Method for manufacturing silicon carbide single crystals, regression model generation apparatus, composition estimation apparatus, program, recording medium, regression model generation method, and composition estimation method

Abstract

Provided is a SiC single crystal production method in which, in a solution growth method, a used solution used in production of SiC single crystals is reused as a raw material for production of SiC single crystals again. This silicon carbide single crystal production method involves (a) a step for preparing a solution 20 containing silicon and carbon, and (b) a step for bringing silicon carbide seed crystals 30 into contact with the solution 20 to grow silicon carbide single crystals on crystal growth planes 30a of the seed crystals 30, wherein a solidified product of a solution used in a silicon carbide single crystal production method previously carried out is used as a raw material for the solution 20 in step (a).

Description

Technical Field

[0001] The present invention relates to a method for manufacturing a single crystal made of silicon carbide.

Background Art

[0002] For example, an inverter circuit is used as a circuit for controlling a motor included in an automobile, home appliances, etc. In this inverter circuit, power semiconductor devices typified by a power MOSFET (Metal Oxide Semiconductor Field Effect Transistor) and an IGBT (Insulated Gate Bipolar Transistor) are used.

[0003] Such power semiconductor devices are required to have, for example, a high breakdown voltage, a low on-resistance, and a low switching loss in addition to a high breakdown voltage. Here, the current mainstream of power semiconductor devices is a field effect transistor formed on a semiconductor substrate mainly composed of silicon, but this power semiconductor device is approaching its theoretical performance limit.

[0004] Regarding this point, semiconductor devices including field effect transistors formed on a semiconductor substrate mainly composed of a semiconductor material having a larger bandgap than silicon (hereinafter referred to as wide bandgap power semiconductor devices) have attracted attention.

[0005] This is because a large bandgap means having a high breakdown strength, making it easier to achieve a high breakdown voltage.

[0006] And when the semiconductor material itself has a high breakdown strength, even if the drift layer for maintaining the breakdown voltage is made thin, the breakdown voltage can be ensured. Therefore, for example, by making the drift layer thin and increasing the impurity concentration, the on-resistance of the power semiconductor device can be reduced.

[0007] In other words, wide-bandgap power semiconductor devices excel in their ability to achieve both improved breakdown voltage and reduced on-resistance, which are typically in a trade-off relationship. Therefore, wide-bandgap power semiconductor devices are expected to be high-performance semiconductor devices.

[0008] Semiconductor materials with a larger band gap than silicon include, for example, silicon carbide (SiC), gallium nitride (GaN), gallium oxide (Ga2O3), and diamond. The following explanation will focus on silicon carbide.

[0009] Single crystals made of silicon carbide (hereinafter also referred to as SiC single crystals) can be manufactured by methods such as sublimation, high-temperature gas growth, and solution growth.

[0010] Sublimation is a method of evaporating silicon and carbon from raw materials at high temperatures of over 2000°C, and then condensing a SiC single crystal onto a silicon carbide seed crystal, which is at a lower temperature than the raw materials. However, because the growth interface is between the gas phase and the solid phase, and the temperature difference between them is large, the grown crystal has a high dislocation density, which presents challenges in terms of crystal quality.

[0011] The high-temperature gas growth method involves introducing a gaseous raw material at temperatures above 2000°C and growing a SiC single crystal on a silicon carbide seed crystal through a chemical reaction. Similar to the sublimation method, this method involves a growth interface between the gas phase and the solid phase, and due to the large temperature difference, the grown crystal has a high dislocation density, resulting in challenges regarding crystal quality.

[0012] In contrast to these methods, the solution growth method involves immersing a silicon carbide seed crystal in a solution containing silicon and carbon, creating a temperature gradient within the solution, and generating a carbon supersaturated state near the growth interface to grow a SiC single crystal. In this method, the growth interface is between the liquid and solid phases, and the temperature difference is smaller compared to the above methods, bringing it closer to thermal equilibrium. As a result, the grown crystal has a low dislocation density and good crystal quality.

[0013] However, solution growth methods have the drawback of slower crystal growth rates compared to sublimation methods. To address this, increasing carbon solubility can be considered to improve the growth rate, and one known method for doing so is to use transition elements such as chromium (Cr) in the solution.

[0014] Furthermore, it is known that adding aluminum (Al), which has low surface tension, to a Si-Cr solution can suppress roughness at the crystal growth interface and improve crystal quality (see, for example, Non-Patent Document 1).

[0015] Furthermore, since yttrium (Y) is less easily incorporated into SiC single crystals compared to Cr and Ti, a method is known that reduces the metal incorporation by including Y in the solution, thereby enabling a reduction in the concentration of metal impurities in SiC single crystals (see, for example, Patent Document 1).

[0016] Furthermore, a method is known for growing high-quality SiC single crystals by suppressing the inclusion of polycrystalline material by incorporating rare earth elements that have a melting point lower than that of silicon (see, for example, Patent Document 2). [Prior art documents] [Patent Documents]

[0017] [Patent Document 1] Japanese Patent Publication No. 2019-19037 [Patent Document 2] Japanese Patent Publication No. 2019-104661 [Patent Document 3] Japanese Patent Publication No. 2018-16498 [Non-patent literature]

[0018] [Non-Patent Document 1] T. Mitani, et al.: J. Cryst. Growth, 401 (2014) p681-685. [Overview of the Initiative]

Problems to be Solved by the Invention

[0019] In the production of SiC single crystals by the solution growth method described above, generally, a crucible made of graphite (C) or silicon carbide (SiC) is used. Using this crucible as a carbon source, a solution containing carbon and silicon at a predetermined concentration is held in the crucible.

[0020] After the SiC single crystal has grown sufficiently, the manufacturing apparatus for the SiC single crystal is cooled from the high-temperature heating state to room temperature, and the manufacturing process ends. At this time, the used solution solidifies upon cooling, but due to the difference in the thermal contraction rate with the crucible during cooling, deformation or collapse of the crucible may occur. Further, in the production of SiC single crystals, since the dissolution of carbon (C) progresses in the crucible, the internal shape of the crucible changes, and in addition to the consumption of silicon (Si) and carbon (C) in the used solution, the composition changes due to the precipitation of hetero-crystals or the like during cooling. Hetero-crystals refer to 4H-SiC or SiC with different polytypes that occur in addition to the SiC crystals grown from the seed crystal.

[0021] Therefore, the used solution and crucible used in the production of SiC single crystals are discarded, and in the next production of SiC single crystals, a new solution and crucible are prepared. As described above, in the solution growth method, elements such as Cr and Al are added to improve the solubility of carbon, etc. However, since a new solution is used each time of production, these added elements are discarded together with the used solution.

[0022] However, added elements such as Cr and Al are components that are not consumed during the production of SiC single crystals. If these added elements can be reused as raw materials for the production of SiC single crystals again, resources can be effectively utilized and the production cost can also be reduced.

[0023] Therefore, an object of the present invention is to provide a method for producing SiC single crystals that reuses a used solution used in the production of SiC single crystals as a raw material for the production of SiC single crystals again in the solution growth method.

[0024] Furthermore, in a method for producing silicon carbide single crystal ingots using the sublimation recrystallization method, a method is known in which, after producing an ingot, recycled raw material powder is obtained by regenerating the raw material residue remaining in the crucible (see Patent Document 3). [Means for solving the problem]

[0025] A method for producing silicon carbide single crystals in one embodiment comprises the steps of (a) preparing a solution containing silicon and carbon, and (b) contacting a silicon carbide seed crystal with the solution and growing a silicon carbide single crystal on the crystal growth surface of the seed crystal, wherein the raw material for the solution in step (a) is a solidified product of a solution previously used in a silicon carbide single crystal production method.

[0026] In one embodiment, the regression model generation device is included in a composition estimation system that estimates the composition of a used solution used in single crystal growth by the solution growth method. This regression model generation device includes a regression model generation unit that, upon input of a pre-use composition value indicating the composition of the solution before use and manufacturing conditions for growing a single crystal, generates a regression model that outputs an estimated value for the composition of the used solution.

[0027] One embodiment of the program is a program that causes a computer to perform a process to estimate the composition of a used solution used in single crystal growth by the solution growth method. This program includes a regression model generation process that, upon input of a pre-use composition value indicating the composition of the solution before use and manufacturing conditions for growing the single crystal, generates a regression model that outputs an estimated value for the composition of the used solution.

[0028] The program described above is recorded on a computer-readable storage medium.

[0029] One embodiment of the regression model generation method is a method in which a computer generates a regression model for estimating the composition of a used solution used in single crystal growth by the solution growth method. This regression model generation method includes a regression model generation step in which a computer generates a regression model that outputs an estimated value for the composition of the used solution when a pre-use composition value indicating the composition of the solution before use and manufacturing conditions for growing the single crystal are input.

[0030] In one embodiment, the composition estimation device is included in a composition estimation system for estimating the composition of a used solution used in single crystal growth by the solution growth method. This composition estimation device includes an estimation unit that estimates the composition of the used solution based on a pre-use composition value indicating the composition of the solution before use, manufacturing conditions for growing the single crystal, and a regression model. Here, the regression model is a function that takes the pre-use composition value and manufacturing conditions as input and outputs an estimated value for the composition of the used solution.

[0031] One embodiment of the program is a program that causes a computer to perform a process to estimate the composition of a used solution used in single crystal growth by the solution growth method. This program includes a process to estimate the composition of the used solution based on the pre-use composition value, which indicates the composition of the solution before use, the manufacturing conditions for growing the single crystal, and a regression model. Here, the regression model is a function that takes the pre-use composition value and manufacturing conditions as input and outputs an estimated value for the composition of the used solution.

[0032] The program described above is recorded on a computer-readable storage medium.

[0033] One embodiment of the composition estimation method is a method by which a computer estimates the composition of a used solution used for single crystal growth by the solution growth method. This composition estimation method comprises a step in which a computer estimates the composition of the used solution based on a pre-use composition value indicating the composition of the solution before use, manufacturing conditions for growing the single crystal, and a regression model. Here, the regression model is a function that takes the pre-use composition value and manufacturing conditions as input and outputs an estimated value for the composition of the used solution. [Effects of the Invention]

[0034] According to one embodiment of the method for producing silicon carbide single crystals, the amount of raw materials used can be reduced and silicon carbide single crystals can be produced efficiently by reusing the raw materials used in the production of SiC single crystals in the solution growth method. [Brief explanation of the drawing]

[0035] [Figure 1] This is a schematic cross-sectional view illustrating an example of the configuration of a manufacturing apparatus used in a method for manufacturing silicon carbide single crystals according to one embodiment. [Figure 2] This is a schematic cross-sectional view showing the crystal orientation of the silicon carbide seed crystal used in a method for manufacturing silicon carbide single crystals according to one embodiment. [Figure 3] This diagram illustrates an example of how to cut out solidified waste from used solutions that are to be reused. [Figure 4] This figure shows an example of the hardware configuration of a composition estimation device. [Figure 5] This is a functional block diagram showing the functions of the composition estimation device. [Figure 6] This table shows an example of related data. [Figure 7] This diagram schematically illustrates the process of generating a regression model using machine learning with relevant data as training data. [Figure 8] This is a flowchart explaining the process of generating regression models. [Figure 9] This is a flowchart explaining the process of estimating the composition of a used solution. [Modes for carrying out the invention]

[0036] In all the drawings illustrating the embodiments, the same reference numeral is generally used for identical components, and repeated explanations of the same components are omitted. Furthermore, to improve clarity, hatching may be added to plan views, or omitted to cross-sectional views.

[0037] <Consideration of improvements> When growing silicon carbide single crystals (SiC single crystals) using solution growth, it is necessary to create a carbon supersaturated state in the solution used for crystal production in order to precipitate the SiC single crystals. Therefore, in solution growth methods for SiC single crystals, a temperature gradient is created in the solution to create a carbon supersaturated state.

[0038] In this case, a high-temperature region and a low-temperature region are formed in the solution, and a carbon supersaturation state is achieved in the low-temperature region. Therefore, by creating a temperature gradient in the solution so that the region of the solution in contact with the silicon carbide seed crystal is the low-temperature region where a carbon supersaturation state is achieved, it is possible to grow crystals on the seed crystal.

[0039] The carbon supersaturation state in the low-temperature region described above is mainly formed on the surface of a solution containing carbon (C) and silicon (Si). By bringing a seed crystal into contact with this solution surface, a SiC single crystal can be grown as described above. The production of this SiC single crystal is carried out inside an insulating material (insulating container) arranged to surround a crucible containing the solution, while maintaining the temperature inside this insulating material at a temperature suitable for single crystal production.

[0040] As mentioned in the above problem, solutions containing Cr and Al are known to be used here to suppress the growth rate of SiC single crystals and the roughness of the crystal growth interface that is formed. However, when these elements are included, a new solution is used each time manufacturing is carried out, and these elements are discarded along with the used solution.

[0041] Therefore, the present inventors have found a method for manufacturing SiC single crystals that allows for the effective use of resources by reusing these elements, and also enables the production of new SiC single crystals by adjusting the composition so that the elemental concentrations remain the same even after reuse.

[0042] <Embodiment> One embodiment of the present invention will be described in detail below.

[0043] In describing this embodiment, we will first describe the manufacturing apparatus used for manufacturing SiC single crystals.

[0044] [Equipment for manufacturing silicon carbide single crystals] Figure 1 is a schematic diagram illustrating the configuration of the silicon carbide single crystal manufacturing apparatus 101 used in the manufacturing method of SiC single crystals in this embodiment (in use).

[0045] In Figure 1, the single crystal manufacturing apparatus 101 is configured to include a crucible 10, a furnace 11, an induction coil 12, a crucible holding shaft 13, and a crystal holding shaft 14.

[0046] The crucible 10 is a container that holds a crystal production solution 20, which is the raw material for growing SiC single crystals, inside. By making the crucible 10 from, for example, graphite (C) or silicon carbide (SiC), it can be made into a high-temperature solution 20 containing silicon (Si) and carbon (C) when crystal growth is performed.

[0047] The furnace 11 is a component that can accommodate a crucible 10 inside and forms a field for manufacturing single crystals. The furnace 11 is shaped to surround the crucible 10 and can maintain a predetermined high temperature (hot zone) inside. The furnace 11 is equipped with a crucible holding shaft 13, to which the crucible 10 can be attached (described later), from its lower part, and a crystal holding shaft 14, to which a silicon carbide seed crystal 30 can be attached, from its upper part.

[0048] Furthermore, an induction coil 12 through which a high-frequency current flows is provided on the outer periphery of the furnace 11, and the crucible 10 can be heated by induction heating based on the high-frequency current flowing through the induction coil 12. Specifically, the induction coil 12 is positioned on either side of the furnace 11, facing the side of the crucible 10, and the crucible 10 is heated by the induction heating phenomenon caused by passing a high-frequency current through this induction coil 12. Although not shown in Figure 1, the induction coil 12 is configured to allow cooling water to flow through its interior.

[0049] Although not shown in the diagram, a housing for the single crystal manufacturing apparatus is formed to surround the furnace 11 and induction coil 12, and its internal space can be filled with an inert gas, such as argon gas, for single crystal manufacturing. This housing is made of an iron-based material such as SUS.

[0050] Furthermore, as a structure for filling with argon gas, for example, a structure can be adopted in which a quartz tube is passed between the furnace 11 and the induction coil 12, and the upper and lower ends of the quartz tube are sealed with flanges to fill with argon gas.

[0051] The crucible holding shaft 13 is a component that fixes and holds the crucible 10 at its upper end, and is configured to be movable in the vertical direction. Furthermore, it may be configured to be able to rotate in either a clockwise or counterclockwise direction. This allows the crucible 10 attached to the crucible holding shaft 13 to be moved in the vertical direction, and also rotated as needed. A base may be attached to the crucible holding shaft 13, and the crucible 10 may be fixed to this base.

[0052] Furthermore, the crucible holding shaft 13 may be configured to have a hollow structure inside, allowing for the insertion of a thermocouple or serving as a path for the measurement light of a radiation thermometer to pass through, thereby enabling temperature measurement at the bottom of the crucible (near the crucible).

[0053] The crystal holding shaft 14 is a component that allows a silicon carbide seed crystal 30 to be attached to its lower end (tip), and further allows its tip to be positioned inside the furnace 11. The crystal holding shaft 14 is also configured to be movable in the vertical direction. Figure 1 shows the state with the seed crystal 30 attached. By moving the crystal holding shaft 14 vertically, the seed crystal 30 attached to its tip also moves vertically in conjunction, and single crystals are manufactured by adjusting its height position.

[0054] The crystal holding axis 14, like the crucible holding axis 13, may be configured to rotate clockwise or counterclockwise. In other words, the presence or absence of a rotation mechanism in the crystal holding axis 14 and the crucible holding axis 13 is optional.

[0055] [Method for producing silicon carbide single crystals] Next, we will explain the method for manufacturing silicon carbide single crystals, using the silicon carbide single crystal manufacturing apparatus 101 described above, and using a crucible 10 made of graphite (C) or silicon carbide (SiC) as an example.

[0056] (a) Steps to prepare the solution As described above, the solution 20 is placed inside the crucible 10. To do this, first, the raw materials for the solution 20, such as silicon (Si) raw materials, and if necessary, carbon (C) raw materials and additive element raw materials such as chromium (Cr) and aluminum (Al), are weighed in the desired composition ratio and placed in the crucible 10. Although carbon is an essential component of SiC single crystals, it does not have to be added as a raw material at this stage.

[0057] Subsequently, power is supplied to the induction coil 12, and the raw material is heated by induction heating to the temperature for crystal growth and dissolved. If this heating is continued, the carbon constituting the crucible 10 will dissolve into the raw material solution, and a carbon-containing solution 20 can be obtained.

[0058] In preparing this solution, it is preferable to attach a silicon carbide seed crystal 30 to the tip of the crystal holding shaft 14 and introduce this seed crystal 30 into the furnace 11 (above the crucible 10).

[0059] The solution 20 obtained in this way contains Si, C, and other elements such as Cr and Al, which may be added as needed. Furthermore, other elements may be added as needed. It is preferable that the solution does not substantially contain elements that are not necessary for the production of SiC single crystals. Here, "substantially contained" means that the solution 20 may contain other elements that are included as impurities in the raw materials of the Si source, C source, Cr source, Al source, or other element sources that may be added, or other elements that are inevitably included from the production equipment during the production of silicon carbide seed crystals.

[0060] Other elements that are inevitably present include, for example, N, B, and Fe. "Substantially absent" of these elements means that their content is less than 1 atomic percent each.

[0061] Furthermore, in this embodiment, in step (a), the solidified solution used in the previously performed silicon carbide single crystal manufacturing method is used as the raw material for the solution. By using the solidified solution used in this way, unused components such as Cr and Al can be reused, thereby promoting the effective use of resources.

[0062] Except for using solidified used solution as a raw material for this solution 20, the SiC single crystal can be manufactured in the same manner as conventionally known methods for manufacturing SiC single crystals.

[0063] However, the composition of each component in the solidified used solution differs from the composition required to produce the desired SiC crystals. Therefore, it is necessary to adjust the content of each component in the solution preparation process to achieve the desired composition.

[0064] In this embodiment of the SiC single crystal manufacturing method, the missing elements are added to the solidified used solution before starting the growth and development of the SiC single crystal. At this time, the amount of missing elements to be added can be calculated based on the results of a compositional analysis of the solidified used solution.

[0065] The compositional analysis is not particularly limited as long as it is a known compositional analysis method. Preferably, compositional analysis devices such as X-ray fluorescence analyzers (XRF), analytical scanning electron microscopes (SEM-EDS), and inductively coupled plasma emission spectroscopy (ICP-AES) analyzers can be used for this analysis. These compositional analysis devices are preferable because they can simultaneously analyze the contained elements.

[0066] The sample used for this compositional analysis can be obtained by cutting out a portion of the solidified material from the used solution, but it is preferable to use a sample from near the center of the solidified material because the elemental concentrations are more stable. This is because, although this solidified material is obtained by cooling it to room temperature after the production of SiC single crystals, the temperature drops relatively quickly at the interface between the solution and the crucible (especially the interface with the bottom of the crucible) and at the liquid surface of the solution during the cooling process. At that time, miscellaneous SiC crystals may form, causing the elemental concentrations to fluctuate.

[0067] Through this compositional analysis, the resulting spectrum allows us to calculate the proportion of elements such as Si, Cr, and Al contained in the solidified material of the used solution.

[0068] Although carbon (C) is present in the solidified material of the used solution, it is difficult to detect in the above compositional analysis. Furthermore, the primary source of C is the crucible, so it is not necessarily an element whose addition amount must be calculated as a deficient element. Therefore, the proportion of C does not need to be considered in this compositional analysis.

[0069] Then, based on the calculated content ratio, when reusing the solution, missing elements are added to create a solution with the desired composition, for example, by adding elements to achieve the same composition as the new solution used in the earlier production of the SiC single crystal. The new solution corresponds to the solution immediately after the raw materials are dissolved in the crucible.

[0070] During the production of SiC single crystals, Si is consumed as the main component of the single crystal, so its amount in the solution decreases. On the other hand, Cr and Al do not constitute SiC single crystals and are not consumed, so they remain in the solution. As mentioned above, Si is consumed, but Cr and Al are not, so these components become concentrated and remain in the used solution. Although Al is not consumed, the production of SiC single crystals is carried out at high temperatures of 1800°C or higher, so it evaporates and its content decreases, and in some cases the concentration fluctuation may be small.

[0071] Here, the solidified material of the used solution to be reused may be reused entirely or only partially. For example, if the crucible does not collapse when the solidified material of the used solution is used and can be reused for the production of SiC single crystals, then the entire crucible containing the solidified material can be reused as is. Alternatively, since impurities are generated during the solidification process of the solution as described above, the solidified material containing these impurities can be removed before reuse. When removing impurities, for example, the solidified material to be reused can be cut out from the solidified material and placed in a new crucible for reuse.

[0072] Next, elements that are lacking are added to the solidified material to be reused. For example, when only a portion is reused, elements such as Si, Cr, and Al (excluding C) will all be deficient. Therefore, the required amounts of each element are calculated, and after adding the raw materials for the deficient elements, the mixture is heated in a crucible to prepare a solution for producing SiC single crystals.

[0073] To calculate the amount of the missing elements mentioned above, you can perform the following calculation (1) for each component.

[0074] The amount of added component (A) = the amount of component in the solution composition at the start of growth (A') - the amount of component in the solution to be reused (A R ) ···(1) Here, the component mass (A') of the solution composition at the start of growth is the desired (target) composition of the solution to be used in the subsequent SiC single crystal production, so its value is known for each component. On the other hand, the component mass (A') of the solution to be reused is known. R The composition of the SiC single crystal is unknown because it varies depending on the solution used, manufacturing conditions, etc., during its production. Therefore, the content ratio of each component is calculated by performing the compositional analysis described above. Specifically, the mole fraction of each component in the solidified material of the reused solution can be calculated from the spectrum obtained from the compositional analysis, and the mass of each component in the reused solution can be calculated. Note that the carbon (C) component can be ignored when calculating this mole fraction.

[0075] For example, consider a scenario where a new solution of Si-40mol%Cr-2mol%Al with a total mass of 500g is used to produce a SiC single crystal, and after the single crystal production is complete, 200g of the solidified used solution is cut out and reused.

[0076] In this case, since the composition of the solidified solution after use is unknown, a sample for compositional analysis is prepared and the composition is analyzed using an X-ray fluorescence analyzer (XRF) or an analytical scanning electron microscope (SEM-EDS). Here, we will illustrate the case where the mole fractions of each component calculated by compositional analysis are 50 mol% for Cr and 2 mol% for Al (Si-50 mol%Cr-2 mol%Al) (Si is included as the remainder, and C is not considered).

[0077] The content of each component in 200g of the solidified solution can be calculated from its atomic weight and the above mole fraction. Furthermore, using these results, the amount of each component to be added can be calculated as follows. Here, the solution prepared for reuse is shown to have the same composition and quantity (500g) as a new solution, but it can be changed each time to any desired composition and mass.

[0078] <Amount of Si added> • Mass of components in the solution composition at the start of growth (A'): 216.46g • Amount of components in the solution to be reused (A R ): 67.37g · Added mass (A): 149.09g (=A'-A R ) <Amount of Cr added> • Mass of components in the solution composition at the start of growth (A'): 276.37g • Amount of components in the solution to be reused (A R ): 129.93g · Added mass (A): 146.44g (=A'-A R ) <Amount of Al added> • Amount of components in the solution composition at the start of growth (A'): 7.17g • Amount of components in the solution to be reused (A R ): 2.70g ·Additional mass (A): 4.47g (=A'-A R ) By calculating the required amount of each component to be added, and placing the solidified solution to be reused and the raw materials for the added mass of each component into a new crucible, and then performing the SiC single crystal manufacturing operation as described above, the SiC single crystal manufacturing method of this embodiment can be carried out. This makes effective use of elements and reduces manufacturing costs.

[0079] (b) Process of growing silicon carbide single crystals Next, the crystal holding shaft 14, to which the seed crystal 30 is attached, is lowered into the furnace 11 and brought into contact with the solution, thereby growing a SiC single crystal on the crystal growth surface 30a. At this time, first, as shown in Figure 1, the crystal growth surface 30a of the seed crystal 30 attached to the crystal holding shaft 14 is positioned in contact with the solution 20.

[0080] In step (b), the seed crystal 30 only needs to be brought into contact with the solution 20, and in particular, the growth of the single crystal can be started by bringing it into contact with the surface of the solution 20 (without submerging the seed crystal 30 in the solution 20).

[0081] Single crystal growth is carried out by moving the crystal holding axis 14 upward or downward, or by maintaining it in a position where it is in contact with the seed crystal 30, thereby causing a silicon carbide single crystal (SiC single crystal) to grow on the lower surface of the seed crystal 30. On the other hand, to terminate crystal growth, the SiC single crystal and the solution 20 are separated by pulling up the crystal holding axis 14. This terminates the growth of the SiC single crystal.

[0082] As described above, a SiC single crystal can be manufactured by operating the single crystal manufacturing apparatus 101. Although it has been stated that crystal growth is terminated by raising the crystal holding axis 14 to isolate the SiC single crystal from the solution 20, this is not the only method. For example, instead of raising the crystal holding axis 14, the SiC single crystal can be isolated from the solution 20 by lowering the crucible holding axis 13, thereby terminating crystal growth.

[0083] The above-described operation can be achieved by providing a control unit that enables such operation, and in the single crystal manufacturing apparatus 101, SiC single crystals are manufactured by the "solution growth method".

[0084] Specifically, the control unit controls induction heating by the induction coil 12, vertical movement and rotation of the crystal holding shaft 14, vertical movement and rotation of the crucible holding shaft 13, etc., to ensure that the single crystal manufacturing method proceeds smoothly.

[0085] The seed crystal 30 used here can be any known polytype of silicon carbide single crystal, such as 2H, 3C, 4H, or 6H. The difference between polytypes lies in the atomic arrangement, and the effect of this crystal growth method may vary depending on the type of polytype. In this embodiment, for example, it is preferable to use a 4H silicon carbide single crystal as the seed crystal, and the resulting silicon carbide single crystal is suitable for power semiconductor devices.

[0086] Figure 2 is an example of a schematic diagram showing a cross-section of a silicon carbide seed crystal 30. In Figure 2, the seed crystal 30 is a 4H silicon carbide single crystal and an off-substrate with an off-angle θ. Specifically, the crystal growth plane 30a (or the normal 30n to the crystal growth plane 30a) is the c-plane ( <0001> It is preferable that the off-angle θ is tilted from the direction (11-20) to the direction (11-20). This off-angle θ is preferably 0.5° or more and 8° or less, and more preferably 0.5° or more and 5° or less. The seed crystal 30 may also be an on-substrate whose crystal growth plane is a c-plane. Although the on-substrate has an off-angle of 0°, when a seed crystal is made as an on-substrate, an off-angle of less than 0.5° may inevitably be formed. For this reason, the on-substrate substantially has an off-angle of 0 to less than 0.5°.

[0087] To improve the crystal quality of silicon carbide single crystals, it is necessary to reduce the dislocation density. However, through-dislocations, such as through-edge dislocations parallel to the c-axis, present in the silicon carbide seed crystal propagate and remain in the grown silicon carbide single crystal. Therefore, it is generally not easy to improve the crystal quality beyond that of the silicon carbide seed crystal.

[0088] However, it is known that one way to solve these problems is to reduce the dislocation density by bending the direction of dislocation propagation through a-axis dislocations parallel to the c-axis through step flow growth, that is, by converting them to dislocations parallel to the a-axis, thereby excluding the dislocations from the crystal growth direction. To achieve this, it is effective to use a substrate with the off-angle described above as a seed crystal to increase the microstep of the crystal growth surface. In other words, by using a seed crystal with such an off-angle, the dislocation (defect) density can be reduced. On the other hand, even with an on-substrate, step flow growth can be performed by concave growth to reduce the dislocation density.

[0089] Furthermore, the temperature during SiC single crystal growth is preferably between 1800°C and 2200°C, and more preferably between 1900°C and 2100°C. It is even more preferable to grow the crystal at a temperature between 1950°C and 2050°C. The conditions for the temperature gradient near the crystal growth surface 30a, the atmosphere inside the furnace 11 during growth, and the pressure can be set to the same conditions as those used in conventional solution growth methods.

[0090] During the growth of the SiC single crystal, the silicon carbide seed crystal 30 may be rotated by the rotation axis 14. Similarly, the crucible 10 may also be rotated by the rotation axis 13.

[0091] In this way, silicon carbide single crystals can be manufactured by the steps (a) and (b) described above. After the silicon carbide single crystal has grown sufficiently, the growth is stopped to obtain a SiC single crystal ingot. At this time, heating by induction coils, etc., is stopped in the silicon carbide single crystal manufacturing apparatus, and the used solution is cooled to room temperature and solidifies. The solidified material obtained in this way can be reused as a raw material for the solution when manufacturing silicon carbide single crystals again, as described above.

[0092] (Regarding solidified waste from used solutions) As described above, this embodiment is characterized by the reuse of the solidified used solution as a raw material for the solution in the subsequent SiC single crystal manufacturing method. As described above, the solidified used solution may be reused entirely or partially, but it is preferable to remove the unwanted crystals and use only a portion of it. This is because if unwanted crystals are included in the solution for manufacturing the SiC single crystal, crystal growth will proceed not only in the seed crystal 30 but also in the unwanted crystals, which may prevent the efficient manufacturing of the SiC single crystal.

[0093] To remove unwanted crystals from the solidified material of a used solution, for example, the regions containing unwanted crystals in the solidified material can be cut away. Figure 3 shows a cross-sectional view of a crucible 10 containing solidified material 21 obtained by first manufacturing a SiC single crystal and then cooling the used solution.

[0094] In Figure 3, the inner surface of crucible 10 that was in contact with the SiC single crystal manufacturing solution is deformed due to dissolution during single crystal manufacturing and cooling operations after heating. In addition, miscellaneous crystals 22 are formed in the solution. As explained above, miscellaneous crystals 22 tend to form in the vicinity of the contact interface with the bottom surface of crucible 10 in solution 20 and in the region close to crucible 10 at the liquid surface of solution 20.

[0095] Then, as shown in Figure 3, if miscellaneous crystals 22 are formed in the solidified product 21 of the used solution, the miscellaneous crystals 22 are removed by cutting along cutting lines X1 and X2 in the horizontal direction and along cutting lines Y1 and Y2 in the vertical direction, and a region R (the region enclosed by cutting lines X1 and X2 and cutting lines Y1 and Y2) that does not contain miscellaneous crystals 22 is cut out and reused as raw material for the solution in the next SiC single crystal manufacturing method.

[0096] The solidified material 21 cut out here can be reused as is, but when preparing the solution next, it is preferable to cut it into smaller pieces so that it can be easily dissolved by heating. In this case, for example, it is preferable to cut the solidified material 21 so that its maximum dimensions are 2 / 3 or less of the inner diameter of the crucible.

[0097] [Variation] In the above embodiment, the deficient element is analyzed for composition, the amount of the deficiency is calculated from the composition analysis results, and the necessary raw materials are added to produce the SiC single crystal, as described in the next step.

[0098] Incidentally, if the production of SiC single crystals using solidified used solution has already been carried out under the same conditions, the above compositional analysis can be omitted. For example, if a SiC single crystal is being produced using solidified used solution as a raw material under the same production conditions (solution composition and SiC single crystal growth conditions) as the production conditions to be carried out next, the amount of missing elements can be calculated without performing a compositional analysis by using the results of the compositional analysis obtained in the previous step.

[0099] For example, when repeating SiC production under the same conditions as described in the compositional analysis above, 500g of raw materials with a composition of Si-40mol%Cr-2mol%Al is prepared as the solution before SiC production for reuse, and the solidified used solution after the production of SiC single crystals will have a composition of Si-50mol%Cr-2mol%Al. In this case, the amount of solidified material to be reused will not be constant and will likely increase or decrease each time, but since the mole fraction of each component contained in the resulting solidified material is known, the content of each component can be calculated from the reused mass. Therefore, the added mass of each component can be calculated using the above formula (1) without performing a compositional analysis.

[0100] This is particularly useful when manufacturing SiC single crystals by repeating the process multiple times under the same reuse conditions (solution composition and SiC single crystal manufacturing conditions). In this case, there is no particular limit to the number of reuse cycles; the used solution can be reused any number of times.

[0101] <Examples> The following describes specific examples.

[0102] In this specification, "starting solution" refers to a solution used to grow a silicon carbide single crystal by contacting a silicon carbide seed crystal. In other words, it is a solution used for growing single crystals in solution growth. Specifically, the "starting solution" is the solution used for the first time. In contrast, the "reused solution" is a used "starting solution" that has already been used to grow single crystals in solution growth and is recovered and reused for growing single crystals.

[0103] First, a "starting solution" was prepared. Specifically, for example, a "starting solution" was prepared containing 53.5 atomic percent silicon (Si), 40 atomic percent chromium (Cr), and 6.5 atomic percent of other elements (Al, Y, Mo, La, Ce, Pr, Nd, Tb, Gd, W, etc.).

[0104] Subsequently, the prepared "preparation solution" was placed in a graphite crucible to grow single crystals. Then, the solidified material of the "preparation solution" used to grow the single crystals was cut along with the graphite crucible. Specifically, as shown in Figure 3, for example, the solidified material of region R, which does not contain impurities, was cut along cutting lines X1, X2 and Y1, Y2.

[0105] Next, the extracted solidified material was placed in an alumina crucible and remelted by heating at 1500°C for 5 minutes, after which it was allowed to solidify. The remelted solution was designated as the "reused solution," and the composition of Si, Cr, and other elements in this "reused solution" was measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES). Similarly, the composition of Si, Cr, and other elements in the "initial solution" was also measured.

[0106] Table 1 shows the measurement results of the Si, Cr, and other elemental compositions in the "initial solution" and the "reused solution," respectively.

[0107] [Table 1]

[0108] As shown in Table 1, the composition of the "reused solution" and the composition of the "starting solution" are different. In order to use the "reused solution" again for single crystal growth, the composition of the "reused solution" must be made to match the composition of the "starting solution". For this reason, additional samples must be added to the "reused solution". Specifically, additional samples are added to the "reused solution" so that the composition is 53.5 atomic percent Si, 40 atomic percent Cr, and 6.5 atomic percent of other elements.

[0109] Table 2 shows the compositional weights of the additional samples and the recycled samples that make up the "recycled solution".

[0110] [Table 2]

[0111] Here, for example, the weight of the additional Si sample can be calculated using the following formula.

[0112] Weight of additional Si sample = (Weight of Si in the initial composition (g / mol)) - (Weight of recycled Si sample per mole) × (Weight of recycled sample) / (Weight of alloy per mole) The term "alloy" here refers to an alloy containing Si, Cr, and other elements, meaning an alloy with the same composition as the recycled sample. While the formula for calculating the weight of the additional Si sample was explained here, the weight of the additional Cr sample and the weight of additional samples of other elements can be calculated similarly. In this way, additional samples can be added to the recycled sample.

[0113] For example, in the example shown in Table 2, the weight of the recycled sample is 327.4g, and the weight of the additional sample is 301.0g. Therefore, the recycling rate in the example shown in Table 2 is 52%.

[0114] Herein, in this specification, the solution obtained by adding a sample to a recycled sample and melting it is referred to as the "recycled solution." The composition of this "recycled solution" is the same as the composition of the "preparation solution."

[0115] The following describes an example of growing a single crystal using a "recycled solution."

[0116] For example, a graphite crucible with an inner diameter of 90 mm and an outer diameter of 110 mm was filled with a "recycled solution" to a solution height of approximately 25 mm. A seed crystal made of 4H-SiC single crystal with a diameter of φ=40 mm, a thickness of 800 μm, and an off-angle of 1° was prepared, and the (0001) plane was bonded to a graphite holder with a diameter of approximately 40 mm using graphite adhesive, so that the growth plane would be the (000-1) plane. This graphite holder was attached to a graphite rod to form the crystal axis.

[0117] In a high-frequency heating furnace (single crystal growth apparatus), after evacuating the furnace, argon (Ar) was introduced to atmospheric pressure, and single crystals were grown in an Ar atmosphere. The internal configuration of the high-frequency heating furnace is, for example, as shown in Figure 1.

[0118] The growth of single crystals was carried out by heating a graphite crucible and the "recycled material" contained within it, thereby melting the "recycled material." Here, "recycled material" as used herein refers to a solidified "recycled solution." In other words, the "recycled solution" is a solution formed by melting the "recycled material."

[0119] Subsequently, single crystal growth was carried out by immersing the seed crystal in a "recycling solution" and holding it at approximately 1900°C for 16 hours. During this time, the crystal axis and graphite crucible were rotated. The temperature distribution of the "recycling solution" was adjusted so that the sides of the graphite crucible were hotter and the seed crystal was colder. After maintaining this state for 16 hours, the grown single crystal was lifted from the surface of the "recycling solution" and cooled slowly over one hour.

[0120] In this way, single crystals can be grown using the "recycled solution".

[0121] Table 3 shows a comparison of the growth rate and surface condition of single crystals grown with the "recycled solution" and single crystals grown with the "preparation solution".

[0122] [Table 3]

[0123] As shown in Table 3, the growth rate was almost the same for the "recycled solution" and the "prepared solution." The growth rate when using the "recycled solution" was 83 μm / h, confirming that a sufficient growth rate can be maintained even when using the "recycled solution."

[0124] Furthermore, when comparing the quality of the crystal faces, the average step height of the crystal faces measured with a laser microscope was almost the same for both the "recycled solution" and the "prepared solution." However, the step height when using the "recycled solution" was 5 μm or less, confirming that a low step height can be secured even when using the "recycled solution." The average step height is the average value of the step height measured from 36 laser microscope images (over the entire crystal).

[0125] The above confirms that it is possible to grow single crystals while maintaining growth rate and quality, even when using a "recycled solution."

[0126] <Further improvements> As described in the above-mentioned examples, in order to produce a "recycled solution" having the same composition as the "initial solution," it is necessary to add an additional sample to the "reused solution" (reused sample). At this time, it is necessary to measure the composition of the "reused solution," for example, by ICP analysis. This is because, without measuring the composition of the "reused solution," it is not possible to produce a "recycled solution" having the same composition as the "initial solution." In other words, without measuring the composition of the "reused solution," it is not possible to determine the amount of missing elements to be added.

[0127] In this regard, if the composition of the "reused solution" can be determined without performing measurements such as ICP analysis, the efficiency of producing the "recycled solution" can be improved. Therefore, in this embodiment, a method has been devised to determine the composition of the "reused solution" without performing measurements such as ICP analysis. The technical concept of this embodiment, in which this method has been devised, will be explained below.

[0128] Specifically, a further innovation in this embodiment is the construction of a system that estimates the composition of a used solution by using a regression model that outputs an estimated value for the composition of a used solution ("reused solution") when the composition of the solution before use ("pre-use solution") and the manufacturing conditions for growing a single crystal are input.

[0129] In the following, we will mainly describe an example in which the composition estimation system described above is configured using a single computer (composition estimation device), but the composition estimation system in this embodiment can also be implemented as a distributed system consisting of multiple computers.

[0130] <<Configuration of the composition estimation device>> <<<Hardware Configuration>>> First, the hardware configuration of the composition estimation device in this embodiment will be described.

[0131] Figure 4 shows an example of the hardware configuration of the composition estimation device 100 in this embodiment. Note that the configuration shown in Figure 4 is merely one example of the hardware configuration of the composition estimation device 100, and the hardware configuration of the composition estimation device 100 is not limited to the configuration shown in Figure 4; other configurations are also possible.

[0132] In Figure 4, the composition estimation device 100 includes a CPU (Central Processing Unit) 101A that executes a program. This CPU 101A is electrically connected via a bus 113 to, for example, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, and a hard disk drive 112, and is configured to control these hardware devices.

[0133] Furthermore, the CPU 101A is connected to input and output devices via the bus 113. Examples of input devices include the keyboard 105, mouse 106, communication board 107, and scanner 111. On the other hand, examples of output devices include the display 104, communication board 107, and printer 110. In addition, the CPU 101 may be connected to, for example, a removable disk drive 108 or a CD / DVD-ROM drive 109.

[0134] The composition estimation device 100 may be connected to a network, for example. For example, if the composition estimation device 100 is connected to other external devices via a network, the communication board 107, which constitutes part of the composition estimation device 100, is connected to a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet.

[0135] RAM 103 is an example of volatile memory, while the recording media of ROM 102, removable disk device 108, CD / DVD-ROM device 109, and hard disk device 112 are examples of non-volatile memory. These volatile and non-volatile memories constitute the storage device of the composition estimation device 100.

[0136] The hard disk drive 112 stores, for example, an operating system (OS) 201, a group of programs 202, and a group of files 203. The programs included in the group of programs 202 are executed by the CPU 101A using the operating system 201. In addition, the RAM 103 temporarily stores at least a portion of the operating system 201 programs and application programs to be executed by the CPU 101A, as well as various data necessary for processing by the CPU 101A.

[0137] The ROM 102 stores the BIOS (Basic Input Output System) program, and the hard disk drive 112 stores the boot program. When the composition estimation device 100 is started, the BIOS program stored in the ROM 102 and the boot program stored in the hard disk drive 112 are executed, and the operating system 201 is started by the BIOS program and the boot program.

[0138] The program group 202 stores programs that implement the functions of the composition estimation device 100, and these programs are read and executed by the CPU 101A. The file group 203 stores information, data, signal values, variable values, and parameters that represent the results of processing by the CPU 101A as individual file items.

[0139] Files are recorded on recording media such as the hard disk drive 112 or memory. Information, data, signal values, variable values, and parameters recorded on the recording media such as the hard disk drive 112 or memory are read by the CPU 101A into main memory or cache memory and used in the operation of the CPU 101A, which is represented by extraction, search, reference, comparison, calculation, processing, editing, output, printing, and display. For example, during the operation of the CPU 101A as described above, information, data, signal values, variable values, and parameters are temporarily stored in main memory, registers, cache memory, buffer memory, etc.

[0140] The functions of the composition estimation device 100 may be realized by firmware stored in ROM 102, or by software only, hardware only (represented by elements, devices, substrates, and wiring), a combination of software and hardware, or a combination with firmware. The firmware and software are recorded as a program on a recording medium such as a hard disk drive 112, removable disk, CD-ROM, or DVD-ROM. The program is read and executed by the CPU 101A. In other words, the program makes the computer function as the composition estimation device 100.

[0141] Thus, the composition estimation device 100 is a computer equipped with a processing unit CPU 101A, a storage device such as a hard disk drive 112 and memory, input devices such as a keyboard 105, a mouse 106, and a communication board 107, and output devices such as a display 104, a printer 110, and a communication board 107. The functions of the composition estimation device 100 are realized using the processing unit, storage device, input devices, and output devices.

[0142] <<<Functional Block Configuration>>> Next, the functional block configuration of the composition estimation device 100 will be described.

[0143] Figure 5 is a functional block diagram showing the functions of the composition estimation device 100.

[0144] The composition estimation device 100 includes an input unit 301, a regression model generation unit 302, an estimation unit 303, an output unit 304, and a data storage unit 305.

[0145] The input unit 301 is configured to input related data.

[0146] Here, "related data" refers to data consisting of pre-use composition data, manufacturing condition data, and used composition data, and is data that links the pre-use composition data, manufacturing condition data, and used composition data. Pre-use composition data consists of pre-use composition values ​​that indicate the composition of the solution before use ("pre-charged solution"). Manufacturing condition data consists of data that indicates the manufacturing conditions when producing single crystals by the solution growth method. Used composition data consists of used composition values ​​that indicate the composition of the used solution ("reused solution").

[0147] For example, Figure 6 is a table showing an example of related data.

[0148] Figure 6 shows that the relevant data consists of pre-use composition data, manufacturing condition data, and used composition data.

[0149] Pre-use composition data consists of the elemental composition values ​​(pre-use composition values) contained in the solution before use. For example, in the example shown in Figure 6, the solution before use contains at least Si, Cr, and Al, and the composition values ​​of each element are specified. Focusing on the related data for "No. 1" in Figure 6, the pre-use composition data includes a Si composition value of "53.5", a Cr composition value of "40", and an Al composition value of "6.5". Similarly, focusing on the related data for "No. 2" in Figure 6, the pre-use composition data includes a Si composition value of "60", a Cr composition value of "40", and an Al composition value of "0".

[0150] Next, the manufacturing condition data includes items such as "growth temperature," "growth time," "crucible diameter," "solution height," and "rotation speed (crucible and crystal axis)."

[0151] Next, the used composition data consists of the elemental composition values ​​(used composition values) contained in the used solution. For example, in the example shown in Figure 6, the used solution contains at least Si, Cr, and Al, and the composition values ​​of each element are specified. Focusing on the related data for "No. 1" shown in Figure 6, the used composition data includes the composition value of Si "43.2", the composition value of Cr "45.5", and the composition value of Al "11.3".

[0152] The related data, configured in this way, is input from the input unit 301 and then stored in the data storage unit 305. This data storage unit 305 functions as a database for storing multiple related data. The input unit 301 is also configured to receive various types of data other than related data. For example, data that can be input to the input unit 301 may include individual pre-use composition data and individual manufacturing condition data. These various types of data are also input from the input unit 301 and then stored in the data storage unit 305.

[0153] The regression model generation unit 302 has the function of generating a regression model based on the relevant data stored in the data storage unit 305. In other words, the regression model generation unit 302 is configured to generate a regression model that relates the pre-use composition value, manufacturing conditions, and used composition value. Specifically, as shown in Figure 7, the regression model generation unit 302 is configured to generate a regression model using machine learning with the relevant data as training data, where the input is the pre-use composition value and manufacturing conditions, and the output is the used composition value (estimated value).

[0154] Here, a "regression model" is defined as a function that, when given pre-use composition values ​​and manufacturing conditions, outputs an estimated used composition value corresponding to these pre-use composition values ​​and manufacturing conditions. In other words, a "regression model" is defined as a function that, when given pre-use composition values ​​and manufacturing conditions whose correspondence to the used composition value is unknown, outputs an estimated used composition value that is presumed to be realized with these pre-use composition values ​​and manufacturing conditions. Thus, a regression model can be said to be a function used to estimate the composition value of a used solution when growing a single crystal using a pre-use solution whose correspondence to the composition of the used solution is unknown.

[0155] The estimation unit 303 is configured to estimate the composition of the used solution based on the pre-use composition value, which indicates the composition of the solution before use, the manufacturing conditions for growing the single crystal, and a regression model. In other words, the estimation unit 303 is configured to obtain an estimated value of the used composition value, which indicates the composition of the used solution, using the regression model generated by the regression model generation unit 302. For example, when using a pre-use solution whose correspondence with the composition of the used solution is unknown, and growing a single crystal using predetermined manufacturing conditions, the estimation unit 303 has the function of estimating the composition value of the used solution based on a regression model whose accuracy has been improved by machine learning.

[0156] The output unit 304 outputs the estimated value of the used composition value estimated by the estimation unit 303.

[0157] In this way, the composition estimation device 100 is configured.

[0158] In this embodiment, the composition estimation device 100 is equipped with a regression model generation unit 302 for generating a regression model and an estimation unit 303 for estimating the composition value of the used solution. In other words, in this embodiment, a composition estimation system for estimating the composition of a used solution used to grow a single crystal by contacting a seed crystal is realized in a single composition estimation device 100.

[0159] However, the technical concept in this embodiment is not limited to this configuration. It is also possible to configure a composition estimation system using a distributed system in which a regression model generation device equipped with a regression model generation unit 302 and an estimation device equipped with an estimation unit 303 are configured on separate computers, and the regression model generation device and the estimation device are connected using a network, for example.

[0160] <<Operation of the composition estimation device>> The composition estimation device 100 is configured as described above, and its operation will be explained below. The operation of the composition estimation device 100 consists of "regression model generation" and "estimated composition of used solution." Therefore, these operations will be explained below.

[0161] <<<Regression Model Generation Process>>> Figure 8 is a flowchart illustrating the process of generating the regression model.

[0162] In Figure 8, first, the input unit 301 receives multiple related data, which are associated with pre-use composition data, manufacturing condition data, and used composition data (S101). Then, the multiple related data entered into the input unit 301 are stored in the data storage unit 305 (S102).

[0163] Next, the regression model generation unit 302 generates a regression model based on the relevant data stored in the data storage unit 305 (S103). Specifically, the regression model generation unit 302 generates a regression model using machine learning with the relevant data as training data, where the inputs are the pre-use composition values ​​and manufacturing conditions, and the output is an estimated value of the used composition value (see Figure 7).

[0164] The regression model generated by the regression model generation unit 302 is then stored in the data storage unit 305 (S104). In this way, the regression model generation operation is performed.

[0165] <<<Estimation of the composition of used solution>>> Next, we will explain the process of estimating the composition of the used solution.

[0166] Figure 9 is a flowchart illustrating the process of estimating the composition of the used solution. The regression model is already stored in the data storage unit 305.

[0167] In Figure 9, the input unit 301 first inputs the composition of the solution before use (pre-use composition value), which is unknown to the composition of the used solution, and also inputs the manufacturing conditions for the solution growth method using this pre-use solution (S201).

[0168] Subsequently, the estimation unit 303 substitutes the pre-use composition value and manufacturing conditions described above into the regression model (S202). As a result, the regression model outputs an estimated value of the used composition value, which indicates the composition of the used solution, according to the input pre-use composition value and manufacturing conditions. Thus, the estimation unit 303 can obtain the estimated value (S203). Then, the output unit 304 outputs the estimated value obtained by the estimation unit 303 (S204).

[0169] In this way, the composition estimation device 100 can obtain an estimated value of the composition of the used solution (used composition value) that is likely to be realized when growing a single crystal by a solution growth method under predetermined manufacturing conditions, using a pre-use solution whose correspondence with the composition of the used solution is unknown (pre-use composition value). As a result, according to this embodiment, the composition of the used solution ("reused solution") that is likely to be realized can be determined without performing measurements such as ICP analysis, thereby improving the efficiency of producing the "recycled solution". In this respect, the technical concept of this embodiment is useful from the viewpoint of reducing the manufacturing cost of single crystals.

[0170] <Composition Estimation Program> The composition estimation method performed by the composition estimation device 100 described above can be realized by a composition estimation program that causes a computer to perform the composition estimation process.

[0171] For example, in the composition estimation device 100 consisting of a computer as shown in Figure 4, the composition estimation program of this embodiment can be introduced as one of the program group 202 stored in the hard disk drive 112. Then, by having the computer, which is the composition estimation device 100, execute this composition estimation program, the composition estimation method of this embodiment can be realized.

[0172] A composition estimation program, which causes a computer to perform various processes to create data related to composition estimation, can be recorded on a computer-readable recording medium and distributed. Recording media include, for example, magnetic storage media such as hard disks and flexible disks, optical storage media such as CD-ROMs and DVD-ROMs, and hardware devices such as non-volatile memory such as ROMs and EEPROMs.

[0173] The present invention has been described in detail above based on its embodiments, but it goes without saying that the present invention is not limited to the above embodiments and can be modified in various ways without departing from its essence. [Explanation of Symbols]

[0174] 10 Crucible 11 Furnace 12 Induction Coil 13 Crucible holding shaft 14 Crystal holding shaft 20 solution 21 Solidified material 22 Miscellaneous crystals 30 seed crystals 30a Crystal growth surface 100 Composition estimation device 101 Single crystal manufacturing apparatus 101A CPU 102 ROM 103 RAM 104 displays 105-key keyboard 106 mice 107 Communication board 108 Removable disk device 109 CD / DVD-ROM device 110 Printers 111 Scanner 112 Hard disk drive 201 Operating Systems 202 Program Groups 203 files 301 Input section 302 Regression Model Generation Unit 303 Estimation Department 304 Output section 305 Data Storage Unit Regions that do not contain R-type crystals

Claims

1. (a) A step of preparing a solution containing silicon and carbon, (b) A step of contacting a silicon carbide seed crystal with the solution and growing a silicon carbide single crystal on the crystal growth surface of the seed crystal, A method for producing silicon carbide single crystals comprising: As the raw material for the solution in step (a) above, the solidified solution used in the previously performed silicon carbide single crystal manufacturing method is used. The solidified material from the used solution is used after removing any contained impurities. A method for producing silicon carbide single crystals.

2. In the method for producing silicon carbide single crystals according to claim 1, In step (a) above, the composition is adjusted by adding elements that are lacking in the solidified material. A method for producing silicon carbide single crystals.

3. In the method for producing silicon carbide single crystals according to claim 2, Silicon is added as the missing element. A method for producing silicon carbide single crystals.

4. In the method for producing silicon carbide single crystals according to claim 2, The amount of the deficient element to be added is calculated based on the results of the compositional analysis of the solidified material. A method for producing silicon carbide single crystals.

5. In the method for producing silicon carbide single crystals according to claim 4, The aforementioned compositional analysis is performed using an analytical instrument that utilizes X-ray fluorescence (XRF), analytical scanning electron microscope (SEM-EDS), or ICP emission spectrometry. A method for producing silicon carbide single crystals.

6. In the method for producing silicon carbide single crystals according to claim 2, The removal of the aforementioned miscellaneous crystals is performed by cutting out the inside of the solidified material of the used solution. A method for producing silicon carbide single crystals.

7. In the method for producing silicon carbide single crystals according to claim 2, If silicon carbide single crystals have already been produced using the solidified used solution under the same conditions, The amount of the deficient element to be added is calculated based on the compositional analysis results obtained in the production of the silicon carbide single crystal under the same conditions. A method for producing silicon carbide single crystals.

Images