A process for thermal field gradient stabilization control in crystal growth

Abstract

This invention relates to the field of single crystal preparation technology and discloses a process for stabilizing the thermal field gradient during crystal growth. The process includes: acquiring the original weighing signal using a weighing sensor and simultaneously obtaining the argon pressure signal; subtracting the mass shift caused by dynamic lift based on a gas flow interference compensation model to obtain an effective weighing signal; determining the interface thermal compensation value based on the crystallization rate fluctuation value; and outputting an angular acceleration pulse to the crucible rotation drive mechanism to induce a vertical pumping flow, thereby transporting the isothermal melt to the solid-liquid interface to compensate for the axial temperature gradient deviation in situ. This invention utilizes the transient momentum conversion mechanism of the melt flow field to avoid heat transfer lag in the thermal field components, actively convection in the flow field to offset the latent heat fluctuation of the phase change, maintains the steady state of the thermal field gradient at the solid-liquid interface, and coordinates the Lorentz force of the magnetic field to constrain the radial fluctuation of the liquid surface, ensuring a stable growth environment.

Description

Technical Field

[0001] This invention belongs to the field of single crystal preparation technology, and particularly relates to a process for stabilizing and controlling the thermal field gradient during crystal growth. Background Technology

[0002] In the current process of preparing large-size semiconductors and photovoltaic single crystals using the Czochralski method, the axial temperature gradient at the solid-liquid interface determines the growth morphology and microlattice quality of the crystal. Maintaining the thermal field balance inside the melt by generating radiative heat flux through a graphite heater is a common practice in the industry. Especially in the constant diameter growth stage, the main heater needs to adjust its output power in real time according to the crystal weight feedback to compensate for the latent heat of phase transition caused by the fluctuation of the crystallization rate at the interface.

[0003] Because the graphite thermal field components inside the single-crystal furnace have a large physical mass, there is significant heat transfer resistance and thermal inertia lag in the heat conduction from the heater to the solid-liquid interface. This physical characteristic causes the system's response to temperature fluctuations to typically be on the order of minutes, while the release of latent heat of phase change at the solid-liquid interface is a transient physical process on the order of seconds. This spatiotemporal misalignment of heat transfer means that conventional power feedback regulation mechanisms are always in a passive catch-up state, unable to offset the instantaneous latent heat distortion in situ, thus inducing interface convexity jumps and generating primary point defect agglomeration. For example, Chinese invention patent application CN119530971A discloses a single-crystal... The silicon growth control methods and monocrystalline silicon growth control devices monitor the real-time temperature gradient at multiple sites inside the molten silicon and adjust the heating power or gas flow rate of the monocrystalline furnace accordingly to maintain a constant ratio of pulling speed to temperature gradient. When dealing with transient fluctuations, the adjustment command is ultimately reflected by changing the heater output or flow field boundary conditions. However, this does not escape the physical constraints of macroscopic heat transfer hysteresis in the thermal field components. Based on the external heat source adjustment approach, when facing second-level or even millisecond-level latent heat fluctuations in phase change, there is a control blind zone caused by response time lag. Frequent fluctuations in heating power induce macroscopic convection turbulence in the melt, making it difficult to simultaneously achieve accurate compensation of the microscopic thermal gradient at the interface and steady-state maintenance of the melt flow field.

[0004] Therefore, the technical problem to be solved by this invention is how to remove the aerodynamic interference component generated by the weighing signal during the lifting process, and use the flow field momentum response of the melt itself as an agile heat transfer medium to construct a compensation mechanism that can transiently offset the latent heat fluctuation of the interface phase change, so as to achieve stable control of the axial thermal field gradient of the solid-liquid interface. Summary of the Invention

[0005] This invention proposes a process for stabilizing and controlling the thermal field gradient during crystal growth, comprising the following steps: Step 101: During the single crystal growth process in the Czochralski single crystal furnace, the original weighing signal reflecting the dynamic change of the crystal rod mass is collected in real time using a weighing sensor, and the real-time gas pressure signal of the argon purging flow field in the furnace chamber and the instantaneous flow signal of the exhaust port are simultaneously acquired using a pressure transmitter. Step 102: Based on the dynamic pressure distribution of the flow field determined by the real-time air pressure signal and the instantaneous flow rate signal, establish the airflow interference compensation model in the single crystal furnace, use the airflow interference compensation model to calculate the dynamic lift force acting on the surface of the crystal rod, subtract the mass offset caused by the dynamic lift force from the original weighing signal, and obtain an effective weighing signal that can accurately reflect the evolution of crystallization quality at the solid-liquid interface. Step 103: Calculate the rate of change of the effective weighing signal with respect to the growth running time to determine the real-time crystallization rate, extract the instantaneous fluctuation value of the real-time crystallization rate relative to the preset target growth rate, and calculate the interface heat compensation value required to offset the abnormal fluctuation of the latent heat release rate of phase change at the solid-liquid interface based on the instantaneous fluctuation value. Step 104: Convert the interface thermal compensation value into the angular velocity correction parameter of the crucible rotation drive mechanism, output discontinuous angular acceleration pulses to the crucible rotation drive mechanism, and use the angular acceleration pulses to induce a pumping flow with vertical upward momentum at the bottom of the melt. The pumping flow transports the isothermal melt at the bottom of the crucible to the solid-liquid interface along the central axis. The convective heat exchange of the melt itself is used to compensate the axial temperature gradient deviation at the solid-liquid interface in situ, and maintain the steady-state distribution of the micro-thermal field gradient at the solid-liquid interface.

[0006] Preferably, step 102 specifically includes: monitoring the real-time pressure at the argon gas inlet in the furnace chamber and the differential pressure signal at the exhaust port; using the Reynolds number determined by the pressure and differential pressure signals to determine the real-time lift coefficient in the airflow interference compensation model; calculating the false offset caused by the dynamic lift symmetry weighing sensor based on the real-time lift coefficient; and subtracting the false offset from the effective weighing signal to eliminate the interference of airflow turbulence in the furnace on the crystallization rate perception.

[0007] Preferably, the momentum intensity of the angular acceleration pulse is positively correlated with the interface thermal compensation value. By adjusting the amplitude or duration of the angular acceleration pulse, the axial delivery rate of the pumped flow is controlled, so that the heat flux pumped to the solid-liquid interface and the heat fluctuation generated by the real-time crystallization rate are dynamically offset in the time domain. The macroscopic heat transfer hysteresis of the hot field components of the single crystal furnace is avoided by the transient change of the momentum of the melt flow field.

[0008] Preferably, while outputting the angular acceleration pulse, the magnetic induction intensity of the superconducting magnetic field around the single crystal furnace is adjusted. The Lorentz damping force generated inside the melt by the magnetic induction intensity is used to constrain the radial fluctuations caused by the pump flow on the surface of the melt, thereby reducing the mechanical disturbance of the pump flow to the meniscus shape of the liquid surface. This enhances the heat exchange efficiency of the micro-interface while maintaining the macroscopic stability of the melt surface inside the single crystal furnace.

[0009] Preferably, in step 103, the interface thermal compensation value is adjusted by monitoring the trend of the change of the first derivative of the real-time crystallization rate with respect to time; when the real-time crystallization rate is detected to be increasing and the derivative is greater than the preset positive threshold, the rotation base speed of the crucible rotation drive mechanism is reduced to reduce the amount of hot melt at the bottom being transported to the solid-liquid interface, lock the thermal gradient equilibrium point in advance, and suppress further shift of the interface undercooling.

[0010] Preferably, during step 104, the output power of the main heater of the single crystal furnace is kept constant, and the axial temperature gradient is finely adjusted only by adjusting the duty cycle and pulse interval of the angular acceleration pulse. The transient nature of the melt dynamics response is used to offset the microscopic thermal field instability caused by the latent heat fluctuation of the phase change, thus ensuring the temperature gradient at the solid-liquid interface. With crystallization rate ratio Maintain within the critical threshold range for point defect formation.

[0011] Preferably, the triggering frequency of the angular acceleration pulse is constrained by the upper limit boundary of the Reynolds number of the melt at the solid-liquid interface. By limiting the minimum time interval between two adjacent angular acceleration pulses, the flow pattern of the pumped flow is restricted to the laminar flow range, preventing the thermal field distribution from becoming disordered due to the melt changing from laminar to turbulent flow, and ensuring that the heat field pumped to the solid-liquid interface is axially symmetric in the direction of the crystal rod's central axis.

[0012] Preferably, in the shoulder-growing stage, the constant-diameter stage, and the tapping stage of single crystal growth, different area coefficients in the airflow interference compensation model are set by a preset compensation matrix to compensate for the nonlinear effects caused by changes in crystal rod diameter and length extension on the windward area of ​​airflow force, ensuring that the calculation accuracy of the peeling dynamic lift force at different growth stages remains within 0.1% of the effective weighing signal.

[0013] Preferably, the amplitude of the angular acceleration pulse is corrected in a closed loop by monitoring the meniscus height and contact angle at the solid-liquid interface; when the meniscus height deviates from the preset threshold range, the angular velocity change direction of the angular acceleration pulse is adjusted to change the scouring intensity of the pumping flow on the local boundary layer of the interface, thereby achieving secondary fine control of the local temperature gradient at the solid-liquid interface and ensuring the constant diameter control accuracy of crystal growth.

[0014] Compared with existing technologies, the thermal field gradient stabilization control process in crystal growth of this invention has the following advantages: 1. In the stable control of the thermal field gradient, the transient transformation of the flow field momentum of the melt itself is used to effectively avoid the physical lag effect of macroscopic heat transfer of graphite thermal field components in single crystal furnace. By obtaining the rate of change of the continuous weighing signal reflecting the crystallization state of the interface, the rotational acceleration pulse of the crucible motor is controlled to generate a vertical suction flow pointing towards the solid-liquid interface inside the melt. The forced convection of the melt itself is used as an agile heat exchange medium to offset the phase change latent heat fluctuation caused by the transient change in growth rate in real time. This achieves millisecond-level in-situ compensation for the axial thermal field gradient of the interface, thereby solving the minute-level thermal response lag problem that exists when the temperature is stabilized by adjusting the heater power in traditional processes.

[0015] 2. A real-time decoupling mechanism for furnace chamber aerodynamic interference is introduced to ensure the authenticity of the crystallization rate sensing signal. By synchronously acquiring the argon purging pressure signal in the furnace chamber and stripping the aerodynamic resistance component from the continuous weighing signal, pure data characterizing the change in interface crystallization quality is extracted. This effectively suppresses the false offset of the weighing signal caused by the fluctuation of airflow in the furnace, ensuring that the subsequent crucible rotation acceleration adjustment logic is always based on the real latent heat release rate of phase transformation, avoiding the false triggering of adjustment commands and their unnecessary disturbance to the stability of the melt.

[0016] 3. Construct a synergistic constraint mechanism of momentum pulse and electromagnetic damping to enhance microscopic heat transfer efficiency while maintaining macroscopic liquid surface steady state. Simultaneously modulate the Lorentz damping force of the peripheral magnetic field while outputting crucible rotation acceleration pulse. Utilize electromagnetic induction effect to suppress the gravitational wave oscillation of the melt surface induced by sudden rotation speed changes. Ensure that the vertical suction flow only plays a thermal counterbalancing role at the microscopic scale below the solid-liquid interface without disrupting the geometric stability of the meniscus during single crystal pulling, thus achieving a logical closed loop of precise thermal field control and growth interface morphology maintenance. Attached Figure Description

[0017] Figure 1 This is a flowchart of the thermal field gradient stabilization control and signal decoupling process of the present invention; Figure 2 This is a diagram of the collaborative control architecture of melt flow field momentum compensation and electromagnetic damping of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, bottom, transverse, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood in conjunction with the specific circumstances.

[0021] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] A process for stabilizing and controlling the thermal field gradient during crystal growth includes the following steps: Step 101: During the single crystal growth process in the Czochralski single crystal furnace, the original weighing signal reflecting the dynamic change of the crystal rod mass is collected in real time using a weighing sensor, and the real-time gas pressure signal of the argon purging flow field in the furnace chamber and the instantaneous flow signal of the exhaust port are simultaneously acquired using a pressure transmitter. Step 102: Based on the dynamic pressure distribution of the flow field determined by the real-time air pressure signal and the instantaneous flow rate signal, establish the airflow interference compensation model in the single crystal furnace, use the airflow interference compensation model to calculate the dynamic lift force acting on the surface of the crystal rod, subtract the mass offset caused by the dynamic lift force from the original weighing signal, and obtain an effective weighing signal that can accurately reflect the evolution of crystallization quality at the solid-liquid interface. Step 103: Calculate the rate of change of the effective weighing signal with respect to the growth running time to determine the real-time crystallization rate, extract the instantaneous fluctuation value of the real-time crystallization rate relative to the preset target growth rate, and calculate the interface heat compensation value required to offset the abnormal fluctuation of the latent heat release rate of phase change at the solid-liquid interface based on the instantaneous fluctuation value. Step 104: Convert the interface thermal compensation value into the angular velocity correction parameter of the crucible rotation drive mechanism, output discontinuous angular acceleration pulses to the crucible rotation drive mechanism, and use the angular acceleration pulses to induce a pumping flow with vertical upward momentum at the bottom of the melt. The pumping flow transports the isothermal melt at the bottom of the crucible to the solid-liquid interface along the central axis. The convective heat exchange of the melt itself is used to compensate the axial temperature gradient deviation at the solid-liquid interface in situ, and maintain the steady-state distribution of the micro-thermal field gradient at the solid-liquid interface.

[0023] Preferably, step 102 specifically includes: monitoring the real-time pressure at the argon gas inlet in the furnace chamber and the differential pressure signal at the exhaust port; using the Reynolds number determined by the pressure and differential pressure signals to determine the real-time lift coefficient in the airflow interference compensation model; calculating the false offset caused by the dynamic lift symmetry weighing sensor based on the real-time lift coefficient; and subtracting the false offset from the effective weighing signal to eliminate the interference of airflow turbulence in the furnace on the crystallization rate perception.

[0024] Preferably, the momentum intensity of the angular acceleration pulse is positively correlated with the interface thermal compensation value. By adjusting the amplitude or duration of the angular acceleration pulse, the axial delivery rate of the pumped flow is controlled, so that the heat flux pumped to the solid-liquid interface and the heat fluctuation generated by the real-time crystallization rate are dynamically offset in the time domain. The macroscopic heat transfer hysteresis of the hot field components of the single crystal furnace is avoided by the transient change of the momentum of the melt flow field.

[0025] Preferably, while outputting the angular acceleration pulse, the magnetic induction intensity of the superconducting magnetic field around the single crystal furnace is adjusted. The Lorentz damping force generated inside the melt by the magnetic induction intensity is used to constrain the radial fluctuations caused by the pump flow on the surface of the melt, thereby reducing the mechanical disturbance of the pump flow to the meniscus shape of the liquid surface. This enhances the heat exchange efficiency of the micro-interface while maintaining the macroscopic stability of the melt surface inside the single crystal furnace.

[0026] Preferably, in step 103, the interface thermal compensation value is adjusted by monitoring the trend of the change of the first derivative of the real-time crystallization rate with respect to time; when the real-time crystallization rate is detected to be increasing and the derivative is greater than the preset positive threshold, the rotation base speed of the crucible rotation drive mechanism is reduced to reduce the amount of hot melt at the bottom being transported to the solid-liquid interface, lock the thermal gradient equilibrium point in advance, and suppress further shift of the interface undercooling.

[0027] Preferably, during step 104, the output power of the main heater of the single crystal furnace is kept constant, and the axial temperature gradient is finely adjusted only by adjusting the duty cycle and pulse interval of the angular acceleration pulse. The transient nature of the melt dynamics response is used to offset the microscopic thermal field instability caused by the latent heat fluctuation of the phase change, thus ensuring the temperature gradient at the solid-liquid interface. With crystallization rate ratio Maintain within the critical threshold range for point defect formation.

[0028] Preferably, the triggering frequency of the angular acceleration pulse is constrained by the upper limit boundary of the Reynolds number of the melt at the solid-liquid interface. By limiting the minimum time interval between two adjacent angular acceleration pulses, the flow pattern of the pumped flow is restricted to the laminar flow range, preventing the thermal field distribution from becoming disordered due to the melt changing from laminar to turbulent flow, and ensuring that the heat field pumped to the solid-liquid interface is axially symmetric in the direction of the crystal rod's central axis.

[0029] Preferably, in the shoulder-growing stage, the constant-diameter stage, and the tapping stage of single crystal growth, different area coefficients in the airflow interference compensation model are set by a preset compensation matrix to compensate for the nonlinear effects caused by changes in crystal rod diameter and length extension on the windward area of ​​airflow force, ensuring that the calculation accuracy of the peeling dynamic lift force at different growth stages remains within 0.1% of the effective weighing signal.

[0030] Preferably, the amplitude of the angular acceleration pulse is corrected in a closed loop by monitoring the meniscus height and contact angle at the solid-liquid interface; when the meniscus height deviates from the preset threshold range, the angular velocity change direction of the angular acceleration pulse is adjusted to change the scouring intensity of the pumping flow on the local boundary layer of the interface, thereby achieving secondary fine control of the local temperature gradient at the solid-liquid interface and ensuring the constant diameter control accuracy of crystal growth.

[0031] Example 1: In the continuous operation of a large-size ingot growth process in a Czochralski single crystal furnace, latent heat of crystallization is continuously released at the solid-liquid interface during phase transition. The latent heat release rate is transiently distorted by the perturbation of the crystallization rate. Conventional control logic relies on extracting the interface morphology signal and feeding back to adjust the power of the main heater. Due to the heat transfer resistance and large mass thermal inertia of the graphite thermal field components inside the single crystal furnace, the change in the power of the main heater reaches the solid-liquid interface through radiation and conduction paths with a physical delay of minutes. The hysteresis compensation mode based on the external heat source cannot offset the thermal field distortion caused by the transient change in local crystallization rate in real time. This leads to the nonlinear accumulation of latent heat, which disrupts the thermal balance of the solid-liquid interface and increases the density of native point defects inside the crystal lattice.

[0032] To offset this thermal distortion, during the growth process, a weighing sensor was used to collect the original weighing signal reflecting the dynamic changes in the crystal rod mass at a sampling rate of 100Hz. A pressure transmitter was used to simultaneously acquire the real-time pressure signal of the argon purge flow field inside the single crystal furnace and the instantaneous flow rate signal of the exhaust port. Based on the dynamic pressure distribution of the flow field determined by the aforementioned pressure and instantaneous flow signals, an airflow disturbance compensation model was established within the single crystal furnace. The dynamic lift force acting on the crystal rod surface was calculated using this model. The mass offset caused by the dynamic lift force was subtracted from the original weighing signal to obtain an effective weighing signal that eliminates aerodynamic disturbance components and reflects the evolution of crystallization mass at the solid-liquid interface. The first-order grading of the effective weighing signal with respect to the growth running time was then calculated. The derivative is used to determine the real-time crystallization rate. The instantaneous fluctuation value of the real-time crystallization rate relative to the preset target growth rate is extracted. Based on the instantaneous fluctuation value, the interfacial heat compensation value required to offset the fluctuation of the latent heat release rate of phase change at the solid-liquid interface is calculated. When the interfacial heat compensation value exceeds the preset dead zone, the current output power of the main heater is locked constant. The interfacial heat compensation value is converted into the angular velocity correction parameter of the crucible rotation drive mechanism. A discontinuous angular acceleration pulse that is positively correlated with the interfacial heat compensation value is output to the crucible rotation drive mechanism. The angular acceleration pulse is used to break the original rigid body rotation state of the melt and induce a pumping flow with vertical upward momentum at the bottom of the melt. This guides the isothermal melt at the bottom of the crucible to be forcibly transported to the solid-liquid interface along the central axis.

[0033] While outputting angular acceleration pulses to the crucible rotation drive mechanism, the magnetic induction intensity of the superconducting magnetic field surrounding the single crystal furnace is positively modulated based on the acceleration integral value acquired during the pulse execution cycle. The Lorentz damping force generated within the melt by the modulated magnetic induction intensity constrains the radial gravity wave oscillations caused by the pumping flow on the melt surface, reducing the mechanical disturbance of the vertical pumping flow to the meniscus shape. The airflow interference compensation model separates the aerodynamic disturbance component and outputs an effective weighing signal. The pumping flow excited by the angular acceleration pulse transfers the heat exchange carrier from the external graphite thermal field components to the melt itself. Forced convection is used to offset the latent heat fluctuations of the phase change at the solid-liquid interface in situ, cutting off the physical interference path between the pulling speed fluctuations and the power fluctuations of the thermal field components. This offsets local thermal field distortion while maintaining a stable melt surface, maintaining a steady-state distribution of the local thermal field gradient at the solid-liquid interface, and ensuring the axial temperature gradient at the solid-liquid interface... With crystallization rate ratio Maintaining within a preset value range, the positive modulation process of the superconducting magnetic field magnetic induction intensity is specifically executed by multiplying the absolute integral value of the calculated angular acceleration pulse amplitude over a single execution time period by a preset single crystal furnace electromagnetic coupling constant coefficient. The main control system calculates the required superconducting coil excitation current increment in real time based on the product result and superimposes it onto the basic sustaining current. The work done by the Lorentz force, which is synchronously strengthened over time, instantly suppresses the radial gravitational fluctuation energy excited on the liquid surface by the sudden change in rotation speed.

[0034] Example 2: On a 12-inch Czochralski single crystal furnace verification platform equipped with a graphite hot zone component, the basic crystal pulling rate was set to 0.8 mm / min. Random turbulent pressure disturbances with a center frequency of 50 Hz and an amplitude of 0.5 kPa were superimposed in the argon purge circuit. Mechanical vibrations with a signal-to-noise ratio of 20 dB were applied to the single crystal furnace base to simulate an operating environment including aerodynamic and structural noise interference. The experiment was divided into a control group and an experimental group based on the present invention. The control group used a heater power negative feedback mechanism. The experimental group based on the introduced crystallization rate mutation amplitude was divided into low-order variation group, medium-order variation group, high-order variation group, and over-limit variation group. The weighing sensor collected the original weighing signal at a sampling rate of 100 Hz. Affected by the argon turbulence disturbance, the original weighing signal contained a high-frequency mass wave with an amplitude of ±2.5 g. The system determines the dynamic pressure distribution of the flow field based on the real-time air pressure signal synchronously acquired by the pressure transmitter and the instantaneous flow signal at the exhaust port. It then establishes an airflow interference compensation model within the single crystal furnace, calculates the dynamic lift force acting on the crystal rod surface using the airflow interference compensation model, subtracts the mass offset caused by the dynamic lift force from the original weighing signal to obtain the effective weighing signal, calculates the first derivative of the effective weighing signal with respect to the growth running time to determine the real-time crystallization rate, triggers a 5% transient jump in the crystallization rate in the intermediate-order variation group, extracts the instantaneous fluctuation value of the real-time crystallization rate relative to the preset target growth rate, and calculates the interface heat compensation value required to offset the abnormal fluctuation of the latent heat release rate of phase change at the solid-liquid interface based on the instantaneous fluctuation value. When the interface heat compensation value exceeds the preset dead zone, the current output power of the main heater is locked constant.

[0035] The interface thermal compensation value is converted into an angular velocity correction parameter for the crucible rotation drive mechanism. A discontinuous angular acceleration pulse, positively correlated with the interface thermal compensation value, is output to the crucible rotation drive mechanism. For the low-order, medium-order, and high-order variation groups, corresponding to crystallization rate abrupt changes of 2%, 5%, and 8%, respectively, a peak value of 0.5 rad / s is output. 2 1.2 rad / s 2 With 2.0 rad / s 2The discontinuous angular acceleration pulses disrupt the original rigid rotational state of the melt, inducing a pumping flow with vertically upward momentum at the bottom of the melt. This guides the isothermal melt at the bottom of the crucible to be forcibly transported along the central axis towards the solid-liquid interface. The pumping flow reaches the solid-liquid interface within 1.5s, 1.2s, and 0.8s, respectively. The convective heat exchange within the melt itself compensates for the axial temperature gradient deviation at the solid-liquid interface in situ, limiting the temperature deviation to 0.5 s. Within a second-level period of physical displacement of the macroscopic fluid, the fluid momentum change excited by the angular acceleration pulse at the bottom of the crucible propagates upward in the silicon melt medium in the form of a pressure wave at a local sound speed of several kilometers per second. This momentum wave instantaneously reaches the solid-liquid interface and penetrates the solute boundary layer within milliseconds, triggering local forced counter-current turbulence in the micro-region. Thus, before the isothermal melt solid at the bottom arrives, the hysteresis state of the thermal boundary layer where latent heat of phase change accumulates is broken in advance, physically achieving millisecond-level in-situ diffusion pre-compensation for the latent heat fluctuation of phase change. In the over-limit variation group, the crystallization rate is set to a sudden change of 15%, with an output of 4.5 rad / s. 2 The angular acceleration pulse excited by the pumping flow generates a radial gravity wave with an amplitude exceeding 5.2 mm at the liquid surface. This exceeds the constraint threshold of the Lorentz damping force generated inside the melt by the magnetic induction intensity of the superconducting magnetic field outside the single crystal furnace, causing the meniscus shape of the liquid surface to break. The data from the over-limit variation group defines the optimal numerical window of the momentum intensity of the angular acceleration pulse, proving that the output angular acceleration pulse needs to balance the mechanical disturbance of the vertical pumping flow on the meniscus shape of the liquid surface.

[0036] The density distribution of primary point defects inside the grown crystal rods of each group was measured. The heat transfer hysteresis of the main heater in the control group was 4.5 minutes. The latent heat of phase transformation at the solid-liquid interface was accumulated, and the axial temperature gradient was measured. With crystallization rate ratio When the deviation from the set value is 15.6%, vacancy-type defect clusters exist in the crystal lattice cross-section. The low-order, mid-order, and high-order variation groups of this invention utilize the pumping flow induced at the bottom of the melt by angular acceleration pulses to transfer the heat exchange carrier to the melt itself, thereby offsetting the fluctuation of latent heat of phase change at the solid-liquid interface. The fluctuation was kept within 2.3%, and the measured defect density was two orders of magnitude lower than that of the control group. The experimental data confirmed that by extracting the dynamic pressure distribution of the flow field to subtract the mass offset and using the vertical pump flow induced by angular acceleration pulse to compensate for the temperature gradient deviation, the physical interference path between the pull-up speed fluctuation and the power fluctuation of the thermal field component was cut off, and the steady-state distribution of the micro thermal field gradient at the solid-liquid interface was maintained.

[0037] Example 3: In the crystal pulling process of a Czochralski single crystal furnace, the argon purge gas continuously introduced into the furnace chamber forms nonlinear aerodynamic turbulence around the growing crystal rod. This turbulence generates time-varying dynamic lift on the surface of the crystal rod, which, when superimposed on the original weighing signal, causes a baseline drift in the crystallization rate. A single static compensation coefficient cannot track the dynamically evolving aerodynamic resistance that accompanies the increase in crystal rod length. Real-time gas pressure signals within the furnace chamber and instantaneous flow signals at the exhaust port are obtained. Based on the ideal gas law, the argon gas density inside the furnace is calculated using the real-time gas pressure signals and the absolute temperature of the furnace chamber. The ratio of the instantaneous flow signal to the geometric cross-sectional area of ​​the annular channel inside the furnace chamber is also calculated. Determine the axial velocity of argon gas, calculate half the product of argon gas density and the square of the axial velocity of argon gas to obtain the dynamic pressure distribution of the flow field, simultaneously read the current crystal pulling stroke data of the pulling mechanism, calculate the current windward cross-sectional area of ​​the crystal rod based on the current crystal pulling stroke data, use the current crystal pulling stroke data and the axial velocity of argon gas as a joint index to query the preset aerodynamic coefficient matrix, retrieve the corresponding dynamic lift coefficient, calculate the continuous product of the dynamic pressure distribution of the flow field, the current windward cross-sectional area and the dynamic lift coefficient to output the dynamic lift acting on the surface of the crystal rod, subtract the dynamic lift and aerodynamic disturbance component from the original weighing signal collected by the weighing sensor, and output the effective weighing signal.

[0038] The real-time crystallization rate is determined by obtaining the first derivative of the effective weighing signal with respect to the growth run time. The instantaneous fluctuation value is extracted by calculating the numerical difference between the real-time crystallization rate and the preset target growth rate. This value is then input into a preset proportional-integral (PI) arithmetic unit to calculate the interface heat compensation value to offset the fluctuation in the latent heat release rate at the solid-liquid interface. During the calculation, the system calls the product of the pre-set physical constant of the latent heat of phase change in single-crystal silicon crystallization and the current target crystal pulling quality growth rate as a reference heat flux parameter. The dimensionless normalized adjustment coefficient output by the PI arithmetic unit is multiplied by this reference heat flux parameter. Through the underlying multiplication logic, a dimensionality reduction transformation from a purely mathematical control domain to the physical heat dimension (Joules / Watts) is achieved. A dead zone threshold is established to determine the response boundary. When the interface heat compensation value exceeds the dead zone threshold, the main heater is locked. With the current output power constant, the target angular acceleration peak value is determined by multiplying the interface thermal compensation value and the crucible geometric coupling coefficient. Based on the target angular acceleration peak value and a fixed pulse width of 0.5s, a discontinuous angular acceleration pulse is generated. The angular acceleration pulse is sent to the crucible rotation drive mechanism to drive the crucible rotation speed to produce a transient step. The speed step induces a pumping flow with vertical upward momentum at the bottom of the melt, guiding the isothermal melt at the bottom of the crucible to be transported along the central axis to the solid-liquid interface. The above-mentioned joint calculation logic relying on gas pressure and flow rate separates the mechanical drift noise caused by the flow field. The pumping flow induced by the angular acceleration pulse carries the bottom heat flux and directly offsets the temperature deviation at the solid-liquid interface in the time domain, cutting off the temperature gradient oscillation path caused by the conduction hysteresis of the main heater, and maintaining the steady-state distribution of the local thermal field gradient at the solid-liquid interface.

[0039] When converting the interface thermal compensation value into angular velocity correction parameters, the main control system reads the preset silicon melt density. Compared with specific heat capacity at constant pressure Simultaneously extract the average bottom temperature returned by a multi-point thermocouple array arranged on the outer wall of the crucible bottom. Combined with the critical temperature constant of phase transition at the solid-liquid interface Calculate the axial temperature difference that characterizes the vertical thermal potential energy. Based on the interfacial heat compensation value required to offset the latent heat fluctuation of phase change Using the formula Calculate the volume of compensating melt that needs to be delivered to the interface region within a single control cycle. The pumping cross-sectional area calculated based on the current crucible inner diameter is retrieved from memory. This will compensate for the melt volume. Divide by pump cross-sectional area Obtain the target vertical pumping displacement Vertical pump displacement of the target Multiply by the dimensionless momentum transfer coefficient determined in the cold calibration experiment of standard viscosity silicone oil. The required step angular velocity increment for obtaining the crucible The step angular velocity increment Dividing by the set fixed pulse width parameter, the output is the actual angular acceleration pulse amplitude that controls the drive mechanism. To overcome the inevitable heat loss along the way caused by the bottom melt passing through the large axial thermal gradient zone and strong convection environment in the single crystal furnace during the vertical pumping of the melt to the solid-liquid interface, the main control system introduces a nonlinear heat dissipation compensation factor that is dynamically determined by the lifting stroke and the real-time heat power of the main heater when calculating the compensation melt volume. The actual required output compensation melt volume is adjusted to the product of the theoretical ideal adiabatic solution volume and the compensation factor, using engineering feedforward to offset the causal deviation between the ideal adiabatic formula and the real strong convection thermodynamic dissipation environment.

[0040] While monitoring the upper limit boundary of the melt Reynolds number at the solid-liquid interface to constrain the pulse triggering frequency, the real-time rotational angular velocity of the crystal rod transmitted back by the pulling mechanism is collected. Compared with the current reference angular velocity of the crucible Calculate the absolute value of the difference between the two as the relative rotational angular velocity. The actual diameter data of the crystal rod during the current crystal pulling stroke is obtained using a CCD vision diameter measurement system above the single crystal furnace to determine the characteristic radius of the solid-liquid interface. Based on the temperature readings at the solid-liquid interface edge, the melt kinematic viscosity parameters in the preset temperature-dynamic viscosity matrix are retrieved. Through formula Calculate the transient Reynolds number of the local flow field at the current moment. The upper limit threshold of the critical laminar Reynolds number determined by fluid dynamics simulation is burned into the system controller. Furthermore, the constant value is set between 1800 and 2100 based on different thermal field sizes, when the transient Reynolds number is calculated. Reaching the upper threshold of the critical laminar Reynolds number When the flow rate reaches 90%, the controller adaptively extends the time interval between adjacent angular acceleration pulse output commands by superimposing a time delay constant. This prevents the superposition of high-frequency pumping momentum from causing the local flow field morphology to change from laminar to disordered turbulent state. Within this pulse control cycle, the system uses a CCD vision system to extract the pixel offset of the meniscus height relative to the preset standard constant diameter baseline height in real time. This value is multiplied by the reference angular acceleration pulse amplitude of the previous control cycle, and the output is used to adjust the pulse correction amplitude for the next cycle. At the same time, when the vision determines that the meniscus height offset direction is below the baseline, the control program automatically reverses the vector direction of the correction angular acceleration, and establishes a complete closed loop for constant diameter control accuracy by utilizing the secondary change of the fluid boundary layer scouring intensity.

[0041] Example 4: When the system faces new furnace dimensions or changes in the configuration of hot zone components, the quantitative perception of the dynamic pressure distribution of the flow field relies on the aerodynamic reference data of a specific furnace chamber space. A scaled-down solid furnace chamber model is constructed, and a flow controller is connected to the exhaust port. Multiple sets of counterweights with known frontal cross-sectional areas are suspended on the lifting mechanism. In a static, unheated state, a step-changing argon gas flow is introduced into the solid furnace chamber model. A tension sensor is used to synchronously record the normal force offset of the counterweights at each flow node. Based on the normal force offset, the frontal cross-sectional area of ​​the counterweights, and the set axial velocity of the argon gas, the dynamic lift coefficient at the corresponding crystal pulling stroke data node is calculated. Dynamic lift coefficient Satisfy the formula ,in, This is the normal force offset. The density of argon gas. The axial flow velocity of argon gas. The windward cross-sectional area of ​​the counterweight is used as the basis for incorporating the dynamic lift coefficient throughout the entire crystal pulling process into the aerodynamic coefficient matrix. Based on this, a gradient of 0.1 rad / s is output to the crucible rotation drive mechanism under no-load thermal field conditions. 2 The test angular acceleration pulse is obtained by measuring the peak vertical displacement of the center point of the melt surface using a Doppler velocimeter. The geometric coupling coefficient of the crucible is determined based on the ratio mapping slope between the test angular acceleration pulse and the peak vertical displacement.

[0042] Based on the normal force offset obtained from the counterweight and the step argon gas flow, a quantitative benchmark is provided for the aerodynamic coefficient matrix. This quantitative benchmark eliminates the baseline drift error of the gas flow interference compensation model under different furnace loading batches. The crucible geometric coupling coefficient determined under the no-load hot field state establishes a mathematical mapping path between the angular acceleration pulse energy and the melt pumping flow rate. In order to eliminate the gap between the physical and chemical properties of the standard test fluid and the actual working fluid, before introducing the dimensionless momentum transfer coefficient, this mapping path uses a built-in algorithm to retrieve in real time the ratio of the kinematic viscosity of the high-temperature silicon melt obtained from the lookup table to the known dynamic viscosity of the calibrated silicone oil, and uses this ratio as a scaling operator to multiply by the aforementioned cold-state test coefficient k, thereby transforming it into an equivalent transfer benchmark under the high-temperature magnetohydrodynamic environment. This mathematical mapping path blocks the transfer loss introduced by nondeterministic estimation. The calculation process of the target angular acceleration peak value based on the interface thermal compensation value obtains the quantitative input source and defines the operational boundary of the solid-liquid interface heat flux counteraction action.

[0043] Example 5: When the system faces the deployment of new thermal field components or the initialization of control algorithms, the calculation of the interface thermal compensation value relies on the numerical reference of the proportional coefficient and integral time constant inside the proportional-integral arithmetic unit. During the no-load test phase, a silicon rod model with a built-in temperature sensor is suspended by the lifting mechanism and heated to the set temperature. A step power disturbance with an amplitude of 5% is artificially input into the main heater. The temperature sensor collects real-time temperature offset data at the simulated position of the solid-liquid interface at a frequency of 50Hz. Based on the real-time temperature offset data, the temperature response time and natural oscillation period are extracted. The temperature response time and natural oscillation period are substituted into the closed-loop tuning algorithm to derive the initial proportional coefficient and initial integral time constant. On this basis, a test angular acceleration pulse is output to the crucible rotation drive mechanism, and the temperature rebound overshoot corresponding to the pulse is recorded. With the temperature rebound overshoot less than 0.1℃ as the optimization target boundary, the descending simplex algorithm is started to iteratively optimize the initial proportional coefficient and initial integral time constant. The convergent proportional coefficient and convergent integral time constant that meet the optimization target boundary are written into the proportional-integral arithmetic unit.

[0044] Based on the closed-loop tuning algorithm, the initial proportional coefficient and initial integral time constant are derived to provide basic control parameters for the proportional-integral arithmetic unit that match the physical characteristics of the thermal field. The test angular acceleration pulse is introduced and the temperature rebound overshoot is used as the optimization target boundary to start the iterative optimization of the descending simplex algorithm to eliminate the system overshoot oscillation caused by empirical assignment. This process constructs a correlation mapping path between instantaneous fluctuation values ​​and interface thermal compensation values. This mapping path constrains the momentum boundary of the angular acceleration pulse output by the crucible rotation drive mechanism, limiting the occurrence of heat flux overload at the solid-liquid interface. The convergence proportional coefficient and convergence integral time constant control the compensation range of the latent heat fluctuation of phase change and maintain the steady-state distribution of the axial temperature gradient at the solid-liquid interface.

[0045] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A process for stabilizing and controlling the thermal field gradient during crystal growth, characterized in that, Includes the following steps: Step 101: During the single crystal growth process in the Czochralski single crystal furnace, the original weighing signal reflecting the dynamic change of the crystal rod mass is collected in real time using a weighing sensor, and the real-time gas pressure signal of the argon purging flow field in the furnace chamber and the instantaneous flow signal of the exhaust port are simultaneously acquired using a pressure transmitter. Step 102: Based on the dynamic pressure distribution of the flow field determined by the real-time air pressure signal and the instantaneous flow rate signal, establish the airflow interference compensation model in the single crystal furnace, use the airflow interference compensation model to calculate the dynamic lift force acting on the surface of the crystal rod, subtract the mass offset caused by the dynamic lift force from the original weighing signal, and obtain an effective weighing signal that can accurately reflect the evolution of crystallization quality at the solid-liquid interface. Step 103: Calculate the rate of change of the effective weighing signal with respect to the growth running time to determine the real-time crystallization rate, extract the instantaneous fluctuation value of the real-time crystallization rate relative to the preset target growth rate, and calculate the interface heat compensation value required to offset the abnormal fluctuation of the latent heat release rate of phase change at the solid-liquid interface based on the instantaneous fluctuation value. Step 104: Convert the interface thermal compensation value into the angular velocity correction parameter of the crucible rotation drive mechanism, output discontinuous angular acceleration pulses to the crucible rotation drive mechanism, and use the angular acceleration pulses to induce a pumping flow with vertical upward momentum at the bottom of the melt. The pumping flow transports the isothermal melt at the bottom of the crucible to the solid-liquid interface along the central axis. The convective heat exchange of the melt itself is used to compensate the axial temperature gradient deviation at the solid-liquid interface in situ, and maintain the steady-state distribution of the micro-thermal field gradient at the solid-liquid interface.

2. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, Step 102 specifically includes: monitoring the real-time pressure at the argon gas inlet in the furnace chamber and the differential pressure signal at the exhaust port; using the Reynolds number determined by the pressure and differential pressure signals to determine the real-time lift coefficient in the airflow interference compensation model; calculating the false offset caused by the dynamic lift symmetry weighing sensor based on the real-time lift coefficient; and subtracting the false offset from the effective weighing signal to eliminate the interference of airflow turbulence in the furnace on the crystallization rate perception.

3. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, The momentum intensity of the angular acceleration pulse is positively correlated with the interface thermal compensation value. By adjusting the amplitude or duration of the angular acceleration pulse, the axial delivery rate of the pumped flow can be controlled, so that the heat flux pumped to the solid-liquid interface and the heat fluctuation generated by the real-time crystallization rate can be dynamically offset in the time domain. The macroscopic heat transfer hysteresis of the hot field components of the single crystal furnace can be avoided by the transient change of the momentum of the melt flow field.

4. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, While outputting angular acceleration pulses, the magnetic induction intensity of the superconducting magnetic field around the single crystal furnace is adjusted. The Lorentz damping force generated inside the melt by the magnetic induction intensity is used to constrain the radial fluctuations caused by the pump flow on the surface of the melt, thereby reducing the mechanical disturbance of the pump flow to the meniscus shape of the liquid surface. This enhances the heat exchange efficiency of the micro-interface while maintaining the macroscopic stability of the melt surface inside the single crystal furnace.

5. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, In step 103, the interface thermal compensation value is adjusted by monitoring the trend of the change of the first derivative of the real-time crystallization rate with respect to time. When the real-time crystallization rate is detected to be increasing and the derivative is greater than the preset positive threshold, the rotation base speed of the crucible rotation drive mechanism is reduced to reduce the amount of hot melt at the bottom being transported to the solid-liquid interface, lock the thermal gradient equilibrium point in advance, and suppress further shift of the interface undercooling.

6. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, During step 104, the output power of the main heater of the single crystal furnace is kept constant. The axial temperature gradient is finely adjusted only by adjusting the duty cycle and pulse interval of the angular acceleration pulse. The transient nature of the melt dynamic response is used to offset the microscopic thermal field instability caused by the latent heat fluctuation of phase transformation, so as to ensure that the ratio of temperature gradient G to crystallization rate V at the solid-liquid interface, G / V, is maintained within the critical threshold range for point defect formation.

7. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, The triggering frequency of the angular acceleration pulse is constrained by the upper limit boundary of the Reynolds number of the melt at the solid-liquid interface. By limiting the minimum time interval between two adjacent angular acceleration pulses, the flow pattern of the pumped flow is restricted to the laminar flow range, preventing the thermal field distribution from becoming disordered due to the melt changing from laminar to turbulent flow, and ensuring that the heat field pumped to the solid-liquid interface is axially symmetric in the direction of the crystal rod's central axis.

8. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, In the shoulder-forming stage, constant diameter stage, and taper stage of single crystal growth, different area coefficients in the airflow interference compensation model are set by a preset compensation matrix to compensate for the nonlinear effects caused by changes in crystal rod diameter and length extension on the windward area of ​​airflow force, ensuring that the calculation accuracy of the peeling dynamic lift force at different growth stages remains within 0.1% of the effective weighing signal.

9. The thermal field gradient stabilization control process in crystal growth according to claim 1, characterized in that, By monitoring the meniscus height and contact angle at the solid-liquid interface, the amplitude of the angular acceleration pulse is corrected in a closed loop. When the meniscus height deviates from the preset threshold range, the direction of the angular velocity change of the angular acceleration pulse is adjusted to change the scouring intensity of the pumping flow on the local boundary layer of the interface, thereby achieving secondary fine control of the local temperature gradient at the solid-liquid interface and ensuring the constant diameter control accuracy of crystal growth.

Images