Data-driven CVD-SiC deposition rate prediction model and method

Abstract

The invention discloses a data-driven CVD-SiC deposition rate prediction model and method, and belongs to the technical field of ceramic matrix composites and semiconductor materials.The data-driven CVD-SiC deposition rate prediction model comprises a reactor, graphite flakes and a deposition sample, the reactor and the deposition sample are both arranged in the reactor, the reactor is tightly attached to the inner wall of the reactor, and the graphite flakes are arranged in the reactor; according to the data-driven CVD-SiC deposition rate prediction model and the prediction method, the data-driven CVD-SiC deposition rate prediction model of the structure is adopted, the prediction method is combined, a more concise chemical reaction model and multi-field coupling of heat and mass transfer are established, and the deposition rate of the CVD-SiC is predicted. The deposition rate of the deposition surface under different process conditions can be accurately predicted, meanwhile, an accurate correlation function of the process conditions and the deposition rate is established through a machine learning method, the efficient and accurate relation between experimental parameters and the deposition rate is constructed, and combination of numerical simulation and machine learning is achieved.

Description

Technical Field

[0001] The present invention relates to the technical fields of ceramic matrix composites and semiconductor materials, and particularly relates to a data-driven CVD-SiC deposition rate prediction model and method. Background Art

[0002] Silicon carbide (SiC) materials have properties such as high strength, high hardness, corrosion resistance, wide (tunable) bandgap, and low density, and are widely used in wear-resistant materials, electrothermal materials, armor protection materials, high-temperature and high-strength composite materials, and power device materials, etc. They are a new generation of wide-bandgap semiconductor materials represented by SiC. Generally, a thin film needs to be covered on the substrate to meet the conditions of device functions. Ceramic matrix composites are a type of composite material with a ceramic as the matrix and combined with other fibers. They have good properties such as high strength, high modulus, low density, high temperature resistance, wear resistance, and corrosion resistance. Especially the high temperature resistance of ceramic matrix composites has attracted attention in the application research under high temperature environments. Generally, a complete and uniform coating also needs to be coated on the outer surface of ceramic matrix composites to play a role in heat insulation and oxidation resistance.

[0003] Wide-bandgap semiconductor material SiC epitaxial layers and ceramic matrix composite SiC coatings with high purity, low density defects, and high-quality structured surfaces are gradually entering the application stage from the laboratory. Chemical Vapor Deposition (CVD) technology is the preferred method for preparing SiC. This technology uses one or several gaseous compounds or simple substances containing the required elements to carry out chemical reactions on the fiber surface to generate thin films or coatings. Compared with other inorganic material preparation methods, CVD can not only prepare high-quality and high-purity coatings, but also realize interface deposition of complex-shaped components through process control groups, and control the composition and material distribution. However, the CVD deposition process involves complex chemical reaction thermodynamics and kinetics, and there are many factors affecting its deposition rate and deposition uniformity. Simply relying on experiments to optimize process parameters will consume a large amount of manpower and material resources.

[0004] Currently, chemical vapor deposition (CVD) is usually used to prepare high-quality SiC epitaxial layers. However, due to the relatively complex deposition process, important factors affecting its deposition rate and deposition uniformity include many process parameters such as gas flow rate, reaction temperature, pressure, gas flow, precursor ratio, and reactor geometry. Using traditional experimental methods to optimize and improve the deposition process is expensive and time-consuming, which greatly restricts the application and development of SiC epitaxial layers and coatings. Based on this, the present invention provides a data-driven CVD-SiC deposition rate prediction model and method to solve the above problems. Summary of the Invention

[0005] The object of the present invention is to provide a data-driven CVD-SiC deposition rate prediction model and method. By establishing a more concise chemical reaction model and multi-field coupling of heat transfer and mass transfer, and using the Randomforest algorithm or Xgboost algorithm, it is possible to quickly and accurately predict the deposition rate under different process and reactor conditions through a small amount of experimental results, and optimize the process parameters of CVD-SiC manufacturing production quickly and at low cost.

[0006] To achieve the above object, the present invention provides a data-driven CVD-SiC deposition rate prediction model, including a reactor, a graphite sheet and a deposition specimen. Both the reactor and the deposition specimen are arranged inside the reactor. The graphite sheet is attached to the inner wall of the reactor, and the deposition specimen is suspended at the middle position inside the reactor by carbon fiber. The inside of the reactor is a reaction chamber.

[0007] Preferably, the reactor includes a heat insulation layer, an inlet, an outlet and a porous disk. The inlet is arranged at one end of the heat insulation layer, the porous disk is arranged at one end inside the heat insulation layer close to the inlet, and the outlet is arranged at the other end of the heat insulation layer.

[0008] The present invention also provides a data-driven CVD-SiC deposition rate prediction method, including the following steps:

[0009] S1. Establish a geometric model for deposition rate prediction according to actual test conditions and equipment;

[0010] S2. After establishing the geometric model, establish a fluid heat transfer model and a laminar flow model, calculate the temperature field and velocity field, and connect them;

[0011] S3. Establish a dilute mass transfer model and a chemical reaction model, input the precursor gas reaction into the model, and calculate the chemical reaction and concentration field distribution;

[0012] S4. The results in S2 and S3 form a multi-physical field model, and the deposition rate is obtained by multi-physical field coupling solution to get the simulated value;

[0013] S5. Use the established numerical simulation model to calculate the deposition rate of different process parameters;

[0014] S6. Use the K-nearest neighbor algorithm to fill the missing feature values in the simulated values;

[0015] S7. Integrate the results in S4, S5 and S6, and build a prediction model through a machine learning algorithm.

[0016] Preferably, the process of establishing the geometric model for deposition rate prediction in S1 is as follows:

[0017] S11. Determine the actual size of the vertical hot-wall reactor and the actual shape and size of the substrate;

[0018] S12. Establish a geometric model of the reactor based on the actual size determined in S11 and divide the mesh to form a geometric region;

[0019] S13. Add material properties to the geometric region and the boundaries of the reactor model.

[0020] Preferably, the process of calculating the temperature field and the flow field in S2 is as follows:

[0021] S21. Calculate the velocity field of the laminar flow model according to the Navier - Stokes equation and the continuity equation. The process is as follows:

[0022]

[0023] where ρ represents the density of the fluid, u represents the velocity field of the fluid, p represents the pressure field of the fluid, μ represents the dynamic viscosity of the fluid, f represents the external force, represents the gradient operator;

[0024] S22. Calculate the temperature field of the heat transfer model using the energy conservation equation. The process is as follows:

[0025]

[0026] where T represents the temperature field, k represents the thermal conductivity, c p represents the heat capacity, and Q represents the heat source;

[0027] S23. Since the velocity field u calculated by the laminar flow model is automatically transferred to the heat transfer model, and the temperature field T calculated by the heat transfer model is used to calculate the changes in the density of the fluid and the dynamic viscosity of the fluid, the laminar flow model and the heat transfer model are coupled into a non - isothermal flow model.

[0028] Preferably, the process of calculating the chemical reaction and the concentration field distribution in S3 is as follows

[0029] S31. Establish a chemical reaction model, input the relevant reactions of the precursor, and determine the gas - phase reaction rate constant, the surface reaction rate constant, and the adsorption reaction rate constant according to the three - parameter Arrhenius formula. The process is as follows:

[0030] K = A * T n * exp(-E / RT);

[0031]

[0032] where K represents the surface reaction rate constant, A represents the reaction frequency factor, T nrepresents the reaction temperature index, E represents the reaction activation energy index, exp(·) represents the exponential function with base e, R represents the gas constant, L1 represents the adsorption reaction rate constant, τ represents the adhesion coefficient, M k represents the molar mass of the adsorbate, v a represents the effective surface vibration frequency 10 13 E a represents the adsorption energy, k represents the Boltzmann constant;

[0033] S32. Calculate the collision integral Ω using the minimum energy value of the main substance characteristic length and the Lennard-Jones interaction potential D ;

[0034] S33. Calculate the diffusivity of the substance using the following formula:

[0035]

[0036] S34. Establish a dilute substance transfer model using the Fick diffusion model and the additional convective transport mechanism. The mixed density ρ1 is:

[0037]

[0038] where p1 represents the total pressure of the system, x i represents the mole fraction of the i-th substance, R represents the gas constant;

[0039] S35. Calculate the concentration field of the formula using the formula:

[0040]

[0041] where j represents the mass flux density, ω i represents the mass fraction of the i-th substance, R i represents the generation or consumption rate of the i-th substance, j i represents the mass flux density of the i-th substance, D i represents the diffusion coefficient of the i-th substance, c i represents the concentration of the i-th substance.

[0042] Preferably, the process of obtaining the simulation value in S4 is as follows:

[0043] S41. Couple the laminar flow model and the dilute substance transfer model to form a reaction flow-dilute substance model, and combine the contents in S2 and S3 to form a unified multi-physics field simulation model;

[0044] S42. Perform multi-physics field coupling solution on the multi-physics field simulation model for the deposition rate to obtain the simulation value;

[0045] S43. Use the dichotomy method to adjust the adsorption reaction rate constant in the chemical reaction model so that the simulated value fits the actual value obtained from the experiment.

[0046] Preferably, in S5, parameter values ​​of different process conditions are input into the calibrated numerical model, and the deposition rates thereof are predicted respectively, and finally several groups of simulation data are obtained.

[0047] Preferably, when a feature missing value appears in the simulation value in S6, K nearest neighbors are selected and the weighted sum of the distances is used as the predicted value of the missing value, and the process is as follows:

[0048]

[0049] Where L P Indicates distance, H i =[x i1 ,x i2 ,…,x in ]∈R n represents the i-th sample, H j represents the jth sample, x if represents the fth eigenvalue of the i-th sample, x jf represents the fth eigenvalue of the jth sample, x im represents missing values, x jm represents the mth eigenvalue of the jth sample.

[0050] Preferably, the process of building a prediction model in S7 is as follows:

[0051] S71, summarizing the experimental true value, the deposition rates of different process parameters obtained in S5, and the supplemented simulation value obtained in S6 to obtain a database of CVD process parameters-deposition rates;

[0052] S72. Use the random forest algorithm and Xgboost algorithm to build a prediction model of process parameter-deposition rate. The process is as follows:

[0053]

[0054] Where X i =[x i1 ,x i2 ,…,x in ] T ∈R n represents the CVD process parameters, t i =[t i1 ,t i2 ,…,t im ] T ∈R mrepresents the deposition rate measured experimentally, MTS represents methyltrichlorosilane, KH2 represents the ratio of H2 to MTS, and Dia represents the reactor diameter.

[0055] Therefore, the data-driven CVD-SiC deposition rate prediction model and method with the above structure of the present invention have the following advantages:

[0056] 1. By establishing a more concise chemical reaction model and multi-field coupling of heat transfer and mass transfer, the deposition rate of the deposition surface under different process conditions can be accurately predicted. Through this numerical simulation model, a large number of process parameters uniformly distributed in the parameter space can be simulated, thereby expanding the database of process conditions - deposition rate;

[0057] 2. Using machine learning methods such as the Randomforest algorithm and the Xgboost algorithm to establish an accurate correlation function between process conditions and deposition rate, this function can quickly and accurately predict the deposition rate under different process conditions, constructing an efficient and accurate connection between experimental parameters and deposition rate, and realizing the combination of numerical simulation and machine learning.

[0058] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0059] Figure 1 is the flowchart of a data-driven CVD-SiC deposition rate prediction model and method of the present invention;

[0060] Figure 2 is the two-dimensional symmetric schematic diagram of the geometric model of a data-driven CVD-SiC deposition rate prediction model and method of the present invention;

[0061] Figure 3 is the temperature field distribution diagram of a data-driven CVD-SiC deposition rate prediction model and method of the present invention;

[0062] Figure 4 is the comparison diagram of the simulation value and the true value of a data-driven CVD-SiC deposition rate prediction model and method of the present invention;

[0063] Figure 5 is the comparison diagram of the calculated values of the random forest and Xgboost algorithms and the true value of a data-driven CVD-SiC deposition rate prediction model and method of the present invention;

[0064] Reference Signs

[0065] 1. Reactor; 11. Inlet; 12. Outlet; 13. Porous disk; 14. Insulation layer; 2. Graphite sheet; 3. Deposition specimen; 4. Carbon fiber; 5. Reaction chamber. Detailed Implementation Modes

[0066] Embodiment

[0067] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0068] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0069] It should be noted that like reference numerals and letters denote like items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0070] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the inventive product is usually placed during use. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention.

[0071] In the description of the present invention, it should also be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0072] The following will describe in detail some implementation modes of the present invention with reference to the accompanying drawings. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0073] Such as Figures 1-5As shown in the figure, a data-driven CVD-SiC deposition rate prediction model includes a reactor 1, a graphite sheet 2, and a deposition specimen 3. The reactor 1 and the deposition specimen 3 are both arranged inside the reactor 1. The graphite sheet 2 is attached to the inner wall of the reactor 1. The deposition specimen 3 is suspended at the middle position inside the reactor 1 through a carbon fiber 4. The inside of the reactor 1 is a reaction chamber 5.

[0074] The reactor 1 includes a heat insulation layer 14, an inlet 11, an outlet 12, and a porous disk 13. The inlet 11 is arranged at one end of the heat insulation layer 14. The porous disk 13 is arranged inside the heat insulation layer 14 near the inlet 11. The outlet 12 is arranged at the other end of the heat insulation layer 14.

[0075] During operation, an eddy current effect is generated by an externally applied induction coil to heat the entire reaction chamber 5. The reaction gas enters the reaction chamber 5 through the inlet 11, reacts on the surface of the deposition specimen, and the remaining reaction gas is discharged from the outlet 12.

[0076] The present invention also provides a data-driven CVD-SiC deposition rate prediction method, including the following steps: The present invention uses a MTS-H2-Ar system precursor (CH3SICl3-H2-Ar).

[0077] S1. Establish a geometric model for deposition rate prediction according to the actual test conditions and equipment.

[0078] S11. Determine the actual size of the vertical hot-wall reactor and the actual shape and size of the substrate.

[0079] S12. Establish a geometric model of the reactor according to the actual size determined in S11 and divide the network to form a geometric region.

[0080] S13. Add material properties to the geometric region and the boundary of the reactor model. The boundary of the reactor model is set to 900 - 1300 °C.

[0081] S2. After establishing the geometric model, establish a fluid heat transfer model and a laminar flow model, calculate the temperature field and the velocity field, and connect them.

[0082] S21. Calculate the velocity field of the laminar flow model according to the Navier-Stokes equation and the continuity equation. The process is as follows:

[0083]

[0084] where ρ represents the density of the fluid, u represents the velocity field of the fluid, p represents the pressure field of the fluid, μ represents the dynamic viscosity of the fluid, f represents the external force, represents the gradient operator;

[0085] S22. Calculate the temperature field of the heat transfer model using the energy conservation equation. Use the solid heat transfer boundary condition between the graphite sheet 2 and the adiabatic layer 14, add the corresponding thermal conductivity and heat capacity values, select the ideal gas for the fluid type, and its properties are all from the selected materials. The process is as follows:

[0086]

[0087] Where T represents the temperature field, k represents the thermal conductivity, c p represents the heat capacity, and Q represents the heat source;

[0088] S23. Since the velocity field u calculated by the laminar flow model is automatically transferred to the heat transfer model, and the temperature field T calculated by the heat transfer model is used to calculate the density of the fluid and the change in the dynamic viscosity of the fluid, the laminar flow model and the heat transfer model are coupled into a non-isothermal flow model.

[0089] S3. Establish a dilute mass transfer model and a chemical reaction model, and input the 4 gas-phase reactions, 1 surface reaction, and 2 adsorption reactions of the CH3SICl3 - H2 - Ar gas into the model as shown in Table 1, and calculate the chemical reaction and concentration field distribution;

[0090] Table 1

[0091]

[0092] S31. Establish a chemical reaction model, input the relevant reactions of CH3SICl3 - H2 - Ar, and determine the gas-phase reaction rate constant, surface reaction rate constant, and adsorption reaction rate constant according to the three-parameter Arrhenius formula. The process is as follows:

[0093] K = A * T n * exp(-E / RT);

[0094]

[0095] Where K represents the surface reaction rate constant, A represents the reaction frequency factor, T n represents the reaction temperature exponent, E represents the reaction activation energy exponent, exp(·) represents the exponential function with base e, R represents the gas constant, L1 represents the adsorption reaction rate constant, τ represents the adhesion coefficient, M k represents the molar mass of the adsorbate, v a represents the effective surface vibration frequency 10 13 E a represents the adsorption energy, and k represents the Boltzmann constant;

[0096] S32. Use the characteristic length σ (10 -10 m) of the main substance and the Lennard-Jones interaction potential ε / k b(K) The minimum energy value is used to calculate the collision integral Ω as shown in Table 2 D ;

[0097] Table 2

[0098]

[0099] S33. Calculate the diffusivity of the substance using the following formula:

[0100]

[0101] S34. Establish a dilute substance transfer model using the Fick diffusion model and an additional convective transport mechanism. The mixed density ρ1 is:

[0102]

[0103] where p1 represents the total pressure of the system, x i represents the mole fraction of the i-th substance, and R represents the gas constant;

[0104] S35. Calculate the concentration field of the formula using the formula:

[0105]

[0106] where represents the mass flux density, ω i represents the mass fraction of the i-th substance, R i represents the generation or consumption rate of the i-th substance, j i represents the mass flux density of the i-th substance, D i represents the diffusion coefficient of the i-th substance, c i represents the concentration of the i-th substance.

[0107] S4. The results in S2 and S3 form a multi-physical field model, and the deposition rate is obtained by multi-physical field coupling solution for the simulation value;

[0108] S41. Couple the laminar flow model and the dilute substance transfer model to form a reaction flow-dilute substance model, and combine the content in S2 and S3 to form a unified multi-physical field simulation model;

[0109] S42. Perform multi-physical field coupling solution for the deposition rate of the multi-physical field simulation model to obtain the simulation value;

[0110] S43. Use the bisection method to adjust the SiCl2 adsorption coefficient in the chemical reaction model to make the simulation value fit the true value obtained from the experiment.

[0111] S5. Using the established numerical simulation model, the deposition rates of different process parameters are calculated. The parameter values ​​of different process conditions are input into the calibrated numerical model as shown in Table 3 and Table 4, and the deposition rates are predicted respectively. Finally, 90 sets of simulation data are obtained. The results after processing are as follows;

[0112] Table 3

[0113]

[0114] Table 4

[0115]

[0116] S6. Use K nearest neighbor algorithm to fill the missing values ​​of features in the simulation value;

[0117] When the simulated value has missing features, K nearest neighbors are selected and the weighted sum of the distances is used as the predicted value of the missing value. The process is as follows:

[0118]

[0119] Where L P Indicates distance, H i =[x i1 ,x i2 ,…,x in ]∈R n represents the i-th sample, H j represents the jth sample, x if represents the fth eigenvalue of the i-th sample, x jf represents the fth eigenvalue of the jth sample, x im represents missing values, x jm represents the mth eigenvalue of the jth sample.

[0120] S7, integrating the results of S4, S5 and S6, and building a prediction model through machine learning algorithm;

[0121] S71, summarizing the experimental true value, the deposition rates of different process parameters obtained in S5, and the supplemented simulation value obtained in S6 to obtain a database of CVD process parameters-deposition rates;

[0122] S72. Use the random forest algorithm and Xgboost algorithm to build a prediction model of process parameter-deposition rate. The process is as follows:

[0123]

[0124] Where X i =[x i1 ,x i2 ,…,xin T ∈R n represents the CVD process parameters, t i = [t i1 , t i2 , …, t im T ∈R m represents the deposition rate measured experimentally, MTS represents methyltrichlorosilane, KH2 represents the ratio of H2 to MTS, and Dia represents the reactor diameter.

[0125] Therefore, the present invention adopts a data-driven CVD-SiC deposition rate prediction model and method with the above structure. By establishing a more concise chemical reaction model and multi-field coupling of heat transfer and mass transfer, the deposition rate of the deposition surface under different process conditions can be accurately predicted. Through this numerical simulation model, a large number of process parameters uniformly distributed in the parameter space can be simulated, thereby expanding the database of process conditions - deposition rate;

[0126] At the same time, machine learning methods such as the Randomforest algorithm and the Xgboost algorithm are used to establish an accurate correlation function between process conditions and deposition rate. This function can quickly and accurately predict the deposition rate under different process conditions, constructing an efficient and accurate connection between experimental parameters and deposition rate, and realizing the combination of numerical simulation and machine learning.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.​​

Claims

1. A data-driven CVD-SiC deposition rate prediction model, characterized by: It includes a reactor, a graphite sheet and a deposition sample. The reactor and the deposition sample are both arranged inside the reactor. The reactor is close to the inner wall of the reactor. The deposition sample is suspended in the middle position of the reactor through carbon fiber. The interior of the reactor is a reaction chamber.

2. A data-driven CVD-SiC deposition rate prediction model according to claim 1, characterized in that: The reactor comprises an insulating layer, an inlet, an outlet and a porous disc. The inlet is arranged at one end of the insulating layer, the porous disc is arranged at one end of the insulating layer near the inlet, and the outlet is arranged at the other end of the insulating layer.

3. A data-driven CVD-SiC deposition rate prediction method, characterized in that: A data-driven CVD-SiC deposition rate prediction model according to any one of claims 1 to 2 is used, comprising the following steps: S1. Establish a geometric model for deposition rate prediction based on actual test conditions and equipment; S2. After establishing the geometric model, establish the fluid heat transfer model and laminar flow model, calculate the temperature field and velocity field, and link them together; S3. Establish a dilute species transfer model and a chemical reaction model, input the precursor gas reaction into the model, and calculate the chemical reaction and concentration field distribution; The results in S4, S2 and S3 form a multi-physics model, and the multi-physics coupling is performed to solve the deposition rate to obtain the simulation value; S5. Calculate the deposition rate of different process parameters using the established numerical simulation model; S6. Use K nearest neighbor algorithm to fill the missing values ​​of features in the simulation value; S7: Integrate the results of S4, S5 and S6, and build a prediction model through machine learning algorithm.

4. A data-driven CVD-SiC deposition rate prediction method according to claim 3, characterized in that: The process of establishing the geometric model for deposition rate prediction in S1 is as follows: S11, determining the actual size of the vertical hot wall reactor and the actual shape and size of the substrate; S12, establishing a geometric model of the reactor according to the actual size determined in S11 and dividing the network to form geometric regions; S13. Add material properties to the geometry regions and boundaries of the reactor model.

5. A data-driven CVD-SiC deposition rate prediction method according to claim 4, characterized in that: The process of calculating the temperature field and flow field in S2 is as follows: S21. Calculate the velocity field of the laminar flow model according to the Navier-Stokes equation and the continuity equation. The process is as follows: Where ρ represents the density of the fluid, u represents the velocity field of the fluid, p represents the pressure field of the fluid, μ represents the dynamic viscosity of the fluid, and f represents the external force. represents the gradient operator; S22. Use the energy conservation equation to calculate the temperature field of the heat transfer model. The process is as follows: Where T represents the temperature field, k represents the thermal conductivity, c p represents heat capacity, Q represents heat source; S23. Since the velocity field u calculated by the laminar flow model is automatically transferred to the heat transfer model, and the temperature field T calculated by the heat transfer model is used to calculate the density of the fluid and the change in the dynamic viscosity of the fluid, the laminar flow model and the heat transfer model are coupled into a non-isothermal flow model.

6. A data-driven CVD-SiC deposition rate prediction method according to claim 5, characterized in that: The process of calculating chemical reactions and concentration field distribution in S3 is as follows S31. Establish a chemical reaction model, input the relevant reactions of the precursor, and determine the gas phase reaction and surface reaction rate constants and adsorption reaction rate constants according to the three-parameter Arrhenius formula. The process is as follows: K=A*T n *exp(-E / RT); Where K represents the surface reaction rate constant, A represents the reaction frequency factor, T n represents the reaction temperature index, E represents the reaction activation energy index, exp(·) represents the exponential function with base e, R represents the gas constant, L1 represents the adsorption reaction rate constant, τ represents the adhesion coefficient, M k represents the molar mass of the adsorbate, v a Indicates the effective surface vibration frequency 10 13 , E a represents the adsorption energy, k represents the Boltzmann constant; S32. Calculate the collision integral Ω using the characteristic length of the main material and the minimum energy value of the Lennard-Jones interaction potential D ; S33. Use the following formula to calculate the diffusion rate of the substance: S34. Use Fick's diffusion model and additional convection transfer mechanism to establish a dilute species transfer model, and the mixed density ρ1 is: Where p1 represents the total pressure of the system, x i represents the mole fraction of the ith substance, and R represents the gas constant; S35. Calculate the concentration field using the formula: Where j represents the mass flux density, ω i represents the mass fraction of the ith substance, R i represents the generation or consumption rate of the i-th substance, j i represents the mass flux density of the ith substance, D i represents the diffusion coefficient of the ith substance, c i represents the concentration of the ith substance.

7. A data-driven CVD-SiC deposition rate prediction method according to claim 6, characterized in that: The process of obtaining the simulated value in S4 is as follows: S41, coupling the laminar flow model and the dilute species transfer model to form a reactive flow-dilute species model, and combining the contents of S2 and S3 to form a unified multi-physics simulation model; S42, performing multi-physics field coupling on the multi-physics field simulation model to solve the deposition rate and obtain a simulation value; S43. Use the dichotomy method to adjust the adsorption reaction rate constant in the chemical reaction model so that the simulated value fits the actual value obtained from the experiment.

8. A data-driven CVD-SiC deposition rate prediction method according to claim 7, characterized in that: In S5, the parameter values ​​of different process conditions are input into the calibrated numerical model, and the deposition rates are predicted respectively, and finally several groups of simulation data are obtained.

9. A data-driven CVD-SiC deposition rate prediction method according to claim 8, characterized in that: In S6, when the simulated value has a missing value, K nearest neighbors are selected and the weighted sum of the distances is used as the predicted value of the missing value. The process is as follows: Where L P Indicates distance, H i =[x i1 ,x i2 ,…,x in ]∈R n represents the i-th sample, H j represents the jth sample, x if represents the fth eigenvalue of the i-th sample, x jf represents the fth eigenvalue of the jth sample, x im represents missing values, x jm represents the mth eigenvalue of the jth sample.

10. A data-driven CVD-SiC deposition rate prediction method according to claim 9, characterized in that: The process of building a prediction model in S7 is as follows: S71, summarizing the experimental true value, the deposition rates of different process parameters obtained in S5, and the supplemented simulation value obtained in S6 to obtain a database of CVD process parameters-deposition rates; S72. Use the random forest algorithm and Xgboost algorithm to build a prediction model of process parameter-deposition rate. The process is as follows: Where X i =[x i1 ,x i2 ,…,x in ] T ∈R n represents the CVD process parameters, t i =[t i1 ,t i2 ,…,t im ] T ∈R m represents the experimentally measured deposition rate, MTS represents methyltrichlorosilane, KH2 represents the ratio of H2 to MTS, and Dia represents the reactor diameter.

Images