Single crystal growth control method and system based on temperature field data fitting

Abstract

The present application relates to the technical field of data fitting, and proposes a single crystal growth control method and system based on temperature field data fitting, the method comprising: based on the multi-source temperature measurement data of a single crystal growth furnace, fitting the continuous distribution state of the temperature field in the furnace at the current time to obtain a temperature field fitting model of the single crystal growth furnace; in the same spatial coordinate system, registering the continuous temperature field fitting surface and the ideal temperature field surface of the single crystal growth furnace to obtain a temperature field deviation scalar field by taking the difference; extracting the deviation characteristics of different regions in the temperature field deviation scalar field and combining them into a deviation feature vector; and according to the deviation feature vector, performing closed-loop regulation and control on the single crystal growth furnace; the present application can improve the efficiency of single crystal growth control.

Description

Technical Field

[0001] This invention relates to the field of data fitting technology, and in particular to a method and system for controlling single crystal growth based on temperature field data fitting. Background Technology

[0002] Single-crystal materials are core foundational materials for high-end manufacturing industries such as semiconductors and photovoltaics. The quality of single-crystal growth directly determines the performance and yield of downstream devices. The temperature field distribution within the furnace during single-crystal growth directly determines the crystal forming quality and the level of internal defects. However, due to the limitations of the high-temperature, enclosed environment, the growth interface cannot be directly measured. Existing technologies rely solely on single-point or local discrete temperature measurement methods for temperature monitoring, lacking an integrated processing and error elimination mechanism for multi-source temperature measurement data across the entire furnace, including thermocouples, infrared thermometry, heater power, crystal pulling speed and position, crucible lifting speed and rotation speed, etc. It is difficult to accurately fit the continuous temperature field distribution within the furnace based on discrete temperature measurement points, and it is impossible to fully characterize the axial and radial temperature gradient of the furnace body. This easily leads to problems such as temperature field characterization distortion and incomplete capture of local temperature change trends, making it difficult to form a reliable global temperature field model to support refined control.

[0003] Current single-crystal growth temperature control methods mostly employ a fixed-parameter, coarse-grained adjustment mode, failing to construct a standardized ideal temperature field reference surface. This makes it impossible to achieve precise registration between the actual and ideal temperature fields, quantify the overall deviation, or extract and quantify the deviation characteristics of different temperature field regions. Furthermore, the control process heavily relies on operator experience and judgment, resulting in delayed responses to anomalies. It lacks the ability to predict and proactively intervene in quality based on the temperature field-crystal quality correlation mapping, making it difficult to implement targeted closed-loop adjustments based on the degree of temperature field deviation. This easily leads to imbalances in the solid-liquid interface temperature gradient and local deviations from the optimal growth range, resulting in defects such as crystal dislocations and multi-type inclusions, thus hindering the improvement of single-crystal yield and quality consistency. Therefore, how to achieve precise temperature field reconstruction based on multi-source data fusion and implement closed-loop intelligent control accordingly has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method and system for controlling single crystal growth based on temperature field data fitting, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a single crystal growth control method based on temperature field data fitting, comprising: S1. Based on the multi-source temperature measurement data of the single crystal growth furnace, the continuous distribution state of the temperature field inside the furnace at the current moment is fitted to obtain the temperature field fitting model of the single crystal growth furnace. S2. Under the same spatial coordinate system, the continuous temperature field fitting surface is registered with the ideal temperature field surface of the single crystal growth furnace, and the difference is calculated to obtain the temperature field deviation scalar field. S3. Extract the deviation features of different regions in the temperature field deviation scalar field and combine them into a deviation feature vector; S4. Based on the deviation feature vector, perform closed-loop control on the single crystal growth furnace.

[0006] In a preferred embodiment, the multi-source temperature measurement data includes: With a preset sampling period, the temperature values ​​of each temperature measurement point in the single crystal growth furnace are collected synchronously to obtain a set of discrete temperature measurement data; Each temperature data point is assigned a timestamp and the corresponding three-dimensional spatial coordinates of the temperature measurement point to obtain a set of temperature field sampling data. By removing gross errors caused by instantaneous sensor jumps in the temperature field sampling data set and eliminating high-frequency electronic noise, a discrete temperature field dataset is obtained.

[0007] In a preferred embodiment, the step of fitting the continuous distribution of the temperature field inside the furnace at the current moment based on the multi-source temperature measurement data of the single crystal growth furnace to obtain the temperature field fitting model of the single crystal growth furnace includes: The discrete temperature field dataset is distributed to multiple annular temperature zones divided along the axial and radial directions of the furnace body, wherein each annular temperature zone contains temperature measurement points located in the same axial and radial segments; For any annular temperature zone, the temperature variation trend along the axial direction and the temperature variation trend along the radial direction are determined by using the temperature values ​​of the internal temperature measuring points. The axial and radial temperature variation trends of adjacent annular temperature zones at the interface are smoothly connected to generate a continuous temperature field fitting model for the single crystal growth furnace.

[0008] In a preferred embodiment, determining the axial and radial temperature variation trends of any annular temperature zone using the temperature values ​​from internal temperature measuring points includes: The temperature measuring points contained within the annular temperature zone are arranged sequentially along the axial coordinate and along the radial distance, respectively. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged sequentially along the axial coordinate, the axial temperature change trend is determined to be either a heating trend or a cooling trend. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged in radial distance sequence, the radial temperature change trend is determined to be either a heating trend from the center outward or a cooling trend from the center outward.

[0009] In a preferred embodiment, the ideal temperature field surface includes: Obtain temperature field morphology records of multiple successful crystal growth batches in history, and extract temperature field morphology data corresponding to the current growth stage from them; The fusion weight of the temperature field morphology data for the current batch is determined based on the quality parameters of the crystals obtained from each historical batch. The temperature field morphology data of each batch are weighted and fused according to the fusion weight to obtain the ideal temperature field surface.

[0010] In a preferred embodiment, the step of registering the continuous temperature field fitting surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculating the difference includes: Extract the solid-liquid interface transition feature points on the continuous temperature field fitting surface and the solid-liquid interface reference feature points on the ideal temperature field surface, respectively. Align the solid-liquid interface transition feature point with the solid-liquid interface reference feature point so that the two surfaces coincide in the same spatial coordinate system, thus obtaining the registered continuous temperature field fitting surface and the ideal temperature field surface.

[0011] In a preferred embodiment, obtaining the temperature field deviation scalar field includes: The temperature values ​​at each coordinate point on the registered continuous temperature field fitting surface are compared with the temperature values ​​at the corresponding coordinate points on the ideal temperature field surface to generate a three-dimensional deviation data set. The temperature difference value of each point in the three-dimensional deviation data set is bound to the corresponding coordinate point to obtain the temperature field deviation scalar field, wherein each spatial location in the temperature field deviation scalar field stores the temperature difference scalar value at that location.

[0012] In a preferred embodiment, extracting the deviation features of different regions in the temperature field deviation scalar field includes: Based on the relative distance between each coordinate point in the temperature field deviation scalar field and the position of the solid-liquid interface in the single crystal growth furnace, the temperature field deviation scalar field is divided into a central interface region, a transition region, and an edge region. The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are extracted as deviation features of different regions.

[0013] In a preferred embodiment, the combination is a deviation feature vector, including: The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are sorted according to the degree of influence of each region on the crystal growth quality. The sorted deviation features are concatenated sequentially to obtain the deviation feature vector.

[0014] To address the aforementioned problems, the present invention also provides a single crystal growth control system based on temperature field data fitting, the system comprising: The temperature field fitting module is used to fit the continuous distribution of the temperature field inside the furnace at the current moment based on the multi-source temperature measurement data of the single crystal growth furnace, so as to obtain the temperature field fitting model of the single crystal growth furnace. The deviation scalar field construction module is used to register the continuous temperature field fitting surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculate the difference to obtain the temperature field deviation scalar field. The deviation feature construction module is used to extract deviation features from different regions in the temperature field deviation scalar field and combine them into a deviation feature vector. The closed-loop control module is used to perform closed-loop control on the single crystal growth furnace based on the deviation feature vector.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention integrates multi-source temperature measurement data from a single-crystal growth furnace and completes error elimination and noise filtering. Based on partition fitting and smooth connection of neighboring temperature zone trends, it constructs a global continuous temperature field fitting model, which can accurately restore the complete axial and radial temperature distribution and change trend inside the furnace. It breaks through the limitation of traditional single-point discrete temperature measurement that cannot characterize the continuous form of the global temperature field, and realizes a refined and accurate characterization of the temperature field state inside the furnace, providing a high-precision basic model support for subsequent quantitative analysis of temperature field deviation.

[0016] 2. This invention constructs a scalar field of temperature field deviation by weighted fusion of ideal temperature field surfaces and precise registration of hyperboloids to obtain the temperature field deviation point by point. It also extracts deviation features by region and constructs deviation feature vectors, which can realize the quantitative characterization and feature condensation of temperature field deviation. Based on the feature vectors, targeted closed-loop control can be implemented to accurately correct the solid-liquid interface and temperature field deviation in each zone of the furnace, stably maintain the optimal temperature field environment for single crystal growth, effectively suppress the generation of defects such as crystal dislocations and inclusions, and significantly improve the yield of single crystal growth products and the consistency of quality between batches. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a single crystal growth control method based on temperature field data fitting, provided in an embodiment of the present invention. Figure 2 A functional block diagram of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 3 A three-dimensional spatial distribution diagram of temperature measurement points for a single crystal growth control system based on temperature field data fitting, provided in an embodiment of the present invention; Figure 4 A furnace temperature field cloud map of the single crystal constant diameter stable growth stage of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 5A comparison diagram of axial temperature curves at various stages of a single crystal growth control system based on temperature field data fitting, provided as an embodiment of the present invention. Figure 6 A comparison diagram of radial temperature curves at various stages of a single crystal growth control system based on temperature field data fitting, provided as an embodiment of the present invention. Figure 7 A continuous temperature field fitting surface diagram of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 8 A radial temperature gradient distribution diagram of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 9 An axial temperature gradient distribution diagram of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 10 A solid-liquid interface feature point and temperature field morphology diagram of a single crystal growth control system based on temperature field data fitting is provided in an embodiment of the present invention. Figure 11 A box plot of multi-success batch quality parameters for a single crystal growth control system based on temperature field data fitting is provided in one embodiment of the present invention. Figure 12 This is a schematic diagram of the annular temperature zone fitting of a single crystal growth control system based on temperature field data fitting, provided as an embodiment of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] This application provides a single crystal growth control method based on temperature field data fitting. The execution entity of this single crystal growth control method based on temperature field data fitting includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the single crystal growth control method based on temperature field data fitting can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0021] Reference Figure 1 The diagram shown is a flowchart illustrating a single-crystal growth control method based on temperature field data fitting according to an embodiment of the present invention. In this embodiment, the single-crystal growth control method based on temperature field data fitting includes: S1. Based on the multi-source temperature measurement data of the single crystal growth furnace, the continuous distribution state of the temperature field inside the furnace at the current moment is fitted to obtain the temperature field fitting model of the single crystal growth furnace. In this embodiment of the invention, the multi-source temperature measurement data includes: With a preset sampling period, the temperature values ​​of each temperature measurement point in the single crystal growth furnace are collected synchronously to obtain a set of discrete temperature measurement data; Each temperature data point is assigned a timestamp and the corresponding three-dimensional spatial coordinates of the temperature measurement point to obtain a set of temperature field sampling data. By removing gross errors caused by instantaneous sensor jumps in the temperature field sampling data set and eliminating high-frequency electronic noise, a discrete temperature field dataset is obtained.

[0022] The multi-source temperature measurement data based on the single crystal growth furnace is used to fit the continuous distribution of the temperature field inside the furnace at the current moment, resulting in a temperature field fitting model for the single crystal growth furnace, including: The discrete temperature field dataset is distributed to multiple annular temperature zones divided along the axial and radial directions of the furnace body, wherein each annular temperature zone contains temperature measurement points located in the same axial and radial segments; For any annular temperature zone, the temperature variation trend along the axial direction and the temperature variation trend along the radial direction are determined by using the temperature values ​​of the internal temperature measuring points. The axial and radial temperature variation trends of adjacent annular temperature zones at the interface are smoothly connected to generate a continuous temperature field fitting model for the single crystal growth furnace.

[0023] For any annular temperature zone, determining the axial and radial temperature variation trends using the temperature values ​​from internal temperature measuring points includes: The temperature measuring points contained within the annular temperature zone are arranged sequentially along the axial coordinate and along the radial distance, respectively. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged sequentially along the axial coordinate, the axial temperature change trend is determined to be either a heating trend or a cooling trend. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged in radial distance sequence, the radial temperature change trend is determined to be either a heating trend from the center outward or a cooling trend from the center outward.

[0024] The single crystal growth furnace contains five temperature measurement layers (Z=0mm, 150mm, 300mm, 450mm, 600mm) arranged axially and four temperature measurement rings (R=0mm, 50mm, 100mm, 150mm) arranged radially, totaling 125 temperature measurement points. One temperature measurement point is arranged in each layer along the central axis (R=0), while eight temperature measurement points are evenly distributed in each of the ring layers (R=50mm, 100mm, 150mm) at 45° circumferential intervals, forming 20 annular temperature zones. The system synchronously acquires multi-source data through 16 thermocouples, 2 infrared thermometers, a lifting and crucible encoder, and 3 power sensors. After real-time filtering by FPGA, sliding window 3σ anomaly removal, and Kalman filtering, a discrete temperature field dataset is obtained.

[0025] The method involves fitting the continuous distribution of the temperature field within the single-crystal growth furnace at the current moment using multi-source temperature measurement data. This results in a temperature field fitting model for the furnace. The process includes: distributing the discrete temperature field dataset to 20 annular temperature zones, where each annular temperature zone contains temperature measurement points located in the same axial and radial segments; determining the axial and radial temperature variation trends of any annular temperature zone using the temperature values ​​of its internal measurement points; providing prior initial values ​​and boundary constraints based on a finite element reference temperature field library; performing full-field fitting correction on the sensor residuals using a Gaussian kernel radial basis function; synthesizing the correction field with the prior field; and smoothly connecting the axial and radial temperature variation trends of adjacent annular temperature zones at their interface to generate a continuous temperature field fitting model for the single-crystal growth furnace. The computation time per frame is approximately 12ms, and the fitting update frequency is no less than 1Hz.

[0026] For any annular temperature zone, determining the axial and radial temperature change trends using the temperature values ​​of internal temperature measuring points includes: arranging the temperature measuring points within the annular temperature zone sequentially along the axial coordinate axis and sequentially along the radial distance; determining whether the axial temperature change trend is an increasing or decreasing trend based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged sequentially along the axial coordinate axis; and determining whether the radial temperature change trend is a heating trend from the center outwards or a cooling trend from the center outwards based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged sequentially along the radial distance.

[0027] The beneficial effects are as follows: by synchronously collecting temperatures at various temperature measurement points within the furnace at fixed intervals and binding them with timestamps and three-dimensional coordinates, and then eliminating gross errors and high-frequency noise, a truly reliable discrete temperature field dataset is obtained, ensuring the purity and accuracy of the basic temperature measurement data. By dividing the discrete temperature field data into annular temperature zones corresponding to the furnace's axial and radial directions, and organizing the temperature measurement points according to the axial and radial directions respectively, the temperature change trend is determined, accurately characterizing the temperature distribution pattern of each temperature zone. Furthermore, the temperature trends at the interfaces of adjacent annular temperature zones are smoothly connected, and a complete and continuous temperature field fitting model within the furnace is obtained. This model can realistically reproduce the continuous temperature field distribution state inside the single crystal growth furnace, providing accurate and reliable data model support for precise temperature control during the single crystal growth process.

[0028] S2. Under the same spatial coordinate system, the continuous temperature field fitting surface is registered with the ideal temperature field surface of the single crystal growth furnace, and the difference is calculated to obtain the temperature field deviation scalar field. In this embodiment of the invention, the ideal temperature field surface includes: Obtain temperature field morphology records of multiple successful crystal growth batches in history, and extract temperature field morphology data corresponding to the current growth stage from them; The fusion weight of the temperature field morphology data for the current batch is determined based on the quality parameters of the crystals obtained from each historical batch. The temperature field morphology data of each batch are weighted and fused according to the fusion weight to obtain the ideal temperature field surface.

[0029] The process of registering the continuous temperature field fitted surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculating the difference includes: Extract the solid-liquid interface transition feature points on the continuous temperature field fitting surface and the solid-liquid interface reference feature points on the ideal temperature field surface, respectively. Align the solid-liquid interface transition feature point with the solid-liquid interface reference feature point so that the two surfaces coincide in the same spatial coordinate system, thus obtaining the registered continuous temperature field fitting surface and the ideal temperature field surface.

[0030] The obtained temperature field deviation scalar field includes: The temperature values ​​at each coordinate point on the registered continuous temperature field fitting surface are compared with the temperature values ​​at the corresponding coordinate points on the ideal temperature field surface to generate a three-dimensional deviation data set. The temperature difference value of each point in the three-dimensional deviation data set is bound to the corresponding coordinate point to obtain the temperature field deviation scalar field, wherein each spatial location in the temperature field deviation scalar field stores the temperature difference scalar value at that location.

[0031] Historical archives of temperature field morphology from eight successful single-crystal growth batches (B01–B08) were retrieved and categorized according to growth process stages. Temperature field morphology data corresponding to the current growth stage were extracted from these records. The crystal quality parameter scores for these eight batches ranged from 75.5 to 95.0, with the axial position of the solid-liquid interface distributed between 303 mm and 356 mm, with an average of approximately 334.9 mm. The fusion weight of the current batch's temperature field morphology data was determined based on the dislocation density, axial resistivity inhomogeneity, radial resistivity inhomogeneity, diameter fluctuation standard deviation, and oxygen content uniformity deviation of the crystals obtained from each historical batch. Historical batches with better quality performance were assigned higher fusion weights (e.g., batch B08 with a quality parameter score of 95.0 received a weight of 0.950, and batch B01 with a quality parameter score of 75.5 received a weight of 0.755), ensuring a strict correlation between weight allocation and actual crystal quality. The temperature field morphology data from each batch were then weighted and fused according to the aforementioned fusion weights to obtain the ideal temperature field surface.

[0032] The step of registering the continuous temperature field fitting surface and the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculating the difference includes: extracting solid-liquid interface transition feature points on the continuous temperature field fitting surface and solid-liquid interface reference feature points on the ideal temperature field surface; using the three-dimensional spatial coordinates of the solid-liquid interface reference feature points as a reference, matching the solid-liquid interface transition feature points with the solid-liquid interface reference feature points so that the two surfaces coincide in position in the same spatial coordinate system, thereby obtaining the registered continuous temperature field fitting surface and the ideal temperature field surface.

[0033] The process of obtaining the temperature field deviation scalar field includes: calculating the difference between the temperature values ​​at each coordinate point on the registered continuous temperature field fitting surface and the temperature values ​​at the corresponding coordinate points on the ideal temperature field surface, thereby generating a three-dimensional deviation data set; and binding the temperature difference values ​​at each point in the three-dimensional deviation data set to the corresponding coordinate points to obtain the temperature field deviation scalar field, wherein each spatial location in the temperature field deviation scalar field stores the temperature difference scalar value at that location.

[0034] The beneficial effects include: by registering and subtracting the actual temperature field from the ideal temperature field, the temperature deviation between the two is accurately captured, clearly presenting the gap between the current temperature field and the ideal state, providing a clear direction for temperature control; by combining historical high-quality temperature field data and rationally allocating fusion weights, the characteristics of high-quality temperature fields are integrated into the construction of the ideal temperature field, ensuring the scientific validity and practicality of the ideal temperature field and effectively avoiding the limitations of single batch data. Simultaneously, through precise comparison and point-by-point verification, the reliability of the temperature field deviation data is ensured, clearly reflecting the temperature differences in each region, providing precise data support for temperature control during single crystal growth.

[0035] S3. Extract the deviation features of different regions in the temperature field deviation scalar field and combine them into a deviation feature vector; In this embodiment of the invention, extracting the deviation features of different regions in the temperature field deviation scalar field includes: Based on the relative distance between each coordinate point in the temperature field deviation scalar field and the position of the solid-liquid interface in the single crystal growth furnace, the temperature field deviation scalar field is divided into a central interface region, a transition region, and an edge region. The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are extracted as deviation features of different regions.

[0036] The formula for calculating the temperature difference distribution pattern is as follows: ; The standard deviation of the temperature difference within the central interface region. This represents the total number of coordinate points within the central interface area. For the first Temperature difference at each coordinate point This is the arithmetic mean of the temperature differences at all coordinate points within the central interface area.

[0037] The combination is a deviation feature vector, including: The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are sorted according to the degree of influence of each region on the crystal growth quality. The sorted deviation features are concatenated sequentially to obtain the deviation feature vector.

[0038] The location information of all coordinate points in the temperature field deviation scalar field is obtained, and the specific location of the solid-liquid interface in the single crystal growth furnace is determined. Each coordinate point in the temperature field deviation scalar field is located one by one, and the straight-line distance between each coordinate point and the solid-liquid interface is calculated. Three distinct regions are defined according to the distance: the region closest to the solid-liquid interface is set as the central interface region, the region at a moderate distance from the solid-liquid interface is set as the transition region, and the region farthest from the solid-liquid interface is set as the edge region, ensuring that each coordinate point can be accurately assigned to the corresponding region.

[0039] Following the sorted order of deviation characteristics, the temperature difference distribution pattern in the central interface area, the temperature difference gradient direction in the transition area, and the temperature difference amplitude in the edge area are integrated sequentially. First, the specific content and presentation form of the deviation characteristics in each region are clarified. Then, the deviation characteristics of different regions are gradually connected in the sorted order to ensure that the deviation characteristics of adjacent regions are connected naturally and without discontinuity. At the same time, the deviation characteristics of each region are bound to the corresponding coordinate position to clarify the spatial position corresponding to each deviation characteristic, forming a complete deviation characteristic system. This system can comprehensively cover the deviation situation of all regions, clearly present the core information such as temperature difference deviation and gradient change in different regions, provide clear deviation basis for subsequent temperature field adjustment, and ensure that the adjustment action can accurately correspond to the deviation problems in different regions.

[0040] Based on the single crystal growth process mechanism, the priority and influence weight of the central interface region, transition region, and edge region on crystal growth quality are determined in advance. The influence of temperature difference characteristics of each region on crystal growth quality, defect generation, and crystal regularity is used as the criterion. The three types of deviation characteristics, namely the temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region, are prioritized. The order is strictly arranged according to the established rule of influence from strong to weak. The influence law corresponding to the actual crystal growth conditions is followed throughout the process to ensure that the sorting result is consistent with the actual process and the hierarchical division is fixed and unique.

[0041] According to the order in which the degree of influence has been sorted, the deviation feature information corresponding to the central interface area, transition area and edge area is retrieved in sequence. The complete attributes of each deviation feature are kept intact and are not split or tampered with. They are connected and combined in an orderly manner according to the established sorting order, and the deviation features of the three types of regions are integrated into a unified whole to form a deviation feature vector with an orderly structure and complete features. This allows the deviation feature vector to fully carry all the temperature field deviation information of different regions that affect crystal growth, providing a standardized feature input carrier for subsequent temperature field correction and process control.

[0042] The beneficial effects are as follows: by dividing the scalar field of temperature field deviation into regions and combining the relative positions of each region with the solid-liquid interface, the central interface region, transition region, and edge region are precisely defined, ensuring the targeting and accuracy of deviation feature extraction and effectively avoiding omissions or misjudgments of deviation features. Specific deviation features for each region are extracted: the central interface region focuses on the temperature difference distribution pattern, the transition region emphasizes the gradient direction, and the edge region focuses on the amplitude range, achieving precise capture of deviation features that closely matches the actual temperature field deviation and avoids the one-sidedness of feature extraction. By sorting the regions according to their influence on crystal growth, the deviation features of each region are systematically spliced ​​to form a deviation feature vector, comprehensively integrating temperature field deviation information and clearly presenting the deviation differences in different regions. This provides accurate and specific deviation basis for subsequent temperature field adjustment and crystal growth process optimization, thereby improving crystal growth quality and ensuring the stability and consistency of crystal growth.

[0043] S4. Based on the deviation feature vector, perform closed-loop control on the single crystal growth furnace.

[0044] The generated deviation feature vectors are obtained, and the deviation information of the central interface area, transition area, and edge area contained in the vectors is decomposed one by one. The temperature difference deviation, gradient direction and amplitude range of each area are clarified, and the specific value and distribution pattern of the deviation in each area are confirmed, so as to provide a precise basis for subsequent regulation.

[0045] The manipulated variables are dynamically adjusted based on the Model Predictive Control (MPC) algorithm. These manipulated variables include the main heater power, the side heater zone power, the crystal pulling speed, the crucible lifting speed, and the crucible / crystal rotation speed ratio. The control cycle is 1 second, the prediction time domain is 30 seconds, and the control time domain covers the first 5 steps. The objective function integrates the weights of temperature deviation, control quantity change, and quality risk. The optimal control sequence is solved through rolling optimization, and the first step is executed. Temperature deviation is the core control objective, the control quantity change weight is used to limit the adjustment amplitude to avoid oscillations, and the quality risk weight is used to ensure growth quality.

[0046] Differentiated control strategies are adopted for different stages of single crystal growth: In the crystal pulling stage, small-step high-frequency temperature fine-tuning is the main approach, with low control gain, sensitive response, and low fault tolerance threshold to ensure dislocation-free growth of the narrow neck; In the shoulder formation stage, temperature and pulling speed are controlled in tandem, and the diameter tracks the set curve to avoid dislocation generation; In the constant diameter stage, a multi-variable collaborative closed-loop control is adopted, prioritizing steady-state accuracy, controlling interface temperature fluctuation within ±1℃, ensuring uniform radial temperature field, and controlling the standard deviation of diameter fluctuation within 0.3mm; In the finishing stage, gradient smoothing control is adopted, with gradual pulling and deceleration combined with a slow temperature rise to prevent thermal shock.

[0047] The system sets four-level control authority: L1 suggestion mode, L2 semi-automatic mode, L3 full-automatic mode and L4 manual mode. The default L1 mode is adopted in the initial operation stage, and is gradually upgraded to L3 mode after stable verification. At the same time, three levels of safety protection are set: when a first-level early warning occurs, the temperature field slightly deviates from the normal range, and the system automatically fine-tunes parameters and strengthens monitoring; when a second-level alarm occurs, the temperature field has a large abnormal deviation, the system increases the regulation intensity and pushes for manual confirmation; when the third-level protection is triggered, the system faces serious abnormalities or high risk of neck breaking, and automatically switches to the safety mode, suspends growth and waits for manual takeover. When a sensor fails, the system automatically degrades to redundant sensor and traditional PID control; an operator can switch to manual control with one key at any time to achieve safety bottoming out.

[0048] After regulation is completed, the temperature data and deviation data of the current temperature field are collected in real time, the original deviation information in the deviation feature vector is compared, and whether the regulation effect meets expectations is judged. If the expectation is not met, the regulation intensity is adjusted according to the actual deviation, and the heating parameters are optimized again until the temperature field deviation is reduced to a reasonable range. The change of the temperature field is continuously monitored, the slight change of the temperature field deviation is captured in real time, the deviation feature vector is updated synchronously, and the regulation strategy is adjusted according to the updated vector to ensure the stability of the temperature field during the single crystal growth process. Through on-site industrial tests, the automatic control coverage rate in the constant diameter stage reaches more than 92%, the continuous operation availability of the system is 99.92%, the steady-state fluctuation of the interface temperature is ≤±1°C, and the standard deviation of diameter fluctuation is reduced from 0.45mm to 0.26mm.

[0049] The beneficial effect is that by accurately dividing temperature field regions, extracting corresponding deviation features and combining them into a feature vector, combined with closed-loop regulation, the temperature field deviation can be accurately captured and the regulation direction can be clarified, effectively avoiding the influence of temperature field fluctuation on crystal growth; at the same time, relying on orderly feature extraction and integration, the temperature field adjustment is more targeted and blind regulation is avoided, which not only ensures the stability and accuracy of the temperature field, but also significantly improves the quality of crystal growth, reduces crystal defects, and makes the temperature field environment fully adapt to the requirements of crystal growth.

[0050] Influence of different growth stages on the temperature field during single crystal growth: single crystal growth is a dynamic process, and the requirements for the temperature field in different growth stages are significantly different, which is specifically reflected in the different requirements for axial temperature gradient and radial temperature gradient. First, in the initial growth stage, a larger radial temperature gradient is required to complete the lateral growth of the crystal, which helps to reduce the occurrence of defects such as voids to a certain extent, while a smaller axial temperature gradient is required. Second, in the stable growth stage, the temperature field requirement changes: a larger axial temperature gradient is required to ensure a faster crystal growth rate, while a smaller radial temperature gradient is required to reduce the introduction of internal stress and dislocations in the crystal. Finally, in the later growth stage, the longitudinal growth of the crystal needs to be promoted, so a large radial temperature gradient is not required.

[0051] Furthermore, both axial and radial temperature gradients must not exceed their respective critical values, otherwise crystal cracking is highly likely. In addition, as the crystal grows, the melt level continuously decreases, and the position and morphology of the solid-liquid interface also continuously change, further requiring the temperature field control system to have real-time response and dynamic adjustment capabilities.

[0052] The division of the annular temperature zone is based on the following: From the perspective of the physical mechanism and engineering practice of single crystal growth, the axial and radial directions are the two most crucial dimensions for describing the temperature field within the furnace, primarily for the following reasons: First, the axial direction is along the crystal growth direction, and the axial temperature gradient determines the crystal growth rate and driving force. A larger axial gradient facilitates the dissipation of latent heat of crystallization and maintains the necessary supercooling; the solid-liquid interface is essentially a specific axial isothermal surface. Second, the radial temperature gradient determines the quality of the crystal cross-section. The radial direction is perpendicular to the growth direction, and the uniformity of the radial temperature gradient is crucial to the quality of the crystal cross-section. If the radial temperature difference is too large, it will lead to uneven radial resistivity, thermal stress, and even cracks. An ideal temperature field should make the radial temperature as uniform as possible. Third, the shape of the solid-liquid interface—convex, flat, or concave—has a significant impact on crystal quality, and this shape is determined by the combined effect of the axial and radial temperature gradients. Finally, for engineering feasibility, single crystal growth furnaces typically employ annular heaters surrounding the crucible, which naturally generate an axisymmetric temperature field. Axial partitioning can correspond to the control areas of different heaters, while radial partitioning can describe the temperature decrease pattern from the center to the edge. Therefore, the axial and radial annular temperature zone division not only conforms to physical laws but also highly matches engineering practice.

[0053] A total of 125 temperature measurement points are used. The number of temperature measurement points affects the entire process of data acquisition, model building, and deviation analysis: First, the number of temperature measurement points directly determines the ability of discrete sampling data to represent the continuous temperature field distribution. The more temperature measurement points and the more reasonable their spatial distribution, the more accurately the fitted continuous temperature field model can reproduce the actual temperature distribution inside the furnace. If there are too few temperature measurement points, the gradual temperature change law in the axial and radial directions of the furnace body cannot be fully represented, resulting in problems such as "distorted temperature field representation and incomplete capture of local temperature change trends." Second, the number of temperature measurement points affects the accuracy of trend determination in the annular temperature zone. After allocating temperature measurement points to each annular temperature zone, this scheme needs to determine the axial and radial temperature change trends based on the "monotonically changing direction of the temperature values ​​at the measurement points." If the number of temperature measurement points in a certain annular temperature zone is insufficient, the direction of temperature change cannot be reliably determined, which may lead to incorrect trend determination. Third, the temperature field deviation scalar field is obtained by point-by-point subtraction of the two registered surfaces. The fineness of the surfaces depends on the density of the initial temperature measurement points. The more temperature measurement points there are, the higher the resolution of the generated deviation scalar field, and the more accurately it can locate subtle areas of temperature field deviation. Furthermore, the number of temperature measurement points affects the effectiveness of the deviation feature vector, which is a combination of deviation features extracted from the central interface region, transition region, and edge region of the temperature field deviation scalar field. If the deviation scalar field is coarse due to insufficient temperature measurement points, the extraction of deviation features in each region will lose accuracy, and the deviation feature vector will not accurately reflect the deviation state of the temperature field. Finally, it is directly related to the closed-loop control effect, which is ultimately based on the deviation feature vector. If the number of temperature measurement points is insufficient, leading to a distorted deviation feature vector, then subsequent control commands will be unable to accurately correct the temperature field deviation within the furnace.

[0054] In an industrial field test and verification embodiment, the single crystal growth control system based on temperature field data fitting was deployed in an 8-inch Czochralski silicon single crystal furnace of a semiconductor materials company via embedded modification. An industrial-grade control computer, a 16-channel thermocouple acquisition module, an FPGA preprocessing unit, two infrared thermometer interfaces, a UPS power supply, and a 15-inch industrial touchscreen were added to the existing control cabinet. The system interacts with the existing PLC, heating power supply, and servo mechanism via Modbus / Profinet. The modification construction period was one week, supporting seamless manual / automatic switching. A total of 25 furnaces were tested, including 5 baseline furnaces, 5 debugging furnaces, and 15 formal testing furnaces. The crystal specification was 8-inch N-type silicon single crystals with a target length of 1500mm.

[0055] Test results show that the average error of the interface center temperature during 200 hours of continuous sampling in the constant diameter stage is 1.5℃, with a 95% confidence interval error of ±2.1℃, and the steady-state fitting accuracy meets the design specifications; the average yield of single crystals increased from 70.8% to 79.9%, an increase of 9.1 percentage points; and the dislocation density increased from 335 dislocations / cm². 2Reduced to 225 / cm 2 The system achieved the following reductions: axial resistivity non-uniformity decreased from 8.1% to 5.9%, and radial resistivity non-uniformity decreased from 5.4% to 4.0%; diameter fluctuation standard deviation decreased from 0.45mm to 0.26mm; average single-furnace growth cycle shortened from 69.2 hours to 65.3 hours, a reduction of 5.6%; unit kilogram power consumption decreased from 261kWh / kg to 244kWh / kg, a reduction of 6.5%; and first-time crystal pulling success rate increased from 62% to 87%. The system has accumulated over 2100 hours of continuous operation without any downtime or control failures. Automatic control coverage reached 92% in the equal-diameter stage, and system availability reached 99.92%. No equipment or quality accidents occurred due to this system, meeting the requirements for industrial mass production applications.

[0056] like Figure 2 The diagram shown is a functional block diagram of a single crystal growth control system based on temperature field data fitting provided in an embodiment of the present invention.

[0057] The single crystal growth control system 100 based on temperature field data fitting described in this invention can be installed in an electronic device. Depending on the functions implemented, the single crystal growth control system 100 based on temperature field data fitting may include a temperature field fitting module 101, a deviation scalar field construction module 102, a deviation feature construction module 103, and a closed-loop control module 104. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0058] In this embodiment, the functions of each module / unit are as follows: The temperature field fitting module 101 is used to fit the continuous distribution state of the temperature field inside the furnace at the current moment based on the multi-source temperature measurement data of the single crystal growth furnace, so as to obtain the temperature field fitting model of the single crystal growth furnace. The deviation scalar field construction module 102 is used to register the continuous temperature field fitting surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculate the difference to obtain the temperature field deviation scalar field. The deviation feature construction module 103 is used to extract deviation features in different regions of the temperature field deviation scalar field and combine them into a deviation feature vector. The closed-loop control module 104 is used to perform closed-loop control on the single crystal growth furnace according to the deviation feature vector.

[0059] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0060] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0061] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0062] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0063] Figure 3 This is a three-dimensional spatial distribution diagram of the temperature measuring points. The diagram shows the complete spatial layout of a total of 125 temperature measuring points inside the single crystal growth furnace. The horizontal axis represents the X and Y radial coordinates, and the vertical axis represents the Z axial coordinate. The furnace body is divided into 5 temperature measuring layers along the axial direction (Z=0mm, 150mm, 300mm, 450mm, 600mm) and 4 temperature measuring rings along the radial direction (R=0mm center, 50mm, 100mm, 150mm). Only one temperature measuring point is arranged on each layer along the central axis of R=0mm (e.g., T001, T026, T051, T076, T101). The temperature measuring points at R=50mm, 100mm, and 150mm are... Eight temperature measurement points are evenly arranged on each annular layer (circumferentially spaced at 45° intervals; for example, when R=50mm, the X coordinates are 50.0, 35.36, 0, -35.36, -50.0, -35.36, 0, 35.36, and the Y coordinates are 0, 35.36, 50.0, 35.36, 0, -35.36, -50.0, -35.36), with 25 points per layer and a total of 125 points across 5 layers. This axisymmetric annular point arrangement structure completely covers the entire space from the furnace center to the furnace wall and from the crystal growth initiation section to the upper melting zone, ensuring that temperature field distortion will not occur due to local temperature measurement omissions during subsequent zonal fitting.

[0064] Figure 4This is a furnace temperature field cloud map of the stable growth stage of single crystal with constant diameter. The map visually presents the continuous temperature field distribution inside the furnace during the constant diameter growth stage. The horizontal axis is the axial Z-coordinate (0–600 mm), and the vertical axis shows the temperature value on the left and the axial height on the right. The solid-liquid interface positions marked in the map correspond to the actual distribution range of historical successful batches (303 mm–356 mm, with an average of approximately 334.9 mm). As can be seen from the temperature range of the cloud map, the highest temperature inside the furnace is located at the top center (Z=0, R=0) at 1550.0°C, and the lowest temperature is located at the bottom edge (Z=600, R=150) at 554.67°C. The maximum temperature difference in the entire furnace reaches 995.33°C. The color gradient of the cloud map clearly distinguishes the temperature difference range between the melt region and the crystal region above and below the solid-liquid interface, providing a visual benchmark for dividing the three major deviation analysis areas of the central interface region, transition region, and edge region, as well as extracting the solid-liquid interface transition feature points.

[0065] Figure 5 This chart compares the axial temperature curves for each stage of growth. It compares the temperature variations along the furnace axis (Z=0–600mm) for four growth stages: seeding, shoulder formation, constant diameter growth, and finishing. Specifically, the temperature in the constant diameter stage (core growth stage) decreases from 1314.80°C at Z=0 to 665.82°C at Z=600; in the seeding stage, it decreases from 1209.61°C to 612.56°C; and in the shoulder formation stage, it decreases from 1262.20°C to 639.19°C. °C, decreasing from 1235.91°C to 625.87°C in the final stage; the figure shows that the axial temperature gradient is relatively gentle in the early stage of growth (crystal introduction / shoulder formation), while it increases significantly after entering the stable growth stage to meet the requirements of latent heat dissipation and supercooling maintenance for the faster growth rate. At the same time, the curves of each stage show that the axial temperature decreases monotonically with the increase of Z, which confirms the determination of the axial cooling trend. The absolute value of the gradient needs to be controlled within the critical value to avoid crystal cracking.

[0066] Figure 6The figure shows a comparison of radial temperature curves for each stage of growth. It compares the temperature distribution curves along the radial direction (R=0–150mm) for four growth stages. In the constant diameter growth stage, the temperature at the center (R=0) is 1140.26°C, decreasing to 1097.11°C at R=50mm, 977.21°C at R=100mm, and 805.77°C at R=150mm. The crystal introduction, shoulder formation, and tailing stages also show a monotonically decreasing trend from the center to the periphery (R=0 is 1050.48°C and R=150 is 742.32°C in the crystal introduction stage). The curve shape of this radial cooling trend indicates that a radial temperature gradient always exists during the growth process. A relatively large radial gradient is required in the early stage of growth to promote the lateral growth of the crystal and reduce void defects. In the stable growth stage, the radial gradient needs to be controlled within a reasonable range to prevent uneven radial resistivity, thermal stress accumulation, or even crack formation due to excessive radial temperature difference.

[0067] Figure 7 This is a continuous temperature field fitting surface diagram, which shows the continuous temperature field fitting model of the single crystal growth furnace generated by partitioning and smoothing 125 discrete temperature measurement points. The horizontal and vertical axes of the surface correspond to the radial and axial coordinates of the furnace body, and the vertical axis represents the temperature value. The surface is constructed based on discrete data of 20 annular temperature zones (5 axial segments × 4 radial segments). Each temperature zone contains 1 (center R=0) or 8 (R=50 / 100 / 150) temperature measurement points. By determining the axial heating / cooling trend and the radial heating / cooling trend from the center outward in each temperature zone, and by smoothing the temperature change trend at the interface of adjacent temperature zones, abrupt changes in the trend are eliminated, and the continuous temperature distribution throughout the furnace is finally restored. Taking the constant diameter stage as an example, the surface smoothly transitions from 1550.0°C at the top center to 554.67°C at the bottom edge, fully characterizing the gradual temperature change law in the axial and radial directions, breaking through the limitation of traditional single-point discrete temperature measurement that cannot depict the continuous morphology of the entire region.

[0068] Figure 8This is a radial temperature gradient distribution diagram, showing the radial temperature gradient distribution during the constant diameter growth stage. The horizontal axis represents the radial distance, and the vertical axis represents the radial temperature gradient value (unit: °C / mm). Based on the temperature measurement data within each annular temperature zone, the radial gradient is calculated as follows: at the Z=0mm layer, the radial gradient is -1.17°C / mm in the R=0 to 50mm segment, -3.26°C / mm in the R=50 to 100mm segment, and -4.66°C / mm in the R=100 to 150mm segment. The corresponding value is [not specified in the original text]. The gradients are -0.59°C / mm, -1.65°C / mm, and -2.36°C / mm, respectively. It can be seen that the absolute value of the radial gradient gradually decreases from the top to the bottom of the furnace, and the absolute value of the gradient increases from the center to the outside within the same axial layer. This distribution indicates that the temperature field exhibits a "cooling from the center to the outside" trend in the radial direction, and the temperature of the outer ring decreases faster. If the radial gradient exceeds the critical value, it will directly cause the degradation of the crystal cross-section quality. Therefore, this figure provides a quantitative basis for evaluating the uniformity of the radial temperature field and adjusting the power distribution of the heater.

[0069] Figure 9 The figure shows the axial temperature gradient distribution during the constant diameter growth stage. The horizontal axis represents the Z-coordinate of the furnace body, and the vertical axis represents the axial temperature gradient value (unit: °C / mm). Calculated using temperature measurement data from the central axis (R=0), the axial gradient is -1.65°C / mm in the Z=0 to 150mm segment, -1.10°C / mm in the Z=150 to 300mm segment, -1.41°C / mm in the Z=300 to 450mm segment, and -0.95°C / mm in the Z=450 to 600mm segment. The gradients in each segment exhibit a non-uniform distribution, reflecting the differentiated requirements for axial driving force and latent heat dissipation at different stages of crystal growth. This figure visually presents the details of the gradient changes near the solid-liquid interface (approximately in the Z=300–356mm range), providing a crucial reference for determining whether the axial temperature field maintains the optimal growth range and avoiding imbalance in the solid-liquid interface temperature gradient.

[0070] Figure 10This diagram illustrates the solid-liquid interface feature points and temperature field morphology. It shows the comparison and registration process between the continuous temperature field fitted surface and the ideal temperature field surface at the solid-liquid interface. The diagram marks the solid-liquid interface transition feature points on the actual surface and the solid-liquid interface reference feature points on the ideal surface. Based on data from eight historical successful batches, the axial position of the actual solid-liquid interface is distributed between 303mm (batch B01) and 356mm (batch B08), with an average of approximately 334.9mm. Due to process differences, the interface position fluctuates by up to 53mm between different batches. The registration operation uses these feature points as a reference to match and force the solid-liquid interface transition feature points on the actual surface with the reference feature points on the ideal surface. This ensures that the two surfaces coincide in the same spatial coordinate system, guaranteeing that the temperature field deviation scalar field obtained by subsequent point-by-point subtraction has an accurate spatial correspondence and avoiding distortion in deviation calculations due to surface misalignment.

[0071] Figure 11 The box plot shows the distribution of crystal quality parameters for eight historical successful batches (B01–B08), used to determine the fusion weight of temperature field morphology data for each batch. The quality parameter scores for the eight batches range from 75.5 to 95.0, with a mean of 82.75, a standard deviation of 6.62, a median of 80.35, and quartiles Q1=79.38 and Q3=83.62. Batch B08 has the best quality (95.0 points) and the highest fusion weight (0.950), while batch B01 has relatively lower quality (75.5 points) and the lowest weight (0.755). The fusion weight is perfectly positively correlated with the quality parameter score (r≈1.0). The box plot visually demonstrates the dispersion and quality levels of the quality parameters. By assigning differentiated weights to each batch based on this plot and then weighting the fusion, individual biases in the temperature field morphology of a single batch can be effectively eliminated, constructing a scientific and robust ideal temperature field surface.

[0072] Figure 12This diagram illustrates the fitting of 20 annular temperature zones within the single crystal growth furnace, divided along the axial direction (Z=0, 150, 300, 450, 600 mm) and radial direction (R=0, 50, 100, 150 mm), along with their fitting logic. Each temperature zone contains temperature measurement points within the same axial and radial segments. For example, the Z=0, R=50 mm zone contains 8 points (T002–T009, uniformly distributed at 45° circumference), while the Z=150, R=0 mm zone contains only 61 center points (T02). The Z=300, R=100mm temperature zone contains 8 points, T060–T067, and so on. For each temperature zone, the axial temperature change trend (heating / cooling direction) and radial temperature change trend (heating / cooling direction from the center outward) are determined based on the internal temperature measurement points. Then, the temperature trend is smoothly connected at the interface of adjacent temperature zones to eliminate abrupt changes in the trend. Finally, a continuous temperature field fitting model covering the entire furnace is generated. This 5×4 annular partitioning method is highly consistent with the axisymmetric physical structure of the furnace body and the engineering control area of ​​the annular heater.

[0073] 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 it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for controlling single-crystal growth based on temperature field data fitting, characterized in that, The method includes: S1. Based on the multi-source temperature measurement data of the single crystal growth furnace, the continuous distribution state of the temperature field inside the furnace at the current moment is fitted to obtain the temperature field fitting model of the single crystal growth furnace. S2. Under the same spatial coordinate system, the continuous temperature field fitting surface is registered with the ideal temperature field surface of the single crystal growth furnace, and the difference is calculated to obtain the temperature field deviation scalar field. S3. Extract the deviation features of different regions in the temperature field deviation scalar field and combine them into a deviation feature vector; S4. Based on the deviation feature vector, perform closed-loop control on the single crystal growth furnace.

2. The single crystal growth control method based on temperature field data fitting as described in claim 1, characterized in that, The multi-source temperature measurement data is obtained through the following methods: With a preset sampling period, the temperature values ​​of each temperature measurement point in the single crystal growth furnace are collected synchronously to obtain a set of discrete temperature measurement data; Each temperature data point is assigned a timestamp and the corresponding three-dimensional spatial coordinates of the temperature measurement point to obtain a set of temperature field sampling data. By removing gross errors caused by instantaneous sensor jumps in the temperature field sampling data set and eliminating high-frequency electronic noise, a discrete temperature field dataset is obtained.

3. The single crystal growth control method based on temperature field data fitting as described in claim 2, characterized in that, The multi-source temperature measurement data based on the single crystal growth furnace is used to fit the continuous distribution of the temperature field inside the furnace at the current moment, resulting in a temperature field fitting model for the single crystal growth furnace, including: The discrete temperature field dataset is distributed to multiple annular temperature zones divided along the axial and radial directions of the furnace body, wherein each annular temperature zone contains temperature measurement points located in the same axial and radial segments; For any annular temperature zone, the temperature variation trend along the axial direction and the temperature variation trend along the radial direction are determined by using the temperature values ​​of the internal temperature measuring points. The axial and radial temperature variation trends of adjacent annular temperature zones at the interface are smoothly connected to generate a continuous temperature field fitting model for the single crystal growth furnace.

4. The single crystal growth control method based on temperature field data fitting as described in claim 3, characterized in that, For any annular temperature zone, determining the axial and radial temperature variation trends using the temperature values ​​from internal temperature measuring points includes: The temperature measuring points contained within the annular temperature zone are arranged sequentially along the axial coordinate and along the radial distance, respectively. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged sequentially along the axial coordinate, the axial temperature change trend is determined to be either a heating trend or a cooling trend. Based on the monotonic change direction of the temperature values ​​of the temperature measuring points arranged in radial distance sequence, the radial temperature change trend is determined to be either a heating trend from the center outward or a cooling trend from the center outward.

5. The single crystal growth control method based on temperature field data fitting as described in claim 1, characterized in that, The ideal temperature field surface includes: Obtain temperature field morphology records of multiple successful crystal growth batches in history, and extract temperature field morphology data corresponding to the current growth stage from them; The fusion weight of the temperature field morphology data for the current batch is determined based on the quality parameters of the crystals obtained from each historical batch. The temperature field morphology data of each batch are weighted and fused according to the fusion weight to obtain the ideal temperature field surface.

6. The single crystal growth control method based on temperature field data fitting as described in claim 5, characterized in that, The process of registering the continuous temperature field fitted surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculating the difference includes: Extract the solid-liquid interface transition feature points on the continuous temperature field fitting surface and the solid-liquid interface reference feature points on the ideal temperature field surface, respectively. Align the solid-liquid interface transition feature point with the solid-liquid interface reference feature point so that the two surfaces coincide in the same spatial coordinate system, thus obtaining the registered continuous temperature field fitting surface and the ideal temperature field surface.

7. The single crystal growth control method based on temperature field data fitting as described in claim 6, characterized in that, The obtained temperature field deviation scalar field includes: The temperature values ​​at each coordinate point on the registered continuous temperature field fitting surface are compared with the temperature values ​​at the corresponding coordinate points on the ideal temperature field surface to generate a three-dimensional deviation data set. The temperature difference value of each point in the three-dimensional deviation data set is bound to the corresponding coordinate point to obtain the temperature field deviation scalar field, wherein each spatial location in the temperature field deviation scalar field stores the temperature difference scalar value at that location.

8. The single crystal growth control method based on temperature field data fitting as described in claim 1, characterized in that, The extraction of deviation features in different regions of the temperature field deviation scalar field includes: Based on the relative distance between each coordinate point in the temperature field deviation scalar field and the position of the solid-liquid interface in the single crystal growth furnace, the temperature field deviation scalar field is divided into a central interface region, a transition region, and an edge region. The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are extracted as deviation features of different regions.

9. The single crystal growth control method based on temperature field data fitting as described in claim 8, characterized in that, The combination is a deviation feature vector, including: The temperature difference distribution pattern of the central interface region, the temperature difference gradient direction of the transition region, and the temperature difference amplitude range of the edge region are sorted according to the degree of influence of each region on the crystal growth quality. The sorted deviation features are concatenated sequentially to obtain the deviation feature vector.

10. A single-crystal growth control system based on temperature field data fitting, characterized in that, The system for implementing the single crystal growth control method based on temperature field data fitting as described in claim 1 includes: The temperature field fitting module is used to fit the continuous distribution of the temperature field inside the furnace at the current moment based on the multi-source temperature measurement data of the single crystal growth furnace, so as to obtain the temperature field fitting model of the single crystal growth furnace. The deviation scalar field construction module is used to register the continuous temperature field fitting surface with the ideal temperature field surface of the single crystal growth furnace in the same spatial coordinate system and then calculate the difference to obtain the temperature field deviation scalar field. The deviation feature construction module is used to extract deviation features from different regions in the temperature field deviation scalar field and combine them into a deviation feature vector. The closed-loop control module is used to perform closed-loop control on the single crystal growth furnace based on the deviation feature vector.

Images