SiC crystal manufacturing apparatus, control device for SiC crystal manufacturing apparatus, method for generating learning model for SiC crystal manufacturing apparatus, and method for controlling SiC crystal manufacturing apparatus

Abstract

To provide an excellent technology related to SiC crystal manufacturing than conventional technologies.SOLUTION: A controller for a SiC crystal manufacturing device has a learning model which outputs, based on a measured value of a first sensor included in the SiC crystal manufacturing device, an estimated value of a second physical amount that cannot be observed upon manufacture of SiC crystals. The controller controls an actuator in a manner such as to decrease a difference between the estimated value outputted by the learning model upon the manufacture of the SiC crystal and a target value of the second physical amount. The controller uses results of a simulation model based on structure data of the SiC crystal manufacturing device as teacher data to cerate a basic learning model. The controller acquires measured values of the first sensor and the second sensor when the actuator is driven under condition in which any SiC crystal cannot be manufactured with the SiC crystal manufacturing device fitted with a second sensor which measures the second physical amount but the second physical amount can be observed, and creates a learning model for correcting output of the basic leaning model by use of the measured values.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present specification relates to a SiC crystal manufacturing apparatus, a control device for the SiC crystal manufacturing apparatus, a method for generating a learning model for the SiC crystal manufacturing apparatus, and a control method for the SiC crystal manufacturing apparatus. SiC is an abbreviation for silicon carbide. [Background technology]

[0002] When manufacturing SiC crystals, the chamber where the crystal grows (especially the crystal growth surface) must be kept at an extremely high temperature (over 2000°C on the crystal growth surface), necessitating the covering of the chamber's inner walls with opaque insulating material. Furthermore, due to spatial interference with the path of supplying raw materials to the crystal growth area, it is not possible to install sensors on the crystal growth surface or other areas in the chamber to measure the conditions during crystal production. Patent Document 1 discloses a technology for displaying a theoretically predicted state of the chamber using a prediction model generated by machine learning using only simulation results based on theoretical formulas as training data and only manual instruction signals to the device as input values. The technology in Patent Document 1 inputs a first state output from a semiconductor crystal manufacturing device into the machine-learned prediction model. An image generated based on a second state output from the prediction model is displayed on the outside of the manufacturing device. The first state refers to a physical quantity within the manufacturing device that can be directly measured by sensors or the like. Examples of the first state include the position of components in the manufacturing device, the temperature or pressure at the measurement location, etc. The second state refers to a physical quantity within the manufacturing device that is extremely difficult to directly measure by sensors or the like. The second state includes, for example, the temperature at each point within a specific region inside the manufacturing equipment, the raw material concentration, the flow rate and vectors indicating its direction, etc. An operator of the manufacturing equipment can view the video and understand the second state inside the manufacturing equipment.

[0003] According to Patent Document 1, the prediction model is obtained through machine learning. The training data is the results of simulating multiple states of the manufacturing equipment using an analytical simulation model. The simulation model is created from a three-dimensional model created from CAD data of the manufacturing equipment, physical property data of the materials used in the reaction, etc.

[0004] Patent Document 2 discloses that data from various sensors in a plasma processing apparatus is used as training data for machine learning. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2018-169818 [Patent Document 2] Japanese Patent Application Publication No. 2020-181959 Summary of the Invention [Problem to be solved by the invention]

[0006] According to the technology of Patent Document 1, training data for creating a predictive model is obtained through simulation. However, simulations may not accurately reflect the characteristics of actual semiconductor crystal manufacturing equipment or may not reflect changes in the equipment over time. In other words, a predictive model created using training data from simulations may not be able to predict the actual state inside a furnace with sufficient accuracy. On the other hand, Patent Document 2 uses sensor data as training data. However, creating a highly accurate learning model requires a huge amount of training data, and obtaining a large amount of training data through experiments is costly both financially and time-wise.

[0007] Furthermore, manufacturing semiconductor crystals, particularly SiC crystals, requires a high-temperature environment of over 2000°C. For this reason, the heated area (the chamber's internal space) must be covered with opaque insulating material, making it impossible to measure the temperature of the crystal growth area (around the susceptor), which is the most important area for growth, using a pyrometer or similar device. Furthermore, if a sensor were installed to measure the temperature of the crystal growth area, the sensor would interfere with the raw material supply path during the actual crystal growth process, making it difficult to directly observe the conditions inside the chamber during the actual crystal growth process.

[0008] Furthermore, simulations using techniques such as FEM have traditionally been used to estimate locations that cannot be directly observed. However, simulations generally require tens of hours of calculation time, and even if a simulation is performed using measurements of chamber conditions that change from moment to moment (temperature, flow rate, pressure, etc.) as input values, the time it takes to obtain simulation results is longer than the time scale (on the order of 100 milliseconds to 1 second) required for feedback control of crystal growth. This makes it difficult to estimate using simulations during the growth process.

[0009] Furthermore, the physical property values ​​used in the simulation are values ​​under predetermined ideal conditions, such as values ​​described in the literature, and do not accurately reflect the components and conditions inside the manufacturing equipment during the actual growth process, including changes over time. Machine learning using simulation data as training data cannot achieve sufficiently accurate estimation of the actual growth process. It is necessary to fine-tune the machine learning model obtained using simulation data as training data using measurement data obtained from the manufacturing equipment used for actual growth.

[0010] This specification proposes a novel, unprecedented technology for manufacturing SiC crystals. Specifically, it provides a method for generating a learning model that can accurately predict physical quantities that cannot be observed during crystal manufacturing, as well as a manufacturing apparatus and control device (a control device for a SiC crystal manufacturing apparatus) that can manufacture high-quality SiC crystals using such a learning model. Note that while Patent Document 1 uses the term "prediction model," in this specification, a model obtained by machine learning will be referred to as a "learning model." Also, for ease of explanation, a "SiC crystal manufacturing apparatus" may be simply referred to as a "crystal manufacturing apparatus." [Means for solving the problem]

[0011] One technique disclosed in this specification provides a SiC crystal manufacturing apparatus. The SiC crystal manufacturing apparatus includes a chamber, an actuator, a first sensor, and a control device. SiC crystals are produced from raw materials (e.g., silicon-containing gas or carbon-containing gas) in the chamber. The chamber may alternatively be called a reactor. The actuator controls the temperature in the chamber and the flow rate / flow velocity / concentration of the raw materials. The first sensor measures a first physical quantity that can be observed when manufacturing SiC crystals, either inside or outside the chamber.

[0012] The control device has a learning model that outputs an estimated value of a second physical quantity that is unobservable when manufacturing SiC crystal from a measurement value of the first sensor. The control device feedback-controls the actuator so that the difference between the estimated value output by the learning model and a target value of the second physical quantity when manufacturing SiC crystal is small.

[0013] The control device includes a basic model generation module, an experimental result acquisition module, a correlation data acquisition module, a correlation determination module, and a model correction module. The basic model generation module creates a basic learning model through machine learning using multiple simulation results (first simulation results) of pairs of a first physical quantity and a second physical quantity created using a simulation model based on structural data of the SiC crystal manufacturing equipment as training data. The experimental result acquisition module acquires multiple experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement value) and the second sensor (second sensor measurement value) when an actuator is driven using a SiC crystal manufacturing equipment equipped with a first sensor and a second sensor that measures the second physical quantity under conditions in which SiC crystals cannot be manufactured but the second physical quantity can be observed. The correlation data acquisition module uses the simulation model to obtain a simulation result (second simulation result) of the second physical quantity when the first physical quantity matches the first sensor measurement value. The correlation determination module determines the correlation between the second simulation result and the second sensor measurement value. The model correction module generates a learning model that corrects the output of the basic learning model using the correlation described above and outputs the corrected result as an estimated value (estimated value of the second physical quantity).

[0014] This specification also provides a method for generating a learning model that obtains an estimate of an unobservable second physical quantity from an observable first physical quantity when manufacturing SiC crystals using a SiC crystal manufacturing apparatus, the method comprising a basic model generation step, an experimental result acquisition step, a correlation data acquisition step, a correlation determination step, and a model correction step.

[0015] In the basic model generation step, a basic learning model is created by machine learning using multiple simulation results (first simulation results) of pairs of a first physical quantity and a second physical quantity created using a simulation model based on structural data of the SiC crystal manufacturing equipment as training data. In the experimental result acquisition step, multiple experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement value) and measurement values ​​of the second sensor (second sensor measurement value) are acquired when an actuator of the SiC crystal manufacturing equipment, which is equipped with a first sensor that measures the first physical quantity and a second sensor that measures the second physical quantity, is driven under conditions where SiC crystals cannot be manufactured but the second physical quantity can be observed.

[0016] In the correlation data acquisition process, a simulation model is used to obtain a simulation result (second simulation result) of a second physical quantity when the first physical quantity matches the first sensor measurement value. In the correlation determination process, the correlation between the second simulation result and the second sensor measurement value is determined. In the model correction process, a learning model is generated that corrects the output of the basic learning model using the aforementioned correlation and outputs the result as an estimate (estimated value of the second physical quantity).

[0017] Furthermore, the technology disclosed in this specification can feedback-control a crystal manufacturing apparatus using a learning model that can accurately predict the second physical quantity. Because a learning model based on machine learning does not require complex calculations, it takes only a few milliseconds to 1 second from input to output (an estimated value of the second physical quantity). Therefore, the learning model can be used for real-time feedback control.

[0018] In particular, the technology disclosed in this specification uses the first sensor used during SiC crystal manufacturing in experiments to obtain data (first and second measurement values) for correcting the output of the basic learning model. Therefore, by correcting the learning model used during SiC crystal manufacturing using the correlation obtained from the experimental results, a highly accurate estimated value (estimated value of the second physical quantity) can be obtained. Furthermore, a control device using such a learning model controls the actuator of the SiC crystal manufacturing device so as to reduce the deviation between the target value and estimated value of the second physical quantity. This control device operates as if it were feeding back the second physical quantity.

[0019] The control device can bring a second physical quantity, which cannot be observed when manufacturing SiC crystal, closer to a target value. An example of a "second physical quantity" is the surface temperature of a susceptor. The surface temperature of a susceptor is an important factor that affects the quality of SiC crystal. If the surface temperature of the susceptor can be brought closer to the target value, high-quality SiC crystal can be obtained. The susceptor is a support plate that supports the substrate when growing the crystal.

[0020] Details and further improvements of the technology disclosed in this specification are described in the following "Description of Embodiments of the Invention." [Brief explanation of the drawings]

[0021] [Figure 1] FIG. 1 is a cross-sectional view of a SiC crystal manufacturing apparatus. [Figure 2] FIG. 1 is a schematic diagram of a SiC crystal manufacturing apparatus. [Figure 3] FIG. 1 is a schematic diagram for explaining a basic learning model. [Figure 4] 10 is a graph showing the relationship between the second simulation result Te and the experimental result (measurement value Ts of the sensor). [Figure 5] FIG. 2 is a schematic diagram for explaining the structure of a learning model. [Figure 6] FIG. 2 is a block diagram of the main control module. [Figure 7] 1 is a flowchart of a learning model generation method. DETAILED DESCRIPTION OF THE INVENTION

[0022] (First embodiment) A SiC crystal manufacturing apparatus 10 will be described with reference to the drawings. Hereinafter, for ease of explanation, the SiC crystal manufacturing apparatus 10 will be simply referred to as crystal manufacturing apparatus 10. The SiC crystal manufacturing apparatus 10 is an apparatus for manufacturing SiC (silicon carbide) single crystals. FIG. 1 shows a cross-sectional view of the crystal manufacturing apparatus 10.

[0023] The crystal manufacturing apparatus 10 comprises a chamber 12, a heater 22, a susceptor 13, a rotation device 14, and a gas supply device 23. These components are housed in a housing 11. The chamber 12 is a reactor that grows (manufactures) SiC single crystals by reacting high-temperature raw material gases. Because SiC is grown at high temperatures of 2000 degrees or more, the inner walls of the chamber 12 are covered with an opaque heat insulating material, although this is not shown in the figure. Alternatively, the chamber 12 itself may be made of an opaque heat insulating material.

[0024] The gas supply device 23 generates a silicon-containing first source gas and a carbon-containing second source gas, and supplies these source gases to the chamber 12 through holes provided in the bottom of the chamber 12. The gas supply device 23 heats the source gases to a predetermined preparation temperature.

[0025] A heater 22 is provided around the periphery of the chamber 12 to heat the source gas inside the furnace. The heater 22 is composed of a high-frequency coil 22a and a high-frequency generator 22b. The heater 22 heats the source gas inside the chamber 12 to 2000°C or higher. The source gas, which has reached a high temperature of 2000°C or higher, crystallizes on the surface 13a of the susceptor 13. The susceptor 13 is a support plate that supports a substrate when growing SiC crystal. The surface 13a of the susceptor 13 corresponds to the crystal growth surface. The susceptor 13 is supported by a rotation device 21. The rotation device 21 rotates the susceptor 13. The rotation device 21 can also move the susceptor 13 up and down.

[0026] A cooler 24 is provided above the chamber 12. The cooler 24 has a circulator 24a that circulates a refrigerant and a cooling pipe 24b through which the refrigerant flows. The cooler 24 cools the remaining source gas discharged from the chamber 12.

[0027] The gas supply device 23, heater 22, rotation device 21, and cooler 24 are controlled by a control device 30. The gas supply device 23, heater 22, rotation device 21, and cooler 24 are actuators for producing SiC crystal (for operating the crystal production apparatus 10), and for convenience, they may be collectively referred to as actuator 20 below. Actuator 20 controls the production conditions, examples of which are listed below. The production conditions are the temperature and pressure within the chamber, the flow rate / flow velocity / gas concentration / gas temperature / gas pressure / gas inlet position of the source gas supplied from the gas supply device 23, the current / control voltage / frequency flowing through coil 22a, the vertical position and movement speed of coil 22a, the vertical position, movement speed and rotation speed of susceptor 13, the temperature of the outer wall of the chamber, the amount of cooling water, the cooling water temperature, the cooling water pressure, and the exhaust pressure / flow rate / temperature.

[0028] 1 shows the flow of the reaction gas during crystal growth. As mentioned above, the source gas reaches a high temperature exceeding 2000°C when it reaches the surface 13a of the susceptor 13.

[0029] Multiple sensors 17a-17e are arranged in crystal production apparatus 10. Sensor 17a is a sensor that measures the temperature of the outer wall of chamber 12. Sensor 17b is a sensor that measures the temperature of heater 22 (or the current flowing through high-frequency coil 22a). Sensor 17c is a sensor that measures the temperature / pressure / flow rate / flow velocity / concentration of the source gas supplied to chamber 12. Sensor 17d is a sensor that measures the temperature / pressure / flow rate / flow velocity / concentration of the residual gas discharged from chamber 12. Sensor 17e is a sensor that measures the temperature of cooler 24. These sensors 17a-17e are sensors that are used to know the state of chamber 12 and the state of the source gas when producing SiC crystal. Hereinafter, these sensors 17a-17e are collectively referred to as first sensor 17. First sensor 17 is a sensor that functions when producing SiC crystal. That is, the physical quantity (such as temperature or flow rate) measured by first sensor 17 corresponds to the first physical quantity observable when manufacturing SiC crystal. First sensor 17 is disposed in a location where the temperature does not exceed the heat resistance temperature of the sensor even during crystal growth. First sensor 17 also includes a sensor that measures the output of actuator 20. When actuator 20 operates to follow a target value, the target value is an approximation of the output of actuator 20, and therefore the target value may also be regarded as a pseudo-measurement value of the first sensor. Crystal manufacturing apparatus 10 also has many other first sensors of various types, but these are not shown in the figure.

[0030] The measurement value of the first sensor 17 is sent to the control device 30. The control device 30 controls the actuators 20 (the rotation device 21, the heater 22, the gas supply device 23, and the cooler 24) based on the measurement value of the first sensor 17. The control device 30 determines a target value for each actuator 20 and controls the actuators 20 so that the output of each actuator 20 follows the target value. For example, the target value and output of the rotation device 21 is the rotation speed of the susceptor 13, and the control device 30 controls the rotation device 21 so that the actual rotation speed of the rotation device 21 follows the target rotation speed. The target values ​​of the gas supply device 23 are the temperature / pressure / flow rate / flow rate / concentration of the source gas, and the control device 30 controls the gas supply device 23 so that the temperature / pressure / flow rate / flow rate / concentration of the source gas generated by the gas supply device 23 follow the target temperature / target pressure / target flow rate / target flow rate / target concentration, respectively.

[0031] The control device 30 controls the actuator 20, causing a SiC single crystal SC to grow on the surface 13a of the susceptor 13. That is, a SiC crystal is produced in the crystal production apparatus 10. To produce a high-quality crystal, the temperature and pressure inside the chamber 12, particularly the temperature of the surface 13a (crystal growth surface) of the susceptor 13, are important. The temperature at the center of the surface 13a of the susceptor 13 is particularly important. In other words, if the temperature and pressure at various locations inside the chamber 12 and the temperature at the surface 13a of the susceptor 13 (temperature at the surface center) can be made to match target values, a high-quality single crystal can be obtained. For convenience, hereinafter, the temperature at the surface 13a of the susceptor 13 (temperature at the surface center) will be simply referred to as the surface temperature.

[0032] However, the internal temperature (particularly the surface temperature) of the chamber 12 during SiC crystal production exceeds 2000°C. When producing SiC crystals, the temperature inside the chamber 12 exceeds the sensor's heat resistance, making it impossible to place a sensor there. Alternatively, placing a sensor in the source gas flow path would obstruct the flow of the source gas, making it impossible to place a sensor there. However, although SiC crystals cannot be produced, it is possible to operate the crystal production apparatus 10 so that the internal temperature of the chamber 12 is below the sensor's heat resistance. When the crystal production apparatus 10 is operated so that the internal temperature of the chamber 12 is below the sensor's heat resistance, it is possible to measure the internal temperature / pressure (particularly the surface temperature) of the chamber 12. Physical quantities that cannot be measured during SiC crystal production (such as the internal temperature / pressure of the chamber 12 and the flow rate / flow velocity of the source gas on the susceptor surface) are referred to as second physical quantities. Although SiC crystals cannot be produced using the second physical quantity, it is possible to measure the internal temperature of the chamber 12 by controlling the actuator 20 so that the internal temperature of the chamber 12 is below the sensor's heat resistance. The second physical quantity is closely related to the quality of the SiC crystals produced. If the second physical quantity can be made to follow the target value, high-quality SiC crystal can be produced.

[0033] As will be described in detail later, the crystal production apparatus 10 includes sensor holders 18a-18d inside the chamber 12 so that the actuator 20 can be controlled to keep the internal temperature of the chamber 12 below the sensor's heat-resistant temperature and measure the second physical quantity at that time. The sensor holder 18a is provided on the surface of the susceptor 13. The sensor holder 18b is provided on the inner side of the chamber 12. The sensor holder 18c is provided on the inner bottom surface of the chamber 12. The sensor holder 18d is provided in the source gas supply path at the bottom of the chamber 12. By placing sensors (sensors measuring temperature and pressure) on the sensor holders 18a-18d and controlling the actuator 20 under conditions that do not exceed the sensor's heat-resistant temperature, the second physical quantity (temperature and pressure) at various locations inside the chamber 12 can be measured. The sensors that measure the second physical quantity are collectively referred to as second sensors 19. The second sensors 19 are not installed when producing SiC crystals, and are therefore depicted by dotted lines in FIG. 1. The sensor fixing portions 18a-18d are merely examples, and the sensor fixing portions may be provided in other locations.

[0034] To make the explanation easier to understand, a schematic diagram of the crystal production apparatus 10 is shown in Figure 2. In Figure 2, multiple actuators 20 are depicted as a single rectangle. A first sensor 17 is placed outside the chamber 12, i.e., in a location that does not exceed the sensor's heat resistance temperature during crystal production. The first sensor 17 also includes a sensor that detects the output of the actuator 20.

[0035] A sensor fixing portion 18 is provided at a location (typically the surface of the susceptor 13) where the temperature exceeds the sensor's heat resistance temperature during crystal production. A second sensor 19 can be attached to the sensor fixing portion 18. However, the second sensor 19 is removed when producing SiC crystal, and is attached when controlling the actuator 20 of the crystal production apparatus 10 under conditions below the sensor's heat resistance temperature. The crystal production apparatus 10 has an attachment portion for attaching the second sensor 19. The attachment portion is formed, for example, by a bracket. The attachment portion has, for example, a screw hole for fastening to the second sensor 19 and an engagement portion for engaging with the second sensor 19.

[0036] The crystal manufacturing apparatus 10 (controller 30) can generate a learning model that outputs an estimated value of the second physical quantity from the measurement values ​​of the first sensor 17 and the second sensor 19. The learning model can estimate the second physical quantity (typically the surface temperature) during crystal manufacturing with high accuracy. In the process of generating the learning model, by using the measurement values ​​obtained when the actuator 20 is controlled so that the temperature inside the chamber 12 is below the sensor heat resistance temperature, a learning model that can obtain an estimated value with high accuracy can be obtained.

[0037] The crystal manufacturing apparatus 10 (controller 30) controls the actuator 20 to operate the crystal manufacturing apparatus 10, and obtains an estimated value of the second physical quantity using the measurement value of the first sensor 17 and the learning model. The controller 30 controls the actuator 20 so as to reduce the deviation between the target value and the estimated value of the second physical quantity. The learning model can output a highly accurate estimated value, allowing for the production of high-quality single crystals.

[0038] The control device 30 will now be described. The control device 30 includes a basic model generation module 31, an experiment result acquisition module 32, a correlation data acquisition module 33, a correlation determination module 34, a model correction module 35, a learning model 36, and a main control module 37. Each module will now be described.

[0039] (Basic Model Generation Module) The basic model generation module 31 generates a basic learning model that serves as the basis for the high-precision learning model 36. Specifically, the basic model generation module 31 generates multiple simulation results (first simulation results) of pairs of a first physical quantity and a second physical quantity using a simulation model created based on structural data of the crystal production apparatus 10. The basic model generation module 31 then creates a basic learning model through machine learning using the first simulation results as training data. The structural data of the crystal production apparatus 10 typically refers to the physical properties (thermal conductivity, rigidity, etc.) of the materials that make up the crystal production apparatus 10. The structural data also includes the physical properties (viscosity, thermal expansion coefficient, boiling point, melting point, etc.) of the source gas. The simulation model can be analytically obtained using, for example, the finite element method (FEM). Simulation of the source gas also uses computational fluid dynamics (CFD: Computer Fluid Dynamics) technology. In FEM, the crystal production apparatus 10 and its internal space are divided into a large number of mesh points, and each mesh point is assigned its structural characteristics (such as thermal conductivity and the properties of the raw material gas), while an energy equation (such as a heat conduction equation or a fluid motion equation) is assigned between adjacent mesh points. A numerical value corresponding to the operation of the actuator 20 is assigned to the mesh point corresponding to the actuator 20, and the energy equation is solved to obtain the change over time in the physical quantity (such as temperature or pressure) at each mesh point. The physical quantity at the mesh point corresponding to the installation position of the first sensor 17 corresponds to the first physical quantity, and the physical quantity at the mesh point corresponding to the installation position of the second sensor 19 (sensor support 18) corresponds to the second physical quantity. A typical example of the second physical quantity is the surface temperature of the susceptor 13.

[0040] Basic model generation module 31 simulates the state of crystal production apparatus 10 when SiC crystal is produced using the simulation model described above. In other words, it simulates the state of crystal production apparatus 10 when actuator 20 is controlled so that crystal can be produced. Basic model generation module 31 obtains pairs (plurality of pairs) of a first physical quantity and a second physical quantity at different times in the simulation as first simulation results.

[0041] The basic model generation module 31 performs machine learning using multiple first simulation results as training data to generate a training model for the crystal production apparatus 10. For ease of explanation, the training model generated in this process is referred to as a basic training model. One first simulation result (a pair of a first physical quantity and a second physical quantity) corresponds to one training data. For machine learning, for example, a deep learning method of a neural network is used. FIG. 3 shows a schematic diagram of the basic training model. The input layer of the neural network has the same number of nodes as the number of first physical quantities in the first simulation results. In the example of FIG. 3, the output layer of the neural network has one node. The basic training model is obtained by training the weights of the neural network so that when the first physical quantity included in the first simulation results is input to the input layer, the second physical quantity included in the first simulation results is obtained from the output layer. In this example, the node value of the output layer is the surface temperature of the susceptor 13.

[0042] By inputting a new, different first physical quantity into the obtained basic learning model, an estimated value of the second physical quantity (susceptor surface temperature) corresponding to that first physical quantity can be obtained. However, the basic learning model does not easily reflect the accurate physical properties of each part of the crystal manufacturing apparatus 10 or the effects of deterioration over time. In other words, the output of the basic learning model cannot be expected to be highly accurate, taking deterioration over time into account. Therefore, the crystal manufacturing apparatus 10 disclosed in this specification identifies a correlation between the actual behavior of the crystal manufacturing apparatus 10 and the simulation results, and uses that correlation to correct the output of the basic learning model.

[0043] (Experimental Result Acquisition Module) The experimental result acquisition module 32 uses the crystal production apparatus 10 equipped with the first sensor 17 and the second sensor 19, and drives the actuator 20 under conditions under which SiC crystals cannot be produced but the second physical quantity can be observed. The second sensor 19, shown by the dashed line in Figures 1 and 2, is actually attached to the crystal production apparatus 10, and an experiment is conducted. The control device 20 drives the actuator 20 so that the internal temperature of the chamber 12 does not exceed the sensor's heat-resistant temperature. The experimental result acquisition module 32 acquires multiple experimental results, each consisting of a pair of the measurement value of the first sensor (first sensor measurement value) and the measurement value of the second sensor (second sensor measurement value). As mentioned above, the second sensor 19 is a sensor that measures the second physical quantity. For ease of explanation, hereinafter, the experiment from which the experimental result acquisition module 32 acquires data may be referred to as a preliminary experiment.

[0044] (Correlation Data Acquisition Module) The correlation data acquisition module 33 uses a simulation model to obtain a simulation result (second simulation result) of a second physical quantity when the first physical quantity matches the first sensor measurement value. The simulation model is the model used by the basic model generation module 31. The correlation data acquisition module 33 uses the simulation model to execute a simulation under the same conditions as the preliminary experiment and obtains an output (second physical quantity). The simulation output when the first physical quantity matches the first sensor measurement value is referred to as the second simulation result.

[0045] (Correlation Determination Module) The correlation determination module 34 determines the correlation between the second simulation result obtained by the correlation data acquisition module 33 and the second sensor measurement value (measurement value of the second sensor 19) obtained in the preliminary experiment. For ease of explanation, the second simulation result will be represented by the symbol Te, and the experimental result (second sensor measurement value) will be represented by the symbol Ts. FIG. 4 shows an example of a graph plotting a set (plural sets) of the second simulation result Te and the second sensor measurement value Ts. In FIG. 4, the horizontal axis represents the second simulation result Te, and the vertical axis represents the experimental result (measurement value Ts).

[0046] When the correlation is found using linear regression from the pair of the second simulation result Te and the second sensor measurement value Ts, a regression line of [second sensor measurement value Ts] = Ca × [second simulation result Te] + Cb is obtained. Here, Ca represents the slope of the regression line, and Cb represents the intercept of the regression line. Figure 4 is merely an example, and the correlation between the second simulation result Te and the experimental result (second sensor measurement value Ts) may be expressed as a polynomial or a nonlinear function.

[0047] The results of the preliminary experiment (second sensor measurement values) reflect the characteristics of the actual crystal production apparatus 10 (physical property values ​​of the structure and materials of the chamber 12, and physical property values ​​after deterioration over time). The second simulation results are data obtained from a simulation model and do not reflect all of the characteristics of the actual crystal production apparatus 10. Correlation from the second simulation results to the second sensor measurement values ​​is useful for correcting the results of the simulation model based on accurate physical property values ​​of the structure of the crystal production apparatus 10.

[0048] (Model Correction Module) The model correction module 35 generates a learning model 36 that corrects the output of the basic learning model using the correlation determined by the correlation determination module 34 and outputs the result as an estimated value (estimated value of the second physical quantity). Figure 5 shows an explanatory diagram of the final learning model 36. The learning model 36 is composed of a basic learning model and a correlation for the output of the basic learning model (the correlation equation obtained by the correlation determination module 34). When a new first physical quantity (the first physical quantity adjusted so that a crystal can be manufactured) is input to the basic learning model, an output (second simulation result Te) is obtained. The second simulation result Te is an estimate of the second physical quantity (e.g., the temperature at the surface 13a of the susceptor 13) during crystal manufacturing using the basic learning model. The final learning model 36 corrects the output of the basic learning model (second simulation result Te) using the correlation equation (final estimated value Tx = Ca × Te + Cb). The learning model 36 outputs the estimated value (final estimated value Tx) corrected using the correlation equation.

[0049] As mentioned above, the learning model 36 reflects the results of preliminary experiments using the actual crystal production apparatus 10. In other words, the learning model 36 reflects the accurate physical property values ​​of each part of the crystal production apparatus 10 and the effects of deterioration over time.

[0050] The advantages of the crystal manufacturing apparatus 10 of the embodiment (particularly the advantages of the technique for generating the learning model 36) will be explained. Generally, a huge amount of training data is required to generate a learning model using machine learning. In the crystal manufacturing apparatus 10, training data is obtained by simulation using an analytical simulation model of the crystal manufacturing apparatus 10. Because the simulation is performed using a computer, a large amount of training data can be automatically obtained by batch processing or the like.

[0051] When generating a learning model using experimental results as training data, multiple experiments must be conducted to obtain a large amount of training data, which is costly both financially and time-wise. The crystal manufacturing apparatus 10 disclosed in this specification uses simulation results to generate a basic learning model, and experimental results (results of preliminary experiments) are used to fine-tune the basic learning model. Since the experimental results are only used for fine-tuning, fewer samples are required than when generating training data for a learning model. The crystal manufacturing apparatus 10 (method of generating a learning model) of the embodiment can generate a learning model (a modified learning model of the crystal manufacturing apparatus 10) that can accurately predict the conditions inside the furnace in a shorter time than conventional methods.

[0052] The crystal manufacturing apparatus 10 also uses the first sensor 17 used in manufacturing the crystal to obtain data for correcting the output of the basic learning model (experimental results obtained by the experiment result acquisition module 32). Since the first sensor 17 used in actually manufacturing the crystal is also used to obtain data for correction, the accuracy of the correlation described above is increased.

[0053] The basic model generation module 31, experimental result acquisition module 32, correlation data acquisition module 33, correlation determination module 34, and model correction module 35 enable the acquisition of a highly accurate learning model (a model that obtains an estimated value of a second physical quantity from a first physical quantity) that reflects the characteristics of the actual crystal manufacturing apparatus 10.

[0054] When actually producing a SiC crystal, the main control module 37 performs feedback control of the actuator 20 based on the measurement value of the first sensor 17. The main control module 37 will now be described. Figure 6 shows a block diagram of the main control module 37. In Figure 6, the crystal production apparatus 10 is depicted in a simplified form as a single rectangle.

[0055] Main control module 37 includes a command generator 38 and a differentiator 39. Main control module 37 controls actuator 20 of crystal manufacturing apparatus 10 so that the second physical quantity (e.g., the surface temperature of susceptor 13) follows a predetermined target value. As mentioned above, the target value is preset to an optimum value for growing SiC crystal.

[0056] The command generating unit 38 generates a command value for the actuator 20 based on the target value of the second physical quantity. Note that the blocks in Fig. 6 do not operate and the actuator 20 is controlled in a predetermined sequence until the internal temperature of the chamber 12 reaches a predetermined initial temperature. Once the internal temperature of the chamber 12 reaches the initial temperature, the blocks in Fig. 6 function and the actuator 20 is controlled so that the second physical quantity follows the target value.

[0057] The crystal manufacturing apparatus 10 is equipped with a first sensor 17, and the measurement value of the first sensor 17 is input to the learning model 36. The learning model 36 outputs an estimated value corresponding to the measurement value of the first sensor 17 (an estimated value of the second physical quantity).

[0058] The command generation unit 38 receives a target value directly as well as the deviation between the target value and the estimated value. The deviation between the target value and the estimated value is generated by a differentiator 39. A control rule that combines feedforward and feedback is implemented in the command generation unit 38. The command generation unit 38 applies a first control rule to the target value to generate a feedforward command. At the same time, the command generation unit 38 applies a second control rule to the deviation between the target value and the estimated value to generate a feedback command. The final command is the sum of the feedforward command and the feedback command. The first and second control rules may be well-known control rules such as a PID (proportional-integral-derivative) control rule, a PI (proportional-integral) control rule, or a PD (proportional-derivative) control rule.

[0059] The command generation unit 38 provides the generated command to the actuator 20 to drive the crystal manufacturing apparatus 10 by feedback control. When the actuator 20 operates according to the command value, the measurement value of the first sensor 17 changes. The measurement value of the first sensor 17 is input to the learning model 36, so the estimated value (estimated value of the second physical quantity) output by the learning model 36 also changes. Although the deviation between the target value and the estimated value changes, the main control module 37 controls the actuator 20 of the crystal manufacturing apparatus 10 so as to reduce the deviation between the target value and the estimated value. The time required to receive the sensor measurement value from the apparatus and control the actuator is approximately the time required for input and output of the dynamic model (several msec to 1 sec).

[0060] As described above, the first physical quantity may include the output of the actuator 20. The command value to the actuator 20 corresponds to an approximation of the output of the actuator 20. Therefore, when the first physical quantity includes the output of the actuator 20, the command value to the actuator 20 may be input to the learning model 36 as a part of the first physical quantity. The dotted arrow line in FIG. 6 represents a case where the command value to the actuator 20 is input to the learning model 36.

[0061] In summary, main control module 37 executes the following process. Main control module 37 utilizes learning model 36. Learning model 36 outputs an estimated value of a second physical quantity that cannot be observed when manufacturing SiC crystal from the measurement value of first sensor 17. Main control module 37 controls actuator 20 so that the deviation between the estimated value output by learning model 36 and the target value of the second physical quantity when manufacturing SiC crystal is reduced. Since the second physical quantity (its estimated value), which is closely related to crystal growth, can be brought closer to the target value, high-quality SiC crystal can be manufactured.

[0062] The advantages of the control device 30 equipped with the learning model 36 will be described. The learning model 36 is composed of a basic learning model based on machine learning and correlation. The basic learning model is a model based on machine learning, and since there is no need to solve complex equations, results can be obtained quickly. Correlation is a relational expression that represents the relationship between the experimental results (measured values) and the corresponding simulation output (second output). Correlation is expressed by a relational expression such as a regression line or an approximate polynomial, and allows for high-speed calculation. In other words, using the learning model 36 allows an estimated value of the second physical quantity to be obtained quickly.

[0063] In addition, the output of the basic learning model is corrected by correlation based on the results of preliminary experiments using the actual crystal manufacturing apparatus 10. The correction increases the accuracy of the estimated value. By using the learning model 36, an accurate estimate of the second physical quantity, which cannot be observed during crystal manufacturing, can be obtained quickly. Therefore, the control device 30 using the learning model 36 is suitable for feedback control. By controlling the actuator 20 of the crystal manufacturing apparatus 10 with the control device 30 that performs feedback control using the learning model 36, high-quality SiC crystals can be obtained.

[0064] The output of the learning model 36 is an estimated value of the second physical quantity, not a measured value. Feedback control using an estimated value instead of a measured value may be called pseudo-feedback control.

[0065] The control device 30 of the embodiment is effective even when used alone. That is, the present specification provides a novel crystal manufacturing apparatus 10 and a novel control device 30 for controlling the crystal manufacturing apparatus.

[0066] (Second Example) This specification also provides a method for generating a learning model suitable for controlling a SiC crystal manufacturing apparatus. This method generates a learning model that obtains an estimate of an unobservable second physical quantity from an observable first physical quantity when manufacturing SiC crystals using the SiC crystal manufacturing apparatus. Figure 7 shows a flowchart of the learning model generation method. This method includes a basic model generation step (step S2), an experimental result acquisition step (step S3), a correlation data acquisition step (step S4), a correlation determination step (step S5), and a model correction step (step S6).

[0067] In the basic model generation step (step S2), a basic learning model is created by machine learning using multiple simulation results (first simulation results) of pairs of first physical quantities and second physical quantities created using a simulation model based on structural data of the SiC crystal manufacturing equipment as training data. Details of the basic model generation step are the same as the processing of the basic model generation module 31 described above.

[0068] The experimental result acquisition process (step S3) uses a SiC crystal manufacturing apparatus equipped with a first sensor that measures a first physical quantity and a second sensor that measures a second physical quantity, and drives the actuator of the SiC crystal manufacturing apparatus under conditions in which SiC crystals cannot be manufactured but the second physical quantity can be observed. This experiment is referred to as a preliminary experiment. The experimental result acquisition process acquires multiple experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement values) and measurement values ​​of the second sensor (second sensor measurement values) in the preliminary experiment. Details of the experimental result acquisition process are the same as the processing of the experimental result acquisition module 32 described above.

[0069] In the correlation data acquisition step (step S4), a simulation result (second simulation result) of the second physical quantity when the first physical quantity matches the first sensor measurement value is obtained using a simulation model. Details of the correlation data acquisition step are the same as the processing of the correlation data acquisition module 33 described above. In the correlation determination step (step S5), a correlation between the second simulation result and the second sensor measurement value is determined. Details of the correlation determination step are the same as the processing of the correlation determination module 34 described above.

[0070] In the model correction step (step S6), a learning model 36 is generated that outputs the result of correcting the output of the basic learning model by the above-mentioned correlation as an estimated value. The details of the model correction step are the same as the processing of the model correction module 35 described above.

[0071] This specification also provides a control method using the learning model 36 obtained by the above-described generation method. In this method, first, a measurement value of a first sensor is obtained from a crystal manufacturing apparatus 10 operating under conditions for manufacturing SiC crystal. Next, the measurement value of the first sensor is input into the learning model to obtain an estimated value of a second physical quantity. Then, the SiC crystal manufacturing apparatus is controlled so as to reduce the deviation between the target value and the estimated value of the second physical quantity. This control method enables the manufacturing of high-quality SiC crystal.

[0072] Next, a typical example of a learning model generation method will be described. The generation method described below is just an example, and the technology disclosed in this specification is not limited to the following example.

[0073] The learning model outputs an estimated value of a second physical quantity at a predetermined position exceeding the sensor's heat resistance temperature from a first physical quantity observable when manufacturing SiC crystals using a SiC crystal manufacturing apparatus. This method includes a basic learning model generation process, an experimental result acquisition process, a correlation data acquisition process, a correlation determination process, and a model correction process. For ease of explanation, the predetermined position exceeding the sensor's heat resistance temperature when manufacturing SiC crystals will be referred to as the "position of interest" below. A typical example of the position of interest is the susceptor surface, and a typical example of the second physical quantity is temperature (or pressure).

[0074] In the basic learning model generation process, a simulation is performed under conditions for manufacturing SiC crystals using a simulation model based on structural data of the SiC crystal manufacturing equipment, and multiple simulation results (first simulation results) for pairs of a first physical quantity and a second physical quantity are obtained. Next, a basic learning model is generated by machine learning using the multiple first simulation results as training data. The basic learning model outputs an estimated value of the second physical quantity when the first physical quantity is input.

[0075] In the experimental result acquisition process, an auxiliary sensor (corresponding to the second sensor 19 described above) that measures the second physical quantity is attached to the position of interest, and the SiC crystal manufacturing equipment is operated so that the temperature at the position of interest is below the sensor's heat resistance temperature, even though SiC crystal cannot be manufactured, and the measurement value of the auxiliary sensor (auxiliary sensor measurement value) is acquired. Operating the SiC crystal manufacturing equipment so that the temperature at the position of interest is below the sensor's heat resistance temperature, even though SiC crystal cannot be manufactured, is referred to as a preliminary experiment.

[0076] In the correlation data acquisition process, a simulation is performed using a simulation model under the same conditions as the preliminary experiment, and a second simulation result corresponding to the auxiliary sensor measurement value obtained in the preliminary experiment is obtained. The second simulation result corresponds to an estimated value of the second physical quantity at the position of interest (an estimated value before the correction described below). In the correlation determination process, the correlation between the second simulation result and the auxiliary sensor measurement value is determined. An example of this correlation is the regression line described above. In the model correction process, the first physical quantity obtained when operating the crystal manufacturing equipment under conditions for manufacturing SiC crystals is input into the basic learning model, and a learning model is generated that corrects the output of the basic learning model at that time using the correlation described above and outputs the result as an estimated value (an estimated value of the second physical quantity).

[0077] The first physical quantity includes the temperature (or pressure) at multiple locations in the SiC crystal manufacturing equipment. In the preliminary experiment, the SiC crystal manufacturing equipment is controlled so that the temperature (or pressure) at each of the multiple locations is different. By conducting the preliminary experiment in this manner, the structural characteristics between each location in the SiC crystal manufacturing equipment and the position of interest are well reflected.

[0078] For convenience, the temperature at the aforementioned position of interest will be referred to as the "temperature of interest" below. The relationship between the temperatures at each location and the temperature of interest depends on the structure of the crystal manufacturing apparatus 10 and is generally the same whether the temperature of interest is high or low. Therefore, the generation method of the above embodiment measures the temperatures at each location and the temperature of interest of the crystal manufacturing apparatus 10 operated so as not to exceed the sensor's heat-resistant temperature through a preliminary experiment. The temperature of interest is also obtained when a simulation is performed under the same conditions as the preliminary experiment. The relationship (the correlation described above) between the temperature of interest obtained in the preliminary experiment (measured value by the second sensor) and the temperature of interest obtained in the simulation (second simulation result) generally holds regardless of the absolute value of the temperature. The correlation between the simulation results and the preliminary experiment results when the temperature is low holds even when the temperature is high. By using such a correlation, it becomes possible to accurately estimate the temperature at the position of interest that exceeds the sensor's heat-resistant temperature during crystal manufacturing.

[0079] Some points to note regarding the technology described so far are described below. The learning models in the examples (basic learning model, final learning model 36) were multi-input, single-output models. The technology disclosed in this specification can also use multi-input, multi-output learning models. That is, the second physical quantity may also be multiple. Furthermore, the technology disclosed in this specification may also employ learning models with different numbers of nodes in each layer (neural network-type learning models) or other learning algorithms (e.g., Gaussian process regression).

[0080] The technology disclosed in this specification creates a basic learning model using training data obtained through simulation. All training data for obtaining the basic learning model is obtained through simulation. A large amount of training data can be obtained in a relatively short time. Meanwhile, the correlation between the results of an experiment using an actual crystal production device and the simulation results corresponding to the experiment is determined. Since the experiment to obtain sensor measurement values ​​is intended to obtain a correlation between the simulation results and the experimental results (results that reflect the state of the actual crystal production device), a smaller amount of training data is sufficient compared to the training data used for the learning model.

[0081] The second physical quantity refers to a physical quantity that cannot be observed during the manufacture of the SiC crystal. Here, "unobservable during the manufacture of the SiC crystal" includes cases where the quantity can be observed at great cost, but such great cost is inappropriate for practical use.

[0082] If the crystal manufacturing apparatus 10 has been operated for a long period of time after the learning model 36 has been created, the experiment result acquisition module 32, correlation data acquisition module 33, correlation determination module 34, and model correction module 35 may be started again. The learning model 36 can be updated according to the latest state of the crystal manufacturing apparatus 10.

[0083] The technology disclosed in this specification is not limited to the crystal manufacturing apparatus shown in FIG. 1, but is also suitable for application to other crystal manufacturing apparatuses (e.g., crystal manufacturing apparatuses based on sublimation methods, solution methods, HT-CVD methods, and CVD methods).

[0084] Although specific examples of the present invention have been described in detail above, these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and variations of the specific examples exemplified above. The technical elements described in this specification or drawings exhibit technical utility alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. Furthermore, the technology exemplified in this specification or drawings can achieve multiple objectives simultaneously, and achieving one of these objectives alone is technically useful. [Explanation of symbols]

[0085] 10: SiC crystal manufacturing apparatus 11: Housing 12: Chamber 13: Susceptor 14: Rotating device 17, 17a-17e: First sensor 18, 18a-18d: Sensor fixing part 19: Second sensor 20: Actuator 21: Rotating device 22: Heater 22a: High frequency coil 22b: High frequency generator 23: Gas supply device 24: Cooler 24a: Circulator 24b: Cooling pipe 30: Control device 31: Basic model generation module 32: Experimental result acquisition module 33: Correlation data acquisition module 34: Correlation determination module 35: Model correction module 36: Learning model 37: Main control module 38: Command generation unit 39: Differential calculator

Claims

1. A SiC crystal manufacturing apparatus, comprising: a chamber in which SiC crystals are produced from a source material; an actuator for controlling the manufacturing conditions for SiC crystal generation; a first sensor that measures a first physical quantity that is observable during the production of the SiC crystal in the chamber; a control device having a learning model that outputs an estimated value of a second physical quantity that is unobservable when manufacturing the SiC crystal from the measurement value of the first sensor, and that feedback controls the actuator when manufacturing the SiC crystal so that a difference between the estimated value output by the learning model and a target value of the second physical quantity becomes small; It is equipped with The control device a basic model generation module that generates a basic learning model by machine learning using training data of a plurality of simulation results (first simulation results) of pairs of the first physical quantity and the second physical quantity, which are generated using a simulation model based on structural data of the SiC crystal manufacturing apparatus; an experimental result acquisition module that acquires a plurality of experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement value) and measurement values ​​of the second sensor (second sensor measurement value) when the actuator is driven under conditions in which the SiC crystal cannot be manufactured but the second physical quantity can be observed in the SiC crystal manufacturing apparatus equipped with the first sensor and a second sensor that measures the second physical quantity; a correlation data acquisition module that obtains a simulation result (second simulation result) of the second physical quantity when the first physical quantity matches the first sensor measurement value using the simulation model; a correlation determination module that determines a correlation between the second simulation results and a second sensor measurement; a model correction module for generating the learning model, which corrects the output of the basic learning model by the correlation and outputs the result as the estimated value; A SiC crystal manufacturing apparatus comprising:

2. the first sensor is disposed in a location where the heat resistance temperature of the first sensor is not exceeded during production of the SiC crystal; a fixing portion of the second sensor is located inside the chamber at a location where the temperature exceeds the heat resistance temperature of the second sensor during manufacturing of the SiC crystal; 2. The SiC crystal manufacturing apparatus of claim 1, wherein the conditions under which the SiC crystal cannot be manufactured but the second physical quantity can be observed include the temperature inside the chamber not exceeding the heat resistance temperature of the second sensor.

3. 3. The SiC crystal manufacturing apparatus according to claim 1, wherein the second physical quantity includes a surface temperature of a susceptor that supports a substrate on which the SiC crystal is grown within the chamber.

4. 4. The SiC crystal manufacturing apparatus of claim 3, wherein the first physical quantity includes any one of a temperature of an outer wall of the chamber, a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the raw material supplied to the chamber, and a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the residual gas discharged from the chamber.

5. 3. The SiC crystal manufacturing apparatus according to claim 1, wherein the first physical quantity includes any one of a temperature of an outer wall of the chamber, a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the raw material supplied to the chamber, and a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the residual gas discharged from the chamber.

6. a chamber in which SiC crystals are produced from a source material; an actuator for controlling the manufacturing conditions for SiC crystal generation; a first sensor that measures a first physical quantity that is observable during manufacturing of the SiC crystal in the chamber, The control device a main control module having a learning model that outputs an estimated value of a second physical quantity that is unobservable when manufacturing the SiC crystal from the measurement value of the first sensor, and that feedback-controls the actuator when manufacturing the SiC crystal so that a difference between the estimated value output by the learning model and a target value of the second physical quantity becomes small; a basic model generation module that generates a basic learning model by machine learning using training data of a plurality of simulation results (first simulation results) of pairs of the first physical quantity and the second physical quantity, which are generated using a simulation model based on structural data of the SiC crystal manufacturing apparatus; a measurement value acquisition module that acquires a plurality of experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement value) and measurement values ​​of the second sensor (second sensor measurement value) when the actuator is driven under conditions in which the SiC crystal cannot be manufactured but the second physical quantity can be observed in the SiC crystal manufacturing apparatus equipped with the first sensor and a second sensor that measures the second physical quantity; a correlation data acquisition module that obtains a simulation result (second simulation result) of the second physical quantity when the first physical quantity matches the first sensor measurement value using the simulation model; a correlation determination module that determines a correlation between the second simulation results and a second sensor measurement; a model correction module for generating the learning model, which corrects the output of the basic learning model by the correlation and outputs the result as the estimated value; A control device for a SiC crystal manufacturing apparatus comprising:

7. the first sensor is disposed in a location where the heat resistance temperature of the first sensor is not exceeded during production of the SiC crystal; a fixing portion of the second sensor is located inside the chamber at a location where the temperature exceeds the heat resistance temperature of the second sensor during manufacturing of the SiC crystal; 7. The control device according to claim 6, wherein the condition under which the SiC crystal cannot be produced but the second physical quantity can be observed includes a condition that the inside of the chamber does not exceed a heat-resistant temperature of the second sensor.

8. 8. The control device according to claim 6, wherein the second physical quantity includes a surface temperature of a susceptor that supports a substrate on which the SiC crystal is grown within the chamber.

9. 9. The control device according to claim 8, wherein the first physical quantity includes any one of a temperature of an outer wall of the chamber, a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the source material supplied to the chamber, and a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the residual gas discharged from the chamber.

10. 8. The control device according to claim 6, wherein the first physical quantity includes any one of a temperature of an outer wall of the chamber, a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the source material supplied to the chamber, and a temperature, a pressure, a flow rate, a flow velocity, or a concentration of a residual gas discharged from the chamber.

11. A method for generating a learning model for obtaining an estimated value of a second physical quantity that is not observable by a second sensor from a first physical quantity that is observable by a first sensor when manufacturing a SiC crystal using a SiC crystal manufacturing apparatus, a basic model generation process for generating a basic learning model by machine learning using training data of a plurality of simulation results (first simulation results) of pairs of the first physical quantity and the second physical quantity generated using a simulation model based on structural data of the SiC crystal manufacturing apparatus; an experimental result acquisition process for acquiring a plurality of experimental results of pairs of measurement values ​​of the first sensor (first sensor measurement value) and measurement values ​​of the second sensor (second sensor measurement value) when an actuator of the SiC crystal manufacturing apparatus, to which the first sensor for measuring the first physical quantity and the second sensor for measuring the second physical quantity are attached, is driven under conditions in which the SiC crystal cannot be manufactured but the second physical quantity can be observed; a correlation data acquisition step of obtaining a simulation result (second simulation result) of the second physical quantity when the first physical quantity coincides with the measurement value of the first sensor using the simulation model; determining a correlation between the second simulation results and second sensor measurements; a model correction step of generating the learning model, which corrects the output of the basic learning model by the correlation and outputs the result as the estimated value; A method for generating a learning model for a SiC crystal manufacturing apparatus, comprising:

12. the first sensor is disposed in a location where the heat resistance temperature of the first sensor is not exceeded during production of the SiC crystal; the fixing portion of the second sensor is located inside a chamber in which the SiC crystal is produced from a raw material, at a location where the temperature exceeds the heat-resistant temperature of the second sensor during production of the SiC crystal; The method according to claim 11, wherein the condition under which the SiC crystal cannot be produced but the second physical quantity can be observed includes a condition under which the temperature inside the chamber does not exceed a heat-resistant temperature of the second sensor. How to generate the above.

13. 13. The method according to claim 11, wherein the second physical quantity includes a surface temperature of a susceptor that supports a substrate on which the SiC crystal is grown in a chamber in which the SiC crystal is produced from a source material.

14. 14. The method according to claim 13, wherein the first physical quantity includes any one of a temperature of an outer wall of the chamber, a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the raw material of the SiC crystal that is supplied to the chamber, and a temperature, a pressure, a flow rate, a flow velocity, or a concentration of the residual gas exhausted from the chamber.

15. 13. The method according to claim 11, wherein the first physical quantity includes any one of the temperature of an outer wall of a chamber in which the SiC crystal is produced from a raw material, the temperature, pressure, flow rate, flow velocity, or concentration of the raw material of the SiC crystal that is supplied to the chamber, and the temperature, pressure, flow rate, flow velocity, or concentration of the residual gas exhausted from the chamber.

16. A control method using the learning model obtained by the generation method according to claim 11 or 12, acquiring a measurement value of the first sensor from the SiC crystal manufacturing apparatus operating under conditions for manufacturing the SiC crystal; inputting the measurement value of the first sensor into the learning model to obtain the estimated value; a control method for controlling the SiC crystal manufacturing apparatus so as to reduce a difference between a target value of the second physical quantity and the estimated value thereof.

Images