of the second encoder unit, and the learning unit performs a first learning process to train the autoencoder to minimize the error between the film thickness data input to the autoencoder and the film thickness data output based on that input, and a second learning process to train the second encoder unit to minimize the error between a first latent variable obtained by inputting the film thickness data to the first encoder unit and a second latent variable obtained by inputting the process parameter corresponding to the film thickness data input to the first encoder unit to the second encoder unit.

[0007] A second embodiment is the film thickness estimation apparatus of the first embodiment, wherein the process parameters include control parameters for controlling a coating apparatus that applies a processing solution to the substrate.

[0008] A third embodiment is the film thickness estimation apparatus of the second embodiment, wherein the process parameters further include apparatus information specific to the coating apparatus.

[0009] A fourth aspect is a film thickness estimation method for estimating the film thickness of a coating applied to a substrate, comprising: a) a learning step of training a base model; and b) an estimation step of estimating film thickness data from process parameters related to the coating process using a submodel based on the trained base model obtained in the learning step, wherein the base model includes an autoencoder having a first encoder unit that takes the film thickness data as input and outputs a first latent variable compressed to a low dimension, and a decoder unit that uses the first latent variable output from the first encoder unit and outputs reconstructed film thickness data, and an autoencoder that takes the process parameters as input and outputs a second latent variable whose number of dimensions matches that of the first latent variable. The submodel comprises a second encoder unit and a decoder unit connected to the output side of the second encoder unit, and the learning step includes a1) a first learning step of training the autoencoder to minimize the error between the film thickness data input to the autoencoder and the film thickness data output based on that input, and a2) a step of training the second encoder unit to minimize the error between a first latent variable obtained by inputting the film thickness data to the first encoder unit and a second latent variable obtained by inputting the process parameter corresponding to the film thickness data input to the first encoder unit to the second encoder unit.

[0010] The fifth aspect is a computer program that causes a computer to execute the film thickness estimation method of the fourth aspect. [Effects of the Invention]

[0011] According to the first to fifth embodiments, the second encoder unit can be trained using the latent variables of the autoencoder as hints, thereby utilizing the training accuracy of the autoencoder. Therefore, by using a submodel that combines the second encoder unit and the decoder unit, film thickness data can be estimated with high accuracy from process parameters.

[0012] According to the film thickness estimation device of the second embodiment, since film thickness data can be estimated by taking into account the control information of the coating device, the accuracy of film thickness estimation can be improved.

[0013] According to the film thickness estimation device of the third embodiment, since film thickness data can be estimated taking into account the characteristics of the coating device, the accuracy of film thickness estimation can be improved. [Brief explanation of the drawing]

[0014] [Figure 1] This diagram schematically shows the overall configuration of a coating apparatus equipped with a film thickness estimation device according to an embodiment. [Figure 2] This figure shows the configuration of the processing liquid supply mechanism included in the coating apparatus shown in Figure 1. [Figure 3] This is a block diagram showing the configuration of the control unit included in the coating apparatus shown in Figure 1. [Figure 4] This is a flowchart showing the film thickness estimation method according to the present invention. [Figure 5] This figure shows the structure of the base model used in the learning process shown in Figure 4. [Figure 6] This figure shows the structure of the submodel used in the estimation process shown in Figure 4. [Modes for carrying out the invention]

[0015] Embodiments of the present invention will be described below with reference to the attached drawings. Note that, for ease of understanding, the dimensions and number of parts in the drawings may be exaggerated or simplified.

[0016] <1. Embodiment> Figure 1 is a schematic diagram showing the overall configuration of a coating apparatus 1 equipped with a film thickness estimation device according to an embodiment. The coating apparatus 1 is a substrate processing apparatus that forms a coating film on a substrate S by discharging a processing liquid onto the upper surface Sf of the substrate S. As will be described later, the control unit 9 of the coating apparatus 1 is equipped with a function as a film thickness estimation device that estimates the film thickness of the coating film.

[0017] The substrate S is, for example, a glass substrate for a liquid crystal display device. Note that the substrate S may be various substrates to be processed for electronic devices such as a semiconductor wafer, a glass substrate for a photomask, a glass substrate for a plasma display, a glass or ceramic substrate for a magnetic / optical disk, a glass substrate for an organic EL, a glass substrate for a solar cell or a silicon substrate, and other flexible substrates and printed boards. The coating device 1 is, for example, a slit coater.

[0018] In FIG. 1, an XYZ coordinate system is defined to explain the arrangement relationship of the elements of the coating device 1. The conveyance direction Dt of the substrate S is the "X direction". The direction in which the substrate S advances in the X direction (the direction toward the downstream of the conveyance direction Dt) is the +X direction, and the opposite direction (the direction toward the upstream of the conveyance direction Dt) is the -X direction. Also, the direction orthogonal to the X direction is the Y direction, and the direction orthogonal to the X direction and the Y direction is the Z direction. In the following description, the Z direction is the vertical direction, and the X direction and the Y direction are the horizontal directions. In the Z direction, the +Z direction is the upward direction, and the -Z direction is the downward direction. Note that these directions are not intended to limit the arrangement of the coating device.

[0019] The coating device 1 includes, in order toward the +X direction, an input conveyor 100, an input transfer unit 2, a floating stage unit 3, an output transfer unit 4, and an output conveyor 110. The input conveyor 100, the input transfer unit 2, the floating stage unit 3, the output transfer unit 4, and the output conveyor 110 form a conveyance path through which the substrate S passes. Also, the coating device 1 further includes a substrate conveyance unit 5, a coating mechanism 7, a processing liquid supply mechanism 8, and a control unit 9.

[0020] The substrate S is conveyed to the input conveyor 100 from a device or the like on the upstream side of the coating device 1. The input conveyor 100 includes a roller conveyor 101 and a rotation drive mechanism 102. The rotation drive mechanism 102 rotates each roller of the roller conveyor 101. By the rotation of each roller of the roller conveyor 101, the substrate S is conveyed downstream (+X direction) in a horizontal posture. The "horizontal posture" means a state in which the main surface (the surface with the largest area) of the substrate S is parallel to the horizontal plane (XY plane).

[0021] The input transfer unit 2 includes a roller conveyor 21 and a rotation / elevation drive mechanism 22. The rotation / elevation drive mechanism 22 rotates each roller of the roller conveyor 21 and elevates the roller conveyor 21. By the rotation of the roller conveyor 21, the substrate S is conveyed downstream (+X direction) in a horizontal posture. Also, by the elevation of the roller conveyor 21, the position of the substrate S in the Z direction is changed. The substrate S is transferred from the input conveyor 100 to the floating stage unit 3 via the input transfer unit 2.

[0022] As shown in FIG. 1, the floating stage unit 3 is substantially flat plate-shaped. The floating stage unit 3 is divided into three parts along the X direction. The floating stage unit 3 includes, in order toward the +X direction, an entrance floating stage 31, an application stage 32, and an exit floating stage 33. The upper surfaces of the entrance floating stage 31, the application stage 32, and the exit floating stage 33 are on the same plane. The floating stage unit 3 further includes a lift pin drive mechanism 34, a floating control mechanism 35, and an elevation drive mechanism 36. The lift pin drive mechanism 34 raises and lowers a plurality of lift pins arranged on the entrance floating stage 31. The floating control mechanism 35 supplies compressed air for floating the substrate S to the entrance floating stage 31, the application stage 32, and the exit floating stage 33. The elevation drive mechanism 36 raises and lowers the exit floating stage 33.

[0023] On the upper surfaces of the entrance floating stage 31 and the exit floating stage 33, a large number of ejection holes for ejecting compressed air supplied from the floating control mechanism 35 are arranged in a matrix. When compressed air is ejected from each ejection hole, the substrate S floats upward with respect to the floating stage unit 3. Then, while the lower surface Sb of the substrate S is separated from the upper surface of the floating stage unit 3, the substrate S is supported in a horizontal posture. The distance (floating amount) between the lower surface Sb of the substrate S and the upper surface of the floating stage unit 3 in the state where the substrate S is floating is, for example, 10 μm or more and 500 μm or less.

[0024] The upper surface of the coating stage 32 is provided with ejection holes for ejecting compressed air supplied from the levitation control mechanism 35 and suction holes for drawing in gas. The ejection holes and suction holes are arranged alternately in the X and Y directions. The levitation control mechanism 35 controls the amount of compressed air ejected from the ejection holes and the amount of air drawn in from the suction holes. This precisely controls the amount of levitation of the substrate S relative to the coating stage 32 so that the position of the upper surface Sf of the substrate S passing above the coating stage 32 in the Z direction is a specified value. The amount of levitation of the substrate S relative to the coating stage 32 is calculated by the control unit 9 based on the detection results of the sensor 61 or sensor 62, which will be described later. Furthermore, the amount of levitation of the substrate S relative to the coating stage 32 is preferably adjustable with high precision by airflow control.

[0025] The substrate S, which has been brought into the levitation stage section 3, is propelled in the +X direction by the roller conveyor 21 and transported onto the inlet levitation stage 31. The inlet levitation stage 31, coating stage 32, and outlet levitation stage 33 support the substrate S in a levitated state. For the levitation stage section 3, for example, the configuration described in Japanese Patent No. 5346643 can be used.

[0026] The substrate transport unit 5 is located below the levitation stage unit 3. The substrate transport unit 5 comprises a chuck mechanism 51 and a suction / travel control mechanism 52. The chuck mechanism 51 is equipped with a suction pad (not shown) provided on a suction member. The chuck mechanism 51 supports the substrate S from below by bringing the suction pad into contact with the peripheral edge of the lower surface Sb of the substrate S. The suction / travel control mechanism 52 applies negative pressure to the suction pad to attract the substrate S to the suction pad. The suction / travel control mechanism 52 also causes the substrate transport unit 5 to reciprocate in the X direction.

[0027] The chuck mechanism 51 holds the substrate S such that its lower surface Sb is positioned higher than the upper surface of the levitation stage 3. With its peripheral edge held by the chuck mechanism 51, the substrate S maintains a horizontal position due to the buoyancy provided by the levitation stage 3.

[0028] As shown in Figure 1, the coating apparatus 1 is equipped with a sensor 61 for measuring the plate thickness. The sensor 61 is positioned near the roller conveyor 21. The sensor 61 detects the position in the Z direction of the upper surface Sf of the substrate S held by the chuck mechanism 51. Furthermore, by positioning a chuck (not shown) that is not holding the substrate S directly below the sensor 61, the sensor 61 can detect the position in the vertical Z direction of the suction surface, which is the upper surface of the suction member.

[0029] The chuck mechanism 51 moves in the +X direction while holding the substrate S that has been brought into the levitation stage section 3. As a result, the substrate S is transported from above the inlet levitation stage 31, through above the coating stage 32, to above the outlet levitation stage 33. Then, the substrate S is moved from the outlet levitation stage 33 to the output transfer section 4.

[0030] The output transfer unit 4 moves the substrate S from a position above the exit floating stage 33 to the output conveyor 110. The output transfer unit 4 includes a roller conveyor 41 and a rotation / lifting drive mechanism 42. The rotation / lifting drive mechanism 42 rotates the roller conveyor 41 and also moves the roller conveyor 41 up and down along the Z direction. As each roller of the roller conveyor 41 rotates, the substrate S moves in the +X direction. Also, as the roller conveyor 41 moves up and down, the substrate S is displaced in the Z direction.

[0031] The output conveyor 110 comprises a roller conveyor 111 and a rotary drive mechanism 112. The output conveyor 110 transports the substrate S in the +X direction by the rotation of each roller of the roller conveyor 111, and discharges the substrate S outside the coating apparatus 1. The input conveyor 100 and the output conveyor 110 are part of the coating apparatus 1. However, the input conveyor 100 and the output conveyor 110 may be incorporated into a separate apparatus from the coating apparatus 1.

[0032] The coating mechanism 7 coats the upper surface Sf of the substrate S with processing liquid. The coating mechanism 7 is positioned above the transport path of the substrate S. The coating mechanism 7 has a nozzle 71. The nozzle 71 is a slit nozzle with a slit-shaped discharge port on its lower surface. The nozzle 71 is connected to a positioning mechanism (not shown). The positioning mechanism moves the nozzle 71 between a coating position above the coating stage 32 (indicated by a solid line in Figure 1) and a maintenance position described later. The processing liquid supply mechanism 8 is connected to the nozzle 71. The processing liquid supply mechanism 8 supplies processing liquid to the nozzle 71, causing the processing liquid to be discharged from a discharge port located on the lower surface of the nozzle 71.

[0033] Figure 2 shows the configuration of the processing liquid supply mechanism 8 provided in the coating apparatus 1 shown in Figure 1. The processing liquid supply mechanism 8 comprises a pump 81, piping 82, a processing liquid replenishment unit 83, piping 84, an on-off valve 85, a pressure sensor 86, and a drive unit 87. The pump 81 is a supply source for supplying the processing liquid to the nozzle 71, and supplies the processing liquid by volume change. For example, a bellows-type pump as described in Japanese Patent Application Publication No. 10-61558 can be used for the pump 81. As shown in Figure 2, the pump 81 has a flexible tube 811 that can elastically expand and contract in the radial direction. One end of the flexible tube 811 is connected to the processing liquid replenishment unit 83 via piping 82. The other end of the flexible tube 811 is connected to the nozzle 71 via piping 84.

[0034] Pump 81 has a bellows 812 that is elastically deformable in the axial direction. The bellows 812 has a small bellows section 813, a large bellows section 814, a pump chamber 815, and an operating disk section 816. The pump chamber 815 is located between the flexible tube 811 and the bellows 812. An incompressible medium is sealed inside the pump chamber 815. The operating disk section 816 is connected to the drive unit 87.

[0035] The processing liquid replenishment unit 83 has a storage tank 831 for storing processing liquid. The storage tank 831 is connected to the pump 81 via piping 82, and an on-off valve 833 is inserted into the piping 82. The on-off valve 833 opens and closes in response to commands from the control unit 9. When the on-off valve 833 is open, processing liquid can be supplied from the storage tank 831 to the flexible tube 811 of the pump 81. When the on-off valve 833 is closed, the supply of processing liquid from the storage tank 831 to the flexible tube 811 of the pump 81 is restricted.

[0036] The piping 84 is connected to the output side of the pump 81. The on-off valve 85 is located on the piping 84. The on-off valve 85 opens and closes the piping 84 in response to a command from the control unit 9. The opening and closing of the on-off valve 85 switches between supplying and stopping the processing liquid to the nozzle 71. The pressure sensor 86 is located on the piping 84. The pressure sensor 86 detects the pressure (discharge pressure) applied to the processing liquid supplied to the nozzle 71 and outputs a signal indicating the detected pressure value to the control unit 9.

[0037] As shown in Figures 1 and 2, a sensor 62 is positioned on the nozzle 71 to which the processing liquid is supplied from the processing liquid supply mechanism 8. The sensor 62 non-contactually detects the height of the substrate S in the Z direction. The sensor 62 is electrically connected to the control unit 9. Based on the detection result of the sensor 62, the control unit 9 measures the distance (separation distance) between the floating substrate S and the upper surface of the coating stage 32. Then, based on the measured separation distance, the control unit 9 adjusts the coating position of the nozzle 71 by the positioning mechanism. For example, an optical sensor or an ultrasonic sensor can be used as the sensor 62.

[0038] The substrate S, which is discharged from the output conveyor 110, is dried in a drying device or the like to form a coating film. Then, as shown in Figure 1, the substrate S with the coated film is transported to the film thickness measuring instrument AP1 as needed, and the film thickness of the coating film is measured. For example, a spectroscopic ellipsometer or an X-ray reflectance measuring device can be used as the film thickness measuring instrument AP1.

[0039] The coating mechanism 7 includes a nozzle cleaning standby unit 72. The nozzle cleaning standby unit 72 performs predetermined maintenance on the nozzle 71 positioned at the maintenance location. The nozzle cleaning standby unit 72 includes a roller 721, a cleaning section 722, and a roller butt 723. The nozzle cleaning standby unit 72 cleans the nozzle 71 and forms a liquid reservoir, preparing the nozzle 71's discharge port for coating.

[0040] Figure 3 is a block diagram showing the configuration of the control unit 9 included in the coating apparatus 1 shown in Figure 1. The control unit 9 controls the operation of each component within the coating apparatus 1. The control unit 9 is a computer comprising a processor 91 and memory 93. The processor 91 has, for example, a CPU (Central Processing Unit). The memory 93 has transient storage devices such as RAM (Random Access Memory). The memory 93 may also have non-transient storage devices such as an HDD (Hard Disk Drive) or SSD (Solid State Drive). The memory 93 is connected to the processor 91 via bus wiring.

[0041] The control unit 9 includes a display device 95 for displaying various information and an input device 97 for receiving user commands. The display device 95 and the input device 97 are connected to the processor 91 via a wiring bus. The display device 95 is, for example, a liquid crystal display. The input device 97 is, for example, a mouse or keyboard. The display device 95 may also function as an input device if it has a touch panel.

[0042] The memory 93 stores the computer program 931. The computer program 931 is provided to the control unit 9 via a non-transient recording medium M. That is, the computer program 931 is recorded on the recording medium M in a readable format by the control unit 9, which is a computer. The recording medium M can be a USB (Universal Serial Bus) memory, an optical disc such as a DVD (Digital Versatile Disc), or a magnetic disc.

[0043] The processor 91 functions as a discharge control unit 910, a film thickness acquisition unit 911, a learning unit 913, a film thickness estimation unit 915, and an optimization unit 917 by executing the computer program 931.

[0044] The discharge control unit 910 controls the operation (supply operation) of the pump 81 that supplies the processing liquid to the nozzle 71 based on preset control parameters. In the coating apparatus 1, in order to coat the processing liquid discharged from the nozzle 71 onto the upper surface Sf of the substrate S with a uniform film thickness, control parameters closely related to the discharge pressure waveform are optimized in advance so that the discharge pressure waveform takes an ideal shape before production (or mass production) of the substrate S is started.

[0045] The control parameters are set values ​​for pump control and include, for example, various parameters that define the movement of the operating disk unit 816 (e.g., acceleration time, steady speed, time to maintain steady speed, deceleration time, etc.).

[0046] The film thickness acquisition unit 911 acquires film thickness data. The film thickness data is information indicating the film thickness of the coating film formed on the substrate S. The film thickness data is data obtained by measurement using the film thickness measuring instrument AP1. The film thickness measuring instrument AP1 measures the film thickness by scanning the coating film formed on the substrate S in one direction, for example. In this case, the distribution of film thickness in one direction (film thickness profile) is measured as film thickness data. When the film thickness measuring instrument AP1 scans the coating film at a constant speed, the film thickness data is acquired as time-series data.

[0047] The learning unit 913 has the function of training the base model Md1. The base model Md1 is a training model incorporated into the computer program 931. The film thickness estimation unit 915 estimates film thickness data from process parameters using the sub-model Md2. The sub-model Md2 is a model based on the base model Md1 trained by the learning unit 913, and takes process parameters as input and outputs film thickness data. The structures of the base model Md1 and the sub-model Md2 will be described later. The control unit 9, which has the learning unit 913 and the film thickness estimation unit 915, functions as a film thickness estimation device that estimates the film thickness of a coating film formed on a substrate S.

[0048] Process parameters are data relating to the coating process for forming a coated film, and are different from film thickness data. Process parameters are, for example, non-time-series data. Process parameters include, for example, control parameters. Process parameters may also include equipment information specific to coating apparatus 1. Equipment information may include, for example, the model name, serial number, equipment model number or type, equipment configuration information, firmware or software version, or maintenance history. Process parameters may also include information other than the above equipment information, such as the temperature or humidity of the environment in which the coating process is performed, information about the processing liquid (composition, etc.), or the type of substrate.

[0049] The optimization unit 917 optimizes the control parameters. That is, the optimization unit 917 brings the control parameters closer to the optimal value so that a coating film of ideal thickness can be formed on the substrate S. Specifically, the optimization unit 917 searches for the optimal solution that maximizes or minimizes a predefined objective function (evaluation criterion) according to an optimization algorithm. The input (variable) of the objective function is the control parameters, and the output of the objective function is, for example, the variation in film thickness (standard deviation) when aiming for uniformity of film thickness. Nonlinear programming, genetic algorithms, Bayesian optimization, etc., can be used as the optimization algorithm.

[0050] The optimization unit 917 selects control parameters and acquires film thickness data corresponding to those control parameters. The film thickness data may be data obtained by actually measuring the coated film formed on the substrate S with the film thickness measuring instrument AP1, or it may be data estimated by the film thickness estimation unit 915. By using the film thickness data estimated by the film thickness estimation unit 915, the amount of actual measurement of film thickness data can be reduced. Therefore, the optimization of control parameters can be performed efficiently.

[0051] Figure 4 is a flowchart showing the film thickness estimation method according to the embodiment. The film thickness estimation method shown in Figure 4 is a process that is executed by the processor 91 of the control unit 9 operating according to the procedure specified in the computer program 931.

[0052] The film thickness estimation method includes a learning step S1 and an estimation step S2. The learning step S1 is a step of training the base model Md1. The learning step S1 is performed by the learning unit 913. The estimation step S2 is a step of estimating film thickness data from given process parameters. The estimation step S2 is performed by the film thickness estimation unit 915.

[0053] <Learning Process> Figure 5 shows the structure of the base model Md1 used in the learning process S1 shown in Figure 4. The base model Md1 has an autoencoder 200. The autoencoder 200 is a neural network that compresses and reconstructs the input film thickness data.

[0054] The autoencoder 200 includes a first encoder unit 201 and a decoder unit 203. The first encoder unit 201 takes film thickness data as input and compresses it into a low-dimensional first latent variable (latent vector). The decoder unit 203 takes the first latent variable as input and outputs the reconstructed film thickness data.

[0055] The configurations of the first encoder section 201 and the decoder section 203 of the autoencoder 200 are designed to maintain symmetry with respect to each other. For example, the first encoder section 201 has a one-dimensional convolutional layer, and is configured to convert the data into a latent variable data size by adjusting the convolution stride, filter number, etc. The decoder section 203 has a one-dimensional inverse convolutional layer and is configured to return the data to the same data size as the original film thickness data x1.

[0056] The base model Md1 further includes a second encoder unit 210. The second encoder unit 210 takes process parameters as input and outputs a second latent variable. The second latent variable has the same number of dimensions as the first latent variable. The second encoder unit 210 can employ multiple (e.g., two) fully connected layers.

[0057] In the learning process S1, the learning unit 913 trains the base model Md1 using the film thickness data x1 and process parameters x2 as a dataset. Specifically, the learning unit 913 first learns to minimize the error between the film thickness data (x1) input to the autoencoder 200 and the reconstructed film thickness data (x1') output based on that input. This learning is referred to below as the "first learning process ST1". The first learning process ST1 optimizes parameters such as weights in the first encoder unit 201 and the decoder unit 203. As the first learning process ST1 progresses, the first encoder unit 201 is able to output a first latent variable z1 that enables the decoder unit 203 to accurately reconstruct film thickness data x1' that is close to the original film thickness data x1.

[0058] Furthermore, the learning unit 913 learns to minimize the error between the first latent variable z1 obtained by inputting film thickness data x1 to the first encoder unit 201 and the second latent variable z2 obtained by inputting process parameters x2 to the second encoder unit 210. This learning is hereafter referred to as the "second learning process ST2". The second learning process ST2 optimizes the parameters of the second encoder unit 210. In other words, as the second learning process ST2 progresses, the second encoder unit 210 becomes able to output a second latent variable z2 that is close to the first latent variable z1 for reconstructing the original film thickness data x1.

[0059] In addition, during the learning process S1, the second learning process ST2 may be executed after the first learning process ST1 has been completely completed. Alternatively, the first learning process ST1 and the second learning process ST2 may be executed alternately.

[0060] Furthermore, the ReLU (Rectified Linear Unit) function can be used as the activation function throughout the entire model during the learning process S1. The Mean Absolute Error (MAE) can be used as the loss function. In the first learning process ST1, the loss function is calculated based on the difference between the output film thickness data x1' and the correct film thickness data x1. In the second learning process ST2, the loss function is calculated based on the difference between the first latent variable z1 at the time of input of process parameter x2 and the first latent variable z1 at the time of input of film thickness data x1. Additionally, batch normalization is applied to the convolutional and deconvolutional layers of the autoencoder 200, as well as the fully connected layer of the second encoder unit 210.

[0061] <Estimation process> Next, the estimation process S2 will be described. Figure 6 shows the structure of the submodel Md2 used in the estimation process S2 shown in Figure 4. The submodel Md2 has a second encoder unit 210 and a decoder unit 203 obtained in the learning process S1. The decoder unit 203 is connected to the output side of the second encoder unit 210.

[0062] In estimation step S2, the film thickness estimation unit 915 calculates film thickness data by inputting process parameters for which the film thickness should be estimated into the submodel Md2. Specifically, the second encoder unit 210 of the submodel Md2 outputs a second latent variable z2 according to the input process parameters. Then, the decoder unit 203 outputs film thickness data x1′ according to the input second latent variable z2.

[0063] <Summary> As described above, the film thickness estimation device implemented by the control unit 9 comprises a learning unit 913 and a film thickness estimation unit 915. The learning unit 913 trains a base model Md1. The film thickness estimation unit 915 uses a submodel Md2 based on the trained base model Md1 obtained by the learning unit 913 to estimate film thickness data from process parameters related to the coating process. The base model Md1 includes an autoencoder 200 and a second encoder unit 210. The autoencoder 200 includes a first encoder unit 201 and a decoder unit 203. The first encoder unit 201 takes film thickness data x1 as input and outputs a first latent variable z1 compressed to a lower dimension. The decoder unit 203 takes the first latent variable z1 output from the first encoder unit 201 as input and outputs reconstructed film thickness data x1′. The second encoder unit 210 takes process parameters x2 as input and outputs a second latent variable z2 whose dimensionality matches that of the first latent variable z1. Submodel Md2 includes a second encoder unit 210 and a decoder unit 203 connected to the output side of the second encoder unit 210. The learning unit 913 executes a first learning process ST1 and a second learning process ST2. In the first learning process ST1, the autoencoder 200 is trained to minimize the error between the film thickness data x1 input to the autoencoder 200 and the film thickness data x1' output based on that input. In the second learning process ST2, the second encoder unit 210 is trained to minimize the error between a first latent variable z1 obtained by inputting the film thickness data x1 to the first encoder unit 201 and a second latent variable z2 obtained by inputting process parameters x2 corresponding to the film thickness data x1 input to the first encoder unit 201 to the second encoder unit 210.

[0064] With the above configuration, the second encoder unit 210 is trained using the first latent variable z1 in the autoencoder 200 as a hint. This allows the second encoder unit 210 to be trained while utilizing the training accuracy of the autoencoder 200. Therefore, by using the submodel Md2, which combines the second encoder unit 210 and the decoder unit 203, film thickness data can be estimated with high accuracy from process parameters.

[0065] Furthermore, process parameter x2 includes control parameters for controlling the coating apparatus 1 that applies the processing solution to the substrate S. With this configuration, the accuracy of film thickness estimation can be improved because the control information of the coating apparatus 1 can be taken into consideration when estimating film thickness data.

[0066] Furthermore, the process parameter x2 also includes device information specific to the coating apparatus 1. With this configuration, the accuracy of the film thickness estimation can be improved because the characteristics of the coating apparatus 1 can be taken into consideration when estimating the film thickness.

[0067] Although this invention has been described in detail, the above description is illustrative in all respects, and the invention is not limited thereto. It is understood that countless variations not illustrated can be conceived without falling outside the scope of this invention. The components described in each of the above embodiments and variations can be combined or omitted as appropriate, as long as they do not contradict each other. [Explanation of Symbols]

[0068] 1: Coating device 9: Control Unit (Film Thickness Estimation Device) 200: Autoencoder 201: First Encoder Unit 203: Decoder section 210: Second encoder section 913: Learning Department 915: Film thickness estimation section 931: Computer program Md1: Base Model Md2: Submodel S: Circuit board

Claims

1. A film thickness estimation device for estimating the film thickness of a coating applied to a substrate, A learning unit that trains the base model, A film thickness estimation unit estimates film thickness data from process parameters related to the coating process using a submodel based on the learned base model obtained by the learning unit, Equipped with, The aforementioned base model is An autoencoder having a first encoder unit that takes the film thickness data as input and outputs a first latent variable compressed to a low dimension, and a decoder unit that takes the first latent variable output from the first encoder unit as input and outputs the reconstructed film thickness data, A second encoder unit that takes the aforementioned process parameters as input and outputs a second latent variable whose number of dimensions matches that of the first latent variable, It has, The aforementioned submodel is The second encoder unit and, The decoder unit connected to the output side of the second encoder unit, It has, The aforementioned learning unit, A first learning process to train the autoencoder to minimize the error between the film thickness data input to the autoencoder and the film thickness data output based on that input, A second learning process to train the second encoder unit to minimize the error between a first latent variable obtained by inputting the film thickness data to the first encoder unit and a second latent variable obtained by inputting the process parameters corresponding to the film thickness data input to the first encoder unit to the second encoder unit, A film thickness estimation device that performs the following actions.

2. A film thickness estimation apparatus according to claim 1, The process parameters include control parameters for controlling a coating apparatus that applies a processing solution to the substrate, wherein the process parameters are a film thickness estimation apparatus.

3. A film thickness estimation apparatus according to claim 2, The process parameters further include device information specific to the coating apparatus, in a film thickness estimation apparatus.

4. A method for estimating the thickness of a coating film applied to a substrate, a) A training process to train the base model, b) An estimation step in which film thickness data is estimated from process parameters related to the coating process using a submodel based on the trained base model obtained in the learning step, Includes, The aforementioned base model is An autoencoder having a first encoder unit that takes the film thickness data as input and outputs a first latent variable compressed to a low dimension, and a decoder unit that uses the first latent variable output from the first encoder unit and outputs the reconstructed film thickness data, A second encoder unit that takes the aforementioned process parameters as input and outputs a second latent variable whose number of dimensions matches that of the first latent variable, It has, The aforementioned submodel is The second encoder unit and, The decoder unit connected to the output side of the second encoder unit, It has, The aforementioned learning process is, a1) A first learning step of training the autoencoder to minimize the error between the film thickness data input to the autoencoder and the film thickness data output based on that input, a2) A method for estimating film thickness, comprising the step of causing the second encoder unit to learn so as to minimize the error between a first latent variable obtained by inputting the film thickness data into the first encoder unit and a second latent variable obtained by inputting the process parameters corresponding to the film thickness data input into the first encoder unit into the second encoder unit.

5. It is a computer program, A computer program for causing a computer to perform the film thickness estimation method described in claim 4.

Images