Learning apparatus, information processing apparatus, substrate processing apparatus, substrate processing system, learning method, and processing condition determination method
Through the method of compressing and learning variable conditions, a machine learning model suitable for semiconductor etching processing is generated, which solves the problem of nozzle action diversity in complex etching processing, realizes efficient processing condition determination and model optimization, and reduces cost and time consumption .
Patent Information
- Application Number
- CN202280093913.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-03-23
- Filing Date
- 2022-12-12
- Publication Date
- 2024-11-05
AI Technical Summary
In the semiconductor manufacturing process, due to the complex movement of nozzles during the etching process, the time series data dimensions increase, making it difficult to optimize the machine learning model, and the suitable processing volume is not limited to one, and there are multiple target processing conditions, resulting in cost and time High consumption.
By acquiring experimental data, compressing the variable conditions to reduce dimensions, generating learning models, using an automatic encoder to learn the variable conditions, generate compressors, and improving model performance through additional learning and distillation models, inferring the processing volume and film thickness in etching processing Poor, determine the appropriate treatment conditions.
A machine learning device and substrate processing system suitable for conditions that vary with time are realized, which can effectively process complex processes, prompt multiple processing conditions, reduce costs and time consumption, and improve model performance and control accuracy.
Smart Images

Figure CN118901121A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a learning device, an information processing device, a substrate processing device, a substrate processing system, a learning method and a method for determining processing conditions, and also relates to a learning device that generates a learning model that simulates processing performed by a substrate processing device according to processing conditions, an information processing device that uses the learning model to determine processing conditions, a substrate processing device having the information processing device, a learning method performed by the learning device, and a method for determining processing conditions performed by the information processing device. Background Art
[0002] In the semiconductor manufacturing process, there is a cleaning process. In the cleaning process, the film thickness of the film formed on the substrate is adjusted by etching the substrate with a chemical solution. In the adjustment of the film thickness, it is important to perform etching to make the surface of the substrate uniform, or to make the surface of the substrate flat by etching. When the etching liquid is sprayed from the nozzle to a part of the substrate, the nozzle needs to be moved radially relative to the substrate. However, the etching process is a complex process in which the amount of processing of the film varies depending on the action of moving the nozzle. In addition, the amount of processing of the film by etching is determined after the substrate is processed. Therefore, the operation of setting the action of moving the nozzle needs to be repeatedly tested by a technician. A lot of cost and time are required before the optimal action of the nozzle is determined.
[0003] Japanese Patent Publication No. 2021-108367 describes a device that uses a learning model that has been machine-learned using learning data with "input" as processing volume (etching volume) and "output" as scanning speed information to determine scanning speed information based on a target processing volume. According to this technology, a scanning speed information is determined based on a target processing volume.
[0004] Prior art literature
[0005] Patent Literature
[0006] Patent document 1: Japanese Patent Application Publication No. 2021-108367. Summary of the invention
[0007] Problem that the invention aims to solve
[0008] On the other hand, it is desirable to make the action of moving the nozzle more complex. The action of moving the nozzle is time series data representing a position that changes over time. When the action of moving the nozzle is made more complex, the sampling interval increases, and thus the dimension of the time series data increases. Generally speaking, when the dimension of the learning data increases, the number of data required for machine learning increases exponentially. Therefore, since the dimension of the learning data increases, it is difficult to optimize the learning model obtained by machine learning. In addition, since the etching process is a complex process, the action of the nozzle suitable for the target processing amount is not limited to one, and sometimes there are multiple ones.
[0009] One of the objects of the present invention is to provide a learning device and a substrate processing system suitable for machine learning of conditions that change with the passage of time for processing a substrate.
[0010] In addition, another object of the present invention is to provide an information processing device, a substrate processing device, a substrate processing system, and a processing condition determination method that can indicate a plurality of processing conditions for a processing result of a complex process for processing a substrate.
[0011] Means used to solve problems
[0012] The present invention comprises: an experimental data acquisition unit, which acquires a processing amount representing the difference in film thickness before and after processing of a film formed on the substrate after driving a substrate processing device that processes a substrate by supplying a processing liquid to the substrate with a coating formed thereon under processing conditions including a variable condition that changes with the passage of time; a compressor, which transforms the variable condition in a manner that reduces the number of dimensions; and a model generation unit, which generates a learning model, the learning model machine learning including a transformation result after the variable condition is transformed by the compressor and learning data for the processing amount corresponding to the processing condition, to infer the processing amount representing the difference in film thickness before and after processing of a film formed on the substrate before being processed by the substrate processing device.
[0013] According to this aspect, the learning data includes the transformation results and the processing amount after the transformation in a dimensionality-reduced manner of the changing conditions that change with the passage of time. Therefore, the dimensionality of the learning data can be reduced. As a result, a learning device suitable for machine learning of conditions that change with the passage of time for processing substrates can be provided.
[0014] Preferably, the present invention further comprises: a variation condition generating unit that generates a plurality of variation conditions; and a compressor generating unit that causes the auto encoder to learn the plurality of variation conditions generated by the variation condition generating unit to generate the compressor.
[0015] According to this aspect, the compressor is generated by causing the auto encoder to learn the variation condition. Therefore, the compressor can be easily generated.
[0016] Preferably, the processing conditions further include fixed conditions that do not change with the passage of time, and the learning data includes the fixed conditions.
[0017] According to this aspect, it is possible to cope with processes with different fixed conditions without generating a plurality of learning models with different fixed conditions.
[0018] Preferably, the present invention further comprises: a performance acquisition unit, which, when a transformation result after the variable conditions assumed by the compressor transformation are assigned to the learning model and the processing amount estimated by the learning model satisfies the allowable conditions, drives the substrate processing device with processing conditions including the assumed variable conditions to obtain a processing amount representing the difference in film thickness before and after processing of a film formed on a substrate processed by the substrate processing device; and an additional learning unit, which uses the transformation result after the variable conditions assumed by the compressor transformation and additional learning data for the processing amount acquired by the performance acquisition unit to enable the learning model to learn.
[0019] According to this aspect, since the learning model is subjected to additional learning, the performance of the learning model can be improved.
[0020] Preferably, the present invention also has: a distillation unit, which generates a new learning model using distillation data including the transformation result and the processing amount inferred by the learning model when the transformation result after the change condition assumed by the compressor transformation is assigned to the learning model and the processing amount inferred by the learning model satisfies the allowable conditions.
[0021] According to this aspect, it is easy to prepare data for learning a new learning model.
[0022] Preferably, the substrate processing apparatus includes a nozzle for supplying the processing liquid to the substrate, and a moving unit for changing a relative position between the nozzle and the substrate, and the changing condition is the relative position between the nozzle and the substrate changed by the moving unit.
[0023] According to this aspect, a learning model is generated that estimates the processing amount of the coating film processed by supplying the processing liquid to the substrate from the nozzle while changing the relative position of the nozzle and the substrate. Therefore, a learning model that estimates the processing amount in the etching process can be generated.
[0024] According to another aspect of the invention, an information processing device is an information processing device that manages a substrate processing device, and the substrate processing device processes a substrate by supplying a processing liquid to a substrate having a coating formed thereon under a processing condition including a variable condition that changes with the passage of time, and the information processing device comprises: a second compressor that transforms the variable condition in a manner that reduces the number of dimensions; and a processing condition determination unit that determines the processing conditions for driving the substrate processing device using a learning model that infers a processing amount representing a difference in film thickness before and after processing of a coating formed on a substrate before being processed by the substrate processing device, the learning model being a machine learning algorithm. An inference model obtained by learning data, the learning data including: a transformation result after transforming a variable condition included in a processing condition of a substrate processing device processing a substrate using a first compressor which is the same as a second compressor; and a processing amount representing a film thickness difference before and after processing of a coating formed on the substrate processed by the substrate processing device; a processing condition determination unit assigns the transformation result after transforming the assumed variable condition by the second compressor to the learning model, and when the processing amount inferred by the learning model satisfies an allowable condition, the processing condition including the assumed variable condition is determined as the processing condition for driving the substrate processing device.
[0025] According to this aspect, when the transformation result after the transformation of the assumed change condition that changes with the passage of time is given to the learning model, and the processing amount estimated by the learning model satisfies the allowable condition, the processing condition including the assumed change condition is determined as the processing condition for driving the substrate processing device. Therefore, it is possible to determine a plurality of assumed change conditions for the processing amount that satisfies the allowable condition. As a result, it is possible to provide an information processing device that can present a plurality of processing conditions for the processing result of a complex process for processing a substrate.
[0026] Preferably, the substrate processing apparatus includes the above-mentioned information processing apparatus.
[0027] According to this aspect, it is possible to provide a substrate processing apparatus capable of presenting a plurality of processing conditions for processing results of a complex process for processing a substrate.
[0028] According to another aspect of the present invention, a substrate processing system manages a substrate processing device for processing a substrate, the substrate processing system having a learning device and an information processing device, the substrate processing device processes the substrate by supplying a processing liquid to a substrate having a coating formed thereon under a processing condition including a variable condition that changes with the passage of time; the learning device having: an experimental data acquisition unit, which, after driving the substrate processing device under the processing condition, acquires a processing amount representing a film thickness difference before and after the processing of the coating formed on the substrate; a first compressor, which transforms the variable condition in a manner of reducing the number of dimensions; and a model generation unit, which generates a learning model, the learning model machine learning including the information processing unit after transforming the variable condition by the first compressor The information processing device comprises: a second compressor, which is the same as the first compressor; and a processing condition determination unit, which determines the processing conditions for driving the substrate processing device using a learning model generated by the learning device. The processing condition determination unit assigns the transformation result after the assumed variable condition is transformed by the second compressor to the learning model, and when the processing amount estimated by the learning model satisfies the allowable condition, the processing condition including the assumed variable condition is determined as the processing condition for driving the substrate processing device.
[0029] According to this aspect, it is possible to provide a substrate processing system that is suitable for machine learning of conditions that change over time for processing a substrate and that can present a plurality of processing conditions for processing results of a complex process for processing a substrate.
[0030] According to another aspect of the invention, a learning method causes a computer to execute: a process of obtaining a processing amount representing a difference in film thickness before and after processing of a film formed on the substrate after driving a substrate processing apparatus that processes a substrate by supplying a processing liquid to the substrate with a coating formed thereon under a processing condition including a variable condition that changes with the passage of time; a compression process of transforming the variable condition in a manner that reduces the number of dimensions; and a process of generating a learning model that learns a transformation result after transforming the variable condition in the compression process and learning data for the processing amount corresponding to the processing condition to infer a processing amount representing a difference in film thickness before and after processing of a film formed on the substrate before being processed by the substrate processing apparatus.
[0031] According to this aspect, it is possible to provide a learning method suitable for machine learning of conditions that change with the passage of time for processing a substrate.
[0032] According to another aspect of the invention, a method for determining processing conditions is executed by a computer that manages a substrate processing device, and the substrate processing device processes a substrate by supplying a processing liquid to a substrate with a coating formed thereon under a processing condition including a variable condition that changes with the passage of time. The method for determining processing conditions includes the following processing: transforming the variable condition in a manner that reduces the number of dimensions; and determining processing conditions for driving the substrate processing device using a learning model that infers a processing amount that represents a difference in film thickness before and after processing of a coating formed on a substrate before being processed by the substrate processing device, wherein the learning model is an inference model obtained by machine learning learning data, wherein the learning data includes: a transformation result after transforming the variable condition included in the processing condition for processing the substrate by the substrate processing device in the same processing as the transformation; and a processing amount that represents a difference in film thickness before and after processing of a coating formed on a substrate after being processed by the substrate processing device, and the processing for determining the processing conditions includes the following processing: assigning a transformation result after transforming the assumed variable condition through the transformation processing to the learning model, and when the processing amount inferred by the learning model satisfies an allowable condition, determining the processing condition including the assumed variable condition as the processing condition for driving the substrate processing device.
[0033] According to this aspect, it is possible to provide a processing condition determination method capable of presenting a plurality of processing conditions for a processing result of a complex process for processing a substrate.
[0034] Effects of the Invention
[0035] A learning device and a substrate processing system suitable for machine learning of conditions that change with the passage of time for processing a substrate can be provided.
[0036] In addition, it is possible to provide an information processing device, a substrate processing device, a substrate processing system, and a processing condition determination method that can present a plurality of processing conditions for a processing result of a complex process for processing a substrate. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 It is a diagram for explaining the structure of a substrate processing system according to one embodiment of the present invention.
[0038] Figure 2 This is a diagram showing an example of the structure of an information processing device.
[0039] Figure 3 This is a diagram showing an example of the configuration of a learning device.
[0040] Figure 4 FIG. 1 is a diagram showing an example of a functional configuration of a substrate processing system according to the present embodiment of the present invention.
[0041] Figure 5 FIG. 1 is a first diagram showing an example of an operation pattern of a nozzle.
[0042] Figure 6 FIG. 2 is a second diagram showing an example of the operation pattern of the nozzle.
[0043] Figure 7 This is a diagram showing an example of film thickness characteristics.
[0044] Figure 8 is a diagram illustrating the predictor.
[0045] Fig. 9 This is a flowchart showing an example of the flow of the automatic encoder generation process.
[0046] Fig.10 This is a flowchart showing an example of the flow of the predictor generation process.
[0047] Fig.11 This is a flowchart showing an example of the flow of the processing condition determination process.
[0048] Fig.12 This is a flowchart showing an example of the flow of the additional learning process. DETAILED DESCRIPTION
[0049] Hereinafter, a substrate processing system according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings. In the following description, a substrate refers to a semiconductor substrate (semiconductor wafer), a FPD (Flat Panel Display) substrate of a liquid crystal display device or an organic EL (Electro Luminescence) display device, an optical disk substrate, a magnetic disk substrate, a magneto-optical disk substrate, a photomask substrate, a ceramic substrate, or a solar cell substrate, etc.
[0050] 1. Overall structure of substrate processing system
[0051] Figure 1 It is a diagram for explaining the structure of a substrate processing system according to one embodiment of the present invention. Figure 1 The substrate processing system 1 includes an information processing apparatus 100, a learning apparatus 200, and a substrate processing apparatus 300. The learning apparatus 200 is, for example, a server, and the information processing apparatus 100 is, for example, a personal computer.
[0052] The learning device 200 and the information processing device 100 are used to manage the substrate processing apparatus 300. In addition, the substrate processing apparatus 300 managed by the learning device 200 and the information processing device 100 is not limited to one, and a plurality of substrate processing apparatuses 300 may be managed.
[0053] In the substrate processing system 1 of the present embodiment, the information processing device 100, the learning device 200 and the substrate processing device 300 are connected to each other through wired or wireless communication lines or communication line networks. The information processing device 100, the learning device 200 and the substrate processing device 300 are respectively connected to the network and can send and receive data to each other. The network uses, for example, a local area network LAN (Local Area Network) or a wide area network (WAN: Wide Area Network). In addition, the network can also be the Internet. In addition, the information processing device 100 and the substrate processing device 300 can also be connected via a dedicated communication line network. The connection method of the network can be a wired connection or a wireless connection.
[0054] In addition, the learning device 200 does not necessarily need to be connected to the substrate processing device 300 and the information processing device 100 via a communication line or a communication line network. In this case, the data generated in the substrate processing device 300 may be transmitted to the learning device 200 via a recording medium. In addition, the data generated in the learning device 200 may also be transmitted to the information processing device 100 via a recording medium.
[0055] The substrate processing apparatus 300 is provided with a display device, a sound output device, and an operation unit (not shown). The substrate processing apparatus 300 operates according to a predetermined processing condition (processing recipe) of the substrate processing apparatus 300 .
[0056] 2. Overview of substrate processing equipment
[0057] The substrate processing device 300 includes a control device 10 and a plurality of substrate processing units WU. The control device 10 controls the plurality of substrate processing units WU. The plurality of substrate processing units WU processes the substrate by supplying a processing liquid to the substrate W on which a coating is formed. The processing liquid includes an etching liquid, and the substrate processing unit WU performs an etching process. The etching liquid is a chemical liquid. For example, the etching liquid is fluoronitric acid (a mixture of hydrofluoric acid (HF) and nitric acid (HNO3)), hydrofluoric acid (HF), buffered hydrofluoric acid (BHF), ammonium fluoride, HFEG (a mixture of hydrofluoric acid and ethylene glycol), or phosphoric acid (H3PO4).
[0058] The substrate processing unit WU includes a rotary chuck SC, a rotary motor SM, a nozzle 311, and a nozzle moving mechanism 301. The rotary chuck SC holds the substrate W horizontally. The rotary motor SM has a first rotary axis AX1. The first rotary axis AX1 extends in the up-down direction. The rotary chuck SC is mounted on the upper end of the first rotary axis AX1 of the rotary motor SM. When the rotary motor SM rotates, the rotary chuck SC rotates around the first rotary axis AX1. The rotary motor SM is a stepping motor. The substrate W held on the rotary chuck SC rotates around the first rotary axis AX1. Therefore, the rotation speed of the substrate W is the same as the rotation speed of the stepping motor. In addition, in the case of providing an encoder that generates a rotation speed signal indicating the rotation speed of the rotary motor, the rotation speed of the substrate W can also be obtained based on the rotation speed signal generated by the encoder. In this case, the rotary motor can use a motor other than a stepping motor.
[0059] The nozzle 311 supplies the etching liquid to the substrate W. The nozzle 311 is supplied with the etching liquid from an etching liquid supply unit (not shown) and discharges the etching liquid to the substrate W being rotated.
[0060] The nozzle moving mechanism 301 moves the nozzle 311 in a substantially horizontal direction. Specifically, the nozzle moving mechanism 301 includes a nozzle motor 303 having a second rotation axis AX2, and a nozzle arm 305. The nozzle motor 303 is arranged in a manner that the second rotation axis AX2 is along a substantially vertical direction. The nozzle arm 305 has a long strip shape extending in a straight line. One end of the nozzle arm 305 is mounted on the upper end of the second rotation axis AX2 so that the longitudinal direction of the nozzle arm 305 is in a direction different from the second rotation axis AX2. At the other end of the nozzle arm 305, the nozzle 311 is mounted in a manner that its nozzle is oriented downward.
[0061] When the nozzle motor 303 is activated, the nozzle arm 305 rotates in a horizontal plane around the second rotation axis AX2. As a result, the nozzle 311 mounted on the other end of the nozzle arm 305 moves (rotates) in a horizontal direction around the second rotation axis AX2. The nozzle 311 sprays the etching liquid toward the substrate W while moving in the horizontal direction. The nozzle motor 303 is, for example, a stepping motor.
[0062] The control device 10 includes a CPU (Central Processing Unit) and a memory, and controls the entire substrate processing apparatus 300 by having the CPU execute a program stored in the memory. The control device 10 controls the rotary motor SM and the nozzle motor 303 .
[0063] The learning device 200 inputs experimental data from the substrate processing apparatus 300 , uses the experimental data to machine-learn a learning model, and outputs the learned learning model to the information processing apparatus 100 .
[0064] The information processing apparatus 100 uses the learned learning model to determine the processing conditions for processing a substrate to be subsequently processed by the substrate processing apparatus 300 , and outputs the determined processing conditions to the substrate processing apparatus 300 .
[0065] Figure 2 is a diagram showing an example of the structure of an information processing device. Figure 2 The information processing device 100 is composed of a CPU 101, a RAM (Random Access Memory) 102, a ROM (Read Only Memory) 103, a storage device 104, an operation unit 105, a display device 106, and an input / output I / F (Interface) 107. The CPU 101, the RAM 102, the ROM 103, the storage device 104, the operation unit 105, the display device 106, and the input / output I / F 107 are connected to a bus 108.
[0066] RAM 102 is used as a work area of CPU 101. System programs are stored in ROM 103. Storage device 104 includes a storage medium such as a hard disk or a semiconductor memory to store programs. Programs may also be stored in ROM 103 or other external storage devices.
[0067] A CD-ROM (Compact Disc-Read Only Memory) 109 can be loaded and unloaded in the storage device 104. The recording medium for storing the program executed by the CPU 101 is not limited to the CD-ROM 109, and may be a medium such as an optical disk (MO (Magnetic Optical Disk) / MD (Mini Disc) / DVD (Digital Versatile Disc)), an IC (Integrated Circuit) card, an optical card, a mask ROM, an EPROM (Erasable Programmable ROM), or a semiconductor memory such as the like. Furthermore, the CPU 101 may download the program from a computer connected to the network and store it in the storage device 104, or the computer connected to the network may write the program to the storage device 104, load the program stored in the storage device 104 to the RAM 102, and execute it by the CPU 101. The program mentioned here includes not only a program directly executable by the CPU 101 but also a source program, a compressed program, an encrypted program, and the like.
[0068] The operation unit 105 is an input device such as a keyboard, a mouse, or a touch panel. The user can give a predetermined instruction to the information processing device 100 by operating the operation unit 105. The display device 106 is a display device such as a liquid crystal display device, which displays a GUI (Graphical User Interface) for accepting user instructions. The input / output I / F 107 is connected to a network.
[0069] Figure 3 is a diagram showing an example of the structure of a learning device. Figure 3 The learning device 200 is composed of a CPU 201 , a RAM 202 , a ROM 203 , a storage device 204 , an operation unit 205 , a display device 206 , and an input / output I / F 207 . The CPU 201 , the RAM 202 , the ROM 203 , the storage device 204 , the operation unit 205 , the display device 206 , and the input / output I / F 207 are connected to a bus 208 .
[0070] RAM 202 is used as a work area of CPU 201. System programs are stored in ROM 203. Storage device 204 includes a storage medium such as a hard disk or a semiconductor memory to store programs. Programs can also be stored in ROM 203 or other external storage devices. CD-ROM 209 can be loaded and unloaded in storage device 204.
[0071] The operation unit 205 is an input device such as a keyboard, a mouse, or a touch panel. The input / output I / F 207 is connected to a network.
[0072] 3. Functional structure of substrate processing system
[0073] Figure 4 FIG. 1 is a diagram showing an example of a functional configuration of a substrate processing system according to the present embodiment of the present invention. Figure 4 The control device 10 of the substrate processing device 300 controls the substrate processing unit WU and processes the substrate W according to the processing conditions. The processing conditions are conditions for processing the substrate W during a predetermined processing time. The processing time is a time determined for processing the substrate. In the present embodiment, the processing time is the time during which the nozzle 311 sprays the etching liquid to the substrate W.
[0074] In this embodiment, the processing conditions include the temperature of the etching solution, the concentration of the etching solution, the flow rate of the etching solution, the rotation speed of the substrate W, and the relative position of the nozzle 311 and the substrate W. The processing conditions include variable conditions that change with the passage of time. In this embodiment, the variable condition is the relative position of the nozzle 311 and the substrate W. The relative position is represented by the rotation angle of the nozzle motor 303. The processing conditions include fixed conditions that do not change with the passage of time. In the embodiment, the fixed conditions are the temperature of the etching solution, the concentration of the etching solution, the flow rate of the etching solution, and the rotation speed of the substrate W.
[0075] The learning device 200 causes the learning model to learn the learning data, and generates an inference model that estimates the etching profile from the processing conditions. Hereinafter, the inference model generated by the learning device 200 is referred to as a predictor.
[0076] The learning device 200 includes a change condition generating unit 251, a compressor generating unit 253, a compressor transmitting unit 255, an experimental data acquiring unit 261, a first compressor 263, a predictor generating unit 265, and a predictor transmitting unit 267. The functions of the learning device 200 are implemented by the CPU 201 of the learning device 200 executing a learning program stored in the RAM 202.
[0077] The variable condition generating unit 251 generates variable conditions. For example, the variable condition generating unit 251 generates a plurality of variable conditions by calculation. In the present embodiment, the substrate W rotates around the first rotation axis AX1, and the nozzle 311 rotates around the second rotation axis AX2. Therefore, the change in the relative position between the nozzle 311 and the substrate W is represented by the change in the position of the nozzle 311. The change in the position of the nozzle 311 is determined by the rotation angle of the nozzle motor 303. In addition, the range of the angle of rotation of the nozzle motor 303 is limited to a prescribed range. Moreover, the processing time is a predetermined period. Therefore, the change in the position of the nozzle 311 can be classified into a plurality of action modes. For example, a plurality of action modes are determined in which at least one of the moving speed of the nozzle 311, the speed change position indicating the position of changing the speed, the moving direction of the nozzle 311, and the number of times the speed of the nozzle 311 is reversed is fixed. The variable condition generating unit 251 determines a function according to each of the plurality of action modes, and generates a variable condition using the function.
[0078] Figure 5 FIG. 1 is a diagram showing an example of a nozzle operation mode. Figure 5 , indicating three changing conditions when the number of times the speed of the nozzle 311 is reversed is different and other variables are the same. The upper section indicates a changing condition of three reversals from the positive side to the negative side, the middle section indicates a changing condition of four reversals from the positive side to the negative side, and the lower section indicates a changing condition of five reversals from the positive side to the negative side.
[0079] Figure 6 FIG. 2 is a second diagram showing an example of the nozzle operation mode. Figure 6 , represents three changing conditions when the accelerations of the nozzle 311 are different and other variables are the same. The upper section represents the changing condition of constant speed, the middle section represents the changing condition of gradually decreasing acceleration, and the lower section represents the changing condition of gradually increasing acceleration.
[0080] Back to Figure 4 The compressor generation unit 253 generates a compressor by machine learning the plurality of variable conditions generated by the variable condition generation unit 251. The compressor transmission unit 255 transmits the compressor generated by the compressor generation unit 253 to the information processing device 100.
[0081] The compressor generation unit 253 performs machine learning using a neural network. In the present embodiment, the compressor generation unit 253 is composed of an autoencoder. The compressor generation unit 253 causes the autoencoder to machine learn the changing conditions generated in the changing condition generation unit 251. Specifically, the compressor generation unit 253 inputs the changing conditions generated by the changing condition generation unit 251 into the autoencoder as generation input data, and determines the parameters in such a way that the output of the autoencoder is equal to the generation input data. The autoencoder is composed of an input layer, an intermediate layer, and an output layer. The part from the input layer to the intermediate layer is equivalent to a compressor, and the part from the intermediate layer to the output layer is equivalent to a decoder. The compressor generation unit 253 generates a neural network as a compressor that introduces the parameters of the part from the input layer to the intermediate layer set in the learned autoencoder. The compressor is an inference program that introduces the parameters of the part from the input layer to the intermediate layer set in the learned autoencoder.
[0082] The compressor generation unit 253 may also standardize the variable conditions and cause the autoencoder to machine learn the standardized variable conditions. The variable conditions represent the relative position of the nozzle 311 relative to the substrate W, for example, an example of setting the range to -150 mm to 150 mm. By standardizing the variable conditions, the ratio is scaled to a range of 0 to 1. By standardizing the variable conditions, the range of the parameters can be unified, which facilitates the machine learning of the autoencoder.
[0083] The compressor generation unit 253 classifies a plurality of change conditions prepared in advance for machine learning into a first change condition for machine learning and a second change condition for compressor evaluation, causes the autoencoder machine to learn the first change condition, and evaluates the performance of the autoencoder using the second change condition.
[0084] The experimental data acquisition unit 261 acquires experimental data from the substrate processing apparatus 300. The experimental data includes processing conditions used when the substrate processing apparatus 300 actually processes the substrate W, and film thickness characteristics of the film formed on the substrate W before and after the processing. The film thickness characteristics are represented by the film thickness of the film formed on the substrate W at each of a plurality of different positions in the radial direction of the substrate W.
[0085] Figure 7 This is a graph showing an example of film thickness characteristics. Figure 7 , the horizontal axis represents the position in the radial direction of the substrate, and the vertical axis represents the film thickness. The origin of the horizontal axis represents the center of the substrate. The film thickness of the film formed on the substrate W before being processed by the substrate processing apparatus 300 is represented by a solid line. The film thickness of the film formed on the substrate W is adjusted by performing the process of applying the etching solution according to the processing conditions using the substrate processing apparatus 300. The film thickness of the film formed on the substrate W after being processed by the substrate processing apparatus 300 is represented by a dotted line.
[0086] The difference between the film thickness of the film formed on the substrate W before being processed by the substrate processing apparatus 300 and the film thickness of the film formed on the substrate W after being processed by the substrate processing apparatus 300 is the processing amount (etching amount). The processing amount represents the thickness of the film reduced by the processing of applying the etching solution by the substrate processing apparatus 300. The radial distribution of the processing amount is called an etching profile. The etching profile includes the processing amount at each of a plurality of positions in the radial direction of the substrate W.
[0087] In addition, it is desired that the film thickness formed by the substrate processing apparatus 300 is uniform over the entire surface of the substrate W. Therefore, a target film thickness is determined for the processing performed by the substrate processing apparatus 300. The target film thickness is indicated by a single dotted line. The deviation characteristic is the difference between the film thickness of the coating formed on the substrate W after being processed by the substrate processing apparatus 300 and the target film thickness. The deviation characteristic includes the difference at each of a plurality of positions in the radial direction of the substrate W.
[0088] Back to Figure 4 The first compressor 263 uses the compressor generated by the compressor generator 253 to transform the change conditions included in the processing conditions of the experimental data input from the experimental data acquisition unit 261 into a low-dimensional data set (dataset). Here, the compressor generated by the compressor generator 253 is a neural network. The first compressor 263 inputs the change conditions to the compressor, that is, the neural network, and outputs the transformation result of the compressor to the predictor generator 265.
[0089] The predictor generation unit 265 receives the conversion result after the change condition is converted from the first compression unit 263, and receives the experimental data from the experimental data acquisition unit 261. The predictor generation unit 265 generates a predictor by causing the neural network to perform supervised learning.
[0090] Specifically, the learning data includes input data and correct answer data. The input data includes the transformation result after the variable conditions are transformed by the first compression unit 263, and the fixed conditions other than the variable conditions of the processing conditions included in the experimental data. The correct answer data includes an etching profile. The etching profile is the difference between the film thickness characteristics of the coating before processing included in the experimental data and the film thickness characteristics of the coating after processing included in the experimental data. The predictor generation unit 265 inputs the input data to the neural network to determine the parameters of the neural network in such a way that the output of the neural network is equal to the correct answer data. The predictor generation unit 265 generates a neural network that introduces the parameters set in the learned neural network as a predictor. The predictor is an inference program that introduces the parameters set in the learned neural network. The predictor generation unit 265 sends the predictor to the information processing device 100.
[0091] Figure 8 is a diagram illustrating the predictor. Figure 8 , the predictor includes an input layer, an intermediate layer and an output layer, and each layer includes a plurality of nodes represented by 0. In addition, although one intermediate layer is shown in the figure, the number of intermediate layers can be more than one. In addition, although 5 nodes are shown in the input layer, 4 nodes are shown in the intermediate layer, and 3 nodes are shown in the output layer, the number of nodes is not limited to this. The output of the upper node is connected to the input of the lower node. The parameter includes a coefficient for weighting the output of the upper node. In addition, the number of intermediate layers is more than one, and its number is not limited.
[0092] When the predictor inputs the transformation result after the variable condition is transformed into a low-dimensional data set and the fixed condition, an etching profile is output. The etching profile is represented by the difference E[n] of the film thickness before and after the treatment at a plurality of positions P[n] (n is an integer greater than 1) in the radial direction of the substrate W. In addition, although the number of output nodes of the three predictors is shown in the figure, in fact, the number of output nodes is n.
[0093] Back to Figure 4 The information processing device 100 includes a processing condition determination unit 151, a compressor receiving unit 153, a predictor receiving unit 155, a second compressor 157, a predictor 159, an evaluation unit 161, and a processing condition sending unit 163. The functions of the information processing device 100 are implemented by the CPU 101 of the information processing device 100 executing a processing condition determination program stored in the RAM 102.
[0094] The compressor receiving unit 153 receives the compressor transmitted from the learning device 200, and outputs the received compressor to the second compressor 157. The predictor receiving unit 155 receives the predictor transmitted from the learning device 200, and outputs the received predictor to the predictor 159.
[0095] The processing condition determination unit 151 determines the processing conditions for the substrate W to be processed using the substrate processing apparatus 300. The processing condition determination unit 151 outputs the variable conditions included in the processing conditions to the second compression unit 157, and outputs the fixed conditions included in the processing conditions to the prediction unit 159. The processing condition determination unit 151 selects one of the plurality of variable conditions prepared in advance using the experimental planning method, the pairwise method, or the Bayesian estimation, and determines the processing conditions including the selected variable conditions and the fixed conditions as the processing conditions for the prediction unit 159 to perform the estimation. The plurality of variable conditions prepared in advance are preferably the plurality of variable conditions generated for the learning device 200 to generate the compressor.
[0096] The second compression unit 157 uses a compressor to compress the change condition input from the processing condition determination unit 151. Specifically, the second compression unit 157 inputs the change condition to the compressor, and outputs the output of the compressor to the prediction unit 159 as a conversion result.
[0097] The prediction unit 159 uses the predictor to estimate the etching profile based on the transformation result of the variable condition of the compressor and the fixed condition. Specifically, the prediction unit 159 inputs the transformation result input from the second compression unit 157 and the fixed condition input from the processing condition determination unit 151 to the predictor, and outputs the etching profile output by the predictor to the evaluation unit 161.
[0098] The evaluation unit 161 evaluates the etching profile input from the prediction unit 159, and outputs the evaluation result to the processing condition determination unit 151. In detail, the evaluation unit 161 obtains the film thickness characteristics of the predetermined substrate W as the processing object before the substrate processing device 300 processes it. The evaluation unit 161 calculates the film thickness characteristics predicted after the etching process based on the etching profile input from the prediction unit 159 and the film thickness characteristics of the substrate W before processing, and compares it with the target film thickness characteristics. If the result of the comparison meets the evaluation criterion, the processing conditions determined by the processing condition determination unit 151 are output to the processing condition sending unit 163. For example, the evaluation unit 161 calculates the deviation characteristics and determines whether the deviation characteristics meet the evaluation criterion. The deviation characteristics are the difference between the film thickness characteristics of the substrate W after the etching process and the target film thickness characteristics. The evaluation criterion can be determined arbitrarily. For example, the evaluation criterion can be set to the maximum value of the difference in the deviation characteristics being below the threshold, or the average of the difference being below the threshold.
[0099] The processing condition transmitting unit 163 transmits the processing condition determined by the processing condition determining unit 151 to the substrate processing apparatus 300. The substrate processing apparatus 300 processes the substrate W according to the processing condition.
[0100] If the evaluation result does not satisfy the evaluation criterion, the evaluation unit 161 outputs the evaluation result to the processing condition determination unit 151. The evaluation result includes the film thickness characteristics predicted after the etching process or the difference between the film thickness characteristics predicted after the etching process and the target film thickness characteristics.
[0101] The processing condition determination unit 151 determines a new processing condition for the prediction unit 159 to make an inference based on the evaluation result input from the evaluation unit 161. The processing condition determination unit 151 selects one of a plurality of variable conditions prepared in advance using an experimental planning method, a pairwise method, or a Bayesian inference, and determines the processing condition including the selected variable condition and the fixed condition as a new processing condition for the prediction unit 159 to make an inference.
[0102] The processing condition determination unit 151 may also use Bayesian estimation to search for processing conditions. When the evaluation unit 161 outputs a plurality of evaluation results, there are a plurality of sets of processing conditions and evaluation results. Based on the tendency of the etching profiles of the plurality of sets, a processing condition is searched for which the film thickness of the coating becomes uniform, or a processing condition is searched for which the difference between the predicted film thickness characteristic after the etching process and the target film thickness characteristic is minimized.
[0103] Specifically, the processing condition determination unit 151 searches for processing conditions in a manner that minimizes the objective function. The objective function is a function that represents the uniformity of the film thickness of the coating, or a function that represents the consistency between the film thickness characteristics of the coating and the target film thickness characteristics. For example, the objective function is a function that represents the difference between the film thickness characteristics predicted after the etching process and the target film thickness characteristics by parameters. The parameters here are the transformation results after the corresponding variable conditions are transformed by the second compression unit 157. The corresponding variable conditions are the variable conditions before the transformation results used for the predictor to estimate the etching profile are transformed. The processing condition determination unit 151 selects the variable conditions corresponding to the parameters determined by the search, i.e., the transformation results, from a plurality of variable conditions, and determines new processing conditions including the selected variable conditions and fixed conditions.
[0104] Alternatively, the processing condition determination unit 151 may select a variable condition obtained by decoding the parameters obtained by searching with the decoder instead of selecting one variable condition from among the variable conditions prepared in advance. The decoder is a part of the auto encoder formed by the compressor generation unit 253 of the learning device 200 .
[0105] Fig. 92 is a flowchart showing an example of the flow of the automatic encoder generation process. The automatic encoder generation process is a process performed by CPU 201 of learning device 200 by executing an automatic encoder generation program stored in RAM 202. The automatic encoder generation program is a part of the learning program.
[0106] refer to Fig. 9 The CPU 201 of the learning device 200 generates a change condition, and the process proceeds to step S02. A plurality of action modes representing the change in the position of the nozzle 311 are predetermined, and one of the plurality of action modes is selected. Furthermore, a function corresponding to the selected action mode is selected, and a change condition is generated by sequentially changing the parameters.
[0107] In step S02, CPU 201 causes the autoencoder to learn, and the process proceeds to step S03. The change condition generated in step S01 is set as the input and output, and the parameters of the neural network are adjusted. The change condition is sampled at a predetermined sampling interval. In this embodiment, the processing time is set to 20 seconds and the sampling interval is set to 0.01 seconds. Therefore, the change condition is data in which 2000 position data representing the position are arranged in time series.
[0108] In step S03, it is determined whether the adjustment is completed. The variable conditions for the evaluation of the compressor are prepared in advance, and the performance of the autoencoder is evaluated with the variable conditions for evaluation. When the evaluation result satisfies the predetermined evaluation criterion, it is determined that the adjustment is completed. If the evaluation result does not satisfy the evaluation criterion (NO in step S03), the process returns to step S01, but if the evaluation result satisfies the evaluation criterion (YES in step S03), the process proceeds to step S04.
[0109] When the process returns to step S01, a new action pattern is generated in step S01. In the loop of steps S01 to S03, CPU 201 causes the auto encoder to perform machine learning using a plurality of change conditions, thereby adjusting the parameters of the auto encoder to appropriate values.
[0110] In step S04, the autoencoder is stored in the storage device 104, and the process proceeds to step S05. The autoencoder is a neural network to which parameters adjusted by machine learning are set.
[0111] In step S05 , the compressor is sent, and the process ends. The CPU 201 controls the input / output I / F 107 to send the portion from the input layer to the intermediate layer in the auto encoder to the information processing device 100 as a compressor.
[0112] Fig.102 is a flowchart showing an example of the flow of the predictor generation process. The predictor generation process is a process performed by the CPU 201 included in the learning device 200 by executing a predictor generation program stored in the RAM 202. The predictor generation program is a part of the learning program.
[0113] refer to Fig.10 , the CPU 201 of the learning device 200 obtains experimental data. The CPU 201 controls the input and output I / F 107 to obtain experimental data from the substrate processing device 300 (step S11). The experimental data can also be obtained by reading the experimental data recorded in a recording medium such as CD-ROM 209 using the storage device 104. The experimental data obtained here are plural. The experimental data includes processing conditions and film thickness characteristics of the film formed on the substrate W before and after processing. The film thickness characteristics are represented by the film thickness of the film formed on the substrate W at different plural positions in the radial direction of the substrate W.
[0114] In the next step S12, experimental data to be processed is selected, and the process proceeds to step S13. In step S13, the variation conditions included in the experimental data are compressed, and the process proceeds to step S14. The variation conditions are input to the compressor generated by the learning device 200, and the output of the compressor is obtained as the conversion result.
[0115] In step S14, the transformation result, the fixed conditions included in the experimental data, and the etching profile are set in the learning data. The etching profile is the difference between the film thickness characteristics of the coating before the treatment included in the experimental data and the film thickness characteristics of the coating after the treatment included in the experimental data. The learning data includes input data and correct answer data. The transformation result of the compressor calculated in step S13 and the fixed conditions included in the experimental data are set in the input data. The etching profile is set in the correct answer data.
[0116] In the next step S15, the CPU 201 causes the predictor to perform machine learning, and the process proceeds to step S16. Input data is input to the predictor, which is a neural network, and parameters are determined so that the output of the predictor is equal to the correct answer data. Thus, the parameters of the predictor are adjusted. The predictor is a neural network having parameters determined by machine learning using learning data.
[0117] In step S16, it is determined whether the adjustment has been completed. Learning materials for the evaluation of the predictor are prepared in advance, and the performance of the predictor is evaluated with the learning data for evaluation. When the evaluation result meets the predetermined evaluation criterion, it is determined that the adjustment is completed. If the evaluation result does not meet the evaluation criterion (NO in step S16), the process returns to step S12, but if the evaluation result meets the evaluation criterion (YES in step S16), the process proceeds to step S17.
[0118] When the process returns to step S12, in step S12, experimental data that has not been selected as the processing object is selected from the experimental data obtained in step S11. In the loop of step S12 to step S16, CPU201 uses a plurality of learning data to make the predictor perform machine learning. Thus, the parameters of the neural network, i.e., the predictor, are adjusted to appropriate values. In step S18, the predictor is sent and the process ends. CPU201 controls the input / output I / F107 to send the predictor to the information processing device 100.
[0119] Fig.11 2 is a flowchart showing an example of the flow of the processing condition determination process. The processing condition determination process is a process executed by the CPU 101 of the information processing device 100 by executing a processing condition determination program stored in the RAM 102 .
[0120] refer to Fig.11 The CPU 101 of the information processing device 100 selects one of the plurality of change conditions prepared in advance (step S21), and proceeds to step S22. The plurality of change conditions are the plurality of change conditions generated to enable the learning device 200 to generate a compressor. One of the plurality of change conditions prepared in advance is selected using an experimental planning method, a pairwise method, or a Bayesian estimation method.
[0121] In step S22, the change condition is compressed, and the process proceeds to step S23. The change condition is compressed using the compressor generated by the learning device 200. Here, the change condition is input to the compressor, and its output is obtained as the conversion result.
[0122] In step S23, a predictor is used to estimate the etching profile based on the transformation result and the fixed conditions, and the process proceeds to step S24. The transformation result and the fixed conditions are input to the predictor, and the etching profile output by the predictor is obtained. In step S24, the film thickness characteristics after processing are compared with the target film thickness characteristics. Based on the film thickness characteristics before processing of the substrate W as the processing object by the substrate processing device 300, and the etching profile estimated in step S23, the film thickness characteristics after processing the substrate W are calculated. In addition, the film thickness characteristics after processing are compared with the target film thickness characteristics. Here, the difference between the film thickness characteristics after processing the substrate W and the target film thickness characteristics is calculated.
[0123] In step S25, it is determined whether the comparison result satisfies the evaluation criteria. If the comparison result satisfies the evaluation criteria (YES in step S25), the process proceeds to step S26, and if not, the process returns to step S21. For example, when the maximum value of the difference is below the threshold, it is determined that the evaluation criteria are satisfied. In addition, when the average of the difference is below the threshold, it is determined that the evaluation criteria are satisfied.
[0124] In step S26, in the candidates of the processing conditions for driving the substrate processing apparatus 300, the processing conditions including the variable conditions previously selected in step S21 are set, and the process proceeds to step S27. In step S27, it is determined whether an end instruction of the search has been accepted. If the end instruction is accepted by the user operating the information processing apparatus 100, the process proceeds to step S28, and if not, the process returns to step S21. In addition, it is also possible to determine whether a predetermined number of processing conditions have been set in the candidates instead of the end instruction input by the user.
[0125] In step S28, one is determined from one or more processing conditions set in the candidates, and the process proceeds to step S29. The user operating the information processing device 100 may also select one from one or more processing conditions set in the candidates. Therefore, the user's range of selection is expanded. In addition, the variable condition with the simplest nozzle movement may be automatically selected from the variable conditions included in the plurality of processing conditions. The variable condition with the simplest nozzle movement can be set, for example, as the variable condition with the least number of speed change points. Thus, it is possible to prompt a plurality of variable conditions for the processing results of the complex nozzle movement for processing the substrate W. As long as a variable condition that is easy to control the nozzle is selected from the plurality of variable conditions, the control of the substrate processing device 300 becomes easy.
[0126] In step S29, the processing conditions including the change conditions determined in step S28 are sent to the substrate processing apparatus 300, and the processing ends. The CPU 101 controls the input / output I / F 107 to send the processing conditions to the substrate processing apparatus 300. When the substrate processing apparatus 300 receives the processing conditions from the information processing apparatus 100, it processes the substrate W according to the processing conditions.
[0127] 4. Specific Examples
[0128] In this embodiment, the variable condition is a time series data in which the processing time of the nozzle action is 20 seconds and the sampling interval is 0.01 seconds. The variable condition is composed of 2000 values. Therefore, the variable condition can express complex nozzle actions. In particular, the variable condition can correctly express the nozzle action in which the number of speed change points for changing the moving speed of the nozzle is relatively increased.
[0129] The compressor generation unit 253 in this embodiment causes the autoencoder to perform machine learning to transform the variation condition into a two-dimensional data set. Therefore, the compressor generated by the compressor generation unit 253 transforms the variation condition consisting of 2000 values into a two-dimensional data set. The inventors have found through experiments that even when the variation condition consisting of 2000 values representing a complex nozzle action is transformed into a two-dimensional data set, the desired result can be obtained as the etching profile predicted by the predictor.
[0130] Therefore, since the amount of data input to the predictor can be reduced, the structure of the predictor can be simplified, and the neural network can be easily learned. In addition, the parameters of the neural network can be adjusted to appropriate values, and the accuracy of the predictor can be improved.
[0131] In addition, since the variable conditions with a dimension of 2000 are transformed into a 2-dimensional data set, there are sometimes multiple variable conditions with the same transformation results among the multiple variable conditions. In this case, the etching profiles predicted by the predictor based on each of the multiple variable conditions with the same transformation results are the same. In this embodiment, since the processing condition determination unit 151 searches for processing conditions, the processing conditions corresponding to the etching profile different from the etching profile are searched, and thus the processing conditions corresponding to the multiple different etching profiles can be selected. Therefore, the processing condition determination unit 151 can effectively search for the processing conditions that predict the etching profile that becomes the target from the multiple processing conditions.
[0132] In addition, although the example of setting the sampling interval to 0.01 seconds has been described, the sampling interval is not limited to this. It can also be set to a longer sampling interval or a shorter sampling interval. For example, the sampling interval can be set to 0.1 seconds or 0.005 seconds. In short, it does not matter what the dimension of the variable condition is, as long as the dimension of the transformation result after the variable condition is transformed by the compressor is 2-dimensional.
[0133] 5. Other Implementation Methods
[0134] (1) In the above-mentioned embodiment, the learning device 200 generates a predictor based on learning data. The learning device 200 may also cause the predictor to perform additional learning. After generating the predictor, the learning device 200 obtains the film thickness characteristics and processing conditions of the coating before and after the substrate W processed by the substrate processing device 300. In addition, the learning device 200 generates learning data based on the film thickness characteristics and processing conditions of the coating before and after the processing, and causes the predictor to perform machine learning, thereby causing the predictor to perform additional learning. Although the structure of the neural network constituting the predictor is not changed through additional learning, the parameters are adjusted.
[0135] Since the predictor is subjected to machine learning using information obtained from the result of the substrate processing apparatus 300 actually processing the substrate W, the accuracy of the predictor can be improved. In addition, the amount of learning data for generating the predictor can be reduced as much as possible.
[0136] Fig.12 2 is a flowchart showing an example of the flow of the additional learning process. The additional learning process is a process performed by the CPU 201 of the learning device 200 by executing the additional learning program stored in the RAM 202. The additional learning program is a part of the learning program.
[0137] refer to Fig.12 , the CPU 201 of the learning device 200 obtains the production data (step S31), and the process proceeds to step S32. The production data includes the processing conditions when the substrate W is processed by the substrate processing device 300 after the predictor is generated, and the film thickness characteristics of the respective films before and after the processing. The CPU 201 controls the input and output I / F 107 to obtain the production data from the substrate processing device 300. The production data can also be obtained by reading the experimental data recorded in the recording medium such as CD-ROM 209 with the storage device 104.
[0138] In step S32, the variable conditions are compressed, and the process proceeds to step S33. The variable conditions are input to the compressor generated by the learning device 200, and the output of the compressor is obtained as the transformation result. In step S33, the transformation result, the fixed conditions included in the processing conditions of the production data, and the etching profile are set in the learning data. The etching profile is the difference between the film thickness characteristics of the coating before processing included in the production data and the film thickness characteristics of the coating after processing included in the production data. The transformation result of the compressor and the fixed conditions included in the processing conditions are set in the input data. The etching profile is set in the correct solution data.
[0139] In the next step S34, the CPU 201 causes the predictor to perform additional learning, and the process proceeds to step S35. The input data is input to the predictor, which is a neural network, and the parameters are determined so that the output of the predictor is equal to the correct answer data. In this way, the parameters of the predictor are further adjusted.
[0140] In step S35, it is determined whether the adjustment is completed. The performance of the predictor is evaluated with the learning data for evaluation. When the evaluation result meets the predetermined additional learning evaluation benchmark, the adjustment is determined to be completed. The additional learning evaluation benchmark is a benchmark that is higher than the evaluation benchmark used when generating the predictor. If the evaluation result does not meet the additional learning evaluation benchmark (NO in step S35), the process returns to step S31. If the evaluation result meets the additional learning evaluation benchmark (YES in step S35), the process ends.
[0141] (2) The learning device 200 may generate a distillation model for performing machine learning on a new learning model using distillation data including the processing conditions determined by the information processing device 100 and the etching profile estimated by the predictor based on the processing conditions. This makes it easy to prepare data for performing learning on the new learning model.
[0142] (3) The learning device 200 generates a compressor by causing the autoencoder to perform machine learning, but the present invention is not limited thereto. The compressor may reduce the dimension of the temporal change in the relative position between the nozzle 311 and the substrate W by other methods.
[0143] (4) In the present embodiment, in the learning data for generating the predictor, the input data includes the transformation result after the variable condition is transformed and the fixed condition. The present invention is not limited to this. The input data may include only the transformation result after the variable condition is transformed, but not the fixed condition.
[0144] (5) In this embodiment, the relative position of the nozzle 311 and the substrate W is shown as an example of a variable condition, but the present invention is not limited to this. When at least one of the temperature of the etching solution, the concentration of the etching solution, the flow rate of the etching solution, and the rotation speed of the substrate W changes with the passage of time, these may be used as the variable condition. In addition, the variable condition is not limited to one, and may include a plurality of types.
[0145] (6) Although the information processing device 100 and the learning device 200 are separated from the substrate processing device 300, the present invention is not limited to this. The information processing device 100 may be incorporated into the substrate processing device 300. Furthermore, the information processing device 100 and the learning device 200 may be incorporated into the substrate processing device 300. In addition, although the information processing device 100 and the learning device 200 are set as independent devices, the information processing device 100 and the learning device 200 may be configured as an integrated device.
[0146] 6. Effects of implementation methods
[0147] The learning device 200 of this embodiment drives the substrate processing device 300 with the processing conditions including the variable conditions, obtains the processing amount representing the difference in film thickness before and after the processing of the film formed on the substrate W, and makes the neural network machine learn the transformation result after the variable conditions are transformed by the compressor as input data, and includes the etching profile corresponding to the processing conditions as the correct answer data, and generates a learning model, i.e., a predictor, for estimating the etching profile. As the learning data, since the transformation result after the transformation in a manner that reduces the dimension of the variable conditions that change with the passage of time is included as input data, the dimension of the learning data can be reduced. Therefore, a learning device suitable for machine learning the conditions that change with the passage of time for processing the substrate W can be generated.
[0148] In addition, the learning device 200 generates a plurality of variable conditions within a range that can drive the substrate processing device 300, and causes the autoencoder to learn the generated plurality of variable conditions to generate a compressor. Therefore, since the compressor is generated by causing the autoencoder to learn the variable conditions, generation of the compressor becomes easy.
[0149] In addition, the processing conditions include variable conditions and fixed conditions that do not change with the passage of time. Therefore, it is possible to handle processing with different fixed conditions without generating a plurality of learning models with different fixed conditions.
[0150] In addition, after generating the predictor, the learning device 200 obtains the processing amount representing the difference in film thickness before and after processing of the film formed on the substrate W processed by the substrate processing device 300 according to the processing conditions, and uses the additional learning data including the transformation result after the change condition assumed by the compressor transformation and the obtained processing amount to make the learning model learn. Therefore, since the learning model is additionally learned, the performance of the learning model can be improved.
[0151] In addition, the learning device 200 generates a new learning model using the distillation data including the transformation result and the processing amount estimated by the learning model when the transformation result after the change condition assumed by the compressor transformation is given to the learning model and the processing amount estimated by the learning model satisfies the permissible condition. Therefore, it is easy to prepare data for learning the new learning model.
[0152] In addition, the substrate processing apparatus 300 includes a nozzle 311 for supplying a processing liquid to the substrate W, and a nozzle moving mechanism 301 for changing the relative position of the nozzle and the substrate W, and the change condition is the relative position of the nozzle 311 and the substrate W changed by the nozzle moving mechanism 301. A learning model is generated to estimate the processing amount of the coating film processed by changing the relative position of the nozzle 311 and the substrate W and supplying the processing liquid from the nozzle 311 to the substrate W. Therefore, a learning model for estimating the processing amount in the etching process can be generated.
[0153] In addition, when the information processing device 100 assigns the transformation result after the assumed variable condition is transformed using the compressor generated by the learning device 200 to the learning model generated by the learning device 200, and the etching profile estimated by the learning model satisfies the permissible condition, the processing condition including the assumed variable condition is determined as the processing condition for driving the substrate processing device 300. Therefore, since the etching profile is estimated based on the processing condition, it is not necessary to conduct experiments to find out the influence of the action of the nozzle with complex actions on the processing result of the etching process. In addition, since a plurality of assumed variable conditions are determined for the processing amount that satisfies the permissible condition, a plurality of variable conditions corresponding to the plurality of etching profiles that satisfy the permissible condition can be determined. Therefore, a plurality of variable conditions can be prompted for the processing result of the complex process of processing the substrate. If a processing condition that makes it easy to control the nozzle action is selected from the plurality of variable conditions, the control of the substrate processing device 300 becomes easy.
[0154] In addition, since the information processing device 100 determines a plurality of processing conditions for the processing amount that satisfies the permissible condition, it is possible to determine a plurality of processing conditions corresponding to a plurality of etching profiles that respectively satisfy the permissible condition. Furthermore, the fixed condition includes the temperature of the etching liquid. Therefore, it is possible to indicate a plurality of etching liquid temperatures for the processing result of a complex process for processing a substrate. In addition, it is possible to select the temperature of the etching liquid that is easy to apply to the etching process from among the plurality of etching liquid temperatures. Furthermore, since it is possible to select the temperature of the etching liquid that is easy to apply, it is easy to adjust the temperature of the etching liquid used for the etching process.
[0155] 7. Correspondence between the components of the technical solution and the parts of the implementation method
[0156] The substrate W is an example of a substrate, the etching liquid is an example of a processing liquid, the substrate processing device 300 is an example of a substrate processing device, the experimental data acquisition unit 261 is an example of an experimental data acquisition unit, the first compression unit 263 is an example of a first compressor, the predictor is an example of a learning model, and the predictor generation unit 265 is an example of a model generation unit. In addition, the information processing device 100 is an example of an information processing device, the variable condition generation unit 251 is an example of a variable condition generation unit, the compressor generation unit 253 is an example of a compressor generation unit, the second compressor 157 is an example of a second compressor, the nozzle 311 is an example of a nozzle for supplying the processing liquid to the substrate, the nozzle moving mechanism 301 is an example of a moving unit, and the prediction unit 159, the evaluation unit 161, and the processing condition determination unit 151 are examples of a processing condition determination unit.
Claims
1. A learning device comprising: an experimental data acquisition unit that acquires a processing amount indicating a difference in film thickness before and after processing of the film formed on the substrate after driving a substrate processing apparatus that processes the substrate by supplying a processing liquid to the substrate on which the film is formed under processing conditions including a variable condition that varies with the passage of time; A compressor transforms the changing condition in a dimensionally reduced manner; as well as The model generating unit generates a learning model, wherein the learning model machine learning includes a transformation result after the variable condition is transformed by the compressor and learning data for the processing amount corresponding to the processing condition, and estimates the processing amount representing the difference in film thickness before and after processing of the coating formed on the substrate before being processed by the substrate processing device.
2. The learning device according to claim 1, in, The learning device also has: A change condition generating unit generates a plurality of change conditions; as well as The compressor generation unit generates the compressor by causing an auto encoder to learn the plurality of variation conditions generated by the variation condition generation unit.
3. The learning device according to claim 1 or 2, in, The processing conditions also include fixed conditions that do not change over time; The learning data includes the fixed condition.
4. The learning device according to any one of claims 1 to 3, in, The learning device also has: a performance acquisition unit that, when a transformation result after the compressor transforms the assumed variable condition, is given to the learning model and the processing amount estimated by the learning model satisfies an allowable condition, drives the substrate processing apparatus under processing conditions including the assumed variable condition, and acquires a processing amount representing a difference in film thickness before and after processing of a coating formed on a substrate processed by the substrate processing apparatus; as well as The additional learning unit causes the learning model to learn using additional learning data including a conversion result obtained by the compressor of the assumed change condition and the processing amount acquired by the actual performance acquisition unit.
5. The learning device according to any one of claims 1 to 4, in, The learning device also has: The distillation unit generates a new learning model using distillation data including the transformation result and the processing amount estimated by the learning model, when the transformation result after the change condition assumed by the compressor transformation is assigned to the learning model and the processing amount estimated by the learning model satisfies the allowable conditions.
6. The learning device according to any one of claims 1 to 5, in, The substrate processing device comprises: a nozzle for supplying the processing liquid to the substrate; as well as A moving part changes the relative position between the nozzle and the substrate, The change condition is a relative position between the nozzle and the substrate that is changed by the moving unit.
7. An information processing device for managing a substrate processing device, in, The substrate processing apparatus processes a substrate by supplying a processing liquid to a substrate having a film formed thereon under processing conditions including a variable condition that varies with the passage of time. The information processing device comprises: a second compressor for transforming the variable condition in a dimensionally reduced manner; as well as a processing condition determination unit that determines a processing condition for driving the substrate processing apparatus using a learning model that estimates a processing amount that indicates a difference in film thickness before and after processing of a coating formed on a substrate before being processed by the substrate processing apparatus, The learning model is an inference model obtained by machine learning learning data, the learning data including: a transformation result obtained by transforming a variable condition included in a processing condition for processing a substrate by the substrate processing apparatus using a first compressor that is the same as the second compressor; and a processing amount indicating a difference in film thickness before and after processing of a coating formed on a substrate processed by the substrate processing apparatus; The processing condition determination unit assigns the transformation result after the assumed change condition is transformed by the second compressor to the learning model, and when the processing amount estimated by the learning model satisfies the allowable condition, the processing condition including the assumed change condition is determined as the processing condition for driving the substrate processing device.
8. A substrate processing apparatus comprising the information processing apparatus according to claim 7.
9. A substrate processing system, which manages a substrate processing device for processing a substrate, in, The substrate processing system comprises a learning device and an information processing device. The substrate processing apparatus processes a substrate by supplying a processing liquid to a substrate having a coating film formed thereon under processing conditions including a variable condition that varies with the passage of time; The learning device comprises: an experimental data acquisition unit that acquires a processing amount indicating a difference in film thickness before and after processing of a coating formed on a substrate after driving the substrate processing apparatus under the processing conditions; a first compressor for transforming the variable condition in a dimensionally reduced manner; as well as a model generating unit for generating a learning model for estimating a processing amount representing a difference in film thickness before and after processing of a coating formed on a substrate before being processed by the substrate processing apparatus by learning the processing data including a transformation result after the variable condition is transformed by the first compressor and the processing amount corresponding to the processing condition; The information processing device comprises: a second compressor, which is identical to the first compressor; and a processing condition determination unit that determines a processing condition for driving the substrate processing device using the learning model generated by the learning device, The processing condition determination unit assigns the transformation result after the assumed change condition is transformed by the second compressor to the learning model, and when the processing amount estimated by the learning model satisfies the allowable condition, the processing condition including the assumed change condition is determined as the processing condition for driving the substrate processing device.
10. A learning method that enables a computer to perform: After driving a substrate processing apparatus for processing a substrate by supplying a processing liquid to a substrate having a film formed thereon under processing conditions including a variable condition that varies with the passage of time, a processing amount indicating a difference in film thickness before and after processing of a film formed on the substrate is acquired; Compression processing of transforming the change condition in a dimensionality-reducing manner; as well as A process of generating a learning model, wherein the learning model machine learning includes a transformation result after transforming the variable condition in the compression process and learning data for the processing amount corresponding to the processing condition, and infers the processing amount representing the difference in film thickness before and after processing of the coating formed on the substrate before being processed by the substrate processing device.
11. A method for determining a processing condition, executed by a computer for managing a substrate processing device, in, The substrate processing apparatus processes a substrate by supplying a processing liquid to a substrate having a film formed thereon under processing conditions including a variable condition that varies with the passage of time. The processing condition determination method includes the following processing: transforming the variation condition in a dimensionally reduced manner; as well as determining a processing condition for driving the substrate processing apparatus using a learning model that estimates a processing amount indicating a difference in film thickness before and after processing of a film formed on a substrate before being processed by the substrate processing apparatus, The learning model is an inference model obtained by machine learning learning data, the learning data including: a transformation result after transforming a variable condition included in a processing condition for processing a substrate by the substrate processing device by the same processing as the transformation; and a processing amount representing a film thickness difference before and after the processing of a coating formed on a substrate processed by the substrate processing device, The process of determining the processing conditions includes the following processing: when the transformation result after the assumed change condition is transformed by the transformation process is assigned to the learning model, and the processing amount estimated by the learning model satisfies the allowable condition, the processing condition including the assumed change condition is determined as the processing condition for driving the substrate processing device.
Citation Information
Patent Citations
Substrate processing apparatus, substrate processing method, substrate processing system, and learning data generation method
JP2021108367A
