Prediction algorithm generation device, information processing device, substrate processing device, prediction algorithm generation method, prediction algorithm generation program, processing condition determination method, processing condition determination program, and prediction algorithm
The prediction algorithm generation device addresses the low generalizability of existing algorithms by classifying and selecting processing conditions based on nozzle movement and residence time, resulting in cost-effective and efficient film thickness predictions for substrate processing.
Patent Information
- Application Number
- PCT/JP2024/016712
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-18
- Filing Date
- 2024-04-30
- Publication Date
- 2024-11-21
AI Technical Summary
Current prediction algorithms for substrate processing in semiconductor manufacturing have low generalizability due to biased learning data, requiring extensive and costly experimentation to cover various processing conditions, which limits their accuracy and applicability across different conditions.
A prediction algorithm generation device that classifies and selects processing conditions based on variable conditions related to nozzle movement and residence time, generating a dataset to train a machine learning model that predicts film thickness changes, allowing for the generation of highly generalizable algorithms at reduced cost.
The approach improves the generalizability of prediction algorithms, reduces the need for extensive dataset generation, and minimizes costs by effectively predicting film thickness changes across various processing conditions, enhancing the efficiency of substrate processing.
Smart Images

Figure JP2024016712_21112024_PF_FP_ABST
Abstract
Description
Prediction algorithm generating device, information processing device, substrate processing device, prediction algorithm generating method, prediction algorithm generating program, processing condition determining method, processing condition determining program, and prediction algorithm
[0001] The present invention relates to a prediction algorithm generating device, an information processing device, a substrate processing device, a prediction algorithm generating method, a prediction algorithm generating program, a processing condition determining method, a processing condition determining program, and a prediction algorithm.
[0002] The etching process is one of the semiconductor manufacturing processes. In the etching process, the thickness of the film formed on the substrate is adjusted by supplying a chemical solution to the substrate. In this film thickness adjustment, it is important to perform the etching process so that etching proceeds uniformly across the entire surface of the substrate, or to flatten the surface of the substrate through the etching process. When ejecting the etching solution from a nozzle onto a portion of the substrate, the nozzle must be moved radially relative to the substrate.
[0003] Patent Document 1 describes a liquid processing apparatus capable of etching a substrate by discharging an etching liquid onto the substrate from an etching nozzle. To achieve a uniform temperature distribution within the wafer while etching the central region of the substrate, the etching liquid is discharged by repeatedly moving the etching nozzle back and forth between a first position on the central side where the discharged etching liquid passes through the center of the wafer and a second position closer to the peripheral edge of the wafer than the central position.
[0004] Etching is a complex process in which the amount of film processed varies depending on the operation of moving the nozzle, the concentration of the etching solution, the temperature, the rotation speed of the substrate, etc. For this reason, it is conceivable to train a learning model using artificial intelligence through machine learning and have the trained learning model infer the amount of processing.
[0005] Japanese Patent Application Laid-Open No. 2015-103656
[0006] It is known that the accuracy of a learning model depends on the learning data used to train the learning model. Generally, the learning data includes processing conditions for actually operating a substrate processing apparatus and processing results obtained by operating the substrate processing apparatus under the processing conditions.
[0007] However, the training data obtained by actually executing a process on a substrate processing apparatus may contain processing conditions that are biased toward certain characteristics and the processing results obtained under those processing conditions. A learning model based on such training data has low generalizability, and when processing conditions with characteristics different from those of the training data are given, the inference accuracy decreases. On the other hand, in order to prepare training data that is free from bias in characteristics, the substrate processing apparatus must actually be operated under a huge number of processing conditions. This results in problems such as significant costs and time being required due to the time spent on setting up the substrate processing apparatus and operating the substrate processing apparatus.
[0008] An object of the present invention is to provide a prediction algorithm generation device that can generate highly generalizable prediction algorithms at low cost.
[0009] Another object of the present invention is to provide a method and a program for generating a prediction algorithm that are capable of generating a highly generalizable prediction algorithm at low cost.
[0010] Another object of the present invention is to provide a highly generalizable prediction algorithm.
[0011] It is still another object of the present invention to provide an information processing apparatus capable of presenting a plurality of processing conditions for a throughput of a complex process for processing substrates, and a substrate processing apparatus equipped with such an information processing apparatus.
[0012] It is still another object of the present invention to provide a processing condition determination method and a processing condition determination program that are capable of presenting a plurality of processing conditions for a throughput of a complex process for processing substrates.
[0013] (1) A predictive algorithm generation device according to one aspect of the present invention includes a processing condition generation unit that generates processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; a classification unit that classifies the processing conditions generated by the processing condition generation unit into a plurality of groups based on variables related to the variable conditions, the processing conditions including variable conditions that vary over time; a processing condition selection unit that selects at least one processing condition from each of the plurality of groups; a processing amount acquisition unit that acquires a processing amount indicating a difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process in accordance with the processing conditions selected by the processing condition selection unit; and a predictive algorithm generation unit that generates a predictive algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amounts acquired in relation to the processing conditions.
[0014] (2) A prediction algorithm according to another aspect of the present invention is generated by the prediction algorithm generation device.
[0015] (3) According to yet another aspect of the present invention, an information processing apparatus manages a substrate processing apparatus using the above-described prediction algorithm, and includes a processing condition determination unit that provides tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determines the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0016] (4) A substrate processing apparatus according to yet another aspect of the present invention includes the above-described information processing apparatus.
[0017] (5) A method for generating a prediction algorithm according to yet another aspect of the present invention includes the steps of generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating, the processing conditions including variable conditions that vary over time, classifying the plurality of processing conditions generated by the processing condition generating step into a plurality of groups based on variables related to the variable conditions, selecting at least one processing condition from each of the plurality of groups, acquiring a processing amount indicating a difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process according to the processing conditions selected by the processing condition selecting step, and generating a prediction algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amount acquired in relation to the processing conditions.
[0018] (6) A prediction algorithm generation program according to another aspect of the present invention is a program for causing a computer to execute the following processes: generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process a coating; the processing conditions include variable conditions that vary over time; classifying the multiple processing conditions generated by the processing for generating processing conditions into multiple groups based on variables related to the variable conditions; selecting at least one processing condition from each of the multiple groups; acquiring a processing amount indicating the difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process according to the processing conditions selected by the processing for selecting processing conditions; and generating a prediction algorithm that predicts the processing amount from the processing conditions using a dataset including the processing conditions and the processing amount acquired in relation to the processing conditions.
[0019] (7) A processing condition determination method according to another aspect of the present invention is a processing condition determination method for managing a substrate processing apparatus using the above-described prediction algorithm, which includes a step of providing tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0020] (8) A processing condition determination program according to another aspect of the present invention is a processing condition determination program for managing a substrate processing apparatus using the prediction algorithm described above in paragraph 9, and causes a computer to execute a process of providing tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0021] According to the present invention, it is possible to provide a prediction algorithm generation device, a prediction algorithm generation method, and a prediction algorithm generation program that can generate a highly generalizable prediction algorithm at low cost. It is also possible to provide a highly generalizable prediction algorithm. It is also possible to provide an information processing device, a processing condition determination method, and a processing condition determination program that can present multiple processing conditions for a processing volume obtained through a complex process.
[0022] FIG. 1 is a diagram for explaining a configuration of a substrate processing system according to an embodiment of the present invention. FIG. 2 is a diagram for explaining a processing amount. FIG. 3 is a diagram for explaining an example of a configuration of an information processing device. FIG. 4 is a diagram for explaining an example of a configuration of a prediction algorithm generation device. FIG. 5 is a diagram for explaining an example of a functional configuration of a substrate processing system according to an embodiment. FIG. 6 is a diagram for explaining a change in the relative position of a nozzle with respect to a substrate in a coating process. FIG. 7 is a diagram for explaining an example of a nozzle operation pattern. FIG. 8 is a diagram for explaining an example of functions of a processing condition generation unit. FIG. 9 is a diagram for explaining divided regions. FIG. 10 is a diagram for explaining a residence time distribution. FIG. 11 is a diagram for explaining an example of a clustering result. FIG. 12 is a diagram for explaining the structure of a learning model in which a prediction algorithm is realized. A flowchart showing an example of the flow of a prediction algorithm generation process. A flowchart showing an example of the flow of a generation processing condition generation process. A flowchart showing an example of the flow of a processing condition determination process. A diagram showing experimental results. A diagram showing experimental results.
[0023] A substrate processing system according to an embodiment of the present invention will be described in detail below with reference to the drawings. In the following description, the term "substrate" refers to a semiconductor substrate (semiconductor wafer), a substrate for a flat panel display (FPD) such as a liquid crystal display or an organic electroluminescence (EL) display, a substrate for an optical disk, a substrate for a magnetic disk, a substrate for a magneto-optical disk, a substrate for a photomask, a ceramic substrate, a substrate for a solar cell, or the like.
[0024] (1) Overall Configuration of Substrate Processing System Fig. 1 is a diagram illustrating the configuration of a substrate processing system according to one embodiment of the present invention. The substrate processing system 1 in Fig. 1 includes an information processing device 100, a prediction algorithm generation device 200, and a substrate processing device 300. The prediction algorithm generation device 200 is, for example, a server, and the information processing device 100 is, for example, a personal computer.
[0025] The prediction algorithm generating device 200 and the information processing device 100 are used to manage the substrate processing device 300. The number of substrate processing devices 300 managed by the prediction algorithm generating device 200 and the information processing device 100 is not limited to one, and a plurality of substrate processing devices 300 may be managed.
[0026] In the substrate processing system 1 according to this embodiment, the information processing apparatus 100, the prediction algorithm generation apparatus 200, and the substrate processing apparatus 300 are connected to one another via a wired or wireless communication line or communication network. The information processing apparatus 100, the prediction algorithm generation apparatus 200, and the substrate processing apparatus 300 are each connected to a network, and are capable of transmitting and receiving data to and from one another. The network may be, for example, a local area network (LAN) or a wide area network (WAN). The network may also be the Internet. The information processing apparatus 100 and the substrate processing apparatus 300 may also be connected via a dedicated communication network. The network connection may be wired or wireless.
[0027] The prediction algorithm generation device 200 does not necessarily need to be connected to the substrate processing apparatus 300 and the information processing apparatus 100 via a communication line or a communication network. In this case, data generated by the substrate processing apparatus 300 may be transferred to the prediction algorithm generation device 200 via a recording medium. Furthermore, data generated by the prediction algorithm generation device 200 may be transferred to the information processing apparatus 100 via a recording medium.
[0028] A display device, a sound output device, and an operation unit (not shown) are provided in the substrate processing apparatus 300. The substrate processing apparatus 300 is operated in accordance with predetermined processing conditions (processing recipe) of the substrate processing apparatus 300.
[0029] (2) Overview of the Substrate Processing Apparatus The substrate processing apparatus 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. Each of the plurality of substrate processing units WU processes a coating formed on a substrate W by supplying a processing liquid to the substrate W on which a coating has been formed in accordance with processing conditions. The processing performed by the substrate processing unit WU by supplying a processing liquid to the substrate W on which a coating has been formed in accordance with processing conditions is called a coating process.
[0030] In this embodiment, the substrate W to be processed has a diameter of 300 mm, but the present invention is not limited to this. The processing liquid includes an etching liquid, and the substrate processing unit WU performs the etching process. The etching liquid is a chemical liquid. The etching liquid is, for example, a mixture of hydrofluoric acid (HF) and nitric acid (HNO 3 )), hydrofluoric acid, buffered hydrofluoric acid (BHF), ammonium fluoride, HFEG (a mixture of hydrofluoric acid and ethylene glycol), or phosphoric acid (H 3 P.O. 4 )
[0031] The substrate processing unit WU includes a spin chuck SC, a spin motor SM, a nozzle 311, and a nozzle moving mechanism 301. The spin chuck SC holds the substrate W horizontally. The substrate W is held on the spin chuck SC so that the center of the substrate W coincides with a first rotation axis AX1 of the spin motor SM. The spin motor SM has a first rotation axis AX1. The first rotation axis AX1 extends in the vertical direction. The spin chuck SC is attached to the upper end of the first rotation axis AX1 of the spin motor SM. When the spin motor SM rotates, the spin chuck SC rotates around the first rotation axis AX1. The spin motor SM is a stepping motor. The substrate W held on the spin chuck SC rotates around the first rotation axis AX1. Therefore, the rotation speed of the substrate W is the same as the rotation speed of the stepping motor. When an encoder that generates a rotation speed signal indicating the rotation speed of the spin motor is provided, the rotation speed of the substrate W may be obtained from the rotation speed signal generated by the encoder. In this case, a motor other than a stepping motor may be used as the spin motor SM.
[0032] The nozzle 311 supplies the etching liquid to the surface (top surface) of the substrate W held by the spin chuck SC. The etching liquid is supplied to the nozzle 311 from an etching liquid supply unit (not shown). The nozzle 311 ejects the etching liquid toward the surface of the substrate W while it is rotating.
[0033] The nozzle movement mechanism 301 moves the nozzle 311 in a substantially horizontal direction. Specifically, the nozzle movement mechanism 301 includes a nozzle motor 303 having a second rotation axis AX2, and a nozzle arm 305. The nozzle motor 303 is disposed so that the second rotation axis AX2 is aligned in a substantially vertical direction. The nozzle arm 305 has a longitudinal shape that extends linearly. One end of the nozzle arm 305 is attached to 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 that of the second rotation axis AX2. The nozzle 311 is attached to the other end of the nozzle arm 305 so that the etching liquid outlet faces downward.
[0034] When the nozzle motor 303 operates, the nozzle arm 305 rotates in a horizontal plane around the second rotation axis AX2. As a result, the nozzle 311 attached to the other end of the nozzle arm 305 moves (pivots) in the horizontal direction around the second rotation axis AX2. While moving in the horizontal direction, the nozzle 311 ejects the etching liquid toward the substrate W. The nozzle motor 303 is, for example, a stepping motor.
[0035] The substrate processing units WU have the same functions and configurations. Hereinafter, the coating processes performed by each of the substrate processing units WU will be collectively referred to as the coating processes performed by the substrate processing apparatus 300. The substrate processing apparatus 300 performs the coating process according to the processing conditions, which changes the thickness of the coating formed on the substrate W. The amount of change in the thickness of the coating before and after the coating process is the processing volume. The processing volume varies depending on the processing conditions used when the substrate processing apparatus 300 performs the coating process.
[0036] Fig. 2 is a diagram for explaining the processing amount. In Fig. 2, the vertical axis represents film thickness, and the horizontal axis represents the radial position of the substrate. The origin of the horizontal axis represents the center of the substrate. The solid line represents the film thickness of the film formed on the substrate W before the coating process is performed by the substrate processing apparatus 300. The substrate processing apparatus 300 performs the coating process in accordance with the processing conditions, thereby adjusting the film thickness of the film formed on the substrate W. The dotted line represents the film thickness of the film formed on the substrate W after the coating process is performed by the substrate processing apparatus 300. The film thickness of the film at each radial position of the substrate W is called the film thickness characteristic.
[0037] The processing amount (etching amount) is the difference between the thickness of the coating formed on the substrate W before the coating process is performed by the substrate processing apparatus 300 and the thickness of the coating formed on the substrate W after the coating process is performed by the substrate processing apparatus 300. In other words, the processing amount indicates the thickness of the coating reduced at each of a plurality of different positions in the radial direction of the substrate W by the coating process by the substrate processing apparatus 300.
[0038] A target film thickness is determined for the coating process performed by the substrate processing apparatus 300. The target film thickness is indicated by a dashed line. The deviation characteristic is the difference between the film thickness of the coating formed on the substrate W after the coating process is performed by the substrate processing apparatus 300 and the target film thickness. The deviation characteristic includes the difference at each of a plurality of different positions in the radial direction of the substrate W.
[0039] 1 , the control device 10 includes a CPU (Central Processing Unit) and a memory, and the CPU executes a program stored in the memory to control the entire substrate processing apparatus 300. The control device 10 controls the spin motor SM and the nozzle motor 303. The prediction algorithm generation device 200 generates a prediction algorithm that predicts the throughput from the processing conditions and outputs the prediction algorithm to the information processing apparatus 100.
[0040] The information processing apparatus 100 uses a prediction algorithm to determine processing conditions for performing a coating process on a substrate that is to be processed by the substrate processing apparatus 300. The information processing apparatus 100 outputs the determined processing conditions to the substrate processing apparatus 300.
[0041] Fig. 3 is a diagram showing an example of the configuration of a prediction algorithm generation device. Referring to Fig. 3, the prediction algorithm generation 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, RAM 202, ROM 203, storage device 204, operation unit 205, display device 206, and input / output I / F 207 are connected to a bus 208.
[0042] The RAM 202 is used as a working area for the CPU 201. The ROM 203 stores a system program. The storage device 204 includes a storage medium such as a hard disk or semiconductor memory, and stores the program. The program may be stored in the ROM 203 or another external storage device. A CD-ROM 209 is detachably attached to the storage device 204.
[0043] 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. The recording medium for storing the program executed by the CPU 201 is not limited to the CD-ROM 209, but may be an optical disk (MO (Magnetic Optical Disc) / MD (Mini Disc) / DVD (Digital Versatile Disc)), an IC card, an optical card, a semiconductor memory such as a mask ROM, or an EPROM (Erasable Programmable ROM). Furthermore, the CPU 201 may download the program from a computer connected to the network and store it in the storage device 204, or the computer connected to the network may write the program to the storage device 204, and the program stored in the storage device 204 may be loaded into the RAM 202 and executed by the CPU 201. The program here includes not only a program that can be directly executed by the CPU 201, but also a source program, a compressed program, an encrypted program, and the like.
[0044] 4 is a diagram showing an example of the configuration of an information processing apparatus. Referring to Fig. 4, information processing apparatus 100 is composed of CPU 101, RAM (random access memory) 102, ROM (read only memory) 103, storage device 104, operation unit 105, display device 106, and input / output I / F (interface) 107. CPU 101, RAM 102, ROM 103, storage device 104, operation unit 105, display device 106, and input / output I / F 107 are connected to bus 108.
[0045] The RAM 102 is used as a work area for the CPU 101. The ROM 103 stores a system program. The storage device 104 includes a storage medium such as a hard disk or semiconductor memory, and stores the program. The program may be stored in the ROM 103 or another external storage device.
[0046] A CD-ROM 109 is detachably attached to 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 an optical disk, a semiconductor memory, or other medium. 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, and the program stored in the storage device 104 may be loaded into the RAM 102 and executed by the CPU 101. The program referred to here includes not only programs that can be directly executed by the CPU 101, but also source programs, compressed programs, encrypted programs, and the like.
[0047] The operation unit 105 is an input device such as a keyboard, a mouse, or a touch panel. A user can give predetermined instructions 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, and displays a GUI (Graphical User Interface) or the like for receiving instructions from the user. The input / output I / F 107 is connected to a network.
[0048] (3) Functional Configuration of the Substrate Processing System FIG. 5 is a diagram showing an example of the functional configuration of a substrate processing system according to one embodiment. Referring to FIG. 5, the control device 10 included in the substrate processing apparatus 300 controls the substrate processing unit WU to perform a coating process on the substrate W in accordance with the processing conditions during the processing time. The processing time is a time set for the coating process on the substrate. In this embodiment, the processing time is the time during which the nozzle 311 discharges the etching solution onto the substrate W. In this embodiment, the processing time is set to 60 seconds. The processing conditions are conditions used when the substrate processing unit WU performs the coating process.
[0049] 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 position of the nozzle 311. The concentration of the etching solution is represented by a mixing ratio of multiple chemical solutions. The position of the nozzle 311 is represented by the relative position between the nozzle 311 and the substrate W at each of multiple points in time while the coating process is being performed. The relative position between the nozzle 311 and the substrate W is represented by the rotation angle of the nozzle motor 303.
[0050] The processing conditions include fixed conditions that do not change over time and variable conditions that change over time. In this 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, and the variable conditions are the relative position of the nozzle 311 and the substrate W during the processing time during which the coating process is performed. In this embodiment, the variable conditions are the relative position of the nozzle 311 with respect to the substrate W at each of a plurality of times obtained by dividing the processing time during which the coating process is performed into 2000 equal parts. Therefore, the variable conditions are made up of 2000 pieces of data.
[0051] The prediction algorithm generation device 200 includes a generation processing condition generation unit 220, a prediction algorithm generation unit 230, and a prediction algorithm transmission unit 240. The functions of the prediction algorithm generation device 200 are realized by a CPU 201 included in the prediction algorithm generation device 200, as the CPU 201 executes a learning model generation program stored in a RAM 202.
[0052] The generation process condition generation unit 220 will be described in detail below, but it generates a data set used by the prediction algorithm generation unit 230 to generate a prediction algorithm. The data set includes process conditions and a processing amount indicating the results of the substrate processing apparatus 300 performing a coating process on a substrate W under the process conditions. The prediction algorithm generation unit 230 receives the data set from the generation process condition generation unit 220. The prediction algorithm generation unit 230 generates a prediction algorithm based on the data set and outputs the generated prediction algorithm to the prediction algorithm transmission unit 240. The prediction algorithm transmission unit 240 transmits the prediction algorithm generated by the prediction algorithm generation unit 230 to the information processing apparatus 100.
[0053] The information processing device 100 includes a prediction algorithm receiving unit 110, a processing condition determination unit 120, a prediction unit 130, an evaluation unit 140, and a processing condition transmission unit 150. The functions of the information processing device 100 are realized by a CPU 101 included in the information processing device 100 as the CPU 101 executes a processing condition determination program stored in a RAM 102.
[0054] The prediction algorithm receiving unit 110 receives the prediction algorithm transmitted from the prediction algorithm generating device 200 and outputs the received prediction algorithm to the prediction unit 130. The processing condition determining unit 120 determines processing conditions for the substrate W to be processed by the substrate processing apparatus 300, and outputs variable conditions and fixed conditions included in the processing conditions to the prediction unit 130.
[0055] The prediction unit 130 predicts the predicted processing amount using a prediction algorithm based on the variable conditions and fixed conditions. The prediction unit 130 inputs the variable conditions and fixed conditions input from the processing condition determination unit 120 into the prediction algorithm, and outputs the predicted processing amount output by the prediction algorithm to the evaluation unit 140.
[0056] The evaluation unit 140 evaluates the predicted throughput input from the prediction unit 130 and outputs the evaluation result to the processing condition determination unit 120. Specifically, the evaluation unit 140 acquires the pre-processing film thickness characteristics of the substrate W to be processed by the substrate processing apparatus 300. The evaluation unit 140 calculates the predicted film thickness characteristics after the etching process from the predicted throughput input from the prediction unit 130 and the pre-processing film thickness characteristics of the substrate W, and compares the calculated film thickness characteristics with the target film thickness characteristics. If the comparison result satisfies the evaluation criteria, the evaluation unit 140 outputs the processing conditions determined by the processing condition determination unit 120 to the processing condition transmission unit 150 as the processing conditions for the substrate W to be processed by the substrate processing apparatus 300. For example, the evaluation unit 140 calculates a deviation characteristic and determines whether the deviation characteristic satisfies the evaluation criteria. The deviation characteristic is the difference between the film thickness characteristics of the substrate W after the etching process and the target film thickness characteristic. The evaluation criteria can be determined arbitrarily. For example, the evaluation criterion may be that the maximum value of the difference in the deviation characteristics is equal to or less than a threshold value, or that the average of the difference is equal to or less than a threshold value.
[0057] The processing condition transmitting unit 150 transmits the processing conditions input from the processing condition determining unit 120 to the control device 10 of the substrate processing apparatus 300. The substrate processing apparatus 300 processes the substrate W in accordance with the processing conditions.
[0058] If the evaluation result does not satisfy the evaluation criteria, the evaluation unit 140 outputs the evaluation result to the processing condition determination unit 120. The evaluation result includes a film thickness characteristic predicted after the etching process or a difference between the film thickness characteristic predicted after the etching process and a target film thickness characteristic.
[0059] In response to the evaluation results input from the evaluation unit 140, the processing condition determination unit 120 determines new processing conditions for the prediction unit 130 to make a prediction. The processing condition determination unit 120 selects one variable condition and one fixed condition from a plurality of previously prepared variable conditions and one fixed condition, using an experimental design method, a pairwise method, Bayesian estimation, or the like, and determines processing conditions including the selected variable condition and fixed condition as new processing conditions for the prediction unit 130 to make a prediction.
[0060] The processing condition determination unit 120 may search for processing conditions using Bayesian estimation. When multiple evaluation results are output by the evaluation unit 140, there will be multiple pairs of processing conditions and evaluation results. Based on the trend of the predicted processing amount in each of the multiple pairs, the processing condition determination unit 120 searches for processing conditions that will result in a uniform film thickness or processing conditions that will minimize the difference between the film thickness characteristics predicted after the etching process and the target film thickness characteristics.
[0061] Specifically, the processing condition determination unit 120 searches for processing conditions that minimize an objective function. The objective function is a function that indicates the uniformity of the film thickness or a function that indicates the agreement between the film thickness characteristics of the film and the target film thickness characteristics. For example, the objective function is a function that indicates, using parameters, the difference between the film thickness characteristics predicted after the etching process and the target film thickness characteristics. The parameters here are corresponding variable conditions. The corresponding variable conditions are variable conditions used by the prediction algorithm to predict the predicted throughput. The processing condition determination unit 120 selects variable conditions, which are parameters determined by the search, from among the multiple variable conditions, and determines new processing conditions that include the selected variable conditions and fixed conditions.
[0062] Here, the fluctuation conditions will be explained. FIG. 6 is a diagram for explaining the change in the relative position of the nozzle with respect to the substrate during coating processing. Referring to FIG. 6, the change in the relative position of the nozzle 311 with respect to the substrate W held on the spin chuck SC is shown. The nozzle 311 moves in an area above the substrate W held on the spin chuck SC. Because the nozzle 311 rotates around the second rotation axis AX2, the trajectory along which the nozzle 311 moves is an arc. The trajectory along which the nozzle 311 moves passes through the substrate center OP, which is the center of the substrate. Therefore, the nozzle 311 moves from the substrate center OP to the entire peripheral edge in the radial direction of the substrate W. Here, one end of the trajectory along which the nozzle 311 moves is indicated by an operating end EP1 located inside the peripheral edge of the substrate W, and the other end is indicated by an operating end EP2 located inside the peripheral edge of the substrate W. The scanning moving from the operating end EP1 of the nozzle 311 to the substrate center OP is indicated by arrow a1, the scanning moving from the substrate center OP of the nozzle 311 to the operating end EP2 is indicated by arrow a2, the scanning moving from the operating end EP2 of the nozzle 311 to the substrate center OP is indicated by arrow a3, and the scanning moving from the substrate center OP of the nozzle 311 to the operating end EP1 is indicated by arrow a4.
[0063] FIG. 7 is a diagram showing an example of a nozzle operation pattern. In FIG. 7, the vertical axis represents the relative position of the nozzle 311 with respect to the substrate W, and the horizontal axis represents elapsed time (seconds). In this embodiment, the scanning period from the start to the end of the nozzle operation of moving the nozzle 311 with respect to the substrate W is equal to the processing time. As described above, the processing time is set to 60 seconds, and therefore the nozzle operation pattern shows the relative position for a period from 0 to 60 seconds. The relative position of the nozzle is represented by a negative value for the substrate center OP, and a positive value for the range from the substrate center OP to the operating end EP1. Since the substrate W has a diameter of 300 mm, the distance from the substrate center OP to the operating end EP1 and EP2 is set to ±150 mm or less. Here, the distance from the substrate center OP to the operating end EP1 is set to −147 mm, and the distance from the substrate center OP to the operating end EP2 is set to +147 mm. In the nozzle operation pattern of Figure 7, the relative position of nozzle 311 when nozzle 311 is located at the substrate center OP is indicated by 0, the relative position of nozzle 311 when nozzle 311 is located at the operating end EP1 is indicated by -147 mm, and the relative position of nozzle 311 when nozzle 311 is located at the operating end EP2 is indicated by 147 mm.
[0064] The nozzle operation pattern shown in FIG. 7 is shown as a scan that reciprocates five times between the operating end EP1 and the operating end EP2. The nozzle operation pattern of the nozzle 311 of the substrate processing apparatus 300 is not limited to a scan that reciprocates five times, but may be a scan that reciprocates one or more times. The first reciprocating scan in the nozzle operation pattern is indicated by the same reference numerals assigned to the relative positions corresponding to the scans indicated by arrows a1 to a4 in FIG. 6 . In the example of the nozzle operation pattern shown in FIG. 7 , an example in which the nozzle 311 simply reciprocates is shown for ease of explanation. However, in actual operation of the nozzle 311, the speed of the nozzle 311 is changed and the nozzle 311 is stopped. Hereinafter, the points on the substrate W where the speed of the nozzle 311 is changed and the nozzle 311 is stopped during processing time are referred to as speed change points. The variation conditions include the number of speed change points, the positions of the speed change points on the substrate, the speed of the nozzle 311 at the positions of the speed change points on the substrate, and the stop time of the nozzle 311.
[0065] Next, a detailed configuration of the generation processing condition generation unit 220 in Fig. 5 will be described. Fig. 8 is a diagram showing an example of the functions of the processing condition generation unit. The generation processing condition generation unit 220 includes a generation variable condition determination unit 221, a conversion unit 222, a count classification unit 223, a classification unit 224, a generation variable condition selection unit 225, a generation fixed condition determination unit 226, a generation processing condition determination unit 227, and a processing amount acquisition unit 228.
[0066] The generating variable condition determination unit 221 determines multiple generating variable conditions scattered throughout the entire range of possible variable conditions. The generating process conditions include generating variable conditions and generating fixed conditions. The generating variable conditions are the position of the nozzle 311 relative to the substrate W, which changes over time. Here, an example is described in which the nozzle 311 travels back and forth across the substrate W one or more times while the substrate processing apparatus 300 performs a coating process, and the speed of the nozzle 311 is changed one or more times during each travel. The variable conditions are determined by the number of times the nozzle 311 travels back and forth, the number of speed change points for the nozzle 311, the speed of the nozzle 311 set at each speed change point, and the travel time of the nozzle 311 moving between the speed change points. The speed of the nozzle 311 includes 0. When the speed of the nozzle 311 is 0, the nozzle 311 is stopped, and the travel time when the speed of the nozzle 311 is 0 is also referred to as a stop time. When the processing time for the coating process is fixed, the scan time T per reciprocation is determined by dividing the processing time by the number of reciprocation passes. The number of speed change points indicates the number of speed changes n. If the scan time is T and the distance L (substrate diameter), and the average speed of the nozzle 311 during the scan time T is Vm, then 2L = T × Vm. If the period during which the nozzle 311 moves at a constant speed is the moving period Ti (1≦i≦n), and the speed of the nozzle 311 during the moving period Ti is Vi, then the following equation (1) holds: 2L = ΣTi × Vi (1) By determining a combination of Ti and Vi that satisfies the above equation (1), the variation conditions for the number of reciprocation passes and the number of speed change points are determined. For example, if either Ti or Vi is determined, the other is determined from equation (1). Ti may be determined first, followed by Vi, or Vi may be determined first, followed by Ti. For example, Ti may be determined by equally dividing the scan time T per round trip, or a plurality of patterns for dividing the scan time T may be prepared in advance. The plurality of patterns may include, for example, a pattern in which Ti gradually increases, a pattern in which Ti gradually decreases, a pattern in which Ti gradually increases and then gradually decreases, etc.
[0067] The generation variable condition determination unit 221 generates a plurality of sets of moving periods Ti and speeds Vi for each of a plurality of sets of the number of round trips and the number of speed change points, thereby generating a plurality of generation variable conditions. When upper limits are set for each of the number of round trips and the number of speed change points, the generation variable condition determination unit 221 generates a plurality of sets of moving periods Ti and speeds Vi for all combinations of the number of round trips and the number of speed change points.
[0068] The plurality of generation variable conditions determined by the generation variable condition determination unit 221 each have a different nozzle operation pattern. For example, the plurality of generation variable conditions generated by the generation variable condition determination unit 221 differ in at least one of the number of reciprocating movements, the number of speed change points, the movement period, and the speed. In this embodiment, the number of generation variable conditions determined by the generation variable condition determination unit 221 is 18,000.
[0069] The conversion unit 222 converts the plurality of generation variable conditions determined by the generation variable condition determination unit 221 into conversion data. The conversion unit 222 outputs a set of the generation variable conditions and the conversion data to the count classification unit 223. The conversion data is a variable related to the generation variable conditions.
[0070] In this embodiment, the converted data is a cumulative residence time distribution. The cumulative residence time distribution is a value obtained by accumulating the residence time of the nozzle at each of a plurality of radial positions on the substrate W within the processing time from the center of the substrate W toward the periphery. The cumulative residence time distribution will now be described.
[0071] FIG. 9 is a diagram illustrating divided regions. Referring to FIG. 9, 15 divided regions b1 to b15 are shown, each obtained by dividing the upper surface of the substrate W into a plurality of concentric circles centered on the substrate center OP. The divided region b15 is a circle, and the divided regions b1 to b14 are annular. The divided regions b2 to b15 each have the same radial length of the substrate W. The radial length of each of the divided regions b2 to b15 is the difference between the radius of the outer periphery and the radius of the inner periphery. The radius of the divided region b1 is the same as the radial length of each of the divided regions b2 to b15. Here, the radius of the divided region b1 is 10 mm, and the difference between the radius of the outer periphery and the radius of the inner periphery of each of the divided regions b1 to b14 is 10 mm. The difference between the radius of the outer periphery and the radius of the inner periphery of each of the divided regions b2 to b15 and the radius of the divided region b1 are greater than the inner diameter of the nozzle 311. The length of each of the divided regions b2 to b15 in the radial direction of the substrate and the radius of the divided region b1 are preferably equal to or greater than the inner diameter of the nozzle 311.
[0072] Fig. 10 is a diagram for explaining the residence time distribution. In Fig. 10, the horizontal axis represents the position in the radial direction of the substrate W. The position of the substrate center OP is indicated by 0 mm, and the end of the substrate W in the radial direction is indicated by 150 mm. Divided regions b1 to b15 are allocated between 0 mm and 150 mm on the horizontal axis.
[0073] The vertical axis indicates the residence time of the nozzle 311 in each of the divided regions b1 to b15. Here, the residence time in each of the divided regions b1 to b15 is shown when the nozzle 311 moves according to the operation pattern shown in FIG. 7. The residence time is the total amount of time that the nozzle 311 is located in each of the multiple divided regions b1 to b15. For example, when the nozzle 311 moves according to the nozzle operation pattern shown in FIG. 7, the nozzle 311 crosses the divided region b2 ten times. The residence time in the divided region b2 is the total amount of time that the nozzle 311 crosses the divided region b2.
[0074] As described above, the movement range of the nozzle 311 is divided into a plurality of divided regions b1 to b15. Therefore, the residence time of the nozzle 311 in each of the plurality of divided regions b1 to b15 is calculated for each of the plurality of divided regions b1 to b15, which includes information on the radial position of the substrate W. Therefore, the residence time in each of the plurality of divided regions b1 to b15 is information including the position in the radial direction of the substrate W. Furthermore, the radial length of the portion of the substrate W traversed by the nozzle 311 in each of the plurality of divided regions b1 to b15 is the same. Therefore, the residence time of the nozzle 311 in each of the plurality of divided regions b1 to b15 can be set to a time that eliminates bias between different positions in the radial direction of the substrate W regarding changes in the relative position of the nozzle 311 with respect to the substrate W.
[0075] The cumulative residence time distribution can be obtained by accumulating the residence times in the residence time distribution shown in Fig. 10 from the center to the periphery of the substrate W. Specifically, the cumulative residence time in divided region b2 is the sum of the residence time in divided region b1 and the residence time in divided region b2, the cumulative residence time in divided region b3 is the sum of the cumulative residence time in divided region b2 and the residence time in divided region b3, and the cumulative residence time in divided region b15 is the sum of the cumulative residence time in divided region b14 and the residence time in divided region b15.
[0076] In this embodiment, the variable condition is converted into conversion data which is the cumulative residence time. The cumulative residence time is determined by dividing the upper surface of the substrate W into 15 divided regions b1 to b15, so the number of conversion data is 15. The number of conversion data decreases as the radial length of the substrate in each of the divided regions b1 to b15 increases. The radial length of the substrate in each of the divided regions b1 to b15 is equal to or greater than the inner diameter of the nozzle 311, so the maximum number of conversion data is determined by the inner diameter of the nozzle 311.
[0077] Returning to FIG. 8 , the number-of-times classification unit 223 receives multiple sets of generation variable conditions and conversion data from the conversion unit 222. The number-of-times classification unit 223 classifies the multiple generation variable conditions into multiple number-of-times groups based on the number of reciprocations of the nozzle 311. In this embodiment, the number-of-times classification unit 223 classifies the multiple generation conversion data into one of a number-of-times group for reciprocations of the nozzle 311 of 0 to 3, a number-of-times group for reciprocations of 3 to 6, a number-of-times group for reciprocations of 6 to 9, and a number-of-times group for reciprocations of 9 or more. Note that the number-of-times groups are classified using thresholds of 3, 6, and 9 reciprocations, but other values may be used as these thresholds. Furthermore, the number of number-of-times groups is not limited to four, and may be one or more.
[0078] Furthermore, the classification unit 224 clusters the generation variable conditions belonging to each of the multiple frequency-based groups into multiple groups based on the conversion data. FIG. 11 is a diagram illustrating an example of the clustering results. FIG. 11 shows the cumulative stay distribution of the nozzle 311 relative to the substrate center OP at each radial position of a cluster with a reciprocation count of 0 to 3. The vertical axis of FIG. 11 shows the cumulative stay time of the nozzle from the substrate center OP, and the horizontal axis shows the radial position of the substrate W. The position of the substrate center OP is indicated by 0 mm, and the radial end of the substrate W is indicated by 150 mm. In the example of FIG. 11 , the vertical axis normalizes the cumulative stay time at each radial position of the substrate W by dividing it by the total cumulative stay time. The reference line SL shown in FIG. 11 represents the cumulative stay distribution of the nozzle 311 when the nozzle 311 is present at each radial position of the substrate W for the same amount of time in the nozzle operation pattern. The nozzle operation patterns included in the generating variable conditions are distributed around the reference line SL, but are not shown in FIG.
[0079] 11 shows an example in which nozzle operation patterns of generation variable conditions are clustered into regions A1 to A4. Regarding regions A1 to A4, an example is shown in which regions A1 and A2 and regions A3 and A4 are symmetrical about the reference line SL, but regions A1 to A4 may be determined by, but are not limited to, the k-means method. A method other than the k-means method may also be used for clustering. Furthermore, although an example is shown in which nozzle operation patterns are clustered into four groups, the number of clusters is not limited to this.
[0080] Returning to FIG. 8 , the generation variable condition selection unit 225 randomly selects one or more generation variable conditions from the generation variable conditions belonging to the clusters clustered by the classification unit 224 and outputs the selected generation variable conditions to the generation process condition determination unit 227. For example, the classification unit 224 randomly selects 60 generation variable conditions from each of the four clusters represented by areas A1 to A4, where the generation variable conditions belonging to groups with a number of round trips of 0 to 3 are shown. As a result, 240 generation variable conditions are selected from the group with a number of round trips of 0 to 3. The generation variable condition selection unit 225 also selects 240 generation variable conditions from each of the groups with a number of round trips of 3 to 6, 6 to 9, and 9 or more using the same method. Thus, the generation variable condition selection unit 225 determines 960 generation variable conditions. As a result, the generation variable condition selection unit 225 uniformly selects generation variable conditions with different cumulative stay time distribution characteristics from multiple generation variable conditions, making it possible to select multiple variable conditions that are not biased toward a specific cumulative stay time distribution.
[0081] The generation fixed condition determination unit 226 generates a plurality of generation fixed conditions scattered throughout the entire range of values that can be set for the fixed conditions, and outputs these to the generation process condition determination unit 227. For example, the generation fixed condition determination unit 226 divides the range that can be taken by the fixed conditions into three equal parts, and generates a fixed condition with a central value of the range as the generation fixed condition. In this embodiment, 27 fixed conditions that combine three substrate rotation speeds, three etching solution flow rates, and three etching solution temperatures are determined as the generation fixed conditions.
[0082] The generation processing condition determination unit 227 receives generation variable conditions as input from the generation variable condition selection unit 225 and generation fixed conditions as input from the generation fixed condition determination unit 226. The generation processing condition determination unit 227 generates multiple processing conditions by combining the generation variable conditions and the generation fixed conditions. The generation processing condition determination unit 227 also determines generation processing conditions for generating a prediction algorithm by narrowing down the number of multiple generation processing conditions, for example, using a pairwise method. In this embodiment, the generation processing condition determination unit 227 receives 960 generation variable conditions as input from the generation variable condition selection unit 225 and 27 generation fixed conditions as input from the generation fixed condition determination unit 226. The generation processing condition determination unit 227 generates 25,920 generation processing conditions by combining the 960 generation variable conditions and the 27 generation fixed conditions. The generation processing condition determination unit 227 also narrows down the 25,920 generation processing conditions to 2,000 generation processing conditions using a pairwise method. As a result, the generation processing condition determination unit 227 determines the 2,000 generation processing conditions as generation processing conditions for generating a prediction algorithm. The generation processing condition determination unit 227 outputs the 2,000 generation processing conditions to the processing amount acquisition unit 228.
[0083] The processing amount acquisition unit 228 receives 2,000 generation process conditions from the generation process condition determination unit 227. The processing amount acquisition unit 228 acquires processing amounts corresponding to each of the 2,000 generation process conditions. Specifically, the processing amount acquisition unit 228 sequentially outputs the 2,000 generation process conditions to the control device 10 of the substrate processing apparatus 300. The control device 10 of the substrate processing apparatus 300 executes a coating process in accordance with the generation process conditions in response to the generation process conditions input from the processing amount acquisition unit 228. The substrate processing apparatus 300 executes the coating process, thereby obtaining a processing amount. The control device 10 of the substrate processing apparatus 300 transmits the processing amount obtained as a result of executing the coating process to the prediction algorithm generation unit 200. The processing amount acquisition unit 228 generates a dataset including the processing amount input from the control device 10 of the substrate processing apparatus 300 and the previously transmitted generation process conditions, and outputs the generated dataset to the prediction algorithm generation unit 230. The processing amount acquisition unit 228 acquires 2000 processing amounts corresponding to the 2000 generation processing conditions input from the generation processing condition determination unit 227, and outputs 2000 data sets to the prediction algorithm generation unit 230.
[0084] The prediction algorithm generation unit 230 receives multiple sets of data sets from the processing amount acquisition unit 228. The data sets include processing conditions for generation and processing amounts. The prediction algorithm generation unit 230 generates a prediction algorithm using the multiple sets of data sets. In this embodiment, the prediction algorithm generation unit 230 generates a prediction algorithm by subjecting the multiple sets of data sets to machine learning in a learning model.
[0085] 12 is a diagram illustrating the structure of a learning model in which a prediction algorithm is implemented. Referring to FIG. 12, the learning model has layers A1 to C1 arranged in this order from the input side to the output side (from upper layer to lower layer). A first convolutional neural network CNN1 is provided in the A1 layer, a first fully connected neural network NN1 is provided in the B1 layer, and a second convolutional neural network CNN2 is provided in the C1 layer.
[0086] The first convolutional neural network CNN1 receives converted data in which the generating variable conditions are converted into a cumulative stay time distribution. The first fully connected neural network NN1 receives the output of the first convolutional neural network CNN1 and the generating fixed conditions. The second convolutional neural network CNN2 receives the output of the first fully connected neural network NN1. Note that instead of the converted data, the generating variable conditions may be input to the first convolutional neural network CNN1.
[0087] The first convolutional neural network CNN1 includes multiple layers. In this embodiment, the first convolutional neural network CNN1 includes three layers. In the first convolutional neural network CNN1, a first layer L1a, a second layer L1b, and a third layer L1c are provided in this order from the input side (upper layer side) to the output side (lower layer side). Note that, in this embodiment, a case where the multiple layers include three layers will be described, but the first convolutional neural network CNN1 may include three or more layers.
[0088] The first layer L1a, the second layer L1b, and the third layer L1c each include a convolutional layer and a pooling layer. Multiple filters are applied in the convolutional layer. The pooling layer compresses the output of the convolutional layer. The number of filters in the convolutional layer of the second layer L1b is set to twice the number of filters in the convolutional layer of the first layer L1a. The number of filters in the convolutional layer of the third layer L1c is set to twice the number of filters in the convolutional layer of the second layer L1b. This allows as many features as possible to be extracted from the cumulative dwell time distribution. Here, the cumulative dwell time distribution indicates the cumulative dwell time at the relative position of the nozzle with respect to the substrate W. Therefore, the first convolutional neural network CNN1 extracts features using multiple filters, thereby extracting multiple features, including a time element, regarding changes in the relative position of the nozzle with respect to the substrate W. Note that, although an example is shown here in which the number of filters in the convolutional layer of the second layer L1b is set to twice the number of filters in the convolutional layer of the first layer L1a, it does not have to be twice as many. The number of filters in the convolutional layer of the second layer L1b only needs to be greater than the number of filters in the convolutional layer of the first layer L1a. Furthermore, the number of filters in the convolutional layer of the third layer L1c does not have to be twice as many as the number of filters in the convolutional layer of the second layer L1b. The number of filters in the convolutional layer of the third layer L1c only needs to be greater than the number of filters in the convolutional layer of the second layer L1b.
[0089] The first fully connected neural network NN1 has multiple layers. In the example of FIG. 12, the first fully connected neural network NN1 has two layers: a b1a layer on the input side and a b1b layer on the output side. In the example of FIG. 12, each layer includes multiple nodes. In the example of FIG. 12, five nodes are shown in the b1a layer and four nodes in the b1b layer, but the number of nodes is not limited to this. The number of nodes in the b1a layer is set to be equal to the sum of the number of nodes on the output side of the first convolutional neural network CNN1 and the number of fixed conditions. The number of nodes in the b1b layer is set to be equal to the number of nodes on the input side of the second convolutional neural network CNN2. The outputs of the nodes in the b1a layer are connected to the inputs of the nodes in the b1b layer. The parameters include coefficients that weight the outputs of the nodes in the b1a layer. One or more intermediate layers may be provided between the b1a layer and the b1b layer.
[0090] The second convolutional neural network CNN2 includes multiple layers. In this embodiment, the second convolutional neural network CNN2 includes three layers. In the second convolutional neural network CNN2, a fourth layer L1d, a fifth layer L1e, and a sixth layer L1f are provided in this order from the input side (upper layer side) to the output side (lower layer side). Note that, although this embodiment describes a case where the second convolutional neural network CNN2 includes three layers as the multiple layers, it may include three or more layers.
[0091] The fourth layer L1d, the fifth layer L1e, and the sixth layer L1f each include a convolutional layer and a pooling layer. Multiple filters are applied in the convolutional layer. The pooling layer compresses the output of the convolutional layer. The number of filters in the convolutional layer of the fifth layer L1e is set to half the number of filters in the convolutional layer of the fourth layer L1d. Furthermore, the number of filters in the convolutional layer of the sixth layer L1f is set to half the number of filters in the convolutional layer of the fifth layer L1e. This allows as many features as possible to be extracted from the throughput. The throughput is a value that differs depending on the radial position of the substrate W. The second convolutional neural network CNN2 uses multiple filters, and therefore extracts multiple features from the throughput, including an element of the radial position of the substrate W. Note that, although an example is shown here in which the number of filters in the fifth layer L1e convolutional layer is set to 1 / 2 the number of filters in the fourth layer L1d, it need not be 1 / 2. The number of filters in the fifth layer L1e convolutional layer may be any number that is smaller than the number of filters in the fourth layer L1d. Furthermore, the number of filters in the sixth layer L1f convolutional layer may not be 1 / 2 the number of filters in the fifth layer L1e. The number of filters in the sixth layer L1f convolutional layer may be any number that is smaller than the number of filters in the fifth layer L1e.
[0092] 13 is a flowchart showing an example of the flow of the prediction algorithm generation process. The prediction algorithm generation process is performed by the CPU 201 included in the prediction algorithm generation device 200 as the CPU 201 executes a prediction algorithm generation program stored in the RAM 202.
[0093] Referring to FIG. 13 , the CPU 201 included in the prediction algorithm generation device 200 executes a generation process condition generation process (step S01). Here, a plurality of generation process conditions for generating a prediction algorithm are generated. The generation process conditions include generation variable conditions and generation fixed conditions. The generation process condition generation process will be described in detail later. Subsequently, in step S02, the CPU 201 selects one of the plurality of generation process conditions generated in step S01 as a processing target, and proceeds to step S03. In step S03, a processing volume corresponding to the generation process condition of the processing target is acquired. In this embodiment, the generation process condition is output to the substrate processing apparatus 300, and a processing volume obtained as a result of the substrate processing apparatus 300 performing a coating process according to the generation process condition is acquired. In step S04, a data set is generated, and the process proceeds to step S05. A data set is generated, including the generation process condition selected as the processing target in step S02 and the processing volume acquired in step S03. In step S05, it is determined whether there is a next generation processing condition to be processed. It is determined whether there is a generation processing condition among the multiple generation processing conditions generated in step S01 that has not been selected as a processing condition in step S02. If such a generation processing condition exists, the process returns to step S02; otherwise, the process proceeds to step S06. By repeating steps S02 to S05, multiple sets of data sets are generated corresponding to the multiple generation processing conditions generated in step S01.
[0094] Next, in step S06, the CPU 201 generates a prediction algorithm using the multiple sets of data sets that have been generated. The data sets include generation processing conditions and processing volume. Here, the prediction algorithm is generated by performing machine learning on the multiple sets of data sets to generate a learning model. The CPU 201 performs machine learning using the generation processing conditions as explanatory variables and the processing volume as a target variable. The learned learning model is the prediction algorithm. The CPU 201 generates the prediction algorithm by performing machine learning on all of the multiple sets of data sets. Thereafter, in step S07, the CPU 201 transmits the generated prediction algorithm to the information processing device 100.
[0095] 14 is a flowchart showing an example of the flow of the generation process condition generation process. Referring to FIG. 14 , the CPU 201 included in the prediction algorithm generation device 200 determines the generation variable conditions (step S11). Here, a plurality of variable conditions scattered throughout the entire range of possible variable conditions are determined as the generation variable conditions. In the next step S12, the CPU 201 converts the determined generation variable conditions into conversion data. Here, the conversion data is a cumulative residence time distribution. The cumulative residence time distribution is determined by converting the variable conditions into the residence time of the nozzle in each of a plurality of divided regions obtained by dividing the top surface of the substrate W into a plurality of regions, and accumulating the residence times sequentially from the center of the substrate W toward the periphery.
[0096] In step S13, the CPU 201 classifies the generation variable conditions into multiple number-based groups. Here, the multiple generation variable conditions are classified into multiple number-based groups based on the number of times the nozzle 311 moves back and forth. In step S14, the CPU 201 selects one number-based group and clusters the multiple generation variable conditions included in the selected number-based group using the conversion data. When the multiple generation variable conditions are clustered using the conversion data, one or more clusters are identified. Each of the one or more clusters includes multiple generation variable conditions. Next, the CPU 201 selects a cluster to be processed from the multiple clusters identified by the clustering (step S15) and selects generation variable conditions from the selected cluster (step S16). In step S16, the CPU 201 randomly selects a predetermined number of generation variable conditions from the multiple generation variable conditions belonging to the cluster selected to be processed. The predetermined number is an integer greater than or equal to 1.
[0097] In step S17, the CPU 201 determines whether there is an unselected cluster among the one or more clusters identified by clustering. If there is an unselected cluster, the process returns to step S15. As a result, steps S15 and S16 are repeated for all of the one or more clusters identified by clustering. In this case, a predetermined number of generation variable conditions are selected from each of the one or more clusters identified by clustering in one frequency-based group, so that generation variable conditions with different cumulative stay time distribution characteristics are selected evenly. If it is determined in step S17 that there is no unselected cluster, the process proceeds to step S18.
[0098] In step S18, the CPU 201 determines whether or not there are any unselected frequency-based groups. If there are any unselected frequency-based groups, the process returns to step S14. As a result, the processes of steps S14 to S17 are executed for all of the multiple frequency-based groups. In this case, the generation processing conditions to be clustered are varied depending on the number of reciprocation movements of the nozzle 311, so this is effective when the processing volume varies depending on the number of reciprocation movements of the nozzle 311. If it is determined in step S18 that there are no unselected frequency-based groups, the process proceeds to step S19.
[0099] In step S19, the CPU 201 generates fixed conditions for generation, and proceeds to step S20. Here, multiple variable conditions scattered throughout the entire range of values that can be set for the fixed conditions are generated as variable conditions for generation. In step S20, the CPU 201 generates multiple process conditions for generation by combining the variable conditions for generation and the fixed conditions for generation.
[0100] 15 is a flowchart showing an example of the flow of processing condition determination processing. The processing condition determination processing is executed by CPU 101 of information processing device 100 as CPU 101 executes a processing condition determination program stored in RAM 102.
[0101] 15 , CPU 101 included in information processing device 100 selects one of a plurality of variable conditions prepared in advance (step S21), and proceeds to step S22. One of the variable conditions prepared in advance is selected using an experimental design method, a pairwise method, Bayesian estimation, or the like.
[0102] In step S22, a prediction algorithm is used to predict the throughput from the variable conditions and fixed conditions, and the process proceeds to step S23. The prediction algorithm receives converted data, obtained by converting the variable conditions into a cumulative residence time distribution, and the fixed conditions. The throughput output by the prediction algorithm is thus obtained. In step S23, the post-processing film thickness characteristic is compared with the target film thickness characteristic. The film thickness characteristic after processing the substrate W is calculated from the pre-processing film thickness characteristic of the substrate W to be processed by the substrate processing apparatus 300 and the etching profile predicted in step S22. The post-processing film thickness characteristic is then compared with the target film thickness characteristic. Here, the difference between the film thickness characteristic after processing the substrate W and the target film thickness characteristic is calculated.
[0103] In step S24, it is determined whether the comparison result satisfies the evaluation criterion. If the comparison result satisfies the evaluation criterion (YES in step S24), the process proceeds to step S25, but if not, the process returns to step S21. For example, it is determined that the evaluation criterion is met if the maximum value of the differences is equal to or less than a threshold value. Also, it is determined that the evaluation criterion is met if the average of the differences is equal to or less than a threshold value.
[0104] In step S25, processing conditions including the variable condition selected in step S21 are set as candidate processing conditions for driving the substrate processing apparatus 300, and the process proceeds to step S26. In step S26, it is determined whether an instruction to end the search has been accepted. If an instruction to end the search has been accepted by the user operating the information processing apparatus 100, the process proceeds to step S27; otherwise, the process returns to step S21. Note that instead of an instruction to end the search being accepted by the user, it may be determined whether a predetermined number of processing conditions have been set as candidates.
[0105] In step S27, one of the one or more candidate processing conditions is determined, and the process proceeds to step S28. The user operating the information processing apparatus 100 may select one of the one or more candidate processing conditions. This broadens the user's range of choices. Furthermore, a variable condition with the simplest nozzle operation may be automatically selected from among the variable conditions included in the plurality of processing conditions. The variable condition with the simplest nozzle operation may be, for example, a variable condition with the fewest speed change points. This allows a plurality of variable conditions to be presented for the throughput of complex nozzle operation for processing the substrate W. Selecting a variable condition with easy nozzle control from among the plurality of variable conditions facilitates control of the substrate processing apparatus 300.
[0106] In step S28, the processing conditions including the variable conditions determined in step S28 are transmitted to the substrate processing apparatus 300, and the processing ends. The CPU 101 controls the input / output I / F 107 to transmit 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, the substrate processing apparatus 300 processes the substrate W in accordance with the processing conditions.
[0107] (4) Experimental Example In the above embodiment, after the variable conditions are converted into conversion data to generate a prediction algorithm, the variable conditions are roughly classified into groups by number of times, and clustering is performed within each group by number of times. Generally, in the coating process of the substrate processing apparatus 300, the behavior of the liquid film on the substrate W during the coating process is a factor that significantly affects the throughput. It is estimated that the behavior of the liquid film on the substrate W during the coating process is highly correlated with the operation of the nozzle 311 during the coating process. The inventor predicted that, among the operations of the nozzle 311, the number of reciprocating movements of the nozzle 311 significantly affects the throughput. To confirm this prediction, the inventor examined the relationship between the cumulative residence time distribution and the throughput using approximately 700 data sets (pairs of processing conditions and throughputs).
[0108] 16 and 17 show experimental results. In FIGS. 16 and 17, approximately 700 data sets are roughly categorized by the number of times the nozzle 311 made a reciprocation. FIG. 16 shows data sets for groups in which the nozzle 311 made 0 to 3 reciprocations, while FIG. 17 shows data sets for groups in which the nozzle 311 made 6 to 20 reciprocations. The upper sections of FIGS. 16 and 17 show cumulative residence time distributions included in clusters CL1 to CL4 identified by clustering using the k-means algorithm based on the cumulative residence time distributions after converting the variable conditions. In FIGS. 16 and 17, the converted data is shown in the upper section, and the throughput is shown in the lower section. The upper section also displays a reference line SL, which represents the cumulative residence time distribution obtained by converting the residence time of the nozzle 311 at each position in the radial direction of the substrate W from the same variable conditions. The vertical axis of the upper graph shows the cumulative residence time distribution of the nozzle 311 relative to the substrate center OP, normalized to 1. The horizontal axis of the upper graph indicates the radial distance from the substrate center OP, and the vertical axis of the lower graph indicates the processing amount, while the horizontal axis indicates the radial distance from the substrate center OP.
[0109] The upper part shows graphs corresponding to the four clusters determined by clustering. In each of the four graphs, it can be seen that multiple cumulative stay times fall within a predetermined range. The lower part shows graphs of the throughput corresponding to the four clusters. It can also be seen that multiple throughputs fall within a predetermined range in each of the four graphs shown in the lower part. As shown in FIGS. 16 and 17 , it was found that there is a correlation between the cumulative stay time distribution and the throughput.
[0110] On the other hand, as shown in FIG. 17 , when the number of reciprocation cycles of the nozzle 311 is six or more, the ranges of processing volume are similar among clusters CL1 to CL4. Therefore, when the number of reciprocation cycles of the nozzle 311 is six or more, the processing volume tends to be similar regardless of the variable conditions. These results indicate that when grouping by the number of reciprocation cycles of the nozzle 311, it is not necessary to further group and subdivide the variable conditions with six or more reciprocation cycles. For less than six reciprocation cycles, it is necessary to group the variable conditions to subdivide the possible range of the variable conditions. Furthermore, for variable conditions with six or more reciprocation cycles, clustering the variable conditions with conversion data generates a data set that shows little difference in processing volume even when the variable conditions are different. In this case, the learning model can be made to learn this tendency.
[0111] (5) Effects of the Embodiment According to the prediction algorithm generation device 200 of the above embodiment, multiple generation variable conditions are generated that are scattered throughout the entire range of possible variable conditions, the multiple generation variable conditions are clustered based on the conversion data, and at least one generation variable condition is selected from each of the multiple clusters. Therefore, multiple generation processing conditions that cover the entire range of possible generation processing conditions and are minimally biased are selected. Furthermore, a prediction algorithm is generated using a dataset that includes the selected generation processing conditions and the generation processing amount. This improves the generalizability of the prediction algorithm. Furthermore, since the number of datasets used to generate a prediction algorithm can be reduced, the cost of generating the dataset can be minimized. As a result, a highly generalizable prediction algorithm can be generated at low cost.
[0112] Furthermore, the prediction algorithm generation device 200 classifies multiple generation variable conditions into multiple number-based groups using the number of times the nozzle 311 moves while the processing liquid is supplied to the substrate during the coating process, and then performs cluster analysis on each number-based group. Since the generation process conditions are classified by the number of times the nozzle moves, which has a large impact on the processing volume, the multiple generation process conditions are classified into multiple groups that subdivide the entire range of possible generation process conditions. This further improves the generalizability of the prediction algorithm.
[0113] Furthermore, since the generation variable conditions are converted into a cumulative residence time obtained by accumulating the residence times of the nozzle 311 at a plurality of different positions in the radial direction of the substrate, it is possible to reduce the number of dimensions of the complex variable conditions, thereby making it possible to easily classify a plurality of processing conditions into a plurality of groups.
[0114] (6) Other Embodiments (6-1) In the above embodiment, an example is shown in which a learning model including a neural network as shown in FIG. 12 is used to generate a prediction algorithm, but the present invention is not limited to this. The learning model may be configured using other models including a neural network. Furthermore, a regression equation, for example, may be used as the prediction algorithm. In this case, the prediction algorithm may be generated by applying a plurality of data sets to a predetermined regression equation and performing a regression.
[0115] (6-2) In the above embodiment, an example is described in which the generation process volume used to generate the prediction algorithm is the process volume obtained by the substrate processing apparatus 300 performing a coating process under generation process conditions. However, the present invention is not limited to this. For example, if past data sets in which coating processes have been performed by the substrate processing apparatus 300 have been accumulated, a process volume corresponding to process conditions that are the same as or similar to the generation process conditions in the accumulated data sets may be used as the generation process volume. Furthermore, if past data sets in which coating processes have been performed by the substrate processing apparatus 300 have been accumulated, a learning model that has been machine-learned from the accumulated data sets may be used. In this case, the process volume predicted by the learning model from the generation process conditions may be used as the generation process volume.
[0116] (6-3) In the above embodiment, the relative position of the nozzle 311 with respect to the substrate W is used as the variable condition, and the physical quantity for which the variable condition is converted is the cumulative residence time. However, the present invention is not limited to this. The flow rate of the etching liquid discharged from the nozzle 311 may also be used as the variable condition. The flow rate is the amount of etching liquid discharged from the nozzle 311 per unit time. Therefore, in addition to the relative position of the nozzle 311 with respect to the substrate W, the flow rate of the etching liquid varies over time. In this case, the physical quantity for which the variable condition is converted can be the cumulative value of the amount of etching liquid supplied to each of the multiple divided regions of the substrate W. Furthermore, other conditions, such as the temperature of the etching liquid, may also be used as the variable condition as long as they change over time.
[0117] (6-4) In the above embodiment, an example was shown in which the information processing device 100, the prediction algorithm generation device 200, and the substrate processing device 300 were separate devices, but the present invention is not limited to this. At least two or more of the information processing device 100, the prediction algorithm generation device 200, and the substrate processing device 300 may be integrated into an apparatus. For example, the substrate processing device 300 may include the prediction algorithm generation device 200 and the information processing device 100.
[0118] (7) Summary of Embodiments (Section 1) A prediction algorithm generation device according to one aspect of the present invention includes: a processing condition generation unit that generates processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; a classification unit that classifies the processing conditions generated by the processing condition generation unit into a plurality of groups based on variables related to the variable conditions; a processing condition selection unit that selects at least one processing condition from each of the plurality of groups; a processing amount acquisition unit that acquires a processing amount indicating a difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process in accordance with the processing conditions selected by the processing condition selection unit; and a prediction algorithm generation unit that generates a prediction algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amounts acquired in relation to the processing conditions.
[0119] According to the prediction algorithm generation device described in paragraph 1, the processing conditions generated by the processing condition generation unit are classified into multiple groups based on variables related to the variable conditions, and at least one processing condition is selected from each of the multiple groups. As a result, when multiple processing conditions are generated, each of which includes multiple variable conditions scattered across the range of possible variable conditions, the multiple processing conditions scattered across the range of possible processing conditions are classified into multiple groups. Since at least one processing condition is selected from each of the multiple groups, multiple processing conditions are selected that cover the range of possible processing conditions and are less biased. Furthermore, a prediction algorithm is generated using a dataset including the selected processing conditions and processing volume. This improves the generalizability of the prediction algorithm. Furthermore, since the number of datasets used to generate the prediction algorithm can be reduced, the cost of generating the dataset can be minimized. As a result, a prediction algorithm generation device capable of generating highly generalizable prediction algorithms at low cost can be provided.
[0120] (2) The prediction algorithm generation device described in 1 may further include a number-based classification unit that classifies the plurality of processing conditions generated by the processing condition generation unit into a plurality of number-based groups based on the number of times the nozzle moves back and forth included in the processing conditions, and the classification unit may classify the plurality of processing conditions included in each of the plurality of number-based groups into the plurality of groups.
[0121] According to the prediction algorithm generation device described in paragraph 2, processing conditions are divided into groups based on the number of times the nozzle moves while supplying processing liquid to the substrate during coating processing. Furthermore, processing conditions within each group are classified into multiple groups. Because processing conditions are classified based on the number of times the nozzle moves, which has a large impact on the processing volume, multiple processing conditions are classified into multiple groups that subdivide the entire range of possible processing conditions. This further improves the generalizability of the prediction algorithm.
[0122] (Clause 3) The prediction algorithm generation device described in clause 1 or 2 further includes a conversion unit that converts the variable conditions into physical quantities at each of a plurality of different positions in the radial direction of the substrate, and the variables may be physical quantities at the relative positions of the nozzle with respect to the substrate converted by the conversion unit.
[0123] According to the prediction algorithm generation device described in paragraph 3, the variable conditions are converted into physical quantities at each of a plurality of different positions in the radial direction of the substrate, which makes it possible to reduce the number of dimensions of the complex variable conditions, thereby making it possible to easily classify a plurality of processing conditions into a plurality of groups.
[0124] (4) In the prediction algorithm generating device described in the 3rd paragraph, the physical quantity may be a residence time of the nozzle.
[0125] According to the prediction algorithm generating device described in Section 4, since the physical quantity is the residence time for which a substance resides at any position on the substrate, it is easy to convert the variable condition into the physical quantity.
[0126] (Item 5) In the prediction algorithm generating device described in item 3, the physical quantity may be an accumulated residence time obtained by accumulating the time the nozzle stays from the center of the substrate toward the outer periphery.
[0127] According to the prediction algorithm generation device described in paragraph 5, the accumulated residence time is the accumulated time that the nozzle stays from the center of the substrate toward the outer periphery, so that it is easy to convert the variable condition into a physical quantity.
[0128] (6) In the prediction algorithm generation device described in any one of paragraphs 1 to 5, the processing amount acquisition unit may provide the processing conditions selected by the processing condition selection unit to the substrate processing apparatus, and acquire the processing amount obtained when the substrate processing apparatus performs the coating process in accordance with the processing conditions.
[0129] According to the prediction algorithm generating device described in paragraph 6, the coating process is performed in the substrate processing apparatus according to the processing conditions selected by the processing condition selection unit, so that an accurate processing amount can be obtained.
[0130] (Item 7) In the prediction algorithm generating device described in any one of Items 1 to 5, the processing amount acquisition unit may provide the processing conditions selected by the processing condition selection unit to a learning model that has been machine-learned from a data set of the processing amounts and the processing conditions at which the substrate processing apparatus performs the coating process according to the processing conditions, and acquire the output of the learning model as the processing amount.
[0131] According to the prediction algorithm generation device described in paragraph 7, it is not necessary to actually operate the substrate processing apparatus in order to generate the data set, which makes it possible to reduce the cost of generating the data set.
[0132] (Item 8) The prediction algorithm generation device described in any one of Items 1 to 5 may further include an experimental dataset acquisition unit that acquires an experimental dataset including the processing conditions and the processing amount obtained when the substrate processing apparatus executes the coating process in accordance with the processing conditions, and the processing amount acquisition unit may extract the experimental dataset including the processing conditions selected by the processing condition selection unit from the plurality of experimental datasets acquired by the experimental dataset acquisition unit.
[0133] According to the prediction algorithm generating device described in Section 8, it is possible to obtain the processing amount corresponding to the processing conditions selected by the processing condition selecting unit using an experimental data set obtained from a coating process previously performed by the substrate processing apparatus, thereby reducing the cost of generating the data set.
[0134] (Item 9) A prediction algorithm according to another aspect of the present invention may be generated by the prediction algorithm generation device according to any one of items 1 to 8.
[0135] According to the prediction algorithm described in Section 9, it is possible to provide a prediction algorithm with high generalizability.
[0136] (Item 10) Another aspect of the present invention is an information processing apparatus that manages a substrate processing apparatus using the prediction algorithm described in item 9, and includes a processing condition determination unit that provides tentative processing conditions to the prediction algorithm and, if the processing volume predicted by the prediction algorithm satisfies an allowable condition, determines the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0137] According to the information processing apparatus described in paragraph 10, when tentative processing conditions are applied to a prediction algorithm and the throughput predicted by the prediction algorithm satisfies the tolerance condition, the tentative processing conditions are determined as the processing conditions for driving the substrate processing apparatus. Therefore, multiple tentative processing conditions can be determined for the throughput that satisfies the tolerance condition. As a result, multiple processing conditions can be presented for the throughput obtained through a complex process.
[0138] (Item 11) A substrate processing apparatus according to another aspect of the present invention includes the information processing apparatus according to item 10.
[0139] According to the substrate processing apparatus described in the eleventh aspect, it is possible to present a plurality of processing conditions for the throughput of a complex process for processing substrates.
[0140] (Item 12) A method for generating a prediction algorithm according to another aspect of the present invention comprises the steps of: generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; classifying the processing conditions generated by the processing condition generating step into multiple groups based on variables related to the variable conditions; selecting at least one processing condition from each of the multiple groups; acquiring a processing amount indicating a difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process according to the processing conditions selected by the processing condition selecting step; and generating a prediction algorithm that predicts the processing amount from the processing conditions using a dataset including the processing conditions and the processing amount acquired in relation to the processing conditions.
[0141] According to the prediction algorithm generation method described in paragraph 12, processing conditions generated by a processing condition generation unit are classified into multiple groups based on variables related to variable conditions, and at least one processing condition is selected from each of the multiple groups. As a result, when multiple processing conditions are generated, each of which includes multiple variable conditions scattered across the range of possible variable conditions, the multiple processing conditions scattered across the range of possible processing conditions are classified into multiple groups. Since at least one processing condition is selected from each of the multiple groups, multiple processing conditions are selected that cover the range of possible processing conditions and are less biased. Furthermore, a prediction algorithm is generated using a dataset including the selected processing conditions and processing volume. This improves the generalizability of the prediction algorithm. Furthermore, since the number of datasets used to generate the prediction algorithm can be reduced, the cost of generating the dataset can be minimized.
[0142] (Item 13) A prediction algorithm generation program according to another aspect of the present invention causes a computer to perform the following steps: generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; classifying the multiple processing conditions generated by the processing for generating processing conditions into multiple groups based on variables related to the variable conditions; selecting at least one processing condition from each of the multiple groups; acquiring a processing amount indicating a difference in film thickness of the coating before and after the substrate processing apparatus performs the coating process in accordance with the processing conditions selected by the processing for selecting processing conditions; and generating a prediction algorithm that predicts the processing amount from the processing conditions using a dataset including the processing conditions and the processing amount acquired in relation to the processing conditions.
[0143] According to the prediction algorithm generation program described in paragraph 13, processing conditions generated by a processing condition generation unit are classified into multiple groups based on variables related to variable conditions, and at least one processing condition is selected from each of the multiple groups. As a result, when multiple processing conditions are generated, each of which includes multiple variable conditions scattered across a range of possible variable conditions, the multiple processing conditions scattered across the range of possible processing conditions are classified into multiple groups. Since at least one processing condition is selected from each of the multiple groups, multiple processing conditions are selected that cover the range of possible processing conditions and are less biased. Furthermore, a prediction algorithm is generated using a dataset including the selected processing conditions and processing volume. This improves the generalizability of the prediction algorithm. Furthermore, since the number of datasets used to generate the prediction algorithm can be reduced, the cost of generating the dataset can be minimized.
[0144] (Item 14) A processing condition determination method according to another aspect of the present invention is a processing condition determination method for managing a substrate processing apparatus using the prediction algorithm described in item 9, comprising the step of providing tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0145] According to the processing condition determination method described in paragraph 14, it is possible to provide a processing condition determination method capable of presenting a plurality of processing conditions for the throughput of a complex process.
[0146] (Item 15) A processing condition determination program according to another aspect of the present invention is a processing condition determination program for managing a substrate processing apparatus using the prediction algorithm described in item 9, which causes a computer to execute a process of providing tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as processing conditions for driving the substrate processing apparatus.
[0147] According to the processing condition determination program described in paragraph 15, it is possible to provide a processing condition determination program that can present a plurality of processing conditions for the throughput of a complex process.
Claims
1. A prediction algorithm generation device comprising: a processing condition generation unit that generates processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; a classification unit that classifies the processing conditions generated by the processing condition generation unit into a plurality of groups based on variables related to the variable conditions; a processing condition selection unit that selects at least one of the processing conditions from each of the plurality of groups; a processing amount acquisition unit that acquires a processing amount indicating a difference in thickness of the coating before and after the substrate processing apparatus performs the coating process in accordance with the processing conditions selected by the processing condition selection unit; and a prediction algorithm generation unit that generates a prediction algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amounts acquired in relation to the processing conditions.
2. A predictive algorithm generating device as described in claim 1, further comprising a number-based classification unit that classifies the plurality of processing conditions generated by the processing condition generating unit into a plurality of number-based groups based on the number of times the nozzle is moved back and forth included in the processing conditions, wherein the classification unit classifies the plurality of processing conditions included in each of the plurality of number-based groups into the plurality of groups.
3. A prediction algorithm generating device as claimed in claim 1 or 2, further comprising a conversion unit that converts the variable conditions into physical quantities at each of a plurality of different positions in the radial direction of the substrate, the variables being physical quantities at the relative positions of the nozzle with respect to the substrate converted by the conversion unit.
4. The prediction algorithm generating device according to claim 3, wherein the physical quantity is a residence time of the nozzle.
5. A prediction algorithm generating device according to claim 3, wherein said physical quantity is an accumulated residence time obtained by accumulating the residence time of said nozzle from the center of the substrate toward the outer periphery.
6. A prediction algorithm generating device according to any one of claims 1 to 5 or 2, wherein the processing amount acquisition unit provides the processing conditions selected by the processing condition selection unit to the substrate processing apparatus, and acquires the processing amount obtained when the substrate processing apparatus executes the coating process in accordance with the processing conditions.
7. A prediction algorithm generating device according to any one of claims 1 to 5 or 2, wherein the processing amount acquisition unit provides the processing conditions selected by the processing condition selection unit to a learning model that has been machine-learned from a data set of the processing amount and the processing conditions when the substrate processing apparatus performs the coating process according to the processing conditions, and acquires an output of the learning model as the processing amount.
8. A prediction algorithm generating device as claimed in any one of claims 1 to 5 or 2, further comprising an experimental dataset acquisition unit that acquires an experimental dataset including the processing conditions and the processing amount obtained when the substrate processing apparatus executes the coating process in accordance with the processing conditions, wherein the processing amount acquisition unit extracts the experimental dataset including the processing conditions selected by the processing condition selection unit from among the multiple experimental datasets acquired by the experimental dataset acquisition unit.
9. A prediction algorithm generated by the prediction algorithm generation device according to any one of claims 1 to 8 or 2.
10. An information processing device that manages a substrate processing apparatus using the prediction algorithm of claim 9, comprising a processing condition determination unit that provides tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determines the tentative processing conditions as the processing conditions for driving the substrate processing apparatus.
11. A substrate processing apparatus comprising the information processing apparatus according to claim 10.
12. A method for generating a prediction algorithm comprising: a step of generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; a step of classifying the plurality of processing conditions generated by the step of generating processing conditions into a plurality of groups based on variables related to the variable conditions; a step of selecting at least one of the processing conditions from each of the plurality of groups; a step of acquiring a processing amount indicating a difference in thickness of the coating before and after the substrate processing apparatus performs the coating process in accordance with the processing conditions selected by the step of selecting the processing conditions; and a step of generating a prediction algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amount acquired in relation to the processing conditions.
13. A prediction algorithm generation program for causing a computer to execute the following processes: generating processing conditions for a substrate processing apparatus to perform a coating process in which a nozzle that ejects a processing liquid is moved over a substrate on which a coating is formed to process the coating; the processing conditions include variable conditions that vary over time; classifying a plurality of the processing conditions generated by the processing for generating processing conditions into a plurality of groups based on variables related to the variable conditions; selecting at least one of the processing conditions from each of the plurality of groups; acquiring a processing amount indicating a difference in thickness of the coating before and after the substrate processing apparatus executes the coating process in accordance with the processing conditions selected by the processing for selecting processing conditions; and generating a prediction algorithm that predicts the processing amount from the processing conditions using a data set including the processing conditions and the processing amount acquired in relation to the processing conditions.
14. A processing condition determination method for managing a substrate processing apparatus using the prediction algorithm of claim 9, comprising the steps of providing tentative processing conditions to the prediction algorithm, and if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as the processing conditions for driving the substrate processing apparatus.
15. A processing condition determination program for managing a substrate processing apparatus using the prediction algorithm of claim 9, the processing condition determination program causing a computer to execute a process of providing tentative processing conditions to the prediction algorithm and, if the processing amount predicted by the prediction algorithm satisfies an allowable condition, determining the tentative processing conditions as the processing conditions for driving the substrate processing apparatus.
Citation Information
Patent Citations
Substrate treatment method and substrate treatment apparatus
JP2020077735A
Substrate processing device, substrate processing method, learning data generating method, learning method, learning device, learned model generating method, and learned model
JP2021190541A
Substrate processing device and substrate processing method
JP2023042256A
Condition-setting assistance method for substrate processing, substrate processing system, storage medium, and learning model
WO2020105517A1
