A process for graphite heater thermal field regulation for vapor deposition

Abstract

The present application relates to the technical field of polysilicon preparation, and discloses a graphite heater thermal field regulation process for gas phase deposition, comprising: collecting a transient electric signal at the end of a power supply bus in a reduction furnace, separating a static composite resistance and a high-frequency dynamic differential impedance excited by commutation trigger ripple through digital filtering; calculating the residual proportion of the two to determine a shunt decoupling factor, and correcting the static composite resistance in situ to obtain an intrinsic thermal resistance value eliminating semiconductor parallel shunt interference; calculating a first-order time differential of the intrinsic thermal resistance value to obtain an intrinsic resistance change rate, and combining a preset standard to obtain a control deviation amount to adjust input power, so that the present application eliminates signal distortion caused by the shunt of the polysilicon shell layer conductive network at high temperature, accurately reconstructs the thermal flux density evolution of the reaction interface, realizes accurate hedging of input power and thermal field variation, and locks the deposition phase interface temperature in the best reaction crystallization window.

Description

Technical Field

[0001] This invention relates to a process for controlling the thermal field of a graphite heater used in vapor deposition, belonging to the field of polycrystalline silicon preparation technology. Background Technology

[0002] Currently, in reduction furnaces used for preparing high-purity silicon via chemical vapor deposition (CVD), low-resistivity graphite heaters are typically used as the heating element. An external transformer supplies a large current to the graphite heater, providing the necessary heat for the reduction reaction and maintaining the reaction surface within a preset crystallization temperature range. The vapor-phase material then undergoes a heterogeneous reduction reaction on the silicon core surface, achieving continuous growth of high-purity silicon. The CVD process is carried out at high temperatures, and complex electrothermal coupling and heat transfer changes occur simultaneously within the furnace. As high-purity silicon is continuously deposited, the diameter of the silicon rod increases, and the furnace spatial boundaries change accordingly. The thermal resistance of the deposition phase interface also undergoes significant nonlinear changes. Simultaneously, radiative heat transfer is correlated with the fourth power of the absolute temperature. Therefore, changes in furnace temperature and heat transfer conditions further increase the difficulty of thermal field control.

[0003] Especially when the outer diameter of the silicon rod continuously increases and the deposition phase interface temperature varies between 600℃ and 800℃, the carrier concentration in the polycrystalline silicon will increase significantly. The originally high-resistivity outer polycrystalline silicon shell will gradually form a parallel conductive path with lower resistance, diverting part of the current that originally flowed through the graphite heater. At this time, the static composite resistance calculated by the ratio of the voltage fundamental wave to the current fundamental wave will be too low, failing to accurately reflect the actual hot resistance of the graphite heater. This can easily mask the true temperature rise caused by local heat accumulation. While physical temperature probes can directly measure the furnace temperature, these probes are exposed to high temperature, high pressure, and corrosive environments for extended periods, making them prone to structural damage and difficult to maintain stable and reliable measurement performance. Existing software control methods and adjustment methods based on overall operating parameters also have shortcomings. For example, Chinese invention patent CN113772674B discloses a control method for a polycrystalline silicon production reduction furnace. By using preset reference curves for voltage, power, and exhaust gas temperature change rate, and combining this with PID control to adjust the operating state of the reduction furnace, the above method requires that the collected voltage and current accurately reflect the state of the heating element. When the reduction furnace is in the middle and later stages of operation, the silicon rod gradually thickens, and the carrier concentration of the polycrystalline silicon shell changes significantly. Because the existing control method does not separate the parallel impedance formed by the graphite heater and the polycrystalline silicon shell, the collected voltage and current will be affected by the current shunting of the polycrystalline silicon shell. As a result, the static composite resistance obtained deviates from the actual hot resistance of the graphite heater, and the original reference curve is also difficult to accurately correspond to the constantly changing electrothermal state in the furnace. In this case, the control system cannot accurately judge the actual heat transfer state of the deposition phase interface and may incorrectly adjust the input power based on the deviation resistance signal, thereby causing the temperature field in the furnace to shift. In severe cases, it may also cause polycrystalline silicon atomization or cause the deposition phase interface to deviate from the suitable crystallization temperature range.

[0004] Therefore, how to utilize the alternating ripple component in the power supply circuit to eliminate the influence of the parallel current shunt in the polycrystalline silicon shell on the resistance measurement results, and to restore the intrinsic hot resistance of the graphite heater online, has become a technical problem that needs to be solved in this field. Summary of the Invention

[0005] To address the problems in the background art, the technical solution of the present invention is as follows: A process for controlling the thermal field of a graphite heater for vapor deposition, comprising the following steps:

[0006] Step S101: Collect the real-time voltage and real-time current of the graphite heater, which serves as the heating element and reaction substrate, in the polycrystalline silicon reduction furnace. Use a phase-locked tracking bandpass digital filter to filter out stray harmonics other than the commutation trigger ripple frequency. Calculate the static composite resistance containing the parallel shunt data of the polycrystalline silicon shell based on the ratio of the real-time voltage to the real-time current at the same sampling time. Separate the high-frequency dynamic differential impedance that is monotonically bound to the commutation trigger ripple frequency at the center main frequency.

[0007] Step S102: Determine the shunt decoupling factor based on the ratio of the difference between the static composite resistance and the high-frequency dynamic differential impedance to the static composite resistance. Use the shunt decoupling factor to correct the static composite resistance to remove the polysilicon shell parallel shunt data and restore the intrinsic hot resistance value after eliminating shunt interference.

[0008] Step S103: Perform first-order time differential calculation on the intrinsic hot resistance value to obtain the intrinsic resistance change rate. Perform differential subtraction between the intrinsic resistance change rate and the preset target resistance change rate to obtain the control deviation. Adjust the trigger conduction angle of the thyristor trigger circuit according to the control deviation to regulate the input power of the graphite heater.

[0009] Preferably, step S101, in which the stray harmonics other than the commutation trigger ripple frequency are filtered out by the phase-locked loop tracking bandpass digital filter, includes: step S1011, continuously sampling the real-time voltage and real-time current at high frequency to obtain sampling data, and setting the sampling frequency to more than 10 times the commutation trigger ripple frequency; step S1012, configuring the phase-locked loop to track the fundamental frequency change of the high-power rectifier grid in real time, binding the center frequency of the digital filter to the 6th harmonic of the fundamental frequency, and setting an upper cutoff frequency and a lower cutoff frequency symmetrical about the center frequency to cut off non-stationary random stray harmonics.

[0010] Preferably, before step S102, the following steps are included: step S1013, retrieving the cumulative running time of the graphite heater and searching for the electrode aging resistance displacement compensation factor corresponding to the cumulative running time; step S1014, using the electrode aging resistance displacement compensation factor to correct the baseline displacement of the static composite resistor and eliminate the steady-state drift of the contact resistance of the electrode contact surface.

[0011] Preferably, step S103, adjusting the triggering conduction angle of the thyristor triggering circuit according to the control deviation, includes: step S1031, obtaining the cumulative power consumption of the graphite heater, and identifying the shell volume thickening stage of the polycrystalline silicon rod based on the cumulative power consumption; step S1032, searching the nonlinear control gain adjustment matrix based on the shell volume thickening stage, using a low gain coefficient of less than 0.5 in the early stage of polycrystalline silicon rod growth, and exponentially increasing the gain coefficient in the middle and late stages of polycrystalline silicon rod growth to adjust the response sensitivity of the control deviation.

[0012] Preferably, while performing the first-order time derivative calculation to obtain the intrinsic resistance change rate in step S103, the method also includes the following steps: Step S1033, simultaneously calculating the second-order time derivative of the intrinsic hot-state resistance to obtain the resistance change acceleration index; Step S1034, introducing the resistance change acceleration index as a safety feedback boundary into the control loop to limit the adjustment range of the trigger conduction angle of the thyristor trigger circuit.

[0013] Preferably, after step S1034, the following steps are also included: Step S10341, comparing the algebraic signs of the control deviation and the resistance change acceleration index, and determining whether the absolute values ​​of the control deviation and the resistance change acceleration index have exceeded the set critical change threshold respectively; Step S10342, when the algebraic signs of the control deviation and the resistance change acceleration index are the same and their absolute values ​​have exceeded the set critical change threshold, loading a preset suppression coefficient in the power regulation control loop to reduce the adjustment range of the heating power.

[0014] Preferably, after step S103, the following steps are also included: step S1035, inverting the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information in the reduction furnace based on the intrinsic thermal resistance; step S1036, adjusting the flow rate and furnace pressure of the mixed gas of trichlorosilane and hydrogen introduced into the reduction furnace based on the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information.

[0015] Preferably, step S1012, which involves configuring a phase-locked loop to track the fundamental frequency change of a high-power rectifier grid in real time, includes: step S10121, monitoring the center frequency fluctuation of the high-power rectifier grid in real time using a frequency sensor; and step S10122, when the center frequency fluctuates within the range of 49.5Hz to 50.5Hz, using a phase-locked loop to drive the center frequency of a digital filter for synchronous dynamic adjustment.

[0016] Preferably, after step S103, the following steps are also included: Step S1037, continuously monitor the intrinsic hot resistance value after correction by the electrode aging resistance compensation factor, and accumulate the total power consumption of the graphite heater in real time; Step S1038, when the intrinsic hot resistance value enters the set steady-state range and the total power consumption reaches the predetermined value of the target diameter of the corresponding polycrystalline silicon rod, stop the introduction of reaction gas into the reduction furnace, and switch to the cooling stage of the reduction furnace.

[0017] Compared with the prior art, the beneficial effects of the present invention are:

[0018] 1. By determining the shunt decoupling factor through the difference between the static composite resistance and the high-frequency dynamic differential impedance, the influence caused by the parallel shunt of the polycrystalline silicon shell can be subtracted from the resistance signal collected from the power supply circuit, thus more accurately restoring the intrinsic hot resistance of the graphite heater. The resulting intrinsic resistance change rate can more realistically reflect the heat transfer change at the deposition phase interface, avoiding misjudgment of the actual heating state of the graphite heater due to low apparent resistance, and improving the accuracy of the thermal field regulation of the reduction furnace.

[0019] 2. Phase-locked tracking bandpass digital filtering enables the filter center frequency to change synchronously with the commutation trigger ripple frequency, and suppresses spurious harmonics and electromagnetic interference outside the center frequency. This reduces glitches and random fluctuations in the original voltage and current signals, prevents noise from being further amplified during differential calculation, and makes the data used for subsequent impedance separation and intrinsic resistance rate calculation more stable, thereby reducing the possibility of high-frequency oscillations in the control loop.

[0020] 3. By using an electrode aging resistance compensation factor to correct the steady-state drift of electrode contact resistance over cumulative operating time, the impact of equipment aging on the static composite resistance baseline can be reduced, ensuring that the intrinsic hot resistance value remains comparable over long-term operation. Furthermore, by identifying the shell volume thickening stage based on cumulative power consumption and employing a nonlinear control gain adjustment matrix matching this stage, power regulation overshoot can be reduced in the early stages of polycrystalline silicon rod growth, and control response capability can be improved in the later stages of growth. This adapts to the increased heat transfer hysteresis after silicon rod thickening, resulting in more stable power regulation at different deposition stages. Attached Figure Description

[0021] Figure 1 This is a flowchart of the intrinsic resistance decoupling and power regulation steps of the graphite heater of the present invention;

[0022] Figure 2 This is an interactive diagram of the closed-loop measurement and control physical module of the polysilicon reduction furnace of the present invention.

[0023] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0024] 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.

[0025] A process for controlling the thermal field of a graphite heater for vapor deposition includes the following steps:

[0026] Step S101: Collect the real-time voltage and real-time current of the graphite heater, which serves as the heating element and reaction substrate, in the polycrystalline silicon reduction furnace. Use a phase-locked tracking bandpass digital filter to filter out stray harmonics other than the commutation trigger ripple frequency. Calculate the static composite resistance containing the parallel shunt data of the polycrystalline silicon shell based on the ratio of the real-time voltage to the real-time current at the same sampling time. Separate the high-frequency dynamic differential impedance that is monotonically bound to the commutation trigger ripple frequency at the center main frequency.

[0027] Step S102: Determine the shunt decoupling factor based on the ratio of the difference between the static composite resistance and the high-frequency dynamic differential impedance to the static composite resistance. Use the shunt decoupling factor to correct the static composite resistance to remove the polysilicon shell parallel shunt data and restore the intrinsic hot resistance value after eliminating shunt interference.

[0028] Step S103: Perform first-order time differential calculation on the intrinsic hot resistance value to obtain the intrinsic resistance change rate. Perform differential subtraction between the intrinsic resistance change rate and the preset target resistance change rate to obtain the control deviation. Adjust the trigger conduction angle of the thyristor trigger circuit according to the control deviation to regulate the input power of the graphite heater.

[0029] Preferably, step S101, in which the stray harmonics other than the commutation trigger ripple frequency are filtered out by the phase-locked loop tracking bandpass digital filter, includes: step S1011, continuously sampling the real-time voltage and real-time current at high frequency to obtain sampling data, and setting the sampling frequency to more than 10 times the commutation trigger ripple frequency; step S1012, configuring the phase-locked loop to track the fundamental frequency change of the high-power rectifier grid in real time, binding the center frequency of the digital filter to the 6th harmonic of the fundamental frequency, and setting an upper cutoff frequency and a lower cutoff frequency symmetrical about the center frequency to cut off non-stationary random stray harmonics.

[0030] Preferably, before step S102, the following steps are included: step S1013, retrieving the cumulative running time of the graphite heater and searching for the electrode aging resistance displacement compensation factor corresponding to the cumulative running time; step S1014, using the electrode aging resistance displacement compensation factor to correct the baseline displacement of the static composite resistor and eliminate the steady-state drift of the contact resistance of the electrode contact surface.

[0031] Preferably, step S103, adjusting the triggering conduction angle of the thyristor triggering circuit according to the control deviation, includes: step S1031, obtaining the cumulative power consumption of the graphite heater, and identifying the shell volume thickening stage of the polycrystalline silicon rod based on the cumulative power consumption; step S1032, searching the nonlinear control gain adjustment matrix based on the shell volume thickening stage, using a low gain coefficient of less than 0.5 in the early stage of polycrystalline silicon rod growth, and exponentially increasing the gain coefficient in the middle and late stages of polycrystalline silicon rod growth to adjust the response sensitivity of the control deviation.

[0032] Preferably, while performing the first-order time derivative calculation to obtain the intrinsic resistance change rate in step S103, the method also includes the following steps: Step S1033, simultaneously calculating the second-order time derivative of the intrinsic hot-state resistance to obtain the resistance change acceleration index; Step S1034, introducing the resistance change acceleration index as a safety feedback boundary into the control loop to limit the adjustment range of the trigger conduction angle of the thyristor trigger circuit.

[0033] Preferably, after step S1034, the following steps are also included: Step S10341, comparing the algebraic signs of the control deviation and the resistance change acceleration index, and determining whether the absolute values ​​of the control deviation and the resistance change acceleration index have exceeded the set critical change threshold respectively; Step S10342, when the algebraic signs of the control deviation and the resistance change acceleration index are the same and their absolute values ​​have exceeded the set critical change threshold, loading a preset suppression coefficient in the power regulation control loop to reduce the adjustment range of the heating power.

[0034] Preferably, after step S103, the following steps are also included: step S1035, inverting the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information in the reduction furnace based on the intrinsic thermal resistance; step S1036, adjusting the flow rate and furnace pressure of the mixed gas of trichlorosilane and hydrogen introduced into the reduction furnace based on the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information.

[0035] Preferably, step S1012, which involves configuring a phase-locked loop to track the fundamental frequency change of a high-power rectifier grid in real time, includes: step S10121, monitoring the center frequency fluctuation of the high-power rectifier grid in real time using a frequency sensor; and step S10122, when the center frequency fluctuates within the range of 49.5Hz to 50.5Hz, using a phase-locked loop to drive the center frequency of a digital filter for synchronous dynamic adjustment.

[0036] Preferably, after step S103, the following steps are also included: Step S1037, continuously monitor the intrinsic hot resistance value after correction by the electrode aging resistance compensation factor, and accumulate the total power consumption of the graphite heater in real time; Step S1038, when the intrinsic hot resistance value enters the set steady-state range and the total power consumption reaches the predetermined value of the target diameter of the corresponding polycrystalline silicon rod, stop the introduction of reaction gas into the reduction furnace, and switch to the cooling stage of the reduction furnace.

[0037] Example 1: Polycrystalline silicon vapor deposition is a thermochemically coupled synchronous growth process. During the reaction, the main challenge within the reduction furnace lies in the nonlinear evolution of the three-dimensional geometric topology as the diameter of the polycrystalline silicon rod continuously increases. This leads to nonlinear dynamic drift in spatial thermal resistance and radiative heat transfer characteristics. This nonlinear evolution of internal thermal equilibrium is a unique intrinsic physical challenge of polycrystalline silicon reduction furnaces. In the later stages of operation, as the silicon rod diameter increases, the increase in semiconductor carriers in the gas-deposited shell at high temperatures causes variations in parallel shunt resistance, resulting in glitch distortion in the original sampling sequence, reduced control sensitivity due to increased thermal resistance, and the inability of static composite resistors to decouple from the outer silicon shell. To address the technical problem of nonlinear shunt interference in the polysilicon shell, the method claimed in this invention collects transient electrical signals from the power supply bus in the reduction furnace. It extracts a static composite resistance containing polysilicon shell parallel shunt data on the first track, separates a high-frequency dynamic differential impedance monotonically bound to the commutation trigger ripple frequency on the second track, calculates the shunt decoupling factor in algebraic space, corrects the static composite resistance, and performs in-situ correction and restoration to obtain the intrinsic thermal resistance value that eliminates semiconductor parallel shunt interference. Finally, it calculates the intrinsic resistance change rate by calculating the first-order time derivative, and combines this with a preset standard to obtain the control deviation to adjust the input power, thereby accurately reconstructing the evolution of the heat flux density at the reaction interface and achieving precise hedging between input power and thermal field variations.

[0038] At the high-power converter power supply bus of the polysilicon reduction furnace, a fixed high-frequency commutation trigger ripple is generated during the commutation AC process. This ripple frequency is deterministically correlated with the power supply bus frequency and the rectifier bridge topology. To address the glitches and distortions in the original sampling sequence caused by the alternating electromagnetic radiation accompanying the high-power rectifier circuit, and the inability of the static composite resistor to decouple the nonlinear shunt interference from the peripheral silicon shell, the system employs a combination of sensor hardware to acquire the original signal stream. The peripheral power supply circuit of the polysilicon reduction furnace is equipped with a Hall current sensor and a high-frequency electromagnetically isolated voltage transformer. Together with a high-frequency electromagnetic isolation voltage transformer, a physical detection carrier is formed to monitor the real-time voltage and current flowing through the graphite heater. A Hall current sensor is configured at the secondary output copper busbar of the high-current transformer at the bottom of the reduction furnace. Its physical measurement accuracy is not less than 0.1%, and the sampling frequency is set to 100kHz to capture transient current fluctuations caused by thyristor commutation. The high-frequency electromagnetic isolation voltage transformer is connected in parallel across the electrodes of the graphite heater. Its bandwidth is not less than 100kHz, and its measurement accuracy is not less than 0.1%. It is used to collect transient voltage waveforms on the graphite heating element substrate in situ.

[0039] To purify the electrical signal, the system inputs the acquired real-time voltage and current sequences into the microprocessor. The microprocessor forms the hardware carrier for data processing. By configuring a phase-locked loop (PLL) tracking bandpass digital filter before signal decoupling, stray noise is eliminated. The microprocessor calls the Chebyshev bandpass digital filter algorithm based on the monotonic tracking frequency of the PLL. The center frequency of the Chebyshev bandpass digital filter algorithm is monotonically bound to the commutation trigger ripple frequency. For a six-pulse rectifier power supply network, its fundamental frequency is 50Hz, and the center frequency of the thyristor commutation trigger ripple is fixed at 300Hz. The fundamental frequency of the power supply bus is extracted in real time through the PLL, and the center frequency of the bandpass filter is monotonically locked at the sixth harmonic of the fundamental frequency, filtering out transient alternating electromagnetic radiation noise and high-frequency stray harmonics other than the center frequency.

[0040] At the same sampling time The filtered real-time voltage is expressed as Real-time current is expressed as The microprocessor performs the first algebraic division operation to directly calculate the ratio of real-time voltage to real-time current at the same sampling moment, thereby extracting the static composite resistance. The static composite resistance is defined as the apparent composite resistance. Apparent composite resistance It is the composite electrical response of the graphite heater substrate resistance and the parallel shunt conductive network of the surrounding deposited polycrystalline silicon shell connected in parallel. Its calculation formula is shown below: ,in, Sampling time Apparent composite resistance, This is the filtered transient voltage value. The filtered transient current value is shown below. Simultaneously, the microprocessor performs a second-track high-frequency response extraction, separating the high-frequency dynamic differential impedance monotonically bound to the commutation trigger ripple frequency at its center frequency. The high-frequency dynamic differential impedance is expressed as... Because polycrystalline silicon outer semiconductors exhibit resistive characteristics under 300Hz high-frequency ripple excitation due to high-resistance cutoff effect, while graphite heating elements have different lattice complex impedance response characteristics, the high-frequency voltage increment excited by the commutation trigger waveform can be extracted. With high-frequency current incremental Calculate the ratio of the two to obtain the high-frequency dynamic differential impedance. The calculation formula is as follows: ,in, Sampling time High-frequency dynamic differential impedance, This refers to the small voltage increment caused by commutation trigger ripple. This refers to the slight increase in current caused by commutation trigger ripple.

[0041] In a specific embodiment, at a certain sampling moment during the middle to late stage of the polycrystalline silicon vapor deposition reaction. At this point, the Hall current sensor and voltage transformer capture the filtered transient voltage at the 300Hz main frequency. 120.05V, transient current The apparent composite resistance is 2401.10A, calculated by algebraic division of the first rail. It is 0.04999 At the moment of thyristor commutation, a small increase in high-frequency voltage is generated due to the modulation effect of the commutation trigger ripple. The voltage is 1.21V, and the synchronously generated high-frequency current increment is small. The high-frequency dynamic differential impedance is 20.15A, calculated using the second track. It is 0.06004 As the thickness of the gas-deposited polycrystalline silicon shell increases under high-temperature conditions, the carrier concentration rises, causing parallel current shunting to occur in the conductive network of the silicon rod shell, thus increasing the directly measured apparent recombination resistance. To address signal distortion caused by current shunting, which deviates from the true hot resistance of the graphite heater, the microprocessor constructs a spatial differential decoupling channel to calculate the apparent composite resistance. With high-frequency dynamic differential impedance The residual ratio is used to determine the shunt decoupling factor. , decoupling factor The calculation rule is as follows: using high-frequency dynamic differential impedance Subtract apparent composite resistance Obtain the space physics residual, and then divide the space physics residual by the apparent composite resistance. After normalization and dimensionless processing, the algebraic calculation formula is as follows: ,in, Sampling time The decoupling factor of the current split, For high-frequency dynamic differential impedance, For the apparent composite resistance, using the data from the numerical deduction example above, the microprocessor calculates the known apparent composite resistance. It is 0.04999 With high-frequency dynamic differential impedance It is 0.06004 Substituting into the above spatial residual orthogonal decoupling algorithm, the algebraic difference between the two is calculated, yielding a spatial physical residual of 0.01005. Divide the residual by the reference apparent composite resistance of 0.04999. The final calculation output yields the current current decoupling factor. It is 0.2010.

[0042] Extracting the shunt decoupling factor The microprocessor then uses this factor to assess the apparent recombination resistance. In-situ correction and restoration are performed to eliminate semiconductor parallel current shunt interference, and the intrinsic thermal resistance of the heating element lattice is reconstructed. Intrinsic thermal resistance The determination procedure is as follows: The apparent composite resistance... As the dividend, the value 1 is used in conjunction with the shunt decoupling factor. The mathematical expression for division using the algebraic difference of and as the divisor is as follows: ,in, To restore the obtained intrinsic thermal resistance, For apparent composite resistance, This is the decoupling factor for the current split.

[0043] Based on the above measurements, the microprocessor will determine the apparent composite resistance. The value is 0.04999 Decoupling factor Substituting the value of 0.2010 into the formula, the denominator is calculated to be 0.7990. Dividing the two terms, the final output is the intrinsic thermal resistance. It is 0.06256 This in-situ correction process successfully reduced the 0.04999 error caused by the shunting of the polycrystalline silicon shell. The apparent value of the variation, reduced to 0.06256 of the intrinsic value of the graphite heater matrix. The intrinsic hot resistance eliminates measurement interference caused by semiconductor current shunting. Polycrystalline silicon vapor deposition is a strongly thermochemically coupled process. The main challenge within the polycrystalline silicon reduction furnace lies in the nonlinear dynamic drift of the furnace's thermal resistance and radiative heat transfer characteristics as the diameter of the polycrystalline silicon rod continuously increases. To transform the intrinsic hot resistance into a control variable that directly characterizes the state of heat flow at the reaction interface, the rate of change of resistance is calculated. The microprocessor then analyzes the intrinsic hot resistance value. Establish a first-order time differential calculation procedure, and allocate a time axis of length for the microprocessor. A sliding sampling window with 512 data points is used to balance the conflict between sampling frequency and high-frequency noise suppression. The rate of resistance change at the current moment is determined by performing linear regression slope fitting on the historical intrinsic thermal resistance sequence extracted within the sliding sampling window. In the specific calculation, the current sampling moment is selected. intrinsic thermal resistance adjacent sampling time intrinsic thermal resistance Divide the algebraic difference between the two by the fixed sampling period time constant. Thus, the rate of change of intrinsic resistance can be calculated. The formula for calculating its first-order differential is as follows: ,in, The intrinsic resistance change rate, The intrinsic thermal resistance at the current moment. This is the intrinsic thermal resistance value at the previous moment. It represents the time increment between adjacent sampling points, which is determined to be 0.00001s based on the sampling frequency of 100kHz.

[0044] To ensure the rate of change of intrinsic resistance Capable of stably corresponding to the crystallization temperature window at the polycrystalline silicon deposition phase interface, the microprocessor internally stores preset standards, including a preset target resistance change rate curve. This curve is a reference sequence obtained by fitting and solidifying the thermal conductivity parameters under historical stable polycrystalline silicon crystallization conditions. The microprocessor calculates the real-time intrinsic resistance change rate. The difference between the control deviation rate and the preset target resistance change rate is compared, and the control deviation is output. The thyristor power controller acts as the terminal actuator and outputs a reverse trigger phase angle adjustment signal according to the control deviation to dynamically adjust the input power of the converter. When the intrinsic resistance change rate is large, indicating that the interface heat flux density is high, the thyristor conduction angle is reduced to lower the input power; when the change rate is low, the conduction angle is increased to increase the input power. Through the above closed-loop control procedure, the spatial heat transfer thermal resistance variation caused by the coarsening of the silicon rod diameter is offset in situ, and the deposition phase interface temperature is always kept within the preset reaction crystallization temperature range.

[0045] Example 2: In this example, a graphite heater serving as both a heating element and a reaction substrate is installed inside the polycrystalline silicon reduction furnace. During operation, the real-time voltage and current of the graphite heater are synchronously acquired at a sampling frequency of 20kHz, and the commutation trigger ripple frequency is tracked by a digital phase-locked loop. In step S101, a phase-locked tracking bandpass digital filter is used to process the real-time voltage and current, ensuring that the filter's center frequency is monotonically bound to the commutation trigger ripple frequency, thus filtering out stray harmonics outside this frequency. The microprocessor calculates the ratio of the real-time voltage to the real-time current at the same sampling moment to obtain the static composite resistance containing the parallel current shunt data of the polycrystalline silicon shell. Simultaneously, the high-frequency voltage increment and high-frequency current increment caused by commutation trigger ripple are extracted from the filtered voltage and current, respectively, and their ratio is calculated to obtain the high-frequency dynamic differential impedance. Step S102: Calculate the static composite resistance. With high-frequency dynamic differential impedance The difference accounts for the static composite resistance The ratio is used to obtain the shunt decoupling factor. The calculation formula is as follows: ;in, The current sampling time is the decoupling factor; subsequently, the decoupling factor is used... Correcting static composite resistance After subtracting the parallel current shunt data of the polycrystalline silicon shell, the intrinsic thermal resistance value after eliminating current shunt interference is obtained. The calculation formula is as follows: Step S103, for intrinsic hot resistance value The rate of change of intrinsic resistance is obtained by performing first-order time differential calculation. Between adjacent sampling times, the following formula is used for calculation: ;in, This is the intrinsic thermal resistance value at the previous adjacent sampling time. The time increment between adjacent sampling points is determined by a 20kHz sampling frequency; the microprocessor calculates the rate of change of intrinsic resistance. A differential subtraction operation is performed with the preset target resistance change rate to obtain the control deviation, and the trigger conduction angle of the thyristor trigger circuit is adjusted according to the control deviation. When the intrinsic resistance change rate is higher than the target resistance change rate, the trigger conduction angle is reduced to reduce the input power of the graphite heater. When the intrinsic resistance change rate is lower than the target resistance change rate, the trigger conduction angle is increased to increase the input power. Steps S101 to S103 constitute a closed-loop control link, and each step is executed continuously by the microprocessor according to a control cycle of 10ms. By separating the parallel impedance formed by the graphite heater and the polycrystalline silicon shell, the intrinsic hot resistance value for power regulation is obtained, so that the input power is adjusted with the change of heat flux density at the deposition interface, and the deposition phase interface is maintained within the preset reaction crystallization temperature range.

[0046] Example 3: In a polycrystalline silicon vapor deposition furnace, the deposition temperature was maintained at 1050℃ to 1150℃, and the furnace pressure was maintained at 0.5MPa to 0.6MPa. This was used to verify the impact of parallel shunt interference on the resistance measurement and closed-loop control stability of the graphite heater under different polycrystalline silicon shell thickening stages. A graphite heater, serving as both the heating element and the reaction substrate, was installed in the test platform and connected to an external test circuit. The physical measurement accuracy of the Hall current sensor and the high-frequency electromagnetic isolation voltage transformer was 0.05%. The raw transient signals output by both sensors were synchronously input to the microprocessor via an analog-to-digital converter with a sampling frequency of 100kHz, thus establishing an experimental group and a control group. The control group directly obtained the static composite resistance by dividing the filtered real-time voltage and real-time current at the same sampling moment. The test group processed the same set of sampled signals using a dual-track impedance separation method: the static composite resistance of the first track was calculated. The second track extracts the high-frequency voltage and current increments from the 300Hz commutation trigger ripple, and calculates the high-frequency dynamic differential impedance. Then, the decoupling factor is determined based on the ratio of the residuals of the two. And use this factor to correct the static composite resistance. The intrinsic thermal resistance value was obtained. .

[0047] When the diameter of the silicon rod is 60 mm, the static composite resistance is measured. The high-frequency dynamic differential impedance is 57.75mΩ. The current is 60.10 mΩ; the difference between 60.10 mΩ and 57.75 mΩ is divided by 57.75 mΩ to obtain the current shunt decoupling factor. The value is 0.0407; this shunt decoupling factor is then used to correct the 57.75mΩ static composite resistance to obtain the intrinsic hot resistance. The static composite resistance is 60.20 mΩ; when the silicon rod diameter increases to 100 mm, the static composite resistance is... Reduced to 50.13mΩ, high-frequency dynamic differential impedance The current is 60.05 mΩ; the shunt decoupling factor is calculated based on the ratio of the residuals of the two. The intrinsic thermal resistance value is 0.1979, obtained after correction. The static composite resistance is 62.50 mΩ; when the silicon rod diameter reaches 140 mm, the static composite resistance is... 48.00mΩ, high-frequency dynamic differential impedance The current is 60.00mΩ; the difference between the two is 12.00mΩ, and this difference accounts for 0.2500 of the static composite resistance. Therefore, the current shunt decoupling factor is... The corrected intrinsic hot-state resistance is 0.2500. It is 64.00mΩ.

[0048] When the diameter of the silicon rod was further increased to 180 mm, the static composite resistance was measured. The high-frequency dynamic differential impedance is 45.29mΩ. The current is 59.50 mΩ; the shunt decoupling factor is obtained based on the ratio of their residuals. The value is 0.3138, and based on this, the static composite resistance is corrected to an intrinsic hot resistance of 66.00 mΩ. As the diameter of the silicon rod increases from 60mm to 180mm, the uncorrected static composite resistance... The monotonically decreasing value from 57.75 mΩ to 45.29 mΩ indicates that the cross-sectional area of ​​the semiconductor conductive path increases after the polycrystalline silicon shell is thickened, and the influence of parallel current shunting on the bus current is enhanced accordingly; correspondingly, the current shunting decoupling factor... The intrinsic hot resistance value was increased from 0.0407 to 0.3138 after correction. Between 60.20 mΩ and 66.00 mΩ, its change no longer decreases synchronously with the static composite resistance, but rather reflects the change in the hot-state resistance of the graphite heater under high-temperature reaction conditions; to determine the shunt decoupling factor The test platform adjusts the silane gas flow rate into the reduction furnace and the conduction angle of the substrate electrode power supply within the operating range suitable for closed-loop control, and measures the decoupling factor at various points. The restoration accuracy was tested at values ​​of 0.0407, 0.2500, and 0.8500; where 0.0407 corresponds to the absolute lower limit condition, 0.2500 corresponds to the normal median condition, and 0.8500 corresponds to the absolute upper limit condition.

[0049] In the decoupling factor The intrinsic thermal resistance is 0.0407 and 0.2500. The static deviation of the restoration is less than 0.5%; when the shunt decoupling factor When the value reaches 0.8500, the proportion of the polysilicon shell shunt current to the total bus current increases, the incremental commutation trigger ripple signal approaches the range edge of the analog-to-digital converter, the dynamic range is compressed, and the intrinsic hot-state resistance... The static deviation of the restoration increases to 1.2%; when the shunt decoupling factor When the value exceeds 0.9000, the control loop triggers out-of-tolerance protection; the microprocessor continuously calculates the shunt decoupling factor during the synchronous coarsening growth of the polysilicon crystal. And accordingly correct the static composite resistance. To reduce the impact of parallel current shunt in the polysilicon shell on the electrical sampling signal; the intrinsic hot resistance value obtained after correction. Input the first-order time differential calculation unit to obtain the intrinsic resistance change rate. This serves as the data basis for the thyristor power controller to adjust the trigger conduction angle.

[0050] Example 4: When the polysilicon reduction furnace is operating in the middle to late stage of polysilicon rod growth, and a commutation abnormality occurs in the external thyristor rectifier circuit, alternating pulse interference will enter the sampling channels of the Hall current sensor and the high-frequency electromagnetic isolation voltage transformer, causing the static composite resistance extracted at the same sampling moment to be affected. With high-frequency dynamic differential impedance It contains harmonic distortion components; if the intrinsic thermal resistance is directly compared... The rate of change of intrinsic resistance obtained by performing the first-order time derivative Step-like jumps that do not conform to the thermal response characteristics of graphite heaters can easily occur, leading to directional adjustment deviations in the power control loop. To address this issue, the microprocessor employs a smoothing filter procedure based on numerical difference variance, storing 512 continuously acquired intrinsic thermal resistance values ​​prior to the current sampling time into a one-dimensional sliding buffer register to form a sliding sampling window. Under normal conditions, the microprocessor calculates the rate of change of intrinsic resistance based on the intrinsic thermal resistance values ​​at adjacent sampling times. When the algebraic variance of the resistance difference between three consecutive sampling points exceeds the resistance-variable slope deviation threshold... At that time, it is determined that the current sampling sequence is affected by alternating pulse interference, and the intrinsic resistance change rate is... Perform a smoothing correction.

[0051] During smoothing correction, extract the sliding sampling window. Intrinsic thermal resistance at the 512th sampling point The historical intrinsic thermal resistance value corresponding to the first sampling point Calculate the resistance difference across the window and multiply it by the linear correction factor for the slope of the resistor. The corrected intrinsic resistance change rate was obtained. The calculation formula is: Among them, the linear correction coefficient for the resistive transformer slope. It is 1.0500. This represents the time increment between two adjacent sampling points; at a sampling frequency of 100kHz, 0.00001s; resistance slope deviation threshold The maximum physical resistance change slope of the graphite heater under full-power step response is determined, and twice the maximum physical resistance change slope is taken as the safety margin. When the algebraic variance of the resistance difference of the continuously sampled data exceeds the threshold, it indicates that the resistance change has exceeded the change range corresponding to the normal heat conduction process of the graphite heater. Based on this, the microprocessor enables smoothing correction to suppress the pseudo resistance glitches formed by harmonic pulses.

[0052] The thyristor power controller receives the corrected intrinsic resistance change rate. The control deviation is obtained by subtracting it from the preset target resistance change rate curve. Then, the triggering angle of the thyristor triggering circuit is adjusted in the opposite direction according to the positive or negative direction of the control deviation; through the sliding sampling window... After differential correction at the endpoints, the control deviation mainly reflects the change in thermal resistance caused by the increase in the surface area of ​​the polycrystalline silicon rod. Based on this, the power supply converter adjusts the input power of the graphite heater to maintain the temperature of the deposition phase interface between 1050°C and 1150°C. In the adjustment process of this embodiment, the thyristor power controller multiplies the control deviation by 0.60 to obtain the damped correction of the trigger angle offset, and lowers the trigger conduction angle of the thyristor trigger circuit from 90 degrees to 85.5 degrees. According to the correspondence between the effective voltage of the AC rectifier circuit and the trigger conduction angle, after the trigger conduction angle is reduced by 4.5 degrees, the effective AC voltage across the graphite heater electrode is reduced by 2.25%, and the input power is reduced by 4.5%.

[0053] Example 5: This example combines Figures 1 to 2 This describes a process for controlling the thermal field of a graphite heater used in vapor deposition, such as... Figure 1 As shown, the process first acquires electrical signals to obtain real-time voltage and current, then performs phase-locked tracking bandpass digital filtering to filter out stray harmonics. Based on the filtered signals, the static composite resistance containing polysilicon shell shunt data and the high-frequency dynamic differential impedance separating the commutation trigger ripple are calculated in parallel. Then, the shunt decoupling factor is determined according to the impedance difference ratio, and the intrinsic hot resistance is restored to subtract the parallel shunt to eliminate interference. Then, the first-order time derivative is performed to calculate the intrinsic resistance change rate, and then the differential operation with the target change rate is performed to obtain the control deviation. Finally, the trigger conduction angle of the trigger circuit is adjusted according to the control deviation to regulate the input power.

[0054] like Figure 2 As shown, the polycrystalline silicon reduction furnace contains a graphite heater serving as the heating element and reaction substrate, and a polycrystalline silicon shell forming a parallel shunt conductive network. The polycrystalline silicon reduction furnace outputs its real-time voltage and real-time current to a physical detection carrier. The physical detection carrier includes a high-frequency electromagnetic isolation voltage transformer with a bandwidth of not less than 100kHz and a Hall current sensor with an accuracy of not less than 0.1%. The physical detection carrier transmits the acquired raw sampling sequence, i.e., transient electrical signal, to a microprocessor serving as a data processing carrier. The microprocessor is equipped with an analog-to-digital converter with a sampling frequency of 100kHz, a phase-locked tracking bandpass digital filter bound to the sixth harmonic of the fundamental frequency, and a spatial differential decoupling channel for eliminating semiconductor parallel shunt interference. After data processing, the microprocessor outputs a trigger phase angle adjustment signal to the external power supply circuit. The external power supply circuit includes a thyristor trigger circuit serving as the terminal actuator and a high-current transformer with a secondary output copper busbar. The external power supply circuit outputs the corresponding input power to the polycrystalline silicon reduction furnace based on the received signal.

[0055] Example 6: During the hot calibration stage before the polysilicon reduction furnace is started, before the reaction gas is introduced, there is a cold impedance difference in the graphite heater substrate, and the contact resistance of the transformer electrodes will also change with the clamping stress of the furnace. In order to eliminate the influence of the above differences on the subsequent resistance measurement, the microprocessor starts the reference parameter calibration program in the pure hydrogen protection section, controls the converter to output a step alternating excitation signal with a fundamental frequency of 50Hz and a commutation trigger ripple frequency of 300Hz. The real-time voltage and real-time current flowing through the graphite heater, which serves as the heating element and the reaction substrate, are synchronously acquired by a high-frequency electromagnetic isolation voltage transformer and a Hall current sensor, respectively. Based on this, the microprocessor calculates the cold initial composite resistance reference value and adjusts the phase tracking step size of the phase-locked loop to keep the gain valley attenuation coefficient at the center frequency of the bandpass digital filter below -40dB, and establishes the zero-point electrical fingerprint feature library of this batch of graphite heaters under the condition of no polysilicon deposition interference.

[0056] The electrode aging resistance compensation factor is determined based on the cumulative operating time of the graphite heater. A preset offline experimental calibration mapping table records the steady-state drift of the contact resistance of the graphite electrode during continuous operation from 0 to 2000 hours in the no-load hot-state test. A calibration node is set every 500 hours, and the ratio of the steady-state drift of the contact resistance at each node to the initial contact resistance is used as the corresponding electrode aging resistance compensation factor. The microprocessor retrieves the current cumulative operating time, performs linear interpolation between adjacent calibration nodes to obtain the compensation factor corresponding to the current operating condition, and then applies the compensation factor to the static composite electrode according to the steady-state drift of the contact resistance corresponding to this compensation factor. The baseline displacement of the resistance is reversed to eliminate the steady-state drift of the contact resistance caused by high-temperature sintering of the electrode contact surface. During the hot calibration process, the flow rate of pure hydrogen circulating in the furnace is increased at a rate of 50 m³ / h, while the temperature of the core heating zone of the graphite heater is raised from 20℃ to 1050℃. The microprocessor obtains the hot resistance evolution curve of the graphite heater based on the fundamental amplitude response of the real-time voltage and real-time current, and extracts the slope of the resistance change corresponding to every 50℃ during the heating stage. The slope of the resistance change is then algebraically divided with the first-order differential coefficient of the standard graphite thermal resistance to obtain the linear correction coefficient of the resistance change slope under the operating conditions of this batch. Its calibration range is 1.0100 to 1.0600; the current batch takes 1.0500 within the calibration range and inputs it into the discrete time domain smoothing filter algorithm.

[0057] The first-order differential coefficient of the thermal resistance of standard graphite is the first-order rate constant of change of the hot resistivity with increasing temperature of high-purity, high-density graphite material in a non-deposition state. Its value is expressed as a percentage of temperature increase per degree Celsius. The microprocessor divides the slope of the resistance change obtained during the actual heating stage by this constant to correct for the effects of fixture assembly stress and geometric differences in the graphite heater on the resistance temperature characteristics; within the resistance slope deviation threshold... During the calibration process, 0.05 times the initial composite resistance reference value in the cold state was used as the deviation verification benchmark in the hot-state calibration stage. This was combined with the maximum physical resistance change slope under the full-power step response of the graphite heater. Perform calibration; after completing the above parameter calibration, the microprocessor will... and The smoothing filter control parameters are written for use in the polysilicon vapor deposition stage; when no reaction gas is introduced, the power supply converter adjusts the input power according to the calibrated control parameters to make the temperature fluctuation of the reduction furnace converge to 0.5℃.

[0058] Example 7: During polycrystalline silicon vapor deposition, as the diameter of the polycrystalline silicon rod continues to increase, the spatial thermal resistance and radiative heat transfer characteristics inside the reduction furnace undergo nonlinear changes. To address this condition, the power supply control device employs a combined discrete calibration method to determine the nonlinear control gain adjustment matrix corresponding to different shell volume thickening stages. A graphite heater serving as both a heating element and a reaction substrate is installed inside the reduction furnace, and a mixture of trichlorosilane and hydrogen is introduced. The furnace pressure is maintained at 0.55 MPa, and the fundamental frequency of the rectified power grid is 50 Hz. A zero-flux current transformer and a high-frequency electromagnetically isolated voltage transformer synchronously acquire the real-time current and voltage at the power supply bus, with a sampling frequency set to 10 kHz. When the polycrystalline silicon rod growth diameter is 30 mm, the thermal inertia inside the reduction furnace is relatively small. The digital-to-analog conversion channel sends the acquired real-time voltage and current to the microprocessor. The first track calculates the ratio of real-time voltage to real-time current at the same sampling moment to obtain the static composite resistance. The impedance is 0.04212Ω. The second rail uses a phase-locked loop to track the ripple component of the rectified power grid and calls a phase-locked tracking Chebyshev Type II bandpass digital filter to lock the center frequency at 300Hz. The upper and lower cutoff frequencies are set to 320Hz and 280Hz, respectively. From this, the high-frequency voltage increment and high-frequency current increment are separated to obtain the high-frequency dynamic differential impedance. It is 0.04212Ω.

[0059] At this time, the polycrystalline silicon shell has not yet formed a clear parallel shunt channel, and the static composite resistance... With high-frequency dynamic differential impedance Equal to each other, the current splitting decoupling factor is calculated. The intrinsic hot resistance value obtained after correction is 0.0000. The intrinsic resistivity is 0.04212Ω; the microprocessor identifies the current shell volume thickening stage based on the cumulative power consumption of the graphite heater and associates this stage with a growth diameter of 30mm; the gain coefficient of the nonlinear control gain adjustment matrix is ​​configured to 0.35; when the process gas feed flow rate fluctuates, the intrinsic resistivity change rate is... The resistance current is 0.01512 Ω / s, which is below the critical sudden change threshold of 0.02000 Ω / s. The heating power control loop maintains a constant output. The linear correction coefficient for the resistive transformer slope is used in the calculation. The time increment between adjacent sampling points is 1.0500. It is 0.00001s.

[0060] When the cumulative power consumption corresponds to the shell volume thickening stage and matches the 70mm growth diameter, the carrier concentration of the polycrystalline silicon material increases at high temperatures, and parallel current shunt channels are formed in the outer polycrystalline silicon shell; the microprocessor collects the real-time voltage and real-time current at the secondary bus of the transformer and calculates the static composite resistance. The impedance is 0.04999Ω, which is obtained after a 300Hz bandpass digital filter to obtain the high-frequency dynamic differential impedance. The current shunt decoupling factor is calculated using the ratio of the difference between the two to the static composite resistance, which is 0.06004Ω. The value is 0.2010. This factor is then used to correct the static composite resistance, thus obtaining the intrinsic hot resistance of the graphite heater. The gain is 0.06256Ω; at this time, the gain coefficient of the nonlinear control gain adjustment matrix is ​​configured to 1.20 to maintain the sensitivity of the control loop to the control deviation.

[0061] The nonlinear control gain adjustment matrix uses a single-row, three-column gain vector consisting of proportional, integral, and derivative gain terms. Each gain term is retrieved based on the shell volume thickening stage corresponding to the cumulative power consumption. In the initial growth stage corresponding to a diameter less than 30 mm, the proportional gain term is 0.35, the integral gain term is 0.10, and the derivative gain term is 0.05. As the cumulative power consumption increases, the shell volume thickening stage changes from the stage corresponding to 30 mm to the stage corresponding to 110 mm and above. Each gain term in the matrix is ​​amplified synchronously according to an exponential growth trend, so that the response sensitivity of the control deviation increases with the increase of the heat transfer area of ​​the deposition interface. When the shell volume thickening stage corresponding to the cumulative power consumption matches the 110 mm growth diameter, the increase in the volume of the polycrystalline silicon rod causes a heat transfer delay, and the surface area expansion causes radiant heat to be fed back to the graphite heater. At this time, the static composite resistance... The high-frequency dynamic differential impedance is 0.03850Ω. The shunt decoupling factor is calculated to be 0.05580Ω. The intrinsic thermal resistance is 0.4494, which is corrected to obtain the intrinsic thermal resistance value. The gain coefficient of the nonlinear control gain adjustment matrix is ​​set to 2.85, which is 0.06992Ω.

[0062] When changes in process gas flow rate cause transient thermal stress shocks, the microprocessor simultaneously calculates the intrinsic thermal resistance. The second-order time derivative yields the acceleration index of the resistance change. The value is 0.00450 Ω / s²; the intrinsic resistivity change rate was also calculated. The value is 0.02550 Ω / s, and this is subtracted from the preset target resistance change rate to obtain the control deviation. The control deviation is then compared with the resistance change acceleration index. When the algebraic signs are the same and their absolute values ​​exceed the critical threshold changes in control deviation (0.02000 Ω / s) and resistance change acceleration (0.00300 Ω / s²) respectively, the power regulation control loop applies a preset suppression coefficient of 0.60 to limit the adjustment range of the thyristor triggering circuit's triggering angle and lower the triggering angle, thus reducing the input power of the power supply circuit by 4.5%. During this growth stage, the microprocessor also considers the intrinsic hot resistance value... A pre-set axisymmetric physical model of a rotating body is input. Based on the constant density of solid-phase growth and the constraint of axial uniform deposition, the model corresponds the time-domain change of intrinsic thermal resistance to the change of the radial cross-sectional area of ​​the silicon rod, and integrates along the axis to obtain the three-dimensional geometric spatial morphology evolution of the silicon rod in the reduction furnace. Subsequently, the three-dimensional geometric boundary is used as the boundary condition of the steady-state thermal conductivity differential equation. The heat generation power density of the heating body obtained by back-derived from the intrinsic thermal resistance is input into the coupled calculation process of porous medium radiation heat transfer and gas convection heat transfer. The three-dimensional temperature distribution in the furnace and the heat flux density at the deposition phase interface are obtained grid by grid, forming full-field thermal reconstruction information.

[0063] The microprocessor compares the reconstructed deposition phase interface temperature and heat flux density with the preset process range. When the deposition phase interface temperature or heat flux density deviates from this range, it adjusts the flow rate of the trichlorosilane-hydrogen mixture and the furnace pressure according to the direction of deviation, so that the vapor deposition process matches the current three-dimensional geometric morphology. During operation, the microprocessor continuously monitors the intrinsic hot resistance value obtained after correcting the static composite resistance baseline displacement by the electrode aging resistance displacement compensation factor. And accumulate the total power consumption of the graphite heater in real time; when the intrinsic thermal resistance value When the furnace enters the preset steady-state range of 0.07500Ω to 0.07600Ω and the total power consumption reaches the predetermined value corresponding to the target diameter of the polycrystalline silicon rod, the power supply control device stops supplying reaction gas into the reduction furnace and switches the reduction furnace to the cooling stage.

[0064] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A process for controlling the thermal field of a graphite heater used in vapor deposition, characterized in that, Includes the following steps: Step S101: Collect the real-time voltage and real-time current of the graphite heater, which serves as the heating element and reaction substrate, in the polycrystalline silicon reduction furnace. Use a phase-locked tracking bandpass digital filter to filter out stray harmonics other than the commutation trigger ripple frequency. Calculate the static composite resistance containing the parallel shunt data of the polycrystalline silicon shell based on the ratio of the real-time voltage to the real-time current at the same sampling time. Separate the high-frequency dynamic differential impedance that is monotonically bound to the commutation trigger ripple frequency at the center main frequency. Step S102: Determine the shunt decoupling factor based on the ratio of the difference between the static composite resistance and the high-frequency dynamic differential impedance to the static composite resistance. Use the shunt decoupling factor to correct the static composite resistance to remove the parallel shunt data of the polysilicon shell and restore the intrinsic hot resistance value after eliminating the shunt interference. Step S103: Perform first-order time differential calculation on the intrinsic hot resistance value to obtain the intrinsic resistance change rate. Perform differential subtraction between the intrinsic resistance change rate and the preset target resistance change rate to obtain the control deviation. Adjust the trigger conduction angle of the thyristor trigger circuit according to the control deviation to regulate the input power of the graphite heater.

2. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, Step S101, which filters out stray harmonics other than the commutation trigger ripple frequency by phase-locked loop tracking bandpass digital filter, includes: Step S1011, continuously sampling the real-time voltage and real-time current at high frequency to obtain sampling data, and setting the sampling frequency to more than 10 times the commutation trigger ripple frequency; Step S1012, configuring the phase-locked loop to track the fundamental frequency change of the high-power rectifier grid in real time, binding the center frequency of the digital filter to the 6th harmonic of the fundamental frequency, and setting an upper cutoff frequency and a lower cutoff frequency symmetrical about the center frequency to cut off non-stationary random stray harmonics.

3. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, Before step S102, the following steps are also included: Step S1013, retrieve the cumulative running time of the graphite heater and retrieve the electrode aging resistance displacement compensation factor corresponding to the cumulative running time; Step S1014, use the electrode aging resistance displacement compensation factor to correct the baseline displacement of the static composite resistor and eliminate the steady-state drift of the contact resistance of the electrode contact surface.

4. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, Step S103, adjusting the triggering conduction angle of the thyristor triggering circuit according to the control deviation, includes: Step S1031, obtaining the cumulative power consumption of the graphite heater, and identifying the shell volume thickening stage of the polycrystalline silicon rod based on the cumulative power consumption; Step S1032, searching the nonlinear control gain adjustment matrix based on the shell volume thickening stage, using a low gain coefficient of less than 0.5 in the early stage of polycrystalline silicon rod growth, and exponentially increasing the gain coefficient in the middle and late stages of polycrystalline silicon rod growth to adjust the response sensitivity of the control deviation.

5. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, While performing the first-order time derivative calculation to obtain the intrinsic resistance change rate in step S103, the following steps are also included: Step S1033, simultaneously calculate the second-order time derivative of the intrinsic hot-state resistance to obtain the resistance change acceleration index; Step S1034, introduce the resistance change acceleration index as a safety feedback boundary into the control loop to limit the adjustment range of the trigger conduction angle of the thyristor trigger circuit.

6. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 5, characterized in that, After step S1034, the following steps are also included: Step S10341, compare the algebraic signs of the control deviation and the resistance change acceleration index, and determine whether the absolute values ​​of the control deviation and the resistance change acceleration index exceed the set critical change threshold respectively; Step S10342, when the algebraic signs of the control deviation and the resistance change acceleration index are the same and the absolute values ​​both exceed the set critical change threshold, load a preset suppression coefficient in the power regulation control loop to reduce the adjustment range of the heating power.

7. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, Step S103 is followed by the following steps: Step S1035, the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information inside the reduction furnace are inverted based on the intrinsic thermal resistance value; Step S1036, the flow rate and furnace pressure of the mixed gas of trichlorosilane and hydrogen introduced into the reduction furnace are adjusted based on the evolution of the three-dimensional geometric space morphology and the full-field thermal reconstruction information.

8. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 2, characterized in that, The step S1012 of configuring a phase-locked loop to track the fundamental frequency change of the high-power rectifier grid in real time includes: step S10121, monitoring the center frequency fluctuation of the high-power rectifier grid in real time through a frequency sensor; step S10122, when the center frequency fluctuates in the range of 49.5Hz to 50.5Hz, using the phase-locked loop to drive the center frequency of the digital filter for synchronous dynamic adjustment.

9. The process for controlling the thermal field of a graphite heater for vapor deposition according to claim 1, characterized in that, Step S103 is followed by the following steps: Step S1037, continuously monitor the intrinsic hot resistance value after correction by the electrode aging resistance compensation factor, and accumulate the total power consumption of the graphite heater in real time; Step S1038, when the intrinsic hot resistance value enters the set steady-state range and the total power consumption reaches the predetermined value of the target diameter of the corresponding polycrystalline silicon rod, stop the introduction of reaction gas into the reduction furnace, and switch to the cooling stage of the reduction furnace.

Images