Intelligent doping device for single crystal furnace in different thermal fields and calculation method

Abstract

The invention relates to the technical field of monocrystalline silicon processing, in particular to an intelligent doping device for a single crystal furnace in different thermal fields and a calculation method.The intelligent doping device comprises a mounting base, a telescopic mechanism is arranged on the mounting base and comprises a fixed cylinder and a telescopic rod, and a material scattering mechanism is arranged on the telescopic rod and comprises a material scattering cylinder and a feeding screw; the scattering barrel is in an inclined state, a feeding port and a discharging port are formed in the outer walls of the upper end and the lower end of the scattering barrel respectively, a guiding pipe which can be connected with the feeding port and injects doping materials into the scattering barrel is arranged beside the mounting base, and a driving mechanism which drives the feeding screw to rotate through translation of the telescopic rod is arranged between the fixing barrel and the scattering barrel. The inclined scattering barrel is matched with the spiral blade to control the falling speed of the doped materials, meanwhile, radial uniform distribution of the doped materials is achieved through reciprocating translation of the telescopic rod, and local accumulation of traditional feeding is avoided.

Description

Technical Field

[0001] This invention relates to the field of single-crystal silicon processing technology, specifically to an intelligent doping device and calculation method for single-crystal furnaces under different thermal fields. Background Technology

[0002] Czochralski (CZ) single-crystal growing technology, as the core process for producing high-purity single-crystal silicon, plays a decisive role in controlling the electrical properties of silicon wafers through its doping process. By precisely adding dopants such as boron, phosphorus, antimony, and gallium, the resistivity of silicon wafers can be effectively adjusted, directly affecting the conversion efficiency and reliability of subsequent solar cells. However, current mainstream doping processes still heavily rely on the experience of operators, revealing numerous technical bottlenecks in complex and ever-changing production environments. Traditional doping processes also have the following drawbacks: Firstly, traditional doping processes mainly rely on human experience to set the amount of dopant added, but they lack effective adaptability to the dynamic changes of the hot field in the single crystal furnace. Furthermore, traditional empirical formulas can only be simplified based on fixed operating conditions and cannot accurately describe the migration law of dopant under different hot field conditions, making it difficult to guarantee doping accuracy. Secondly, the distribution of dopants in the silicon melt is highly susceptible to thermal disturbances. Fluctuations in temperature gradients and centrifugal forces generated by melt rotation can lead to radial and axial concentration deviations in dopants. This non-uniformity causes resistivity fluctuations in different regions of the single-crystal silicon rod, severely reducing silicon wafer yield and increasing production costs. Summary of the Invention

[0003] Therefore, it is necessary to provide an intelligent doping device and calculation method for single crystal furnaces under different thermal fields to address the existing technical problems.

[0004] To address the problems of existing technologies, the present invention adopts the following technical solution: an intelligent doping device for a single crystal furnace under different thermal fields, comprising a mounting base located beside a crucible, a telescopic mechanism on the mounting base, the telescopic mechanism comprising a fixed cylinder horizontally connected to the mounting base and a telescopic rod sliding radially within the fixed cylinder along the crucible, a material dispensing mechanism on the telescopic rod, the material dispensing mechanism comprising a dispensing cylinder connected to the telescopic rod and a feeding screw rotating within the dispensing cylinder, the dispensing cylinder being inclined, and having an inlet and an outlet respectively on its upper and lower outer walls, a guide pipe on the side of the mounting base capable of connecting to the inlet and injecting dopant into the dispensing cylinder, and a drive mechanism between the fixed cylinder and the dispensing cylinder for driving the feeding screw to rotate by the translation of the telescopic rod.

[0005] To demonstrate how the telescopic rod extends and retracts, a strip-shaped through groove is provided on the outer wall of the fixed cylinder. The strip-shaped through groove is parallel to the telescopic rod. A support rod is fixed on the telescopic rod, extending vertically downward through the strip-shaped through groove. The telescopic mechanism also includes a cylinder that is horizontally fixed to the mounting base, and the output end of the cylinder is fixed to the support rod.

[0006] To demonstrate the specific structure of the feeding screw and how it rotates within the spreading cylinder, both ends of the spreading cylinder are open, and each end of the opening is fixed with a circular end cap. Each circular end cap is fitted with a No. 1 ball bearing. The feeding screw includes a rotating rod and a spiral blade. The rotating rod is coaxially disposed within the spreading cylinder, and both ends of the rotating rod are connected to two No. 1 ball bearings respectively. The spiral blade is fixedly connected to the outer wall of the rotating rod near the discharge port, and the spiral blade extends spirally along the axial direction of the rotating rod.

[0007] To demonstrate the specific structure of the drive mechanism, a first strip-shaped bracket extending axially along the telescopic rod is fixedly connected to the outer wall of the upper end of the spreading cylinder. The drive mechanism includes a drive shaft, a gear, a rack, and a transmission component. The drive shaft is vertically rotatably connected to the outer end of the first strip-shaped bracket. The gear is coaxially fixed to the drive shaft. The rack is fixed to the outer wall of the fixed cylinder and is parallel to the telescopic rod. The gear meshes with the rack. The transmission component connects the drive shaft and the rotating rod.

[0008] To demonstrate the specific structure of the transmission component, a second strip support is provided above the first strip support and is fixedly connected to the outer wall of the spreading cylinder. The second strip support extends along the axial direction of the spreading cylinder. The transmission component includes a first transmission shaft, a second transmission shaft, a first bevel gear set, and a second bevel gear set. The first transmission shaft is rotatably connected to the first strip support and is parallel to the first strip support. The second transmission shaft is rotatably connected to the second strip support and is perpendicular to the spreading cylinder. The first bevel gear set drives the first transmission shaft and the drive shaft. The second bevel gear set drives the second transmission shaft and the rotating rod. A universal joint is provided between the first and second transmission shafts to drive them together.

[0009] To ensure that the feeding screw rotates only when the telescopic rod is extended, a No. 2 ball bearing is embedded in the center of the gear, and the No. 2 ball bearing is coaxially sleeved on the drive shaft. A ring of ratchet teeth is fixedly provided on the top of the gear. A circular frame located above the gear is coaxially fixed to the drive shaft. Several elastic pawls are fixedly provided on the circular frame, evenly distributed along its circumference. Each elastic pawl engages with the ratchet teeth in one direction only, driving the drive shaft to rotate only when the telescopic rod is extended.

[0010] To reduce sliding friction during translation of the telescopic rod, linear bearings are coaxially fixedly connected to both ends of the fixed cylinder, and the telescopic rod passes through the two linear bearings coaxially.

[0011] To improve the stability of the telescopic rod during translation, a guide rod parallel to the telescopic rod is provided above it, and the guide rod is fixed to the telescopic rod through a connecting block. A guide sleeve for the guide rod to pass through is fixed on the outer wall of the fixed cylinder.

[0012] To demonstrate how the spreading cylinder is connected to the telescopic rod, a connecting sleeve is formed on the outer wall of the spreading cylinder, and one end of the telescopic rod is coaxially fixed to the connecting sleeve.

[0013] A smart doping calculation method for a Czochralski single crystal furnace under different thermal fields, the method comprising the following steps: I. Acquisition of Thermal Field Parameters and Process Data During crystal pulling in a Czochralski single crystal furnace, thermal field parameters are collected in real time using sensors such as thermocouples, infrared thermometers, and gas flow meters built into the furnace. At the same time, process parameters of the historical crystal pulling process are recorded, such as seed crystal contact time, temperature stabilization time, dopant addition time, corresponding target resistivity, and actual doping concentration. II. Multiphysics Coupling Simulation Based on the collected actual structural parameters of the single crystal furnace, a three-dimensional geometric model was established. Subsequently, a multiphysics coupled model including temperature field, flow field, and dopant diffusion field was constructed to simulate the diffusion behavior of dopants in the melt under different thermal conditions. The governing equations required for multiphysics coupling simulation are as follows: For the temperature field, solve the energy conservation equation ∇·(k∇T) + q = ρc_p∂T / ∂t, where k is the thermal conductivity, P is the density, c_p is the specific heat capacity at constant pressure, q is the internal heat source (heater power), T is the measured temperature, and t is the time period; The flow field is described by the Navier-Stokes equations: ∂u / ∂t + u·∇u = -∇p / ρ + ν∇²u+ gβΔT, where u is the flow velocity, p is the pressure, ν is the kinematic viscosity, g is the gravitational acceleration, β is the coefficient of thermal expansion, and ΔT is the temperature difference. The doping diffusion field is coupled with Fick's law ∂C / ∂t + u·∇C = D∇²C, where C is the dopant concentration and D is the diffusion coefficient; After the multiphysics coupling simulation is completed, the coupling solution is performed by the host computer, and the axial and radial concentration distribution of dopants in the melt under different thermal field parameters is output. III. Intelligent Modeling of Doping Behavior The collected thermal field parameters and their corresponding doping concentration distributions and target resistivity are normalized to eliminate the influence of dimensions. A machine learning algorithm is used to establish the mapping relationship between "thermal field parameters → doping concentration distribution → target resistivity". Finally, the model is trained using historical process data, and the hyperparameters are optimized through cross-validation to ensure that the model has good generalization ability under different thermal field conditions. IV. Intelligent Doping Optimization Calculation Input the thermal field parameters collected during the current crystal pulling process into the multiphysics coupling model to quickly predict the doping diffusion characteristics under the current thermal field, such as effective diffusion time and melt convection intensity. Based on the target resistivity of the silicon wafer, and combined with the material properties of the dopant such as solid solubility and segregation coefficient, the total amount of dopant theoretically required is calculated. The intelligent modeling module outputs the optimal dopant addition amount and timing based on the current thermal field parameters and target resistivity. At the same time, a reinforcement learning mechanism is introduced, using the deviation between the actual doping concentration and the target value as the reward function to continuously optimize the model parameters. V. Feedback Correction and Closed-Loop Control The doping concentration of the current crystal rod is obtained by online spectrometer or periodic destructive sampling. The deviation from the target value is calculated and the deviation data is fed back to the intelligent modeling module to update the training set of the machine learning model and improve the prediction accuracy. If the deviation exceeds the threshold, the thermal field parameters or dopant addition strategy are adjusted to achieve closed-loop control and ensure doping accuracy. VI. Specific Doping Process S1, After the actual total amount of doping is calculated through the above process, a fixed amount of doping material will fall into the spreading cylinder through the feed pipe and the feed inlet; S2, the telescopic rod drives the feeding cylinder to extend towards the crucible. Under the transmission action of the drive mechanism, the drive shaft will drive the rotating rod to rotate, and the rotating rod will drive the spiral blade to rotate. S3, the rotating spiral blades will cause the dopants in the feeding cylinder to gradually move down and eventually fall into the crucible through the discharge port.

[0014] The beneficial effects of this invention compared to the prior art are: Firstly, on the equipment, the inclined feeding cylinder, together with the spiral blades, controls the speed of the drop of the dopant. At the same time, the reciprocating translation of the telescopic rod achieves radial uniform distribution of the dopant, avoiding the local accumulation of traditional feeding. Secondly, in terms of algorithms, the diffusion path is predicted through multiphysics simulation, the intelligent model outputs the optimal strategy, and combined with online detection and feedback correction, the radial and axial concentration deviations are reduced, thus reducing the silicon wafer scrap rate. Third, a full-process framework of "data acquisition - simulation prediction - intelligent decision-making - feedback correction" is constructed. Machine learning models replace human experience to accurately calculate doping amount, reinforcement learning is introduced to continuously optimize model, and feedback correction unit updates data in real time. This eliminates reliance on experience, determines parameters without trial and error, and ultimately improves process stability. Attached Figure Description

[0015] Figure 1 This is a three-dimensional structural schematic diagram of the present invention; Figure 2 yes Figure 1 A magnified view of the area indicated by A1 in the diagram; Figure 3 This is a partial three-dimensional structural schematic diagram of the present invention; Figure 4 yes Figure 3 The enlarged view of the area indicated by A2 in the diagram; Figure 5 It is an exploded three-dimensional structural diagram of the drive shaft and gears; Figure 6 This is a schematic diagram of the planar structure of the present invention; Figure 7 yes Figure 6 The enlarged schematic diagram shown in A3.

[0016] The following are the labels in the diagram: 1. Mounting base; 2. Fixed cylinder; 3. Telescopic rod; 4. Spreading cylinder; 5. Feeding screw; 6. Inlet; 7. Outlet; 8. Guide pipe; 9. Strip groove; 10. Support rod; 11. Cylinder; 12. Circular end cap; 13. Ball bearing No. 1; 14. Rotating rod; 15. Spiral blade; 16. Strip bracket No. 1; 17. Drive shaft; 18. Gear; 19. Rack; 20. Strip bracket No. 2; 21. Drive shaft No. 1; 22. Drive shaft No. 2; 23. Bevel gear group No. 1; 24. Bevel gear group No. 2; 25. Universal joint; 26. Ball bearing No. 2; 27. Racket; 28. Circular frame; 29. ​​Elastic pawl; 30. Linear bearing; 31. Guide rod; 32. Connecting block; 33. Guide sleeve; 34. Connecting sleeve. Detailed Implementation

[0017] To further understand the features, technical means, and specific objectives and functions achieved by the present invention, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0018] refer to Figures 1 to 3The intelligent doping device for a single crystal furnace under different thermal fields shown includes a mounting base 1 located beside the crucible. The mounting base 1 is equipped with a telescopic mechanism, which includes a fixed cylinder 2 horizontally connected to the mounting base 1 and a telescopic rod 3 that slides radially within the fixed cylinder 2 along the crucible. The telescopic rod 3 is equipped with a material spreading mechanism, which includes a material spreading cylinder 4 connected to the telescopic rod 3 and a feeding screw 5 rotating within the material spreading cylinder 4. The material spreading cylinder 4 is inclined, and its upper and lower outer walls are respectively provided with an inlet 6 and an outlet 7. A guide pipe 8 is provided beside the mounting base 1, which can connect with the inlet 6 and inject dopant into the material spreading cylinder 4. A drive mechanism is provided between the fixed cylinder 2 and the material spreading cylinder 4, which drives the feeding screw 5 to rotate by the translation of the telescopic rod 3.

[0019] In practical use, this intelligent doping device combines physical field coupling simulation and machine learning systems to find the relationship between thermal field parameters and doping distribution, ultimately achieving dynamic optimization of the doping amount. Specifically, the intelligent doping process consists of five key steps: I. Acquisition of Thermal Field Parameters and Process Data Using sensors such as thermocouples and infrared thermometers built into the single crystal furnace, thermal field data such as heater power, temperature of each area, gas flow rate and crucible rotation speed are collected in real time. At the same time, historical crystal pulling process parameters, target resistivity and actual doping concentration are recorded. II. Multiphysics Coupling Simulation A three-dimensional model is built based on the actual structure of the single crystal furnace. Then, the temperature field is calculated using the energy conservation equation, the melt flow field is calculated using the Navier-Stokes equation, and the doping diffusion field is calculated using Fick's law. Boundary conditions such as crucible wall and melt surface are then set. Finally, the host computer is used to solve the problem and obtain the concentration distribution of dopant in the melt under different thermal fields. III. Intelligent Modeling of Doping Behavior The collected thermal field parameters and corresponding doping concentration and resistivity data are normalized to eliminate the influence of different units. Then, machine learning algorithms such as gradient boosting trees and deep neural networks are used to build a model of "thermal field parameters → doping concentration distribution → target resistivity". Inputting thermal field parameters will output how much dopant to add. The model is then trained with historical data and the parameters are optimized through cross-validation so that the model can be used under different thermal fields. IV. Intelligent Doping Optimization Calculation During crystal pulling, real-time thermal field parameters are first input into a multiphysics model to quickly predict the diffusion of dopant under the current thermal field. Then, based on the target resistivity required for the silicon wafer and the material properties of the dopant, the theoretically required total amount of dopant is calculated. Finally, through an intelligent model, the optimal dopant addition amount and timing are output, and a reinforcement learning mechanism is used to continuously optimize the model based on the deviation between the actual doping concentration and the target value. V. Feedback Correction and Closed-Loop Control The doping concentration of the current crystal rod is detected using an online spectrometer built into the single crystal furnace or by periodic sampling, and the deviation from the target value is calculated. The deviation data is fed back to the intelligent model to update the training set and improve prediction accuracy. If the deviation is too large, the heater power, crucible rotation speed, or doping strategy is adjusted to form a closed-loop control. After calculating the actual total doping amount through the above process, doping is performed using this doping device. The specific process is as follows: Initially, the telescopic rod 3 is in the retracted state. At this time, the guide cylinder is connected to the feed port 6 on the spreading cylinder 4. A certain amount of dopant is injected into the spreading cylinder 4 through the guide cylinder. After the dopant enters the spreading cylinder 4, it is blocked by the feeding screw 5. Then, the telescopic rod 3 will drive the spreading cylinder 4 to extend towards the crystal rod at the center of the crucible. During this process, the drive mechanism will drive the feeding screw 5 to rotate through the translation of the telescopic rod 3. As the feeding screw 5 rotates, the dopant in the spreading cylinder 4 will move downwards. The material is moved and finally falls into the crucible through the discharge port 7. In summary, the telescopic rod 3 will move back and forth during the doping process, so that the dopant in the feeding cylinder 4 will be gradually guided by the feeding screw 5 to the discharge port 7 and fall out. The feeding screw 5 is used to limit the falling speed of the dopant, and the telescopic rod 3 is used to change the falling direction of the dopant, so as to ensure that the dopant can be evenly dispersed in the crucible. The guide tube 8 and the mounting base 1 are both fixed to the inner wall of the single crystal furnace, and one end of the guide tube 8 passes through the single crystal furnace and extends out of the single crystal furnace.

[0020] refer to Figure 3 and Figure 6 The outer wall of the fixed cylinder 2 is provided with a strip-shaped through groove 9, which is parallel to the telescopic rod 3. A support rod 10 is fixed on the telescopic rod 3, which passes vertically downward through the strip-shaped through groove 9. The telescopic mechanism also includes a cylinder 11 that is horizontally fixed to the mounting base 1, and the output end of the cylinder 11 is fixed to the support rod 10.

[0021] When cylinder 11 is started, cylinder 11 will drive telescopic rod 3 to slide through support rod 10. Finally, telescopic rod 3 will drive material feeding cylinder 4 to extend and retract along the radial direction of crucible. During the extension and retraction of telescopic rod 3, support rod 10 on telescopic rod 3 will slide in strip groove 9, thereby avoiding support rod 10 through strip groove 9.

[0022] refer to Figure 6 and Figure 7As shown, both ends of the spreading cylinder 4 are open structures, and a circular end cap 12 is fixedly provided on each of the openings at both ends of the spreading cylinder 4. A ball bearing 13 is embedded in each circular end cap 12. The feeding screw 5 includes a rotating rod 14 and a spiral blade 15. The rotating rod 14 is coaxially arranged inside the spreading cylinder 4, and both ends of the rotating rod 14 are respectively connected to two ball bearings 13. The spiral blade 15 is fixedly connected to the outer wall of the rotating rod 14 near the discharge port 7, and the spiral blade 15 extends spirally along the axial direction of the rotating rod 14.

[0023] The spiral blade 15 rotates inside the feeding cylinder 4 via the rotating rod 14. When the dopant material falls into the feeding cylinder 4 through the feed port 6, the dopant material will be blocked by the spiral blade 15 and will not fall. When the rotating rod 14 drives the spiral blade 15 to rotate, the dopant material will move downward with the rotation of the spiral blade 15. Finally, the dopant material will be gradually discharged from the discharge port 7 and fall into the crucible.

[0024] refer to Figure 2 , Figure 3 and Figure 6 The outer wall of the upper end of the spreading cylinder 4 is fixedly connected to a first strip bracket 16 extending axially along the telescopic rod 3. The driving mechanism includes a drive shaft 17, a gear 18, a rack 19 and a transmission component. The drive shaft 17 is vertical and rotatably connected to the outer end of the first strip bracket 16. The gear 18 is coaxially fixed to the drive shaft 17. The rack 19 is fixed to the outer wall of the fixed cylinder 2 and is parallel to the telescopic rod 3. The gear 18 and the rack 19 mesh with each other. The transmission component connects the drive shaft 17 and the rotating rod 14.

[0025] When cylinder 11 drives telescopic rod 3 to move horizontally, the spreading cylinder 4 will also move horizontally along with it. During this process, the spreading cylinder 4 will drive drive shaft 17 and gear 18 to move horizontally through first strip bracket 16. Since gear 18 meshes with rack 19, gear 18 will be driven to rotate by rack 19 during the horizontal movement of telescopic rod 3. In this way, gear 18 will drive drive shaft 17 to rotate. After drive shaft 17 rotates, transmission component will drive rotating rod 14 to rotate. Finally, rotating rod 14 will drive the mixed material in spreading cylinder 4 to move downward horizontally through spiral blade 15 and fall out from outlet 7.

[0026] refer to Figure 2 , Figure 3 and Figure 6Above the first strip support 16, there is a second strip support 20 that is fixedly connected to the outer wall of the spreading cylinder 4, and the second strip support 20 extends along the axial direction of the spreading cylinder 4. The transmission components include a first transmission shaft 21, a second transmission shaft 22, a first bevel gear set 23, and a second bevel gear set 24. The first transmission shaft 21 is rotatably connected to the first strip support 16 and is parallel to the first strip support 16. The second transmission shaft 22 is rotatably connected to the second strip support 20 and is perpendicular to the spreading cylinder 4. The first bevel gear set 23 drives the first transmission shaft 21 and the drive shaft 17. The second bevel gear set 24 drives the second transmission shaft 22 and the rotating rod 14. A universal joint 25 is provided between the first transmission shaft 21 and the second transmission shaft 22 to drive them together.

[0027] During the translation of the spreading cylinder 4 driven by the telescopic rod 3, the drive shaft 17 will be driven to rotate by the cooperation of the gear 18 and the rack 19. After the drive shaft 17 rotates, it will drive the first transmission shaft 21 to rotate through the first bevel gear group 23. After the first transmission shaft 21 rotates, the second transmission shaft 22 will be driven to rotate by the first transmission shaft 21 through the universal joint 25. After the second transmission shaft 22 rotates, the rotating rod 14 will be driven to rotate through the second bevel gear group 24, and finally the spiral blade 15 will be driven to rotate through the rotating rod 14.

[0028] refer to Figure 4 and Figure 5 The center of the gear 18 is fitted with a second ball bearing 26, which is coaxially sleeved on the drive shaft 17. A ring of ratchet teeth 27 is fixedly provided on the top of the gear 18. A circular frame 28 located above the gear 18 is coaxially fixed on the drive shaft 17. Several elastic pawls 29 are fixedly provided on the circular frame 28, which are evenly distributed along its circumference. Each elastic pawl 29 engages with the ratchet teeth 27 in one direction and drives the drive shaft 17 to rotate only when the telescopic rod 3 is extended.

[0029] During the process of the telescopic rod 3 driving the spreading cylinder 4 to extend outward, the tip of the elastic pawl 29 will get stuck between the adjacent ratchet teeth 27. The gear 18 driven by the rack 19 will drive the drive shaft 17 to rotate. During the process of the telescopic rod 3 driving the spreading cylinder 4 to retract, the back of the elastic pawl 29 will slide on the ratchet teeth 27. At this time, the gear 18 driven by the gear 18 will not drive the drive shaft 17 to rotate. Ultimately, this ensures that the spiral blade 15 can only rotate in one direction, so that the mixed material in the spreading cylinder 4 can only move downward toward the discharge port 7.

[0030] refer to Figure 6 Both ends of the fixed cylinder 2 are coaxially fixedly connected with linear bearings 30, and the telescopic rod 3 passes through the two linear bearings 30 coaxially.

[0031] The telescopic rod 3 uses two linear bearings 30 to reduce its sliding friction, thereby enabling the telescopic rod 3 to move more smoothly.

[0032] refer to Figure 1 and Figure 6 A guide rod 31 parallel to the telescopic rod 3 is provided above it, and the guide rod 31 is fixedly connected to the telescopic rod 3 through the connecting block 32. A guide sleeve 33 for the guide rod 31 to pass through is fixed on the outer wall of the fixed cylinder 2.

[0033] When the telescopic rod 3 moves horizontally, it will drive the guide rod 31 to slide within the sliding sleeve. This improves the stability of the telescopic rod 3 during horizontal movement through the cooperation between the guide rod 31 and the guide sleeve 33.

[0034] refer to Figure 2 and Figure 6 A connecting sleeve 34 is formed on the outer wall of the spreading cylinder 4, and one end of the telescopic rod 3 is coaxially fixed to the connecting sleeve 34.

[0035] The spreading cylinder 4 is fixedly connected to the telescopic rod 3 by the connecting sleeve 34. When the spreading cylinder 4 is installed on the telescopic rod 3, the spreading cylinder 4 is in an inclined state. The inclined spreading cylinder 4 makes it easier for the mixed material inside to fall out from the discharge port 7.

[0036] A smart doping calculation method for a Czochralski single crystal furnace under different thermal fields, the method comprising the following steps: I. Acquisition of Thermal Field Parameters and Process Data During crystal pulling in a Czochralski single crystal furnace, thermal field parameters are collected in real time using sensors such as thermocouples, infrared thermometers, and gas flow meters built into the furnace. At the same time, process parameters of the historical crystal pulling process are recorded, such as seed crystal contact time, temperature stabilization time, dopant addition time, corresponding target resistivity, and actual doping concentration. II. Multiphysics Coupling Simulation Based on the collected actual structural parameters of the single crystal furnace, a three-dimensional geometric model was established. Subsequently, a multiphysics coupled model including temperature field, flow field, and dopant diffusion field was constructed to simulate the diffusion behavior of dopants in the melt under different thermal conditions. The governing equations required for multiphysics coupling simulation are as follows: For the temperature field, solve the energy conservation equation ∇·(k∇T) + q = ρc_p∂T / ∂t, where k is the thermal conductivity, P is the density, c_p is the specific heat capacity at constant pressure, q is the internal heat source (heater power), T is the measured temperature, and t is the time period; The flow field is described by the Navier-Stokes equations: ∂u / ∂t + u·∇u = -∇p / ρ + ν∇²u+ gβΔT, where u is the flow velocity, p is the pressure, ν is the kinematic viscosity, g is the gravitational acceleration, β is the coefficient of thermal expansion, and ΔT is the temperature difference. The doping diffusion field is coupled with Fick's law ∂C / ∂t + u·∇C = D∇²C, where C is the dopant concentration and D is the diffusion coefficient; After the multiphysics coupling simulation is completed, the coupling solution is performed by the host computer, and the axial and radial concentration distribution of dopants in the melt under different thermal field parameters is output. III. Intelligent Modeling of Doping Behavior The collected thermal field parameters and their corresponding doping concentration distributions and target resistivity are normalized to eliminate the influence of dimensions. A machine learning algorithm is used to establish the mapping relationship between "thermal field parameters → doping concentration distribution → target resistivity". Finally, the model is trained using historical process data, and the hyperparameters are optimized through cross-validation to ensure that the model has good generalization ability under different thermal field conditions. IV. Intelligent Doping Optimization Calculation Input the thermal field parameters collected during the current crystal pulling process into the multiphysics coupling model to quickly predict the doping diffusion characteristics under the current thermal field, such as effective diffusion time and melt convection intensity. Based on the target resistivity of the silicon wafer, and combined with the material properties of the dopant such as solid solubility and segregation coefficient, the total amount of dopant theoretically required is calculated. The intelligent modeling module outputs the optimal dopant addition amount and timing based on the current thermal field parameters and target resistivity. At the same time, a reinforcement learning mechanism is introduced, using the deviation between the actual doping concentration and the target value as the reward function to continuously optimize the model parameters. V. Feedback Correction and Closed-Loop Control The doping concentration of the current crystal rod is obtained by online spectrometer or periodic destructive sampling. The deviation from the target value is calculated and the deviation data is fed back to the intelligent modeling module to update the training set of the machine learning model and improve the prediction accuracy. If the deviation exceeds the threshold, the thermal field parameters or dopant addition strategy are adjusted to achieve closed-loop control and ensure doping accuracy. VI. Specific Doping Process S1, after the actual total amount of doping is calculated through the above process, a certain amount of doping material will fall into the spreading cylinder 4 through the feed inlet 6 and the feed guide pipe 8. S2, the telescopic rod 3 drives the spreading cylinder 4 to extend towards the crucible. Under the transmission action of the drive mechanism, the drive shaft 17 will drive the rotating rod 14 to rotate, and the rotating rod 14 will drive the spiral blade 15 to rotate. S3, the rotating spiral blade 15 will cause the dopants in the feeding cylinder 4 to gradually move down and eventually fall into the crucible through the discharge port 7.

[0037] The above embodiments only illustrate one or more implementations of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A smart doping device for a single crystal furnace under different thermal fields, characterized in that, The system includes a mounting base (1) located beside the crucible. The mounting base (1) is equipped with a telescopic mechanism. The telescopic mechanism includes a fixed cylinder (2) horizontally connected to the mounting base (1) and a telescopic rod (3) sliding radially within the fixed cylinder (2) along the crucible. The telescopic rod (3) is equipped with a feeding mechanism. The feeding mechanism includes a feeding cylinder (4) connected to the telescopic rod (3) and a feeding screw (5) rotating within the feeding cylinder (4). The feeding cylinder (4) is inclined, and its upper and lower outer walls are respectively provided with an inlet (6) and an outlet (7). The mounting base (1) is equipped with a guide pipe (8) that can connect to the inlet (6) and inject dopant into the feeding cylinder (4). A drive mechanism is provided between the fixed cylinder (2) and the feeding cylinder (4) to drive the feeding screw (5) to rotate by the translation of the telescopic rod (3).

2. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 1, characterized in that, The outer wall of the fixed cylinder (2) is provided with a strip-shaped through groove (9), which is parallel to the telescopic rod (3). A support rod (10) is fixed on the telescopic rod (3) and passes vertically downward through the strip-shaped through groove (9). The telescopic mechanism also includes a cylinder (11) that is horizontally fixed to the mounting base (1), and the output end of the cylinder (11) is fixed to the support rod (10).

3. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 1, characterized in that, The two ends of the spreading cylinder (4) are open structures, and a circular end cap (12) is fixedly provided on the openings at both ends of the spreading cylinder (4). Each circular end cap (12) is embedded with a No. 1 ball bearing (13). The feeding screw (5) includes a rotating rod (14) and a spiral blade (15). The rotating rod (14) is coaxially arranged inside the spreading cylinder (4), and the two ends of the rotating rod (14) are respectively connected to two No. 1 ball bearings (13). The spiral blade (15) is fixedly connected to the outer wall of the rotating rod (14) near the discharge port (7), and the spiral blade (15) extends spirally along the axial direction of the rotating rod (14).

4. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 3, characterized in that, The outer wall of the upper end of the spreading cylinder (4) is fixedly connected to a first strip bracket (16) extending axially along the telescopic rod (3). The driving mechanism includes a drive shaft (17), a gear (18), a rack (19), and a transmission component. The drive shaft (17) is vertically connected to the outer end of the first strip bracket (16). The gear (18) is coaxially fixed to the drive shaft (17). The rack (19) is fixed to the outer wall of the fixed cylinder (2) and is parallel to the telescopic rod (3). The gear (18) meshes with the rack (19). The transmission component connects the drive shaft (17) and the rotating rod (14) in a transmission manner.

5. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 4, characterized in that, Above the first strip support (16) is a second strip support (20) fixedly connected to the outer wall of the spreading cylinder (4), and the second strip support (20) extends along the axial direction of the spreading cylinder (4). The transmission components include a first transmission shaft (21), a second transmission shaft (22), a first bevel gear group (23), and a second bevel gear group (24). The first transmission shaft (21) is rotatably connected to the first strip support (16), and the first transmission shaft (21) is connected to the first strip support (16). Parallel to each other, the second drive shaft (22) is rotatably connected to the second strip support (20), and the second drive shaft (22) is perpendicular to the spreading cylinder (4). The first bevel gear group (23) drives the first drive shaft (21) and the drive shaft (17). The second bevel gear group (24) drives the second drive shaft (22) and the rotating rod (14). A universal joint (25) is provided between the first drive shaft (21) and the second drive shaft (22) to drive them together.

6. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 4, characterized in that, The center of the gear (18) is fitted with a No. 2 ball bearing (26), and the No. 2 ball bearing (26) is coaxially sleeved on the drive shaft (17). A ring of ratchet teeth (27) is fixedly provided on the top of the gear (18). A circular frame (28) located above the gear (18) is coaxially fixed on the drive shaft (17). Several elastic pawls (29) are fixedly provided on the circular frame (28) and evenly distributed along its circumference. Each elastic pawl (29) meshes with a single ratchet tooth (27) and drives the drive shaft (17) to rotate only when the telescopic rod (3) is extended.

7. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 1, characterized in that, The fixed cylinder (2) has linear bearings (30) coaxially fixedly connected to both ends, and the telescopic rod (3) passes through the two linear bearings (30) coaxially.

8. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 1, characterized in that, The telescopic rod (3) is provided with a guide rod (31) parallel to it above, and the guide rod (31) is fixedly connected to the telescopic rod (3) through the connecting block (32). The outer wall of the fixed cylinder (2) is fixed with a guide sleeve (33) for the guide rod (31) to pass through.

9. The intelligent doping device for a single crystal furnace under different thermal fields according to claim 1, characterized in that, A connecting sleeve (34) is formed on the outer wall of the spreading cylinder (4), and one end of the telescopic rod (3) is coaxially fixed to the connecting sleeve (34).

10. A smart doping calculation method for a Czochralski single crystal furnace under different thermal fields, employing the smart doping device for a single crystal furnace under different thermal fields as described in claim 6, characterized in that... The calculation method includes the following steps: I. Acquisition of Thermal Field Parameters and Process Data During crystal pulling in a Czochralski single crystal furnace, thermal field parameters are collected in real time using sensors such as thermocouples, infrared thermometers, and gas flow meters built into the furnace. At the same time, process parameters of the historical crystal pulling process are recorded, such as seed crystal contact time, temperature stabilization time, dopant addition time, corresponding target resistivity, and actual doping concentration. II. Multiphysics Coupling Simulation Based on the collected actual structural parameters of the single crystal furnace, a three-dimensional geometric model was established. Subsequently, a multiphysics coupled model including temperature field, flow field, and dopant diffusion field was constructed to simulate the diffusion behavior of dopants in the melt under different thermal conditions. The governing equations required for multiphysics coupling simulation are as follows: For the temperature field, solve the energy conservation equation ∇·(k∇T) + q = ρc_p∂T / ∂t, where k is the thermal conductivity, P is the density, c_p is the specific heat capacity at constant pressure, q is the internal heat source (heater power), T is the measured temperature, and t is the time period; The flow field is described by the Navier-Stokes equations: ∂u / ∂t + u·∇u = -∇p / ρ + ν∇²u + gβΔT, where u is the flow velocity, p is the pressure, ν is the kinematic viscosity, g is the gravitational acceleration, β is the coefficient of thermal expansion, and ΔT is the temperature difference. The doping diffusion field is coupled with Fick's law ∂C / ∂t + u·∇C = D∇²C, where C is the dopant concentration and D is the diffusion coefficient; After the multiphysics coupling simulation is completed, the coupling solution is performed by the host computer, and the axial and radial concentration distribution of dopants in the melt under different thermal field parameters is output. III. Intelligent Modeling of Doping Behavior The collected thermal field parameters and their corresponding doping concentration distributions and target resistivity are normalized to eliminate the influence of dimensions. A machine learning algorithm is used to establish the mapping relationship between "thermal field parameters → doping concentration distribution → target resistivity". Finally, the model is trained using historical process data, and the hyperparameters are optimized through cross-validation to ensure that the model has good generalization ability under different thermal field conditions. IV. Intelligent Doping Optimization Calculation Input the thermal field parameters collected during the current crystal pulling process into the multiphysics coupling model to quickly predict the doping diffusion characteristics under the current thermal field, such as effective diffusion time and melt convection intensity. Based on the target resistivity of the silicon wafer, and combined with the material properties of the dopant such as solid solubility and segregation coefficient, the total amount of dopant theoretically required is calculated. The intelligent modeling module outputs the optimal dopant addition amount and timing based on the current thermal field parameters and target resistivity. At the same time, a reinforcement learning mechanism is introduced, using the deviation between the actual doping concentration and the target value as the reward function to continuously optimize the model parameters. V. Feedback Correction and Closed-Loop Control The doping concentration of the current crystal rod is obtained by online spectrometer or periodic destructive sampling. The deviation from the target value is calculated and the deviation data is fed back to the intelligent modeling module to update the training set of the machine learning model and improve the prediction accuracy. If the deviation exceeds the threshold, the thermal field parameters or dopant addition strategy are adjusted to achieve closed-loop control and ensure doping accuracy. VI. Specific Doping Process S1, after the actual total amount of doping is calculated through the above process, a certain amount of doping material will fall into the spreading cylinder (4) through the feed inlet (6) along the feed pipe (8); S2, the telescopic rod (3) drives the spreading cylinder (4) to extend towards the crucible. Under the transmission action of the driving mechanism, the driving shaft (17) will drive the rotating rod (14) to rotate, and the rotating rod (14) will drive the spiral blade (15) to rotate. S3, the rotating spiral blade (15) will cause the dopants in the feeding cylinder (4) to gradually move down and eventually fall into the crucible through the discharge port (7).

Images