A method and device for controlling the growth of silicon carbide sublimation method crystals, and an electronic device

Abstract

The embodiment of the specification discloses a growth regulation method and device of silicon carbide sublimation method crystal and electronic equipment. The method comprises the following steps: acquiring parameter data of the silicon carbide crystal, simulating the physical field data corresponding to the silicon carbide crystal according to the parameter data; determining the current growth stage of the silicon carbide crystal based on the physical field data, and determining the first prediction model corresponding to the current growth stage; taking the parameter data as the input data, outputting the crystal quality data from the first prediction model; and in response to any item target crystal quality data in the crystal quality data being lower than the quality threshold value corresponding to the current growth stage, querying the optimal parameter to adjust the parameter data to the optimal parameter. In the embodiment, the corresponding optimal parameter can be queried in time for optimization and adjustment according to the abnormal target crystal quality data during the growth process, the crystal quality and growth consistency are significantly improved, the growth stability is better, and the yield is higher.

Description

Technical Field

[0001] The embodiments in this specification pertain to the field of semiconductor material growth control, and particularly relate to a method, apparatus, and electronic device for controlling the growth of silicon carbide sublimation crystals. Background Technology

[0002] Silicon carbide (SiC) is widely used in electronic devices operating under extreme conditions such as high power, high frequency, and high temperature due to its excellent properties, including a wide bandgap, high thermal conductivity, and high breakdown electric field. With the continuous improvement of device performance, the demand for high-quality, large-size SiC single crystal materials is increasing. Among these, the physical vapor transport (PVT) method is currently the mainstream SiC crystal growth process. However, this method suffers from key technical bottlenecks such as complex temperature control, large fluctuations in sublimation rate, and high crystal defect density, which severely restrict the stable improvement of crystal quality.

[0003] Current process parameter settings largely rely on expert experience or limited experimental data, lacking systematic modeling and prediction mechanisms. This makes it difficult to reveal the deep coupling relationship between multi-physics fields (temperature, flow, concentration, etc.) and crystal morphology evolution. While traditional PID (Proportional-Integral-Derivative) control can provide feedback regulation for some single variables (such as temperature or airflow), its core algorithm assumes a linear, predictable system that responds to single-variable disturbances. This makes it ill-suited to the highly nonlinear, strongly coupled, time-delay, and dynamic disturbance problems prevalent in silicon carbide crystal growth. Especially during crystal growth, minute parameter fluctuations often cause drastic changes in crystal structure, morphology, and defect distribution, rendering traditional control strategies inadequate in terms of both stability and predictability. Summary of the Invention

[0004] The embodiments of this disclosure provide a method, apparatus, and electronic device for controlling the growth of silicon carbide sublimation crystals, which are intended to solve one or more of the above-mentioned problems and other potential problems.

[0005] According to a first aspect of this disclosure, a method for controlling the growth of silicon carbide sublimation crystals is provided. The method includes acquiring parameter data of the epitaxial growth process of silicon carbide crystals in a current PVT system, simulating physical field data corresponding to the silicon carbide crystals based on the parameter data, wherein the parameter data includes process parameters, geometric parameters, and control parameters, and the physical field data includes temperature field data, gas flow field data, and spatial distribution data of sublimation / deposition rates; determining the current growth stage of the silicon carbide crystals based on the physical field data, and determining a first prediction model corresponding to the current growth stage; using the parameter data as input data, outputting crystal quality data from the first prediction model, wherein the crystal quality data includes thickness non-uniformity, defect density, and interface integrity; and responding to the presence of any target crystal quality data that is lower than the quality threshold corresponding to the current growth stage, querying the optimal parameters for the data type of the target crystal quality data corresponding to the current growth stage, so as to adjust the parameter data to the optimal parameters.

[0006] According to a second aspect of this disclosure, a growth control device for silicon carbide sublimation crystals is provided. The device includes a physical field simulation module configured to acquire parameter data of the epitaxial growth process of silicon carbide crystals in a current PVT system, and to simulate the physical field data corresponding to the silicon carbide crystals based on the parameter data. The parameter data includes process parameters, geometric parameters, and control parameters, and the physical field data includes temperature field data, gas flow field data, and spatial distribution data of sublimation / deposition rates. A prediction model determination module is configured to determine the current growth stage of the silicon carbide crystals based on the physical field data, and to determine a first prediction model corresponding to the current growth stage. A model processing module is configured to take the parameter data as input data and output crystal quality data from the first prediction model. The crystal quality data includes thickness non-uniformity, defect density, and interface integrity. A parameter adjustment module is configured to, in response to any target crystal quality data being lower than the quality threshold corresponding to the current growth stage, query the optimal parameters corresponding to the data type of the target crystal quality data in the current growth stage, and adjust the parameter data to the optimal parameters.

[0007] According to a third aspect of this disclosure, an electronic device is provided, including one or more processors and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform a method provided according to a first scheme.

[0008] According to a fourth aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method provided according to the first aspect.

[0009] The method provided in this specification allows for the determination of the current growth stage of a silicon carbide crystal based on physical field data. Then, a machine learning model (i.e., the first prediction model) corresponding to the current growth stage is selected to process the parameter data to predict crystal quality data. Finally, the optimal parameters are queried and adjusted based on a comparison between the crystal quality data and a quality threshold. This approach enables more accurate prediction of crystal quality data through machine learning models and allows for timely optimization and adjustment of the optimal parameters based on abnormal target crystal quality data during the growth process. This significantly improves crystal quality and growth consistency, resulting in better growth stability and a higher yield. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0011] Figure 1 A schematic flowchart of a method for controlling the growth of silicon carbide sublimation crystals according to some embodiments of this disclosure is shown.

[0012] Figure 2 A schematic diagram illustrating the principle process of the prediction model during the training phase of some embodiments of this disclosure is shown;

[0013] Figure 3 A schematic diagram of the structure of a growth control device for silicon carbide sublimation crystals according to some embodiments of the present disclosure is shown;

[0014] Figure 4 A schematic block diagram of an electronic device according to some embodiments of the present disclosure is shown. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0016] The terms “comprising” and “having”, and any variations thereof, in this specification, claims, and the foregoing drawings are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. Depending on the context, the word “if” as it applies herein may be interpreted as “when”, “in response to determination”, or “in response to detection”.

[0017] Figure 1 A schematic flowchart of a method 100 for controlling the growth of silicon carbide sublimation crystals according to some embodiments of this disclosure is shown. Method 100 can be executed by a terminal, which may include, but is not limited to, mobile phones, tablets, desktop computers, servers, etc. Figure 1 As shown in box 102, method 100 can obtain parameter data of the epitaxial growth process of silicon carbide crystal in the current PVT system, so as to simulate the physical field data corresponding to the silicon carbide crystal based on the parameter data. The parameter data includes process parameters, geometric parameters and control parameters, and the physical field data includes temperature field data, gas flow field data and spatial distribution data of sublimation / deposition rate.

[0018] In this embodiment, sublimation crystal refers to crystals grown by the sublimation method. For silicon carbide, this is mainly prepared using the Physical Vapor Transport (PVT) method. During the epitaxial growth of silicon carbide crystals, parameter data can be collected using non-contact in-situ sensors. Among the process parameters in the parameter data, the seed crystal temperature is used to control the deposition interface temperature of the crystal, affecting the growth rate and crystal quality; the source material temperature is used to determine the sublimation rate of silicon carbide powder; the temperature gradient, the temperature difference between the source material and the seed crystal, is used to determine the direction and intensity of material migration; the sublimation time is used to affect the crystal growth thickness and thermal field stability; the pressure is used to regulate the vapor density and deposition behavior; and the inert gas flow rate is used to regulate the transport environment and the stability of the vapor composition. Among the geometric parameters in the parameter data, the crucible geometry is used to affect the heat flow distribution and airflow disturbance; the seed crystal size is used to affect the defect propagation mode and deposition surface quality; the powder packing density and powder packing thickness are used to affect the source material consumption rate and impurity release; and the gas gap distance is used to determine the length of the temperature difference zone and the gas phase transport path. Among the control parameters in the parameter data, heater current and heater power are used to regulate the temperature field and energy input; crucible lifting rate is used to dynamically adjust the crystal growth interface and temperature distribution; crucible rotation rate is used to improve growth uniformity; and cooling rate and gas flow rate are used to affect stress release and defect formation in the later stage of crystallization.

[0019] After obtaining the parameter data, a three-dimensional transient coupled model can be constructed using finite element / finite volume simulation software (such as COMSOL Multiphysics, ANSYS Fluent, Crystal Growth Lab, etc.). The three-dimensional transient coupled model includes temperature and gas flow fields. Based on the local temperature and gas supersaturation at each location in the model, the sublimation rate of the raw material region at each location can be determined. Furthermore, based on the gas flux reaching the surface, the deposition rate on the seed crystal surface can be determined, thereby obtaining the sublimation rate distribution and deposition rate distribution.

[0020] In box 104, method 100 can determine the current growth stage of the silicon carbide crystal based on physical field data, and determine the first prediction model corresponding to the current growth stage.

[0021] In this embodiment, the physical field data of silicon carbide crystals differ significantly at different growth stages, thus the current growth stage of the silicon carbide crystal can be determined based on the physical field data. To ensure the accuracy of the model prediction, different prediction models will be trained separately for each growth stage. That is, only the parameter data and crystal quality data corresponding to that growth stage will be used to train a prediction model specifically for quality prediction at that growth stage. Each growth stage will have a corresponding prediction model trained. After determining the current growth stage, the prediction model corresponding to the current growth stage will be used as the first prediction model for subsequent data processing.

[0022] As an example, different physical field data judgment conditions can be set according to the characteristics of different growth stages. The current growth stage can be determined based on the judgment conditions satisfied by the physical field data. Specifically, in the nucleation and initial growth stages, the temperature of the raw material region in the temperature field has just reached the sublimation point (2200–2400℃), while the temperature of the seed crystal region is slightly lower (2000–2200℃), and the axial temperature gradient is large. In the gas flow field, gas convection begins, but the flow rate is low, with diffusion dominating, and the concentration of gaseous components (Si, C) gradually increases on the seed crystal surface. In the spatial distribution data, the raw material surface begins to sublimate slightly, at a low and uneven rate. The gas phase supersaturation on the seed crystal surface just exceeds the critical value, and surface nucleation begins, with an extremely low deposition rate. Judgment conditions could include a discontinuous distribution of deposition rate on the seed crystal surface, gas phase transport driven by the temperature gradient direction, but the deposition area concentrated at the seed crystal center or defects, and the gas flow not yet forming a stable convection structure.

[0023] During the stable growth stage, the axial temperature gradient of the temperature field is stable (typical value: 20–50℃ / cm), the radial temperature distribution is uniform, the thermal field has good symmetry, and the temperature difference between the raw material and the seed crystal is maintained at 50–150℃. The airflow field forms stable natural convection, with the airflow rising from the center of the crucible and descending at the edges, forming a single or multi-vortex structure. The transport of gas phase components is stable, and diffusion and convection are in balance. In the spatial distribution data, the sublimation rate in the raw material region is uniform, maintaining a stable gas phase supply, and the deposition rate on the seed crystal surface is high and radially uniform. Judgment criteria include a radial variation in deposition rate of <10%, time invariance in the temperature and flow fields (quasi-steady state), moderate gas phase supersaturation, and no spontaneous nucleation caused by localized supersaturation.

[0024] During the interface and defect evolution stages, the thickening of the crystal in the temperature field leads to increased thermal resistance, a shift in the thermal field, an increase in temperature on the seed crystal side, a slight decrease in temperature on the feedstock side, and a reduction in the axial gradient. The convection intensity of the airflow field weakens or its structure changes (e.g., from a single vortex to a double vortex), the boundary layer thickens, and low-velocity zones or backflow appear in localized areas. In the spatial distribution data, the difference in deposition rate between the center and the edge increases, and the growth interface changes from a planar surface to a convex or concave surface. Judgment criteria can include a radially non-monotonic deposition rate distribution (e.g., high at the center, low at the edge), a temperature field showing significant interface curvature (which can be determined by the curvature of the isothermal surface), and the appearance of low-velocity zones in the flow field.

[0025] During the later growth and stress accumulation stages, significant thermal stress occurs in the temperature field, leading to an increased radial temperature gradient. The gas phase composition of the gas flow field changes due to feedstock consumption, convection weakens, diffusion dominates, and transport efficiency decreases, potentially resulting in localized supersaturation and non-uniform deposition. Spatially, the overall growth rate decreases (due to feedstock consumption and weakened driving force), and streaks appear in the deposition layer. Judgment criteria include a continuously decreasing growth rate (e.g., from 0.5 mm / h to 0.2 mm / h), a high thermal stress region in the temperature field (which can be calculated using a thermal stress model), and a Si / C ratio in the gas phase deviating from the ideal stoichiometry.

[0026] During the growth and cooling phases, the temperature field gradually decreases overall, aiming for uniform cooling. Convection in the airflow field gradually disappears, tending towards static. In the spatial distribution data, sublimation / deposition essentially ceases. The criteria for judgment can be that the heating power reaches zero, the temperature decreases monotonically, and the deposition rate is zero.

[0027] In box 106, method 100 can take parameter data as input data and output crystal quality data from a first prediction model. The crystal quality data includes thickness inhomogeneity, defect density, and interface integrity.

[0028] In this embodiment, during the training phase, the prediction model is trained using parameter data obtained from historical data, as well as quality data actually measured and calculated by instruments under those parameter data. Therefore, after acquiring the parameter data and determining the first prediction model corresponding to the current growth stage, the first prediction model can output the crystal quality data of the silicon carbide crystal at the current moment based on the parameter data. Thickness inhomogeneity refers to the thickness difference of the crystal at different radial positions, reflecting the flatness of the growth interface and the consistency of the growth rate. It can be determined by measuring the thickness at multiple points on the crystal surface using a high-precision thickness gauge (such as a laser rangefinder or contact micrometer), and by the ratio of the difference between the maximum and minimum thickness to the average thickness. Defect density generally refers to screw dislocation density, specifically non-microtubular screw dislocations, which also affects the reverse blocking performance of the device. It can be determined by the ratio of the number of screw dislocation etch pits to the observed area. Interface integrity refers to the geometric shape and flatness of the growth interface, reflecting the stability of the thermal field and mass transport. It can be obtained by extracting contour lines from XRT images and fitting the shape of the growth interface (planar, convex, concave). When quantitative analysis is required, it can be represented by the height difference between the center and edge of the interface or the radius of curvature.

[0029] In box 108, method 100 may, in response to the presence of any target crystal quality data in the crystal quality data that is lower than the quality threshold corresponding to the current growth stage, query the optimal parameter of the data type of the target crystal quality data corresponding to the current growth stage, so as to adjust the parameter data to the optimal parameter.

[0030] In this embodiment, based on the different growth requirements of the current growth stage, corresponding quality thresholds are pre-set for each crystal quality data (different quality thresholds can be set for crystal quality data of different data types). If the target crystal quality data corresponding to a certain data type is lower than its corresponding quality threshold, it is considered that the crystal growth state has an anomaly in a certain dimension, which may lead to growth failure if no intervention is taken. In the database, the optimal parameters for the current growth stage will be pre-determined for each data type according to its different data types. These optimal parameters are actually the optimal parameter combination for the crystal quality data of each data type. That is, the parameter data of all data types will be adjusted according to the optimal parameters, rather than just the parameter data corresponding to the target crystal quality data. In this way, when the target crystal quality data is abnormal, the parameter data can be adjusted in a timely manner according to the corresponding optimal parameters to achieve dynamic adjustment of the crystal growth process, improve crystal quality and growth consistency, achieve better growth stability, and increase the yield rate.

[0031] Instead of setting a single, universally applicable optimal parameter, we set individual optimal parameters for each data type. This is because optimizing multiple parameters involves trade-offs and prioritization in multi-objective optimization. As an example, the objective function for this stage can be set as follows:

[0032]

[0033] in, For thickness non-uniformity, For defect density, For the sake of interface integrity, These are the weighting coefficients.

[0034] Different growth stages have different requirements for the acceptable range of values ​​for the aforementioned crystal quality data. By adjusting the weighting coefficients, different parameter values ​​can be obtained within the corresponding acceptable ranges. Thus, for each data type, by assigning a higher weight to that data type, the optimal parameters that can better adjust the crystal quality data for that data type can be determined. For example, when the target crystal quality data is of thickness inhomogeneity, the weighting coefficient corresponding to thickness inhomogeneity can be set higher. This results in a smaller calculated value for thickness inhomogeneity, and the optimal parameters obtained will place greater emphasis on adjusting thickness inhomogeneity.

[0035] Based on the obtained parameter values ​​for thickness inhomogeneity, defect density, and interface integrity, the corresponding parameter data can be determined in reverse, and this parameter data is the optimal parameter. If there is a sufficient amount of historical growth data, the historical growth data that is closest to the obtained parameter values ​​for thickness inhomogeneity, defect density, and interface integrity can be queried, and the parameter data in that data can be used as the optimal parameter. If there is a limited amount of historical growth data, the simulated physical field in the finite element / finite volume simulation software can be manually adjusted based on the parameter values ​​for thickness inhomogeneity, defect density, and interface integrity. When the simulated crystal state meets the parameter value requirements, the parameter data displayed in the software at that time can be used as the optimal parameter.

[0036] In one possible implementation, the method further includes:

[0037] Based on crystal quality data within a first preset time period, a quality data curve is constructed to determine the growth quality trend of the silicon carbide crystal; and

[0038] Based on the first data interval and growth quality trend corresponding to the crystal quality data at the current moment, a quality score is determined for each data type, and the growth risk level of the silicon carbide crystal is determined according to the weighted value of each quality score.

[0039] In this embodiment, a quality data curve between crystal quality data and time can be constructed based on the crystal quality data continuously obtained within a first preset time period. According to the quality data curve, the growth quality trend (e.g., rising, falling, flattening, etc.) of the crystal quality data for each data type can be determined. Furthermore, to ensure normal crystal growth, a data range is preset for each type of crystal quality data; when the data falls within this range, the crystal quality is considered normal. Based on the first data interval corresponding to the crystal quality data at the current moment and the growth quality trend, a quality score for each data type can be determined through fuzzy judgment. As an example, suppose the data range is divided into three data intervals: A, B, and C. Interval A indicates that if the data continues to decrease, it will fall below the minimum value of the data range; similarly, interval C indicates that if the data continues to rise, it will exceed the maximum value of the data range. Therefore, if the first data interval is A and the growth quality trend is decreasing, or if the first data interval is C and the growth quality trend is increasing, the growth quality is considered poor, and a fuzzy value of "poor" can be set. If the first data interval is A and the growth quality trend is upward, or if the first data interval is C and the growth quality trend is downward, then the fuzzy value can be set to "normal". If the first data interval is B, the fuzzy value can be set to "good". Different fuzzy values ​​are preset with different quality scores. By weighting the quality scores, a weighted value can be obtained. Different weighted values ​​correspond to different growth risk levels; the lower the weighted value, the higher the growth risk level. Different handling methods can be set for different growth risk levels. For example, when the growth risk level is level one, no action can be taken; when it is level two, an early warning can be issued to remind staff to pay attention; when it is level three, a clear alarm can be triggered, automatic optimization can be stopped, and manual intervention is required for adjustment.

[0040] In one possible implementation, after constructing the quality data curve, the method further includes:

[0041] Calculate the average growth rate of the crystal quality data within a first preset time period, and calculate the estimated quality data after a second preset time period based on the average growth rate and the crystal quality data at the current moment.

[0042] Based on the second data interval corresponding to the estimated quality data, determine the estimated quality score corresponding to the estimated quality data; and

[0043] In response to the estimated quality score being less than the score threshold, the parameter data at the current moment is adjusted to the recommended parameters corresponding to the current growth stage.

[0044] In this embodiment, the average growth rate can be calculated based on the ratio of the difference between the final value and the initial value of the crystal quality data within a first preset time period to the first preset time period. Then, the estimated quality data after a second preset time period is calculated using this average growth rate, and the estimated quality score is determined based on the estimated quality data. Compared to the aforementioned method of determining the quality score based on numerical intervals and growth quality trends, since the estimated quality data is a prediction at a future moment and cannot determine the growth quality trend at that moment, the estimated quality score can be determined solely based on the second data interval. That is, determining which second data interval the estimated quality data falls into. Each second data interval has a pre-set corresponding quality score, thus determining the estimated quality score. If the estimated quality score is less than a preset score threshold, it is considered that if the current state of growth continues, crystal growth may encounter problems. In this case, the optimization parameters are not considered; the parameter data is adjusted to preset recommended parameters to prioritize the normal progress of the growth process.

[0045] In one possible implementation, the method includes:

[0046] Determine the second prediction model corresponding to the next growth stage; and

[0047] In response to changes in the growth stage, target parameter data within a third preset time period from the current moment is obtained from the historical parameter data of the first prediction model, and the target parameter data is input into the second prediction model according to the timestamp.

[0048] In this embodiment, when the growth stage changes based on the physical field data, it is necessary to switch to the second prediction model for the next growth stage to predict the crystal quality data. To avoid significant deviations in the prediction results due to insufficient data in the second prediction model immediately after the model switch, the last part of the target parameter data from the historical parameter data of the first prediction model can be input into the second prediction model according to the timestamp when the growth stage changes. This ensures that the second prediction model has sufficient data in the initial stage, improving the accuracy of subsequent predictions.

[0049] In one possible implementation, the method further includes:

[0050] Based on historical growth data, training samples are determined, including parameter data samples and crystal quality data samples.

[0051] Based on the parameter data samples, predicted crystal quality data is generated from the initial prediction model; and

[0052] Using crystal quality data samples as supervision signals, the initial prediction model is trained for at least one round to obtain the prediction model.

[0053] In this embodiment, as Figure 2 As shown, Figure 2 A schematic diagram of the principle process 200 of the prediction model in the training phase of some embodiments of this disclosure is shown. In process 200, historical growth data can store historical parameter data and historical crystal quality data used for each silicon carbide crystal growth. The historical crystal quality data can be obtained by acquiring images of the silicon carbide crystal surface using optical microscopes, scanning electron microscopes, atomic force microscopes, etc., and then extracting structural features (such as crystal step density, surface roughness, etc.) from the crystal surface images using, for example, a convolutional neural network model based on ResNet-50, and calculating the quality based on the feature values ​​corresponding to the structural features. Next, the historical parameter data can be used as parameter data sample 210-1, and the historical crystal quality data as crystal quality data sample 210-2, and the parameter data sample 210-1 and crystal quality data sample 210-2 corresponding to the same growth are associated. Training samples 210 can be constructed based on parameter data sample 210-1 and crystal quality data sample 210-2. The initial prediction model used for training can be a random forest model, an XGBoost model, a convolutional neural network model, a Transformer model, etc. During the training of the prediction model 220 using training sample 210, the generator 221 can generate predicted crystal quality data 222 based on parameter data sample 210-1. By comparing crystal quality data sample 210-2 and the predicted crystal quality data 222, a generator loss 223 can be obtained. This generator loss 223 is then used to assess the loss of the predicted crystal quality data 222, using a comparison loss function. The generator loss 223 can be a relatively large value, and it can be backpropagated to generator 221 to guide the optimization of generator 221's parameters, thus achieving one round of supervised training for generator 221. This training process can be iterated repeatedly until generator 221 can generate more accurate predicted crystal quality data 222, i.e., until the loss value calculated by the loss function is smaller. After training, the prediction model 220 can output crystal quality data 230.

[0054] Figure 3 A schematic diagram of the structure of a silicon carbide sublimation crystal growth control device 300 according to some embodiments of this disclosure is shown. The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. Figure 3As shown, the device 300 includes a physical field simulation module 301, configured to acquire parameter data of the epitaxial growth process of silicon carbide crystal in the current PVT system, so as to simulate the physical field data corresponding to the silicon carbide crystal based on the parameter data. The parameter data includes process parameters, geometric parameters, and control parameters. The physical field data includes temperature field data, gas flow field data, and spatial distribution data of sublimation / deposition rate. A prediction model determination module 302 is configured to determine the current growth stage of the silicon carbide crystal based on the physical field data, and determine the first prediction model corresponding to the current growth stage. A model processing module 303 is configured to take the parameter data as input data and output crystal quality data from the first prediction model. The crystal quality data includes thickness non-uniformity, defect density, and interface integrity. A parameter adjustment module 304 is configured to respond to the fact that any target crystal quality data in the crystal quality data is lower than the quality threshold corresponding to the current growth stage, query the optimal parameter corresponding to the data type of the target crystal quality data in the current growth stage, and adjust the parameter data to the optimal parameter.

[0055] In one possible implementation, the process parameters include seed crystal temperature, source material temperature, temperature gradient, sublimation time, pressure, and inert gas flow rate; the geometric parameters include crucible geometry, seed crystal size, powder packing density, powder packing thickness, and air gap distance; and the control parameters include heater current, heater power, crucible lifting rate, crucible rotation rate, cooling rate, and gas flow rate.

[0056] In one possible implementation, the growth stages of silicon carbide crystals include a nucleation and initial growth stage, a stable growth stage, an interface evolution and defect evolution stage, a later growth and stress accumulation stage, and a growth termination and cooling stage.

[0057] In one possible implementation, the device further includes: a risk level determination module, configured to construct a quality data curve based on crystal quality data within a first preset time period, so as to determine the growth quality trend of the silicon carbide crystal according to the quality data curve; and to determine a quality score corresponding to each data type based on a first data interval and growth quality trend corresponding to the crystal quality data at the current moment, so as to determine the growth risk level of the silicon carbide crystal according to the weighted value of each quality score.

[0058] In one possible implementation, the risk level determination module is further configured to calculate the average growth rate of crystal quality data within a first preset time period, and to calculate the estimated quality data after a second preset time period based on the average growth rate and the crystal quality data at the current moment; determine the estimated quality score corresponding to the estimated quality data based on the second data interval corresponding to the estimated quality data; and adjust the parameter data at the current moment to the recommended parameters corresponding to the current growth stage in response to the estimated quality score being less than the score threshold.

[0059] In one possible implementation, the device further includes: a parameter input module configured to determine a second prediction model corresponding to the next growth stage; and in response to a change in the growth stage, to obtain target parameter data within a third preset time period from the historical parameter data of the first prediction model, and to input the target parameter data into the second prediction model according to the timestamp.

[0060] In one possible implementation, the apparatus further includes: a model training module configured to determine training samples based on historical growth data, the training samples including parameter data samples and crystal quality data samples; generate predicted crystal quality data from an initial prediction model based on the parameter data samples; and perform at least one round of model training on the initial prediction model using the crystal quality data samples as a supervision signal to obtain a prediction model.

[0061] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this specification is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0062] Figure 4 A block diagram of an electronic device 400 that can implement various embodiments of the present disclosure is shown. For example... Figure 4 As shown, the electronic device 400 includes a processor 410, a disk drive 420, an input / output interface 430, a network interface 440, and a memory 450. The processor 410, disk drive 420, input / output interface 430, network interface 440, and memory 450 can communicate with each other via a communication bus 460.

[0063] The processor 410 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.

[0064] The memory 450 can be implemented in the form of ROM (Read Only Memory), RAM (Read Access Memory), static memory, dynamic storage devices, etc. The memory 450 can store the operating system 451 used to control the operation of the electronic device 400, and the basic input / output system (BIOS) 452 used to control the low-level operations of the electronic device 400. Additionally, it can store a web browser 453, a data storage management system 454, etc. In summary, when the technical solution provided in this application is implemented through software or firmware, the relevant program code is stored in the memory 450 and is called and executed by the processor 410.

[0065] Input / output interface 430 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0066] Network interface 440 is used to connect a communication module (not shown in the figure) to enable communication and interaction between the device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0067] Bus 460 includes a pathway for transmitting information between various components of the device, such as processor 410, disk drive 420, input / output interface 430, network interface 440, and memory 450.

[0068] It should be noted that although the above-described device only shows the processor 410, disk drive 420, input / output interface 430, network interface 440, memory 450, bus 460, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the method of this application, and does not necessarily include all the components shown in the figures.

[0069] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0070] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. Furthermore, although operations are depicted in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0071] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for controlling the growth of silicon carbide sublimation crystals, characterized in that, The method includes: The parameter data of the epitaxial growth process of silicon carbide crystal in the current PVT system are obtained, so as to simulate the physical field data corresponding to the silicon carbide crystal based on the parameter data. The parameter data includes process parameters, geometric parameters and control parameters, and the physical field data includes temperature field data, gas flow field data and spatial distribution data of sublimation / deposition rate. Based on the physical field data, the current growth stage of the silicon carbide crystal is determined, and the first prediction model corresponding to the current growth stage is determined. Using the parameter data as input data, the first prediction model outputs crystal quality data, which includes thickness inhomogeneity, defect density, and interface integrity; and In response to the fact that any target crystal quality data in the crystal quality data is lower than the quality threshold corresponding to the current growth stage, the optimal parameter of the data type of the target crystal quality data corresponding to the current growth stage is queried, so as to adjust the parameter data to the optimal parameter; Based on the crystal quality data within a first preset time period, a quality data curve is constructed to determine the growth quality trend of the silicon carbide crystal according to the quality data curve; and Based on the first data interval corresponding to the crystal quality data at the current moment and the growth quality trend, a quality score corresponding to each data type is determined, so as to determine the growth risk level of the silicon carbide crystal according to the weighted value of each quality score.

2. The method for controlling the growth of silicon carbide sublimation crystals according to claim 1, characterized in that, The process parameters include seed crystal temperature, source material temperature, temperature gradient, sublimation time, pressure, and inert gas flow rate. The geometric parameters include crucible geometry, seed crystal size, powder packing density, powder packing thickness, and air gap distance. The control parameters include heater current, heater power, crucible lifting rate, crucible rotation rate, cooling rate, and gas flow rate.

3. The method for controlling the growth of silicon carbide sublimation crystals according to claim 1, characterized in that, The growth stages of the silicon carbide crystal include the nucleation and initial growth stage, the stable growth stage, the interface evolution and defect evolution stage, the later growth and stress accumulation stage, and the growth end and cooling stage.

4. The method for controlling the growth of silicon carbide sublimation crystals according to claim 1, characterized in that, After constructing the quality data curve, the following is also included: Calculate the average growth rate of the crystal quality data within the first preset time period, and calculate the estimated quality data after the second preset time period based on the average growth rate and the crystal quality data at the current moment. Based on the second data interval corresponding to the estimated quality data, determine the estimated quality score corresponding to the estimated quality data; and In response to the estimated quality score being less than the score threshold, the parameter data at the current moment is adjusted to the recommended parameters corresponding to the current growth stage.

5. The method for controlling the growth of silicon carbide sublimation crystals according to claim 1, characterized in that, The method includes: Determine the second prediction model corresponding to the next growth stage; and In response to changes in the growth stage, target parameter data within a third preset time period from the current moment is obtained from the historical parameter data of the first prediction model, and the target parameter data is input into the second prediction model according to the timestamp.

6. The method for controlling the growth of silicon carbide sublimation crystals according to claim 1, characterized in that, The method further includes: Based on historical growth data, training samples are determined, including parameter data samples and crystal quality data samples. Based on the parameter data sample, predicted crystal quality data is generated from the initial prediction model; and Using the crystal quality data samples as supervision signals, the initial prediction model is trained for at least one round to obtain the prediction model.

7. A growth control device for silicon carbide sublimation crystals, characterized in that, The device includes: The physical field simulation module is configured to acquire parameter data of the epitaxial growth process of silicon carbide crystal in the current PVT system, so as to simulate the physical field data corresponding to the silicon carbide crystal based on the parameter data. The parameter data includes process parameters, geometric parameters and control parameters, and the physical field data includes temperature field data, gas flow field data and spatial distribution data of sublimation / deposition rate. The prediction model determination module is configured to determine the current growth stage of the silicon carbide crystal based on the physical field data, and to determine the first prediction model corresponding to the current growth stage. The model processing module is configured to take the parameter data as input and output crystal quality data from the first prediction model. The crystal quality data includes thickness inhomogeneity, defect density, and interface integrity. The parameter adjustment module is configured to, in response to any target crystal quality data in the crystal quality data being lower than the quality threshold corresponding to the current growth stage, query the optimal parameter corresponding to the data type of the target crystal quality data in the current growth stage, and adjust the parameter data to the optimal parameter. The risk level determination module is configured to construct a quality data curve based on crystal quality data within a first preset time period, so as to determine the growth quality trend of silicon carbide crystals according to the quality data curve; and to determine the quality score corresponding to each data type based on the first data interval and growth quality trend corresponding to the crystal quality data at the current moment, so as to determine the growth risk level of silicon carbide crystals according to the weighted value of each quality score.

8. An electronic device, comprising: One or more processors, and A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the growth control method for silicon carbide sublimation crystal according to any one of claims 1-6.

9. A computer program product, comprising a computer program that, when executed by a processor, implements a method for controlling the growth of silicon carbide sublimation crystals according to any one of claims 1-6.

Images