Information processing method, information processing device, and computer program

Abstract

Provided are an information processing method, an information processing device, and a computer program for adjusting a target trajectory generation algorithm.　In this information processing method: a target trajectory generation process is performed in which a target trajectory generation algorithm is used to generate a target trajectory representing the target value of a variable controlled during operation of a substrate processing device; the target trajectory generation algorithm includes an adjustable parameter that is not changed during the target trajectory generation process; and a parameter adjustment process is performed in which an evaluation function for evaluating the result of the target trajectory generation process is used to adjust the adjustable parameter.

Description

Information processing method, information processing device, and computer program

[0001] The present invention relates to an information processing method, an information processing device, and a computer program.

[0002] In substrate processing, which involves etching, film formation, and other processes on substrates such as semiconductor wafers or glass substrates, it is necessary to control a large number of variables representing process conditions over time, such as the composition of gases used in the process, temperature, pressure, and applied voltage. A combination of target values ​​of multiple variables at multiple points in time during the operation of a substrate processing apparatus performing substrate processing is called a target trajectory. The target trajectory represents the target values ​​of variables controlled during operation of the substrate processing apparatus. In order to operate the substrate processing apparatus appropriately and obtain appropriate substrate processing results, the target trajectory must be properly determined. An appropriate target trajectory is generated by information processing. For example, information processing is performed using a model of the substrate processing apparatus, adjusting the target trajectory so that the substrate processing results obtained from the model are appropriate. Patent Document 1 discloses a technology for optimizing plant control parameters.

[0003] Japanese Patent Application Laid-Open No. 2019-145156

[0004] An algorithm used in information processing to generate an appropriate target trajectory is called a target trajectory generation algorithm. Depending on the target trajectory generation algorithm, desirable conditions may not be met, for example, the information processing may take too long. Therefore, it is necessary to appropriately adjust the target trajectory generation algorithm. However, it is difficult to appropriately adjust the target trajectory generation algorithm depending on the situation.

[0005] The present disclosure provides an information processing method, an information processing device, and a computer program for adjusting a target trajectory generation algorithm.

[0006] An information processing method according to one aspect of the present disclosure performs a target trajectory generation process to generate a target trajectory representing a target value of a variable controlled during operation of a substrate processing apparatus using a target trajectory generation algorithm, the target trajectory generation algorithm including adjustable parameters that are not changed during the target trajectory generation process, and performs a parameter adjustment process to adjust the adjustable parameters using an evaluation function for evaluating the results of the target trajectory generation process.

[0007] According to the present disclosure, it is possible to provide an information processing method, an information processing device, and a computer program for adjusting a target trajectory generation algorithm.

[0008] FIG. 1 is a conceptual diagram showing an example of the configuration of a substrate processing system. FIG. 2 is a block diagram showing an example of the internal configuration of an information processing apparatus according to embodiment 1. FIG. 3 is a conceptual diagram showing an example of the configuration of a target trajectory generation algorithm according to embodiment 1. FIG. 4 is a schematic diagram showing a first example of a model consisting of a plurality of models. FIG. 5 is a schematic diagram showing a second example of a model consisting of a plurality of models. FIG. 6 is a schematic diagram showing a third example of a model consisting of a plurality of models. FIG. 7 is a flowchart showing an example of the procedure of a parameter adjustment process executed by the information processing apparatus according to embodiment 1. FIG. 8 is a schematic diagram showing an example of an output of an adjustment history. FIG. 9 is a flowchart showing the procedure of a process executed by the substrate processing system to control the substrate processing apparatus. FIG. 10 is a conceptual diagram showing a first example of the configuration of a target trajectory generation algorithm according to embodiment 2. FIG. 11 is a flowchart showing an example of the procedure of a parameter adjustment process executed by the information processing apparatus according to embodiment 2. FIG. 12 is a conceptual diagram showing a second example of the configuration of the target trajectory generation algorithm according to embodiment 2. FIG. 13 is a block diagram showing an example of the internal configuration of an information processing apparatus according to embodiment 3. FIG. 14 is a flowchart showing an example of the procedure of a process executed by the information processing apparatus to generate a function form of the target trajectory generation algorithm.

[0009] The present disclosure will be described in detail below with reference to the drawings illustrating embodiments. <Embodiment 1> The manufacturing process of substrates such as semiconductor wafers, glass substrates, and flat panel substrates includes processes for etching or film formation on substrates. The processing of substrates is called substrate processing, and the device that performs the substrate processing is called a substrate processing apparatus. For example, a substrate processing apparatus includes a process chamber, and substrate processing such as etching is performed on a substrate placed in the process chamber. In substrate processing, a number of variables representing processing conditions, such as the composition of etching gas, temperature, pressure, and applied voltage, are controlled over time. A combination of target values ​​of multiple variables at multiple points during operation of the substrate processing apparatus is called a target trajectory. The target trajectory represents the target values ​​of variables controlled during operation of the substrate processing apparatus. The target trajectory includes a time series of target values ​​for each variable. In this embodiment, the target trajectory is generated by information processing.

[0010] 1 is a conceptual diagram showing an example configuration of a substrate processing system 100. The substrate processing system 100 includes a substrate processing apparatus 21, a control device 22 that controls the substrate processing apparatus 21, and an information processing apparatus 1 that performs processing to generate a target trajectory. The substrate processing apparatus 21 performs substrate processing on substrates such as semiconductor wafers, glass substrates, or flat panel substrates. For example, the substrate processing apparatus 21 includes a process chamber, and performs etching as the substrate processing. The control device 22 controls the operation of the substrate processing apparatus 21.

[0011] The information processing apparatus 1 executes an information processing method. The information processing apparatus 1 performs processing to generate an appropriate target trajectory. The control device 22 controls the substrate processing apparatus 21 using the target trajectory generated by the information processing apparatus 1. Furthermore, the information processing apparatus 1 performs processing to adjust a target trajectory generation algorithm for generating the target trajectory so that the target trajectory becomes more desirable. The target trajectory generation algorithm includes adjustable parameters that are not changed during the processing to generate the target trajectory. The information processing apparatus 1 adjusts the target trajectory generation algorithm by evaluating the generation result of the target trajectory using the target trajectory generation algorithm and adjusting the adjustable parameters according to the generation result of the target trajectory.

[0012] FIG. 2 is a block diagram showing an example of the internal configuration of an information processing device 1 according to the first embodiment. The information processing device 1 is configured using a computer such as a personal computer or a server device. The information processing device 1 includes a calculation unit 11, a memory 12, a storage unit 13, a reading unit 14, an operation unit 15, a display unit 16, and an input / output unit 17. The calculation unit 11 is configured using, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or a multi-core CPU. The calculation unit 11 may also be configured using a quantum computer. The memory 12 stores temporary data generated in conjunction with calculations. The memory 12 is, for example, a RAM (Random Access Memory). The storage unit 13 is non-volatile, for example, a hard disk or a non-volatile semiconductor memory. The reading unit 14 reads information from a recording medium 10 such as an optical disc or a portable memory.

[0013] The operation unit 15 receives input of information such as text by receiving operations from the user. The operation unit 15 is, for example, a keyboard, a pointing device, or a touch panel. The display unit 16 displays images. The display unit 16 is, for example, a liquid crystal display or an EL display (Electroluminescent Display). The operation unit 15 and the display unit 16 may be integrated. The input / output unit 17 inputs and outputs data. The input / output unit 17 is, for example, an input / output interface or a communication unit.

[0014] The calculation unit 11 causes the reading unit 14 to read a computer program (program product) 131 recorded on the recording medium 10, and stores the read computer program 131 in the storage unit 13. The calculation unit 11 executes processing to realize the functions of the information processing device 1 in accordance with the computer program 131. The computer program 131 may be stored in the storage unit 13 in advance, or may be downloaded from outside the information processing device 1. In this case, the information processing device 1 does not need to include the reading unit 14.

[0015] The computer program 131 can be deployed to run on a single computer, or on multiple computers located at one site or distributed across multiple sites and interconnected by a communications network. That is, the information processing device 1 may be configured with multiple computers, and the computer program 131 may be executed on multiple computers connected via a communications network. The information processing device 1 may be configured using a cloud server.

[0016] The processing of each step described below for executing the information processing method can be executed by multiple computers. The processing of each step can also be executed by different computers. Data used during the processing can be stored in multiple computers. The processing of each step can also be executed using a virtual machine. The processing of each step can be executed by multiple processing units. The processing of each step can also be executed by different processing units. For example, part of the processing can be executed by one computer, and another part of the processing can be executed by another computer. The information processing device 1 can be configured using a gate-type quantum computer, a quantum annealer, a classical annealer (CMOS annealer), an optical computer, or an analog computer.

[0017] The information processing device 1 includes a target trajectory generation algorithm 3 for generating a target trajectory. The information processing device 1 performs a process for generating a target trajectory (target trajectory generation process) using the target trajectory generation algorithm 3. A computer program and data for realizing the target trajectory generation algorithm 3 are stored in the storage unit 13. For example, the computer program for realizing the target trajectory generation algorithm 3 is included in a computer program 131.

[0018] 3 is a conceptual diagram showing an example of the configuration of the target trajectory generation algorithm 3 according to the first embodiment. An initial value of the target trajectory is input to the target trajectory generation algorithm 3, and a generated target trajectory is output. The target trajectory generation algorithm 3 includes a target trajectory adjustment process 31 that adjusts the target trajectory, an adaptation process 32 that adapts the target trajectory to specific constraints, a model 33 that calculates the operation of the substrate processing apparatus 21, an objective function 34 that calculates an index according to the calculation result of the model 33, and a search process 35 that changes the target trajectory with a specific probability. In the process using the target trajectory generation algorithm 3, the information processing apparatus 1 iterates the processes of the target trajectory adjustment process 31, the adaptation process 32, the model 33, the objective function 34, and the search process 35 to generate a more appropriate target trajectory from the initial value of the target trajectory.

[0019] The adaptation process 32 adjusts the target trajectory so that it conforms to specific constraints. The target trajectory is subject to constraints imposed on the operation of the substrate processing apparatus 21. For example, in a substrate processing apparatus 21 including a process chamber in which substrate processing is performed under reduced pressure, the door opening mechanism of the process chamber and the vacuum pump that depressurizes the interior of the process chamber cannot be turned on at the same time. Thus, there is a constraint that some variables representing processing conditions cannot simultaneously assume specific values. There is also a constraint that the values ​​of various variables, such as temperature, pressure, voltage, and power consumption, must fall within a specified range determined by the performance of the substrate processing apparatus 21. There are also constraints on the time resolution or quantization width of values. For example, even if the value of a variable is defined at a certain time interval, such as 100 ms, the substrate processing apparatus 21 handles the value of the variable at another time interval, such as every 250 ms. For example, even if the value of a variable is expressed in a certain quantization width increment, such as 0.1, the substrate processing apparatus 21 handles the value of the variable at another quantization width increment, such as 0.3. The target trajectory must comply with such constraints on the operation of the substrate processing apparatus 21 .

[0020] Constraints on the target trajectory may be imposed by customer requirements. For example, for safety reasons, the customer may request that the value of a variable, such as temperature or pressure, be limited to a value that is somewhat lower than the upper limit value in terms of the performance of the substrate processing apparatus 21. Constraints on the target trajectory may also be imposed by the performance of the control device 22. For example, there is a constraint that the rate of change of a variable with respect to time when the variable, such as temperature or pressure, is changed by the control of the control device 22 must fall within a specified range determined by the performance of the control device 22.

[0021] In the adaptation process 32, the target trajectory is adjusted so that it conforms to various constraints. For example, among multiple variables that cannot simultaneously be set to specific values, the target values ​​of multiple variables are modified so that one variable becomes a specific value and the other variables become other values. For example, the target values ​​of the variables are modified so that the values ​​of each variable fall within a specified range. For example, the target trajectory is modified so that the time resolution or the quantization width of the values ​​conforms to the constraints. For example, the target values ​​of the variables are modified so that the values ​​of each variable conform to customer requirements. For example, the time series of the target values ​​of the variables is modified so that the rate of change of the variables over time falls within a specified range. The adaptation process 32 is realized, for example, by a function.

[0022] The search process 35 is a process that significantly changes the target trajectory. When optimizing a target trajectory through iterative calculation, the target trajectory may reach a local optimum solution, resulting in the calculation ending and a global optimum solution not being obtained. The search process 35 can significantly change the target trajectory. Here, a significant change refers to a change that is larger than the change in the target trajectory when the target trajectory is adjusted by the target trajectory adjustment process 31. For example, the search process 35 randomly generates a target trajectory that is significantly different from the previous target trajectory with a specific probability. The search process 35 is configured to reduce the probability of a significant change occurring. The search process 35 can generate a target trajectory that is more appropriately adjusted than the local optimum solution. The search process 35 is realized, for example, by a function.

[0023] The adaptation process 32 and the search process 35 do not have to be included in the target trajectory generation algorithm 3. The target trajectory generation algorithm 3 does not have to include both the adaptation process 32 and the search process 35, and may include only one of the adaptation process 32 or the search process 35.

[0024] The model 33 calculates the operation of the substrate processing apparatus 21 according to the target trajectory. For example, the model 33 is a simulation model that simulates the operation of the substrate processing apparatus 21 according to the target trajectory based on physical principles and outputs the simulation results. For example, the model 33 is a trained model that has been trained in advance to output the operation results of the substrate processing apparatus 21 when the target trajectory is input. The trained model has been trained in advance using experimental results or simulation results using the substrate processing apparatus 21. For example, the model 33 is a regression model that outputs the operation results of the substrate processing apparatus 21 according to the target trajectory. The regression model has been generated in advance using experimental results or simulation results.

[0025] For example, model 33 is a table-based model that uses a table that records the operation results of substrate processing apparatus 21 in association with target trajectories, and reads out and outputs the operation results of substrate processing apparatus 21 corresponding to the target trajectories from the table. The table that records the operation results of substrate processing apparatus 21 in association with target trajectories is stored in storage unit 13. For example, model 33 is an inverse model that calculates a target trajectory that will result in a specific operation result of substrate processing apparatus 21.

[0026] The model 33 may be composed of multiple models. FIG. 4 is a schematic diagram showing a first example of the model 33 composed of multiple models. The model 33 includes multiple models such as model A and model B. For example, each model included in the model 33 calculates part of the operation of the substrate processing apparatus 21. For example, the model 33 includes a model that calculates the temperature of a processed substrate, a model that calculates the pressure of a gas used in substrate processing, and a model that calculates a voltage applied during substrate processing. Each model calculates part of the operation of the substrate processing apparatus 21, and a series of operations of the substrate processing apparatus 21 are calculated by the multiple models. The range of the series of operations can be set arbitrarily. By using multiple models, the operation of the substrate processing apparatus 21 can be calculated in detail.

[0027] FIG. 5 is a schematic diagram showing a second example of a model 33 consisting of multiple models. The multiple models included in the model 33 are divided into multiple groups. In the example shown in FIG. 5, multiple models such as model A and model B are included in one group, and multiple other models such as model C and model D are included in another group. One group may contain only a single model. The model 33 includes an interaction model that calculates interactions between multiple groups. The interaction model calculates interactions between multiple groups based on the calculation results of the models included in the multiple groups.

[0028] The interaction model is a model that calculates interactions between multiple functions of the substrate processing apparatus 21. For example, the substrate processing apparatus 21 includes multiple units, and the interactions between the units are calculated using the interaction model. The model 33 may output only the calculation results using the interaction model, or may output both the calculation results of models included in multiple groups and the calculation results using the interaction model. By using the interaction model, it is possible to calculate the operation of the substrate processing apparatus 21 in which multiple functions interact with each other.

[0029] FIG. 6 is a schematic diagram showing a third example of a model 33 composed of multiple models. The model 33 includes multiple groups that are divided into multiple groups, and has a hierarchical structure in which one group includes another group. In the example shown in FIG. 6, a group including models A and B, etc., is included in a group including the group, model C, and interaction model a, etc. The interaction model a calculates the interaction between the group including models A and B, etc., and other models such as model C. Furthermore, the model 33 includes other models such as model D, which is included in a group separate from the group including interaction model a, and includes an interaction model b, which calculates the interaction between the group including interaction model a and models such as model D. In this way, the hierarchical structure of the model 33 makes it possible to calculate the operation of the substrate processing apparatus 21 when multiple functions possessed by the substrate processing apparatus 21 are hierarchically structured.

[0030] The objective function 34 is a function that calculates an index according to the calculation result of the model 33. The calculated index is an index for determining whether the target trajectory is appropriate. For example, the objective function 34 calculates an index that represents the quality of the results of the operation of the substrate processing apparatus 21 calculated by the model 33 according to the target trajectory. For example, the substrate processing apparatus 21 is an apparatus that etches substrates, and the objective function 34 calculates the uniformity of the etching rate within the substrate surface as an index.

[0031] The target trajectory adjustment process 31 is a process for adjusting the target trajectory in accordance with the index calculated by the objective function 34. The target trajectory adjustment process 31 uses a method for searching for a solution of a function, such as the gradient method or the quasi-Newton method, to adjust the target trajectory so that the index calculated by the objective function 34 becomes a more appropriate value. For example, the objective function 34 calculates the uniformity of the etching rate within the substrate surface as an index, and the target trajectory adjustment process 31 adjusts the target trajectory so that the difference in etching rate within the substrate surface becomes smaller.

[0032] The target trajectory generation algorithm 3 includes adjustable parameters. The adjustable parameters are parameters that are not changed during processing using the target trajectory generation algorithm 3, and are so-called hyperparameters. The target trajectory adjustment processing 31, the adaptation processing 32, the model 33, the objective function 34, and the search processing 35 use adjustable parameters. For example, in the target trajectory adjustment processing 31, the learning rate used in the gradient method or the parameters of the quasi-Newton method are adjustable parameters.

[0033] For example, in the adaptation process 32, information that determines which of multiple variables that cannot be simultaneously set to specific values ​​should be set to specific values ​​and which should be set to other values ​​is an adjustable parameter. For example, in the adaptation process 32, parameters that define the shape of a function that corrects the target values ​​of variables so that the values ​​of each variable fall within a specified range are adjustable parameters. For example, in the model 33, parameters such as coefficients used to calculate the operation of the target trajectory adjustment process 31 are adjustable parameters. For example, parameters such as coefficients included in the objective function 34 are adjustable parameters. In the search process 35, parameters that specify the probability of changing the target trajectory or the distribution of random numbers for randomly generating a changed target trajectory are adjustable parameters.

[0034] The target trajectory generation algorithm 3 is differentiable with respect to the adjustable parameters. The functions or equations used in the target trajectory adjustment process 31, the fitting process 32, the model 33, the objective function 34, and the search process 35 are differentiable with respect to the adjustable parameters.

[0035] The information processing device 1 is equipped with an evaluation function 4 for evaluating the result of the target trajectory generation process using the target trajectory generation algorithm 3. The evaluation function 4 calculates an evaluation value for evaluating the result of the target trajectory generation process. For example, the evaluation function 4 calculates, as the evaluation value, the calculation cost required to execute the target trajectory generation process. The calculation cost is the calculation time or memory usage, etc. For example, the evaluation function 4 calculates, as the evaluation value, the error between the target trajectory generated by the target trajectory generation process and the global optimum value of the target trajectory. For example, the evaluation function 4 calculates, as the evaluation value, the accuracy of adjustment of the target trajectory adjusted in the target trajectory generation process.

[0036] The information processing apparatus 1 performs a parameter adjustment process to adjust adjustable parameters of the target trajectory generation algorithm 3 in accordance with the status of the substrate processing apparatus 21. Fig. 7 is a flowchart showing an example of the procedure of the parameter adjustment process executed by the information processing apparatus 1 according to the first embodiment. Hereinafter, step will be abbreviated as S. The calculation unit 11 executes information processing in accordance with the computer program 131, whereby the information processing apparatus 1 executes the following processes.

[0037] The information processing apparatus 1 selects a model 33 to be used in the target trajectory generation process (S101). In S101, the calculation unit 11 selects the model 33 included in the target trajectory generation algorithm 3 depending on the state of the substrate processing apparatus 21 or the operating conditions of the substrate processing apparatus 21. The operation of the substrate processing apparatus 21 may change depending on the state of the substrate processing apparatus 21, such as the specifications, settings, maintenance status, or aging of the substrate processing apparatus 21. Furthermore, the operation of the substrate processing apparatus 21 may change depending on the operating conditions of the substrate processing apparatus 21. For example, a recipe that specifies the processing conditions for a substrate specifies the operating conditions of the substrate processing apparatus 21, and the operation of the substrate processing apparatus 21 may change depending on the recipe.

[0038] The storage unit 13 stores data for realizing a plurality of models 33 that differ in the operation of the substrate processing apparatus 21 to be calculated. A user operates the operation unit 15 to input an instruction to the information processing apparatus 1 to specify a model 33 from the plurality of models 33 that corresponds to the state of the substrate processing apparatus 21 or the operating conditions of the substrate processing apparatus 21. The calculation unit 11 selects a model 33 to be used in the target trajectory generation process from the plurality of models 33 in accordance with the input instruction. The calculation unit 11 may obtain information indicating the state of the substrate processing apparatus 21 or the operating conditions of the substrate processing apparatus 21 via the input / output unit 17, and select a model 33 based on the obtained information.

[0039] When the model 33 is composed of a plurality of models, the calculation unit 11 selects each of the plurality of models. At this time, the calculation unit 11 selects a model to be used in the target trajectory generation process from among the plurality of models that calculate the same part of the operation of the substrate processing apparatus 21. For example, the calculation unit 11 selects one model from the plurality of models that calculate the temperature, selects one model from the plurality of models that calculate the gas pressure, and selects one model from the plurality of models that calculate the applied voltage.

[0040] Models exist for each of multiple specifications or settings of the substrate processing apparatus 21, and a model 33 that matches the actual specifications of the substrate processing apparatus 21 is selected. Alternatively, models 33 exist for various operating conditions of the substrate processing apparatus 21, and a model 33 that matches the actual operating conditions is selected. For example, assume that there are models that can calculate the operation of the substrate processing apparatus 21 over a wide temperature range from low to high, and models that have high calculation accuracy at low temperatures but low calculation accuracy at high temperatures. If the recipe specifies that substrate processing is to be performed over a wide temperature range from low to high, a model 33 that can calculate the operation of the substrate processing apparatus 21 over a wide temperature range is selected. If the recipe specifies that substrate processing is to be performed mainly at low temperatures, a model 33 that has high calculation accuracy at low temperatures is selected. If there are models that have high calculation accuracy at low temperatures and models that have high calculation accuracy at high temperatures, a model 33 obtained by appropriately combining both models may be selected. By selecting a model 33 according to the state or operating conditions of the substrate processing apparatus 21, a model 33 that accurately calculates the operation of the substrate processing apparatus 21 is used.

[0041] In S101, if there is no model that matches the state or operating conditions of the substrate processing apparatus 21, the information processing apparatus 1 may perform a process of selecting an appropriate model 33 by interpolating existing models. For example, assume that a model for a high temperature such as 100°C and a model for a low temperature such as 50°C exist. If the temperature at which the substrate processing apparatus 21 operates is an intermediate temperature such as 70°C, the calculation unit 11 may generate a model to be applied to the intermediate temperature by linearly interpolating the high temperature model and the low temperature model, and select the generated model. Parameters to be interpolated when interpolating the model may be multidimensional.

[0042] The information processing apparatus 1 adjusts the adjustable parameters of the model 33 based on the operation data of the substrate processing apparatus 21 (S102). In S102, operation data representing the actual operation results of the substrate processing apparatus 21 is input to the input / output unit 17, and the calculation unit 11 adjusts the adjustable parameters of the model 33 based on the operation data. For example, the calculation unit 11 adjusts the adjustable parameters so that the operation results included in the operation data are obtained in the calculation of the operation of the substrate processing apparatus 21 using the model 33. The operation of the substrate processing apparatus 21 may change depending on the state of the substrate processing apparatus 21, such as aging. Since the adjustable parameters are adjusted based on the actual operation results of the substrate processing apparatus 21, the model 33 is adjusted so that calculation results corresponding to the actual operation results of the substrate processing apparatus 21 are obtained. Either S101 or S102 may be executed.

[0043] The information processing device 1 generates an initial value of the target trajectory (S103). For example, the initial value of the target trajectory is determined in advance, and the calculation unit 11 generates the initial value of the target trajectory by reading out a predetermined initial value stored in advance in the storage unit 13. The user may input the initial value of the target trajectory to the information processing device 1 by operating the operation unit 15, and the calculation unit 11 may use the input initial value.

[0044] The information processing device 1 performs the adaptation process 32 (S104). In S104, the calculation unit 11 executes the adaptation process 32, which adapts the target trajectory to specific constraints, as described above. If the adaptation process 32 is not included in the target trajectory generation algorithm 3, the information processing device 1 omits S104. The information processing device 1 then calculates the operation of the substrate processing device 21 using the model 33 (S105). In S105, the calculation unit 11 inputs the target trajectory adapted to the specific constraints by the adaptation process 32 to the model 33, and causes the model 33 to calculate the operation of the substrate processing device 21 according to the target trajectory. The calculation unit 11 acquires the calculation results by the model 33. The information processing device 1 then calculates an index according to the calculation results of the model 33 using the objective function 34 (S106). In S106, the calculation unit 11 inputs the calculation results by the model 33 to the objective function 34, and calculates an index using the objective function 34. In S106, the calculation unit 11 may use the objective function 34 to calculate a plurality of indexes.

[0045] The information processing device 1 then determines whether a specific condition is satisfied (S107). In S107, the calculation unit 11 determines whether a condition is satisfied that the value of the index calculated using the objective function 34 is within a predetermined range. For example, the calculation unit 11 determines whether a condition is satisfied that the value of the index is less than a predetermined threshold. The specific condition may be a condition that the amount of change in the index updated by repeatedly calculating it in S106 in the target trajectory generation process is less than a predetermined lower limit. The calculation unit 11 determines whether the condition is satisfied that the amount of change in the index is less than a predetermined lower limit. By determining whether these conditions are satisfied, the information processing device 1 searches for a target trajectory in which the index calculated by the objective function 34 becomes an extreme value.

[0046] The specific condition may be a condition that the number of repetitions of the target trajectory generation process reaches a predetermined upper limit. The calculation unit 11 determines whether the condition that the number of repetitions of the target trajectory generation process reaches a predetermined upper limit is satisfied. In S107, the calculation unit 11 may use multiple conditions. For example, multiple conditions using the multiple indexes calculated in S106 may be used. The calculation unit 11 may determine that the condition is satisfied when at least one or a predetermined number of the multiple conditions are satisfied, or may determine that the condition is satisfied when all of the multiple conditions are satisfied.

[0047] If the specific condition is not satisfied (S107: NO), the information processing device 1 adjusts the target trajectory (S108). The processes of S107 to S108 correspond to the target trajectory adjustment process 31. In S108, the calculation unit 11 adjusts the target trajectory according to the index calculated using the objective function 34 in S106. For example, the calculation unit 11 adjusts the target trajectory using a method for searching for a solution of a function, such as a gradient method or a quasi-Newton method, so that the index calculated by the objective function 34 approaches an extreme value. The calculation unit 11 may adjust the target trajectory based on a specific rule that does not use an index, such as setting a new target trajectory to the average of multiple target trajectories calculated in the past.

[0048] The information processing device 1 then performs search processing 35 (S109). In S109, the calculation unit 11 changes the target trajectory with a specific probability, as described above. The calculation unit 11 may perform processing to change the target trajectory more significantly than the change in the target trajectory in S108 with a certain probability. The calculation unit 11 may perform the processing of S109 so that the larger the change in the target trajectory, the lower the probability of the target trajectory changing, and the smaller the change in the target trajectory, the higher the probability of the target trajectory changing. If the search processing 35 is not included in the target trajectory generation algorithm 3, the information processing device 1 omits S109.

[0049] After S109, the information processing device 1 returns the process to S104. In S104, the information processing device 1 performs adaptation processing 32 on the target trajectory adjusted in S108 or the target trajectory changed in S109. Note that the processes of S107 to S109 may be executed in other orders. That is, the processes may be executed in the order of S107, S109, S108, S107, S109, S108, S109, S107, S109, or S109, S108, S107, S108, or S109, S108, S107. If S108 is executed last, the information processing device 1 returns the process to S104 after S108. If S107 is executed last, the information processing device 1 returns the process to S104 if a specific condition is not satisfied in S107.

[0050] The information processing device 1 repeats the processes of S104 to S109 until a specific condition is satisfied in S107. The processes of S104 to S109 correspond to the target trajectory generation process using the target trajectory generation algorithm 3. By repeating the processes of S104 to S109, the information processing device 1 generates a target trajectory that satisfies the condition in S107.

[0051] If the specific condition is satisfied in S107 (S107: YES), the information processing apparatus 1 calculates an evaluation value for evaluating the result of the target trajectory generation process using the evaluation function 4 (S110). In S110, the calculation unit 11 calculates the evaluation value using the evaluation function 4 based on the result of the target trajectory generation process. For example, the calculation unit 11 calculates, as the evaluation value, the calculation cost required to execute the target trajectory generation process, the error between the generated target trajectory and the global optimum value, a function in which the accuracy of adjustment of the target trajectory is a factor, or the value of the final objective function. The calculation unit 11 may calculate multiple evaluation values ​​using the evaluation function 4. Note that the evaluation value may include the actual processing result in the substrate processing apparatus 21 corresponding to the target trajectory, the simulation result corresponding to the target trajectory, or feedback from the user.

[0052] The information processing device 1 then determines whether a specific condition is satisfied (S111). In S111, the calculation unit 11 determines whether a condition is satisfied that the evaluation value calculated using the evaluation function 4 is within a predetermined range. For example, the calculation unit 11 determines whether a condition is satisfied that the calculation cost is less than a predetermined threshold. In S111, the calculation unit 11 may use multiple conditions using the multiple evaluation values ​​calculated in S110. The calculation unit 11 may determine that a condition is satisfied if at least one or a predetermined number of the multiple conditions are satisfied, or may determine that a condition is satisfied if all of the multiple conditions are satisfied. In S111, the information processing device 1 determines whether a desirable condition for the target trajectory generation algorithm 3 is satisfied, such as minimizing the calculation cost required for the target trajectory generation process.

[0053] If the specific condition is not satisfied (S111: NO), the information processing device 1 adjusts the adjustable parameters of the target trajectory generation algorithm 3 (S112). In S112, the calculation unit 11 adjusts the adjustable parameters according to the evaluation value calculated using the evaluation function 4 in S110. For example, the calculation unit 11 adjusts the adjustable parameters using a method for searching for a solution of a function, such as a gradient method or a quasi-Newton method. Since the target trajectory generation algorithm 3 is differentiable with respect to the adjustable parameters, the calculation unit 11 differentiates the target trajectory generation algorithm 3 with respect to the adjustable parameters. Alternatively, the calculation unit 11 performs higher-order differentiation, such as second-order differentiation, on the target trajectory generation algorithm 3. The calculation unit 11 adjusts the adjustable parameters using the gradient method using the result of the differentiation, or the quasi-Newton method using the result of the higher-order differentiation, so that the evaluation value calculated by the evaluation function 4 approaches an extreme value.

[0054] The calculation unit 11 may adjust the adjustable parameters using a method other than the gradient method or the quasi-Newton method. The calculation unit 11 may adjust the adjustable parameters without using an evaluation value, for example, by randomly changing the adjustable parameters. In an information processing device 1 that does not use differentiation or higher-order differentiation of the target trajectory generation algorithm 3 in S112, the target trajectory generation algorithm 3 may not be differentiable. The adjustable parameters adjusted in S112 may include an initial value of the target trajectory.

[0055] After S112, the information processing device 1 returns the process to S103. From S103 onwards, the information processing device 1 repeats the target trajectory generation process again. The information processing device 1 repeats the processes of S103 to S112 until a specific condition is met in S111. Repeating the processes of S103 to S112 means repeating the target trajectory generation process while changing the adjustable parameters, which is equivalent to adjusting the adjustable parameters using deep evolution. By performing deep evolution, the information processing device 1 can appropriately adjust the target trajectory generation algorithm 3 so that desirable conditions, such as minimizing the calculation cost, are met.

[0056] The information processing device 1 may adjust the adjustable parameters using a method other than the method using deep evolution. For example, the information processing device 1 may adjust the adjustable parameters using Bayesian optimization, an evolutionary algorithm, swarm intelligence, or a local search method. Even when these methods are used, the information processing device 1 can appropriately adjust the target trajectory generation algorithm 3.

[0057] If a specific condition is satisfied in S111 (S111: YES), the information processing device 1 stores the adjusted adjustable parameters and the adjustment history (S113). In S113, the calculation unit 11 stores the adjusted adjustable parameters and the adjustment history in the storage unit 13. The calculation unit 11 may store the adjusted adjustable parameters in addition to past adjustable parameters that have already been stored. The adjustment history associates the time when the parameter adjustment process was performed with the adjustment content of the adjustable parameters. For example, the stored time includes the year, month, date, and time.

[0058] The information processing device 1 outputs the adjustment history of the adjustable parameters (S114). In S114, the calculation unit 11 generates an image showing the contents of the adjustment history stored in the storage unit 13 and displays the generated image on the display unit 16. FIG. 8 is a schematic diagram showing an example of an output of the adjustment history. Whether or not the adjustable parameters in the target trajectory adjustment process 31, the adaptation process 32, the objective function 34, and the search process 35 have been adjusted is output in association with the date and time indicating when the parameter adjustment process was performed. In addition, for each of the multiple models that should constitute the model 33, whether or not it has been selected and whether or not the adjustable parameters have been adjusted are output.

[0059] In S114, the information processing device 1 may output an adjustment history including the values ​​of each adjustable parameter. The information processing device 1 may output an adjustment history that graphically represents changes in each adjustable parameter. The user can check the output adjustment history and learn how the adjustable parameters were adjusted.

[0060] After S114 is completed, the information processing apparatus 1 terminates the parameter adjustment process. The information processing apparatus 1 executes the processes of S101 to S114 as needed. For example, the information processing apparatus 1 executes the processes of S101 to S114 periodically. The information processing apparatus 1 may execute the processes of S101 to S114 whenever the state or operating conditions of the substrate processing apparatus 21 are changed, such as when maintenance is performed on the substrate processing apparatus 21, when the settings of the substrate processing apparatus 21 are changed, or when the recipe is changed. The process of S114 may be executed separately from the parameter adjustment process.

[0061] The information processing device 1 performs a target trajectory generation process using the adjusted adjustable parameters as necessary, and the control device 22 controls the substrate processing device 21 using the generated target trajectory. Figure 9 is a flowchart showing the procedure of the process executed by the substrate processing system 100 to control the substrate processing device 21. The information processing device 1 selects a model 33 to be used in the target trajectory generation process (S21). In S21, the calculation unit 11 selects the model 33 depending on the state of the substrate processing device 21 or the operating conditions of the substrate processing device 21.

[0062] The information processing apparatus 1 then sets adjustable parameters of the target trajectory generation algorithm 3 (S22). In S22, the calculation unit 11 sets the adjusted adjustable parameters stored in the storage unit 13 as adjustable parameters of the target trajectory generation algorithm 3. The storage unit 13 may store a plurality of adjustable parameters adjusted according to the state or operating conditions of the substrate processing apparatus 21. The calculation unit 11 may select an adjustable parameter according to the state or operating conditions of the substrate processing apparatus 21 from the plurality of adjustable parameters, and set the selected adjustable parameter as an adjustable parameter of the target trajectory generation algorithm 3.

[0063] The information processing apparatus 1 then performs a target trajectory generation process using the target trajectory generation algorithm 3 including the set adjustable parameters (S23). In S23, the calculation unit 11 executes the same processes as S103 to S109. An appropriate target trajectory is generated by the process of S23. The information processing apparatus 1 outputs the generated target trajectory, which is input to the control device 22. The control device 22 controls the substrate processing apparatus 21 using the target trajectory (S24). In S24, the control device 22 operates the substrate processing apparatus 21 and controls the substrate processing apparatus 21 so that each variable representing the processing condition of the substrate processing apparatus 21 matches the target value included in the target trajectory at each time point. After S24 is completed, the substrate processing system 100 terminates processing.

[0064] The processes of S21 to S24 are executed as needed. Once the target trajectory is generated, the process of S24 may be executed repeatedly using the generated target trajectory without executing the processes of S21 to S23 again. Since the substrate processing apparatus 21 is controlled using an appropriately determined target trajectory, the substrate processing apparatus 21 can be operated appropriately. As a result, appropriate substrate processing results can be obtained.

[0065] As described above in detail, in this embodiment, the information processing apparatus 1 generates a target trajectory using the target trajectory generation algorithm 3, and adjusts adjustable parameters of the target trajectory generation algorithm 3 using the evaluation function 4. Adjusting the adjustable parameters makes it possible to appropriately adjust the target trajectory generation algorithm 3 so as to satisfy desirable conditions, such as minimizing calculation costs. By using the adjusted target trajectory generation algorithm 3, the information processing apparatus 1 can generate an appropriate target trajectory under desirable conditions. By controlling the substrate processing apparatus 21 using the generated target trajectory, appropriate substrate processing results can be obtained.

[0066] <Embodiment 2> The configuration of a substrate processing system 100 according to Embodiment 2 is the same as that of Embodiment 1, and the configuration of an information processing apparatus 1 according to Embodiment 2 is the same as that of Embodiment 1. Fig. 10 is a conceptual diagram showing a first example of the configuration of a target trajectory generation algorithm 3 according to Embodiment 2. The target trajectory generation algorithm 3 has a hierarchical structure. The target trajectory generation algorithm 3 according to Embodiment 2 includes a modification process 301, an adaptation process 302, a plurality of lower-layer target trajectory generation algorithms 303, an objective function 304, and a search process 305. The adaptation process 302 or the search process 305 does not necessarily have to be included in the target trajectory generation algorithm 3. The modification process 301 adjusts the target trajectory and changes adjustable parameters of the lower-layer target trajectory generation algorithm 303. The objective function 304 corresponds to a second objective function.

[0067] In the target trajectory generation algorithm 3 according to the second embodiment, the model 33 in the target trajectory generation algorithm 3 according to the first embodiment is replaced with a plurality of models, each of which is replaced with a lower-layer target trajectory generation algorithm 303. The lower-layer target trajectory generation algorithm 303 includes a target trajectory adjustment process 311, an adaptation process 312, a model 313, an objective function 314, and a search process 315. The model 313 calculates part of the operation of the substrate processing apparatus 21. The objective function 314 corresponds to the first objective function. The adaptation process 312 or the search process 315 does not necessarily have to be included in the target trajectory generation algorithm 303. While FIG. 10 shows an example in which two target trajectory generation algorithms 303 are included in the target trajectory generation algorithm 3, the target trajectory generation algorithm 3 may include more than two target trajectory generation algorithms 303, or may include a single target trajectory generation algorithm 303.

[0068] 11 is a flowchart showing an example of the procedure of the parameter adjustment process executed by the information processing device 1 according to the second embodiment. The information processing device 1 selects a model 313 to be used in the target trajectory generation process (S301), similar to S101, and adjusts adjustable parameters of the model 313 based on the operation data of the substrate processing device 21, similar to S102 (S302). The information processing device 1 generates an initial value of the target trajectory (S303), similar to S103, and performs the adaptation process 32 (S304). If the adaptation process 302 is not included in the target trajectory generation algorithm 3, the information processing device 1 omits S304.

[0069] The information processing device 1 then performs a target trajectory generation process (S305). In S305, the calculation unit 11 executes the same processes as S104 to S109 using the lower-layer target trajectory generation algorithm 303. In the same process as S105, the calculation unit 11 calculates a part of the operation of the substrate processing device 21 using a model 313, and in the same process as S106, calculates a first index using an objective function 314. If the adaptation process 312 is not included in the target trajectory generation algorithm 303, the information processing device 1 omits the same process as S104. If the search process 315 is not included in the target trajectory generation algorithm 303, the information processing device 1 omits the same process as S109.

[0070] In S305, the calculation unit 11 performs a plurality of target trajectory generation processes using a plurality of target trajectory generation algorithms 303. The plurality of target trajectory generation processes may be executed sequentially or in parallel. The calculation unit 11 acquires the target trajectories generated using each of the target trajectory generation algorithms 303 and the calculation results of each of the models 313.

[0071] The information processing device 1 then calculates a second index according to the processing result of the target trajectory generation processing using the objective function 304 (S306). The second index is an index for adjusting adjustable parameters of the lower-layer target trajectory generation algorithm 303. The calculated second index may include an index for adjusting the target trajectory generated by the lower-layer target trajectory generation algorithm 303. The calculation unit 11 calculates the second index for each target trajectory generation algorithm 303.

[0072] The information processing device 1 then determines whether a specific condition is satisfied (S307). In S307, the calculation unit 11 determines whether a condition that the value of the second index is within a predetermined range is satisfied. The specific condition may be any other condition. If the specific condition is not satisfied (S307: NO), the information processing device 1 adjusts the adjustable parameters and target trajectory of the lower-layer target trajectory generation algorithm 303 (S308). The processes of S307 to S308 correspond to the change process 301. In S308, the calculation unit 11 adjusts the adjustable parameters and target trajectory of the lower-layer target trajectory generation algorithm 303 in accordance with the second index. The calculation unit 11 adjusts the adjustable parameters of multiple target trajectory generation algorithms 303. The calculation unit 11 may adjust the adjustable parameters and target trajectory of the lower-layer target trajectory generation algorithm 303 based on a specific rule that does not use the second index. The calculation unit 11 may adjust only the adjustable parameters of the lower-layer target trajectory generation algorithm 303.

[0073] The information processing device 1 then performs search processing 305 (S309), similar to S109. If search processing 305 is not included in the target trajectory generation algorithm 3, the information processing device 1 omits S309. After S309, the information processing device 1 returns the process to S304. Note that the processes of S307 to S309 may be executed in a different order. That is, the processes may be executed in the order of S307, S309, S308; S308, S307, S309; ​​S308, S309, S307; S309, S307, S308; or S309, S308, S307. If S308 is executed last, the information processing device 1 returns the process to S304 after S308. When S307 is executed last, the information processing device 1 returns the process to S304 if the specific condition is not satisfied in S307.

[0074] The information processing device 1 repeats the processes of S304 to S309 until a specific condition is satisfied in S307. By repeating the processes of S304 to S309, the information processing device 1 adjusts the adjustable parameters of the lower-layer target trajectory generation algorithm 303 so as to satisfy the condition in S307.

[0075] Repeating the processes of S304 to S309 means repeating the target trajectory generation process while changing the adjustable parameters of the lower-layer target trajectory generation algorithm 303, and is equivalent to adjusting the adjustable parameters of the lower-layer target trajectory generation algorithm 303 using deep development. By performing deep development, the information processing device 1 can appropriately adjust the lower-layer target trajectory generation algorithm 303 so that specific conditions are satisfied. The information processing device 1 may adjust the adjustable parameters of the lower-layer target trajectory generation algorithm 303 using a method other than the method using deep development.

[0076] If the specific condition is satisfied in S307 (S307: YES), the information processing device 1 calculates an evaluation value using the evaluation function 4 (S310) and determines whether the specific condition is satisfied (S311). If the specific condition is not satisfied (S311: NO), the information processing device 1 adjusts the adjustable parameters of the target trajectory generation algorithm 3 (S312), similar to S111. In S312, the calculation unit 11 adjusts the adjustable parameters included in the change process 301, the adaptation process 302, the objective function 304, and the search process 305. The calculation unit 11 may adjust the adjustable parameters of the target trajectory generation algorithm 303.

[0077] After S312, the information processing device 1 returns the process to S303. The information processing device 1 repeats the processes of S303 to S312 until a specific condition is satisfied in S311. Repeating the processes of S303 to S312 corresponds to adjusting the adjustable parameters of the target trajectory generation algorithm 3 having a hierarchical structure using deep development. In this way, the information processing device 1 can appropriately adjust the target trajectory generation algorithm 3 having a hierarchical structure. The information processing device 1 may adjust the adjustable parameters using a method other than the method using deep development.

[0078] If the specific condition is met in S311 (S311: YES), the information processing device 1 stores the adjusted adjustable parameters and the adjustment history (S313), similar to S113, and outputs the adjustment history of the adjustable parameters (S314), similar to S114. After S314 ends, the information processing device 1 ends the parameter adjustment process. The information processing device 1 executes the processes of S301 to S314 as needed. The process of S314 may be executed separately from the parameter adjustment process.

[0079] In the above description, an example has been shown in which the target trajectory generation algorithm 3 has a two-layer hierarchical structure. However, the target trajectory generation algorithm 3 may have a hierarchical structure with more than two layers. Even when the target trajectory generation algorithm 3 has a hierarchical structure with more than two layers, the information processing device 1 can adjust the adjustable parameters of the target trajectory generation algorithm 3 having a hierarchical structure by using multiple deep layer expansions. The information processing device 1 may also adjust the adjustable parameters using a method other than the method using deep layer expansion. For example, the information processing device 1 adjusts the adjustable parameters included in the target trajectory generation algorithm of each layer so that the target trajectory generation algorithms of multiple layers other than the lowest layer are Pareto optimal. Even when this method is used, the information processing device 1 can appropriately adjust the target trajectory generation algorithm 3.

[0080] 12 is a conceptual diagram showing a second example of the configuration of the target trajectory generation algorithm 3 according to the second embodiment. The target trajectory generation algorithm 3 has a hierarchical structure. The target trajectory generation algorithm 3 according to the second embodiment includes a change process 301, an adaptation process 302, an objective function 304, a search process 305, and a lower-layer target trajectory generation algorithm 303 and a model 306. The target trajectory generation algorithm 3 has a configuration in which some of the multiple models in the target trajectory generation algorithm 3 according to the first embodiment are replaced by the lower-layer target trajectory generation algorithm 303, and the other model 306 remains. The target trajectory generation algorithm 3 may include multiple target trajectory generation algorithms 303 or multiple models 306. The target trajectory generation algorithm 303 may have a hierarchical structure.

[0081] The information processing apparatus 1 executes the processes of S301 to S314 using the target trajectory generation algorithm 3 according to the second example. In S305, the information processing apparatus 1 performs the target trajectory generation process using the lower-layer target trajectory generation algorithm 303, and calculates the operation of the substrate processing apparatus 21 using the model 306. In S306, the information processing apparatus 1 calculates a second index using the objective function 304 in accordance with the result of the target trajectory generation process using the target trajectory generation algorithm 303 and the calculation result using the model 306. In the second example as well, the information processing apparatus 1 performs the parameter adjustment process through the processes of S301 to S314.

[0082] After adjusting the adjustable parameters of the target trajectory generation algorithm 3, the substrate processing system 100 performs the processes of S21 to S24 as necessary. That is, the information processing device 1 performs the target trajectory generation process using the adjusted adjustable parameters, and the control device 22 controls the substrate processing device 21 using the generated target trajectory. The substrate processing device 21 is controlled using the appropriately determined target trajectory, and appropriate substrate processing results are obtained.

[0083] In the second embodiment as well, the information processing apparatus 1 can appropriately adjust the target trajectory generation algorithm 3 by adjusting the adjustable parameters of the target trajectory generation algorithm 3. By using the adjusted target trajectory generation algorithm 3, the information processing apparatus 1 can generate an appropriate target trajectory under desirable conditions. By controlling the substrate processing apparatus 21 using the generated target trajectory, appropriate substrate processing results can be obtained.

[0084] <Embodiment 3> Embodiment 3 illustrates an embodiment in which a target trajectory generation algorithm or model is generated using a language generation model. FIG. 13 is a block diagram showing an example of the internal configuration of an information processing device 1 according to embodiment 3. The configuration of the substrate processing system 100 other than the information processing device 1 is the same as that of embodiment 1. The information processing device 1 includes a language generation model 5. The language generation model 5 is a trained model that is trained to output a sentence corresponding to the input prompt and information when a prompt consisting of a command statement requesting a response and information accompanying the prompt are input. For example, the language generation model 5 is a large-scale language model. More specifically, the language generation model 5 is BERT, GPT-4, Bard, LLaMA, or the like. The language generation model 5 is trained in advance. The language generation model 5 may be trained in advance with knowledge related to the substrate processing device 21.

[0085] The language generation model 5 is realized by the calculation unit 11 executing information processing in accordance with the computer program 131. The storage unit 13 stores data for realizing the language generation model 5. The language generation model 5 may be configured using dedicated hardware other than the calculation unit 11 and the storage unit 13. The language generation model 5 may be realized using a quantum computer. Alternatively, the language generation model 5 may be provided outside the information processing device 1, and the information processing device 1 may execute processing using the external language generation model 5. The language generation model 5 may be realized using multiple computers connected via a communication network, or may be realized using the cloud. The configuration of the information processing device 1 other than the language generation model 5 is the same as in embodiment 1.

[0086] 14 is a flowchart showing an example of the processing procedure for generating a function form of the target trajectory generation algorithm 3, performed by the information processing device 1. The information processing device 1 acquires the specifications of the substrate processing device 21 (S41). For example, data representing the specifications of the substrate processing device 21 is input to the input / output unit 17, and the calculation unit 11 acquires the input data to acquire the specifications of the substrate processing device 21. For example, the specifications of the substrate processing device 21 are represented by a specification document, or experimental results or simulation results of the operation of the substrate processing device 21.

[0087] The information processing device 1 inputs a request for generating a function form of the target trajectory generation algorithm 3 to the language generation model 5 (S42). In S42, the calculation unit 11 generates a prompt including the specifications of the substrate processing device 21 and the request for generating a function form of the target trajectory generation algorithm 3, and inputs the generated prompt to the language generation model 5. The calculation unit 11 may generate the prompt using data input by the user operating the operation unit 15.

[0088] The function form of the target trajectory generation algorithm 3 is the function form of the function used by the target trajectory generation algorithm 3, such as the function used in the target trajectory adjustment process 31, the function used by the model 33 to calculate the operation of the substrate processing apparatus 21, and the objective function 34. For example, the prompt includes a request to generate the function form of the function used by the model 33 to calculate the operation of the substrate processing apparatus 21 in accordance with specifications. For example, the prompt includes a request to generate the function form of the objective function 34 that calculates a specific index such as the uniformity of the etching rate within the surface of a substrate processed by the substrate processing apparatus 21. For example, the prompt includes a request to generate the function form of a function used to adjust the target trajectory by a specific method such as a gradient method in the target trajectory adjustment process 31. The prompt may include a request to specify the function form to some extent, such as expressing the function form as a combination of linear functions.

[0089] The language generation model 5 performs calculations in response to the input of the prompt and outputs a function form of the target trajectory generation algorithm 3. The function form includes undetermined adjustable parameters. The language generation model 5 outputs a function form of the target trajectory generation algorithm 3 according to the specifications of the substrate processing apparatus 21.

[0090] The information processing device 1 acquires the function form of the target trajectory generation algorithm 3 (S43). In S43, the calculation unit 11 acquires the function form of the target trajectory generation algorithm 3 output by the language generation model 5. After S43 ends, the information processing device 1 ends the processing. After executing S41 to S43, the information processing device 1 generates the target trajectory generation algorithm 3 using the obtained function form of the target trajectory generation algorithm 3.

[0091] 15 is a flowchart showing an example of the procedure of a process performed by the information processing device 1 to generate a model 33. The information processing device 1 acquires specifications of the substrate processing device 21 (S51). The information processing device 1 inputs a request for generating a model 33 to the language generation model 5 (S52). In S52, the calculation unit 11 generates a prompt including the specifications of the substrate processing device 21 and the request for generating the model 33, and inputs the generated prompt to the language generation model 5. The calculation unit 11 may generate the prompt using data input by the user operating the operation unit 15.

[0092] For example, the prompt includes a request to generate a functional form of a function that the model 33 uses to calculate the operation of the substrate processing apparatus 21 in accordance with the specifications. The functional form of the model 33 is the functional form of a function that the model 33 uses to calculate the operation of the substrate processing apparatus 21. For example, the model 33 is a simulation model that simulates the operation of the substrate processing apparatus 21 based on physical principles. For example, the prompt includes a request to generate a functional form of a simulation model for performing a simulation in accordance with the specifications of the substrate processing apparatus 21.

[0093] The language generation model 5 performs calculations in response to the input of a prompt and outputs a model 33 that calculates the operation of the substrate processing apparatus 21. The model 33 includes undetermined adjustable parameters. The language generation model 5 outputs the model 33 according to the specifications of the substrate processing apparatus 21.

[0094] The information processing device 1 acquires the model 33 (S53). In S53, the calculation unit 11 acquires the model 33 output by the language generation model 5. After S53 is completed, the information processing device 1 terminates the processing. After executing S51 to S53, the information processing device 1 generates the target trajectory generation algorithm 3 using the acquired model 33.

[0095] In the third embodiment, the information processing apparatus 1 can generate a function form or model 33 of the target trajectory generation algorithm 3 using the language generation model 5, and generate the target trajectory generation algorithm 3. After generating the target trajectory generation algorithm 3, the information processing apparatus 1 can adjust adjustable parameters of the target trajectory generation algorithm 3 by executing the processes of S101 to S114 or S301 to S314. The information processing apparatus 1 can perform the processes of S21 to S24 using the target trajectory generation algorithm 3 whose adjustable parameters have been adjusted. The control device 22 can control the substrate processing apparatus 21 using the generated target trajectory.

[0096] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. In other words, embodiments obtained by combining technical means modified appropriately within the scope of the claims are also included in the technical scope of the present invention.

[0097] The matters described in each embodiment can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations, regardless of the reference format. Furthermore, the claims do not use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. They may be written using a multiple claim format or a format in which multiple claims (multi-multi claim) reference at least one other multiple claim.

[0098] REFERENCE SIGNS LIST 1 Information processing device 10 Recording medium 11 Calculation unit 13 Storage unit 131 Computer program 21 Substrate processing device 22 Control device 3, 303 Target trajectory generation algorithm 33, 306, 313 Model 34, 304, 314 Objective function 4 Evaluation function 5 Language generation model

Claims

1. An information processing method comprising: performing a target trajectory generation process using a target trajectory generation algorithm to generate a target trajectory representing a target value of a variable controlled during operation of a substrate processing apparatus; the target trajectory generation algorithm including adjustable parameters that are not changed during the target trajectory generation process; and performing a parameter adjustment process to adjust the adjustable parameters using an evaluation function for evaluating the results of the target trajectory generation process.

2. The information processing method according to claim 1, wherein the parameter adjustment process uses the evaluation function to calculate an evaluation value for evaluating a result of the target trajectory generation process, changes the adjustable parameters according to the evaluation value, performs the target trajectory generation process using the target trajectory generation algorithm including the changed adjustable parameters, and repeats the change of the adjustable parameters, the target trajectory generation process, and the calculation of the evaluation value until a predetermined condition is satisfied.

3. The information processing method according to claim 1, wherein the target trajectory generation algorithm is differentiable with respect to the adjustable parameter, and the parameter adjustment process adjusts the adjustable parameter using a result of differentiation or higher-order differentiation of the target trajectory generation algorithm with respect to the adjustable parameter.

4. The information processing method according to claim 3, wherein the parameter adjustment process adjusts the adjustable parameters using deep evolution.

5. The information processing method according to claim 1, wherein the parameter adjustment process adjusts the adjustable parameters using Bayesian optimization, evolutionary algorithm, swarm intelligence, or neighborhood search.

6. The information processing method according to claim 1, wherein the target trajectory generation algorithm includes a model that calculates the operation of the substrate processing apparatus according to the target trajectory, an objective function that calculates an index according to the calculation result of the model, and a target trajectory adjustment process that adjusts the target trajectory according to the index, wherein the model is a simulation model that outputs a simulation result of the operation of the substrate processing apparatus, an inverse model that calculates a target trajectory that will result in a specific operation result of the substrate processing apparatus, or a table-based model that uses a table that outputs the operation result of the substrate processing apparatus according to the target trajectory, and wherein the target trajectory generation process repeats processing using the model, calculation of the index using the objective function, and the target trajectory adjustment process until a predetermined condition is satisfied.

7. The information processing method according to claim 6, wherein the target trajectory generation algorithm includes a plurality of models each of which calculates a part of the operation of the substrate processing apparatus, and the plurality of models are combined to calculate a series of operations of the substrate processing apparatus.

8. An information processing method according to claim 6, wherein the target trajectory generation algorithm includes a plurality of models divided into a plurality of groups, and the target trajectory adjustment process adjusts the target trajectory by combining the calculation results within each group and the calculation results of the interactions between the groups.

9. The information processing method according to claim 6, wherein the target trajectory generation algorithm includes a plurality of models divided into a plurality of groups, and the plurality of models have a hierarchical structure in which one group includes other groups.

10. The information processing method according to claim 6, wherein the model to be used in the target trajectory generation process is selected from among a plurality of models that calculate the same part of the operation of the substrate processing apparatus, depending on the state of the substrate processing apparatus or the operating conditions of the substrate processing apparatus.

11. The information processing method according to claim 6, wherein the model is generated using a language generation model.

12. The information processing method according to claim 1, wherein the target trajectory generation algorithm includes at least one of a process for adapting the target trajectory to specific constraints and a process for changing the target trajectory with a specific probability.

13. The information processing method according to claim 6, wherein the parameter adjustment process includes adjusting the adjustable parameters included in the model based on actual operation results of the substrate processing apparatus.

14. The information processing method according to claim 2, wherein the evaluation function calculates, as the evaluation value, the error between the target trajectory generated by the target trajectory generation process and an optimal value, the calculation cost of the target trajectory generation process, or the accuracy of adjustment of the target trajectory adjusted during the target trajectory generation process.

15. The target trajectory generation algorithms have a hierarchical structure, and the target trajectory generation algorithm in the lowest layer includes: a model that calculates an operation of the substrate processing apparatus according to a target trajectory; a first objective function that calculates a first index according to a calculation result of the model; and a target trajectory adjustment process that adjusts the target trajectory according to the first index, and in the target trajectory generation process using the target trajectory generation algorithm in the lowest layer, processing using the model, calculation of the first index using the first objective function, and the target trajectory adjustment process are repeated until a predetermined condition is satisfied, and in the target trajectory generation process using the target trajectory generation algorithm in the lowest layer, the target trajectory generation algorithm in a layer other than the lowest layer includes: a target trajectory generation algorithm in a lower layer, a second objective function that calculates a second index according to a processing result of the target trajectory generation process using the target trajectory generation algorithm in the lower layer, and a modification process that modifies an adjustable parameter included in the target trajectory generation algorithm in the lower layer according to the second index, and in the target trajectory generation process using the target trajectory generation algorithm in a layer other than the lowest layer, the target trajectory generation process using the target trajectory generation algorithm in the lower layer, calculation of the second index using the second objective function, and the modification process are repeated until a predetermined condition is satisfied. The information processing method according to claim 1 .

16. The information processing method according to claim 15, wherein in a target trajectory generation process using a target trajectory generation algorithm other than that of the lowest layer, the adjustable parameters are adjusted using deep layer expansion.

17. The information processing method according to claim 15, wherein in a target trajectory generation process using a target trajectory generation algorithm in multiple layers other than the lowest layer, adjustable parameters included in the target trajectory generation algorithm in each layer are adjusted so that the target trajectory generation algorithm in multiple layers other than the lowest layer is Pareto optimal.

18. The information processing method according to claim 1, wherein the function form of the target trajectory generation algorithm is generated using a language generation model.

19. An information processing apparatus comprising: a calculation unit, which performs a target trajectory generation process using a target trajectory generation algorithm to adjust a target trajectory representing a target value of a variable controlled during operation of a substrate processing apparatus, the target trajectory generation algorithm including adjustable parameters that are not changed during the target trajectory generation process; and which performs a parameter adjustment process to adjust the adjustable parameters using an evaluation function that evaluates the result of the target trajectory generation process.

20. A computer program that causes a computer to execute a process of performing a target trajectory generation process using a target trajectory generation algorithm to adjust a target trajectory that represents a target value of a variable controlled during operation of a substrate processing apparatus, wherein the target trajectory generation algorithm includes adjustable parameters that are not changed during the target trajectory generation process, and performing a parameter adjustment process to adjust the adjustable parameters using an evaluation function that evaluates the results of the target trajectory generation process.

Images