Computer program, information processing method, and information processing device

Abstract

Provided is an information processing method in which parameters related to substrate processing conditions are optimized by using a prediction model that calculates a predicted state after substrate processing on the basis of the initial state of a substrate and the parameters related to substrate processing conditions, wherein: acquired is optimization information that is similar to substrate processing content and is obtained when parameters for machining the substrate from a first pre-machined state to a first post-machined state have been optimized; and the acquired optimization information and the prediction model are used to optimize parameters for machining the substrate from a second pre-machined state to a second post-machined state.

Description

Computer program, information processing method, and information processing device

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

[0002] Substrate processing such as etching or film formation on substrates such as semiconductor wafers or glass substrates is performed according to a process recipe that defines the processing details. Conventionally, shape simulation has been performed using a computer to calculate a predicted shape of a substrate after the substrate processing. In the shape simulation, the substrate processing is simulated using a plurality of processing parameters related to the substrate processing according to the process recipe. In addition, a process recipe for obtaining a desired substrate shape is also searched for by adjusting the processing parameters so that a specific predicted shape is obtained by the shape simulation. Patent Document 1 discloses a technology for performing shape simulation.

[0003] Patent No. 6890632

[0004] There are two types of shape simulation methods: a method of predicting the post-processing shape using a machine learning model, and a method of predicting the post-processing shape using CAE analysis (non-machine learning). The former method has a faster processing speed than the latter, but because it has many processing parameters, it has a problem that it takes a long time to predict the substrate shape even when using a machine learning model. This problem is not limited to shape simulation, but is a technical challenge that simulations that predict any state of a substrate after substrate processing, such as the state of a film after processing (film quality, composition), shape distribution such as the in-plane uniformity of the processed shape, etc.

[0005] An object of the present disclosure is to provide a computer program, an information processing method, and an information processing device that can efficiently calculate various parameters related to a substrate processing process from the pre-processing state of a substrate and an expected post-processing state when creating a process recipe required for substrate processing.

[0006] A computer program according to one aspect of the present disclosure causes a computer to perform a process of optimizing parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on the parameters related to the conditions of substrate processing and the initial state of the substrate. The computer program acquires optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state, and causes the computer to perform a process of optimizing the parameters for processing the substrate from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model.

[0007] An information processing method according to one aspect of the present disclosure is an information processing method that optimizes parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on the parameters related to the conditions and the initial state of the substrate, wherein the method obtains optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state, and optimizes the parameters for processing the substrate from a second pre-processing state to a second post-processing state using the obtained optimization information and the prediction model.

[0008] An information processing device according to one aspect of the present disclosure is an information processing device that includes a processing unit that optimizes parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on parameters related to the conditions of the substrate processing and the initial state of the substrate, wherein the processing unit acquires optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state, and optimizes the parameters for processing the substrate from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model.

[0009] According to the present disclosure, it is possible to provide a computer program, an information processing method, and an information processing device that can efficiently calculate various parameters related to a substrate processing process from the pre-processing state of a substrate and an expected post-processing state when creating a process recipe required for substrate processing.

[0010] 1 is a block diagram showing an example of the configuration of an information processing device according to embodiment 1. FIG. 2 is a conceptual diagram showing an information processing method according to embodiment 1. FIG. 3 is a conceptual diagram showing a method for optimizing plasma parameters according to embodiment 1. FIG. 4 is a flowchart showing a procedure for creating optimization information according to embodiment 1. FIG. 5 is a conceptual diagram showing an example of optimization information. FIG. 6 is a flowchart showing a procedure for optimizing plasma parameters according to embodiment 1. FIG. 7 is a block diagram showing an example of the configuration of an information processing device according to embodiment 2. FIG. 8 is a conceptual diagram showing a data structure of an optimization information DB according to embodiment 2. FIG. 9 is a conceptual diagram showing an information processing method according to embodiment 2. FIG. 10 is a flowchart showing a procedure for optimizing plasma parameters according to embodiment 2. FIG. 11 is a block diagram showing an example of the configuration of an information processing device according to embodiment 3. FIG. 12 is a conceptual diagram showing a method for optimizing plasma parameters according to embodiment 3. FIG. 13 is a flowchart showing a procedure for optimizing plasma parameters according to embodiment 3.

[0011] A computer program, an information processing method, and an information processing device according to embodiments of the present disclosure will be described below with reference to the drawings. Note that the present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope equivalent to the claims. In addition, at least some of the embodiments described below may be combined in any manner.

[0012] (Embodiment 1) <Outline of Information Processing Method> The information processing method according to this embodiment enables efficient calculation of various plasma parameters related to a substrate processing process from the pre-processing state of the substrate and the expected post-processing state when creating a process recipe required for substrate processing using plasma. The pre-processing state and post-processing shape include the shape of the substrate before and after processing, the film state (film quality, composition), etc. The post-processing shape includes the shape of a structure such as a hole or wiring, and the average shape of multiple structures. The post-processing shape also includes the shape distribution of the substrate after processing, for example, in-plane uniformity. In particular, the information processing method according to embodiment 1 utilizes optimization information (a history of plasma parameter optimization processes) of similar substrate states from the past to more quickly find and efficiently optimize optimal plasma parameters.

[0013] Substrate processing refers to processes such as etching and film formation performed on substrates such as semiconductor wafers, glass substrates, or flat panel substrates. A process recipe is information that defines the details of the substrate processing. A substrate processing apparatus, such as a semiconductor manufacturing apparatus, that performs substrate processing includes a process chamber and performs substrate processing on a substrate placed in the process chamber based on the process recipe. The process recipe includes setting values ​​(hereinafter referred to as recipe setting values) for controlling the operation of the substrate processing apparatus. The recipe setting values ​​are numerical values ​​that can be input by an operator operating the substrate processing apparatus, and include gas flow rate, gas type, RF applied voltage, frequency, chamber pressure, temperature, processing time, etc.

[0014] Plasma parameters are multiple parameters required for simulations that predict the post-processing state of a substrate from its pre-processing state. Plasma parameters include flux-related information and reaction-related information. Flux-related information includes information such as the type, flow rate, energy, and angle dependency of particles incident on the wafer. Reaction-related information includes information such as the removal rate, deposition rate, modification rate, and sputtering rate of the target medium corresponding to the flux information. Plasma parameters not only enable prediction of the plasma state inside the chamber, but also make it possible to indirectly evaluate the relationship between the knob information of the substrate processing equipment and the plasma state. Furthermore, the interaction and reaction information between particles and the medium included in the plasma parameters is a uniquely determined parameter and serves as an indicator for other process predictions and development.

[0015] 1 is a block diagram showing an example of the configuration of an information processing device 1 according to embodiment 1. The information processing device 1 is a computer that implements the information processing method according to embodiment 1, and includes a processing unit 11, a display unit 12, an operation unit 13, and a storage unit 14. Each unit is connected via a bus.

[0016] The information processing device 1 may be a standalone computer or a server device connected to a network. The information processing device 1 may also be a computer in an on-premise environment or a computer such as a server in a cloud environment. The information processing device 1 may be configured to perform distributed processing using multiple computers, may be realized by multiple virtual machines provided in a single server, or may be realized using a cloud server.

[0017] The processing unit 11 is a processor having arithmetic circuits such as a CPU (Central Processing Unit), a multi-core CPU, a GPU (Graphics Processing Unit), a GPGPU (General-purpose computing on graphics processing units), a TPU (Tensor Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), and an NPU (Neural Processing Unit), internal storage devices such as a ROM (Read Only Memory) and a RAM (Random Access Memory), an I / O terminal, a timer, etc. The processing unit 11 performs the information processing method according to the first embodiment by executing a computer program (program product) 141 stored in the storage unit 14 described below. Note that each functional unit of the information processing device 1 may be realized by software, or part or all of it may be realized by hardware.

[0018] The display unit 12 is, for example, a display device such as a liquid crystal panel or an organic EL (Electro Luminescence) display.

[0019] The operation unit 13 is an input device such as a hardware keyboard, a pointing device, a touch panel, etc. A user of the information processing device 1 can input information such as text to the information processing device 1 using the operation unit 13. The operation unit 13 may be configured integrally with the display unit 12.

[0020] The storage unit 14 includes, for example, a main storage unit and an auxiliary storage unit. The main storage unit is a temporary storage area such as a static random access memory (SRAM), a dynamic random access memory (DRAM), or a flash memory, and temporarily stores data required for the processing unit 11 to execute arithmetic processing. The auxiliary storage unit is a storage device such as a hard disk or an electrically erasable programmable read-only memory (EEPROM). The storage unit 14 stores a computer program 141 and a prediction model 142 executed by the processing unit 11. The storage unit 14 also stores a recipe table (not shown) that associates pre-processing states and post-processing states before and after typical or standard substrate processing with recipe setting values.

[0021] The auxiliary storage unit may be an external storage device connected to the information processing device 1. The computer program 141 and the prediction model 142 may be written to the storage unit 14 during the manufacturing stage of the information processing device 1, or may be distributed by an external server and acquired by the information processing device 1 via communication and stored in the storage unit 14. The computer program 141 and the prediction model 142 may be readably recorded on a recording medium 10 such as a magnetic disk, an optical disk, or a semiconductor memory, or may be read from the recording medium 10 by a reading device and stored in the storage unit 14.

[0022] The prediction model 142 is a numerical calculation model that predicts a post-processing state based on image data of the substrate's pre-processing state and multiple plasma parameters related to the substrate processing. The prediction model 142 may be either a machine learning model or a state simulator (CAE analysis simulator using a non-machine learning model) that analytically predicts the post-processing state. The state simulator includes a shape simulator that predicts the shape of the substrate after processing. Furthermore, all or part of the prediction model 142 may be configured by hardware. In the first embodiment, the prediction model 142 is described as a machine learning model that is realized in software by the arithmetic processing of the processing unit 11.

[0023] The prediction model 142 includes, for example, a convolutional neural network (CNN) that has been trained using deep learning. The prediction model 142 has an input layer to which image data of the substrate in its pre-processing state and a plurality of plasma parameters are input, an intermediate layer that extracts feature quantities of the image data of the pre-processing state and the plasma parameters, and an output layer that outputs image data of the predicted post-processing state.

[0024] The pre-processing state is data representing the state of the substrate before substrate processing, and the post-processing state is data representing the state of the substrate after substrate processing. The pre-processing state and the post-processing state are, for example, image data representing the state of the substrate, such as a cross-sectional view of the substrate. The pre-processing state and the post-processing state may be data including a feature value representing the size or physical structure of the substrate. The size of the substrate is, for example, the length, width, or height of the substrate, or the size of a specific portion of the substrate. The feature value representing the physical structure is, for example, the critical dimension (CD) value of a pattern formed on the substrate or the aspect ratio of a groove formed on the substrate.

[0025] The prediction model 142 is generated by machine learning using training data in a computer-based learning device. The training data records the pre-processing state of the substrate, the values ​​of plasma parameters, and the post-processing state of the substrate, all correlated with one another. The training data includes a large number of data sets in which the pre-processing state, the values ​​of plasma parameters, and the post-processing state are correlated. The learning device inputs the pre-processing state and plasma parameters recorded in the training data into a model (pre-learning prediction model 142) that is the basis of the prediction model 142. The model performs calculations in response to the input pre-processing state and plasma parameters, and outputs a post-processing state, which is a predicted state of the substrate. The learning device adjusts the calculation parameters of the model to reduce the error between the post-processing state output by the model and the post-processing state associated with the input pre-processing state and plasma parameters. For example, the parameter adjustment is performed using an error backpropagation method.

[0026] The learning device performs machine learning by repeating the above process using multiple data sets included in the training data and adjusting the model parameters. By adjusting the calculation parameters in this manner, a prediction model 142 is generated. Data indicating the configuration and parameters of the model that is the basis of the prediction model 142 (pre-learning prediction model 142) is stored as the prediction model 142 in the storage unit 14 of the information processing device 1.

[0027] <Information Processing Method> FIG. 2 is a conceptual diagram showing an information processing method according to the first embodiment, FIG. 3 is a conceptual diagram showing a plasma parameter optimization method according to the first embodiment, and FIG. 4 is a flowchart showing a process for creating optimization information according to the first embodiment.

[0028] First, the information processing device 1 performs a process of optimizing plasma parameters using a conventional method and creating optimization information. Specifically, the processing unit 11 of the information processing device 1 acquires the pre-processing state and the post-processing state of the substrate by the substrate processing (step S111). For example, as shown in FIG. 2 , the processing unit 11 acquires the pre-processing state A1 and the post-processing state A2.

[0029] Next, the processing unit 11 acquires recipe setting values ​​corresponding to the acquired pre-processing state and post-processing state (step S112). Specifically, the processing unit 11 refers to the recipe table stored in the storage unit 14 and acquires recipe setting values ​​corresponding to the acquired pre-processing state and post-processing state. Note that if there are no recipe setting values ​​corresponding to the acquired pre-processing state and post-processing state, the processing of step S112 ends without acquiring recipe setting values.

[0030] Next, processing unit 11, functioning as parameter setting unit 22, sets initial values ​​of provisional plasma parameters (step S113). The provisional plasma parameters are plasma parameters that are provisionally set and input to prediction model 142 when optimizing the plasma parameters using prediction model 142. The method for selecting the initial values ​​is not particularly limited, and may be a predetermined value set in advance or an arbitrary value.

[0031] The processing unit 11 then inputs the pre-processing state and the temporary plasma parameters acquired in step S111 into the prediction model 142 to estimate a post-processing state (step S114). The processing unit 11 then calculates an evaluation value of the post-processing state calculated by the prediction model 142 (step S115). The evaluation value is calculated using an evaluation function 21. The evaluation function 21 is a function that evaluates the difference between the target post-processing state acquired in step S111 and the post-processing state predicted by the prediction model 142. The evaluation function 21 is, for example, a function that returns the pixel error between the image of the post-processing state acquired in step S111 and the image of the post-processing state output by the prediction model 142.

[0032] The processing unit 11 then determines whether the evaluation value is less than the predetermined threshold (step S116). If it is determined that the evaluation value is less than the predetermined threshold (step S116: YES), the processing unit 11, functioning as the parameter setting unit 22, changes the provisional plasma parameters (step S117) and returns the process to step S114. The processing unit 11 continues searching for provisional plasma parameters until the evaluation value becomes equal to or greater than the predetermined threshold. If it is determined that the evaluation value is equal to or greater than the predetermined threshold (step S116: NO), the processing unit 11 stores the history of the plasma parameter search process performed in steps S113 to S117 as optimization information in the storage unit 14 (step S118), and then ends the process. As shown in FIG. 2 , the search for plasma parameters determines a plasma parameter A whose evaluation value is equal to or greater than the predetermined threshold. That is, the plasma parameter A is determined when the substrate is processed from the pre-processing state A1 to the post-processing state A2.

[0033] 5 is a conceptual diagram illustrating an example of optimization information, which is information that associates a number indicating a step in the plasma parameter optimization process with the plasma parameters trialed during optimization and an evaluation value of the post-processing state estimated using the plasma parameters.

[0034] Next, a process will be described in which the generated optimization information is used to optimize plasma parameters for processing substrates having similar substrate processing contents from a pre-processing state to a post-processing state.

[0035] 6 is a flowchart showing the procedure for optimizing plasma parameters according to the first embodiment. The processing unit 11 of the information processing device 1 acquires pre-processing and post-processing states of a substrate during substrate processing (step S131). For example, as shown in FIG. 2, the processing unit 11 acquires pre-processing state B1 and post-processing state B2. Here, it is assumed that the pre-processing state A1 used when generating the optimization information is the same as or similar to the pre-processing state B1 acquired in step S131. Then, the processing unit 11 acquires recipe setting values ​​corresponding to the acquired pre-processing and post-processing states (step S132), similar to step S111.

[0036] Next, the processing unit 11 acquires the optimization information created in the previous process (step S133). That is, the processing unit 11 reads the optimization information from the storage unit 14.

[0037] The processing unit 11 then sets initial values ​​of the provisional plasma parameters (step S134). The processing unit 11 then inputs the pre-processing state and the provisional plasma parameters acquired in step S131 into the prediction model 142 to estimate the post-processing state (step S135). The processing unit 11 then calculates an evaluation value of the post-processing state calculated by the prediction model 142 (step S136).

[0038] Then, processing unit 11 determines whether the evaluation value is less than a predetermined threshold (step S137). If it determines that the evaluation value is less than the predetermined threshold (step S137: YES), processing unit 11, functioning as parameter setting unit 22, changes the temporary plasma parameters based on the optimization information acquired in step S133 and the evaluation value acquired in step S136 (step S138), and returns the process to step S135. For example, processing unit 11 references the optimization information and identifies a method for changing the temporary plasma parameters when the temporary plasma parameters and the evaluation value are similar and the evaluation value is next improved. Then, processing unit 11 changes the temporary plasma parameters using the identified change method.

[0039] The search for the provisional plasma parameters continues until the evaluation value is equal to or greater than the predetermined threshold. If it is determined that the evaluation value is equal to or greater than the predetermined threshold (step S137: NO), the processing unit 11 stores the history of the plasma parameter search process performed in steps S134 to S138 as optimization information in the storage unit 14 (step S139). As shown in FIG. 2 , the search for the plasma parameters determines the plasma parameter B that results in the evaluation value being equal to or greater than the predetermined threshold. In other words, the plasma parameter B is determined when the substrate is processed from the pre-processing state B1 to the post-processing state B2. The optimization information stored in step S139 can be used for optimizing the plasma parameters for other substrate processing.

[0040] Next, the processing unit 11 outputs the optimized plasma parameters (step S140). For example, the processing unit 11 displays the plasma parameters on the display unit 12.

[0041] The operator of the substrate processing apparatus or the processing section 11 determines recipe setting values ​​by adjusting or correcting them based on the optimized plasma parameters (step S141). The processing section 11 performs plasma processing of the substrate by controlling the operation of the substrate processing apparatus (not shown) using the determined recipe setting values ​​(step S142), and then the processing is completed.

[0042] According to the information processing device 1 of the first embodiment configured as described above, when creating a process recipe required for substrate processing, various plasma parameters related to the substrate processing process can be efficiently calculated from the pre-processing state of the substrate and the expected post-processing state by utilizing the optimization information.

[0043] Then, the optimized plasma parameters are used to determine optimal recipe setting values, and plasma processing can be performed using the substrate processing apparatus.

[0044] Furthermore, by using the history of plasma parameters and evaluation values ​​that have been tried when optimizing plasma parameters for similar substrate processing content, the processing unit 11 can optimize plasma parameters more flexibly and efficiently than when using only the optimal values ​​of the plasma parameters.

[0045] <Example of use> Currently, when engineers model the state of a circuit board, they analyze each state feature individually. However, quantifying each state feature and comprehensively searching for corresponding parameters requires a huge amount of time for analysis and consideration. Furthermore, with conventional optimization methods, even if similar state calculations have been performed in the past, optimization calculations are performed from scratch, and analyzing the factors behind specific state feature values ​​requires a significant amount of optimization time.

[0046] Therefore, the information processing device 1 according to the first embodiment calculates plasma parameters for obtaining a desired average post-processing state from the pre-processing state, and stores the optimization information. Then, for each post-processing state that includes a state error (variation) from the average post-processing state, optimization is performed using the optimization information. This allows the plasma parameters corresponding to the state error (variation) to be evaluated.

[0047] By using the information processing method according to the first embodiment, it is possible to efficiently calculate plasma parameters for obtaining a similar post-processing state by using optimization information obtained in a previous optimization process. Furthermore, by using optimization information for a state averaged within measurement data under the same experimental conditions, it is possible to quantify the factors (plasma parameters) that cause state error (variation) from the averaged state. Furthermore, this method reduces the problem of redundancy in search solutions by utilizing arbitrary optimization information for target optimization. This makes it useful for evaluating robustness to a target post-processing state by ensuring similarity in the search parameter space when optimizing under the same process conditions.

[0048] (Embodiment 2) The information processing apparatus 1 according to embodiment 2 differs from embodiment 1 in that it searches for optimization information for similar substrate processing content from multiple pieces of optimization information stored in the past, and optimizes plasma parameters. Since the other configurations of the information processing apparatus 1 are the same as those of the information processing apparatus 1 according to embodiment 1, the same reference numerals are used for similar parts, and detailed descriptions thereof will be omitted.

[0049] 7 is a block diagram showing an example of the configuration of an information processing device 1 according to embodiment 2. The information processing device 1 according to embodiment 2 includes an optimization information DB 243 that stores multiple pieces of optimization information obtained by previously performed optimization processes of plasma parameters.

[0050] 8 is a conceptual diagram showing the data structure of the optimization information DB 243 according to the second embodiment. The optimization information DB 243 stores record numbers, optimized plasma parameters, recipe setting values, pre-processing states, post-processing states, and optimization information in association with each other. The optimization information is similar to that shown in FIG. 5 and includes a history of multiple plasma parameters tried during the optimization process and evaluation values ​​of predicted states obtained using each plasma parameter.

[0051] Fig. 9 is a conceptual diagram showing an information processing method according to embodiment 2, and Fig. 10 is a flowchart showing a procedure for optimizing plasma parameters according to embodiment 2. As in embodiment 1, the processing unit 11 acquires pre-processing and post-processing states of a substrate by substrate processing (step S131), and acquires recipe setting values ​​corresponding to the acquired pre-processing and post-processing states (step S132).

[0052] If necessary, the user can arbitrarily specify values ​​of some of the plasma parameters among the plurality of plasma parameters, and can input the specified values ​​of the plasma parameters using the operation unit 13. When values ​​of some of the plasma parameters are specified using the operation unit 13, the processing unit 11 acquires the specified values ​​(step S233).

[0053] Next, the processing unit 11 searches for and acquires optimization information having processing contents similar to the current substrate processing contents from among the past optimization information recorded in the optimization information DB 243 (step S234).

[0054] For example, the processing unit 11 may search for and acquire optimization information having similar recipe setting values ​​from the optimization information DB 243. Specifically, the processing unit 11 calculates the similarity between the recipe setting value identified in step S132 and the recipe setting value associated with each piece of optimization information recorded in the optimization information DB 243, and selects optimization information associated with the recipe setting value having the highest similarity.

[0055] For example, the processing unit 11 may search for and acquire optimization information having similar pre-processing states and post-processing states from the optimization information DB 243. Specifically, the processing unit 11 calculates the similarity between the pre-processing state and post-processing state acquired in step S131 and the pre-processing state and post-processing state associated with each piece of optimization information recorded in the optimization information DB 243, and selects optimization information associated with a state having a high degree of similarity.

[0056] For example, the processing unit 11 may search for and acquire optimization information having similar plasma parameters from the optimization information DB 243. Specifically, the processing unit 11 calculates the similarity between the designated values ​​of the plasma parameters acquired in step S233 and the corresponding plasma parameters associated with each piece of optimization information recorded in the optimization information DB 243, and selects optimization information associated with plasma parameters having a high degree of similarity.

[0057] Although three methods for searching for optimization information have been exemplified, the processing unit 11 may search for and acquire optimization information by comprehensively evaluating the above three similarities. Alternatively, the processing unit 11 may search for and acquire optimization information by comprehensively evaluating any two of the above three similarities. The method of comprehensive evaluation is not particularly limited, and for example, the processing unit 11 may select optimization information with the largest sum of similarities.

[0058] Thereafter, the processing unit 11 executes the processes of steps S134 to S142 in the same manner as in the first embodiment.

[0059] According to the information processing device 1 of the second embodiment configured as described above, it is possible to search for optimization information with similar substrate processing contents from the optimization information stored in the optimization information DB 243, and to efficiently optimize plasma parameters.

[0060] (Embodiment 3) An information processing device 1 according to embodiment 3 differs from embodiment 1 in that it optimizes plasma parameters using multiple learning models with different prediction processing speeds and state prediction accuracies. Since the other configurations of the information processing device 1 are similar to those of the information processing device 1 according to embodiment 1, similar parts are denoted by the same reference numerals and detailed descriptions thereof will be omitted.

[0061] 11 is a block diagram showing an example configuration of an information processing device 1 according to embodiment 3. The information processing device 1 according to embodiment 3 stores two different models as the prediction model 142. Specifically, the storage unit 14 stores a first learning model 342a and a second learning model 342b. The processing unit 11 according to embodiment 3 optimizes plasma parameters using the different first learning model 342a and second learning model 342b.

[0062] The second learning model 342b is a model with higher substrate state prediction accuracy and slower prediction processing speed than the first learning model 342a. For example, the first learning model 342a is a machine learning model, and the second learning model 342b is a machine learning model with higher substrate state prediction accuracy and slower prediction processing speed than the first learning model 342a. Alternatively, the first learning model 342a may be configured as a machine learning model, and the second learning model 342b may be configured as a state simulator that calculates a predicted state by CAE analysis based on plasma parameters and the initial state of the substrate. In other words, the first learning model 342a may be a machine learning model, and the second learning model 342b may be a non-machine learning model. Hereinafter, in this third embodiment, the first learning model 342a is described as a machine learning model, and the second learning model 342b is described as a non-machine learning model.

[0063] 12 is a conceptual diagram showing a plasma parameter optimization method according to embodiment 3, and FIGS. 13 and 14 are flowcharts showing the plasma parameter optimization process procedure according to embodiment 3. Here, for similar substrate processing contents, a plasma parameter optimization process is performed using a first learning model 342a and a second learning model 342b, and the memory unit 14 is assumed to store optimization information obtained in the process of the plasma parameter optimization process using the first learning model 342a and optimization information obtained in the process of the plasma parameter optimization process using the second learning model 342b.

[0064] The processing unit 11 uses the first learning model 342a as the prediction model 142 and executes the same processes as steps S131 to S138 in the first embodiment (steps S331 to S338).

[0065] In step S337, the processing unit 11 determines whether the evaluation value of the processed state estimated using the first learning model 342a is less than a first predetermined threshold (step S337). If it is determined that the evaluation value of the processed state estimated using the first learning model 342a is equal to or greater than the first predetermined threshold (step S337: NO), the processing unit 11 executes a plasma parameter optimization process using the second learning model 342b.

[0066] Specifically, the processing unit 11 acquires optimization information for the second learning model 342b (step S339), and takes over the plasma parameters optimized using the first learning model 342a as provisional plasma parameters (step S340).

[0067] Next, the processing unit 11 estimates the post-processing state using the second learning model 342b based on the pre-processing state acquired in step S331 and the temporary plasma parameters inherited in step S340 (step S341). The processing unit 11 calculates an evaluation value of the post-processing state calculated by the second learning model 342b (step S342).

[0068] The processing unit 11 then determines whether the evaluation value is less than a second predetermined threshold (step S343). The second predetermined threshold is preferably set to a value greater than the first predetermined threshold. If it determines that the evaluation value is less than the second predetermined threshold (step S343: YES), the processing unit 11, functioning as the parameter setting unit 22, changes the provisional plasma parameters based on the optimization information acquired in step S339 and the evaluation value calculated in step S342 (step S344), and returns the process to step S341. For example, the processing unit 11 references the optimization information to identify a method for changing the provisional plasma parameters when the provisional plasma parameters and the evaluation value are similar and the evaluation value is next improved. The processing unit 11 then changes the provisional plasma parameters using the identified change method.

[0069] The search for the provisional plasma parameters is continued until the evaluation value becomes equal to or greater than the second predetermined threshold. If it is determined that the evaluation value is equal to or greater than the second predetermined threshold (step S343: NO), the processing unit 11 stores the history of the plasma parameter search process performed in steps S334 to S338 as optimization information for the first learning model 342a, and the history of the plasma parameter search process performed in steps S340 to S344 as optimization information for the second learning model 342b in the storage unit 14 (step S345).

[0070] Thereafter, the processing unit 11 executes the processes of steps S140 to S142 (steps S346 to S348) in the same manner as in the first embodiment.

[0071] According to the information processing device 1 of embodiment 3 configured as described above, plasma parameters can be optimized more quickly and with higher accuracy than conventional techniques that simply use machine learning models.

[0072] Means for solving the problems of the present disclosure are appended below. (Supplementary Note 1) A computer program causing a computer to execute a process of optimizing parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on the parameters related to the substrate processing conditions and an initial state of the substrate, the computer program causing the computer to execute a process of: acquiring optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state; and optimizing the parameters for processing the substrate from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model. (Supplementary Note 2) The computer program according to Supplementary Note 1, wherein the optimization information includes a history of a plurality of parameters obtained in the process of optimizing the parameters for processing the substrate from the first pre-processing state to the first post-processing state and evaluation values ​​for the plurality of parameters. (Supplementary Note 3) The computer program according to Supplementary Note 1 or Supplementary Note 2, wherein the parameters include plasma parameters related to plasma processing conditions. (Supplementary Note 4) The computer program according to any one of Supplements 1 to 3, wherein the parameters include setting values ​​of a process recipe. (Supplementary Note 5) The computer program according to any one of Supplementary Note 1 to Supplementary Note 4, which searches for optimization information having similar substrate processing content from a storage unit that stores the optimization information including information correlating a first pre-processing state, a first post-processing state, and a plurality of the parameters optimized for processing a substrate from the first pre-processing state to the first post-processing state, and optimizes the parameters for processing a substrate from a second pre-processing state to a second post-processing state using the plurality of parameters related to the optimization information obtained by the search. (Supplementary Note 6) The computer program according to Supplementary Note 5, which searches for the optimization information based on a similarity between setting values ​​of a process recipe for the first pre-processing state and the first post-processing state and a process recipe for the second pre-processing state and the second post-processing state.(Supplementary Note 7) The computer program according to Supplementary Note 5, which searches for the optimization information based on the similarity between the first pre-machining state and first post-machining state and the second pre-machining state and second post-machining state. (Supplementary Note 8) The computer program according to Supplementary Note 5, which searches for the optimization information based on the similarity with the plurality of parameters, some of which are set to specific values. (Supplementary Note 9) The computer program according to Supplementary Note 5, which searches for the optimization information based on the similarity between a setting value of a process recipe for the first pre-machining state and first post-machining state and a setting value of a process recipe for the second pre-machining state and second post-machining state, the similarity between the first pre-machining state and first post-machining state and the second pre-machining state and second post-machining state, and the similarity with the parameters, some of which are set to specific values. (Supplementary Note 10) The computer program according to any one of Supplements 1 to 3, wherein the prediction model optimizes the parameters using different first and second learning models. (Supplementary Note 11) The computer program according to Supplementary Note 10, wherein the second learning model has higher substrate state prediction accuracy and a slower prediction processing speed than the first learning model. (Supplementary Note 12) The computer program according to Supplementary Note 10 or Supplementary Note 11, wherein the first learning model is a machine learning model, and the second learning model includes a machine learning model that has higher substrate state prediction accuracy and a slower prediction processing speed than the first learning model. (Supplementary Note 13) The computer program according to Supplementary Note 10 or Supplementary Note 11, wherein the first learning model is a machine learning model, and the second learning model includes a state simulator that calculates a predicted state of the substrate by CAE analysis based on the parameters and an initial state of the substrate. (Supplementary Note 14) The computer program according to any one of Supplements 1 to 13, further comprising: determining process recipe setting values ​​based on the optimized parameters; and controlling substrate processing based on the determined process recipe setting values.

[0073] 1: Information processing device 10: Recording medium 11: Processing unit 12: Display unit 13: Operation unit 14: Storage unit 141: Computer program 142: Prediction model 342a: First learning model 342b: Second learning model 243: Optimization information DB

Claims

1. A computer program that causes a computer to execute a process of optimizing parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on the parameters related to the conditions of the substrate processing and the initial state of the substrate, the computer program causing the computer to execute a process of: acquiring optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state; and optimizing the parameters for processing substrates from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model.

2. The computer program according to claim 1, wherein the optimization information includes a history of the plurality of parameters determined in the process of optimizing the plurality of parameters for processing the substrate from a first pre-processing state to a first post-processing state, and evaluation values for the plurality of parameters.

3. The computer program of claim 1, wherein the parameters include plasma parameters related to conditions of the plasma processing.

4. The computer program of claim 1, wherein the parameters include process recipe settings.

5. A computer program as claimed in claim 1, which searches for optimization information with similar substrate processing content from a storage unit that stores the optimization information including information correlating a first pre-processing state, a first post-processing state, and a plurality of parameters optimized for processing a substrate from the first pre-processing state to the first post-processing state, and optimizes the parameters for processing a substrate from a second pre-processing state to a second post-processing state using the plurality of parameters related to the optimization information obtained by the search.

6. A computer program according to claim 5, which searches for the optimization information based on the similarity between the setting values of the process recipe relating to the first pre-processing state and the first post-processing state and the process recipe relating to the second pre-processing state and the second post-processing state.

7. A computer program according to claim 5, wherein the optimization information is searched for based on the similarity between the first pre-processing state and first post-processing state and the second pre-processing state and second post-processing state.

8. The computer program according to claim 5, wherein the optimization information is searched for based on the similarity to the plurality of parameters, some of which are set to specific values.

9. A computer program as described in claim 5, which searches for the optimization information based on the similarity between the setting values of the process recipe for the first pre-processing state and first post-processing state and the setting values of the process recipe for the second pre-processing state and second post-processing state, the similarity between the first pre-processing state and first post-processing state and the second pre-processing state and second post-processing state, and the similarity with the parameters some of which are set to specific values.

10. The computer program of claim 1, wherein the predictive model optimizes the parameters using different first and second learning models.

11. The computer program according to claim 10, wherein the second learning model has higher accuracy in predicting the state of the substrate and a slower prediction processing speed than the first learning model.

12. The computer program of claim 10, wherein the first learning model is a machine learning model, and the second learning model includes a machine learning model that has higher accuracy in predicting the state of the substrate and a slower prediction processing speed than the first learning model.

13. The computer program of claim 10, wherein the first learning model is a machine learning model, and the second learning model includes a state simulator that calculates a predicted state of the substrate by CAE analysis based on the parameters and the initial state of the substrate.

14. The computer program according to claim 1, further comprising: determining process recipe setting values based on the optimized parameters; and controlling substrate processing based on the determined process recipe setting values.

15. An information processing method for optimizing parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on the parameters related to the conditions of substrate processing and the initial state of the substrate, the method comprising: acquiring optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state; and optimizing the parameters for processing a substrate from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model.

16. An information processing device having a processing unit that optimizes parameters related to substrate processing conditions using a prediction model that calculates a predicted state after substrate processing based on parameters related to the conditions of substrate processing and the initial state of the substrate, wherein the processing unit acquires optimization information obtained when optimizing the parameters for processing substrates having similar substrate processing contents from a first pre-processing state to a first post-processing state, and optimizes the parameters for processing the substrate from a second pre-processing state to a second post-processing state using the acquired optimization information and the prediction model.

Images