A method for controlling the constant diameter drawing of a single crystal silicon rod based on the estimation of the silicon melt level
Patent Information
- Application Number
- CN202611192074.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-07
- Publication Date
- 2026-09-08
AI Technical Summary
[0004]本发明提供一种基于硅熔体液位估计的单晶硅棒等径拉制控制方法，以解决在单晶硅棒等径拉制过程中，硅熔体液位通常难以直接测量，多依赖操作人员观察、经验判断或通过晶体直径、提拉速度、加热功率等间接参数推测熔体状态，导致液位变化无法被连续、准确获取；同时，单一参数反馈方式容易受到热场波动、硅熔体热膨胀、加料变化以及设备差异等因素影响，使液位估计结果存在偏差，难以及时反映真实熔体供需状态；此外，缺少针对不同单晶炉运行特性的自适应修正机制，无法有效消除长期运行过程中的测量漂移和环境扰动影响，导致等径拉制过程中晶体直径稳定性不足，影响单晶硅棒生产质量和过程控制精度的技术问题
[0017]1.通过利用硅料输入质量与单晶硅棒生长消耗质量之间的质量守恒关系，将不可直接测量的液位变化转换为可计算的体积变化，并结合坩埚几何结构完成液位初始估计，避免了传统方法中液位测量困难导致的状态判断滞后问题。
Smart Images
Figure CN122707239A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of liquid level measurement technology, and in particular to a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt liquid level estimation. Background Technology
[0002] Single-crystal silicon rods are a crucial basic material in semiconductor manufacturing. Their dimensional consistency, crystal integrity, and internal quality directly affect subsequent wafer processing and device manufacturing performance. In the production of single-crystal silicon rods, the Czochralski method is typically used for crystal growth. By controlling the silicon melt state, the crystal pulling process, and the thermal field distribution, a stable constant-diameter growth state is maintained for the single-crystal silicon rod. The silicon melt level, as a critical parameter reflecting the melt supply state and the crystal growth balance, is closely related to crystal diameter changes, thermal field stability, and the crystal growth process. Therefore, during the constant-diameter pulling process of single-crystal silicon rods, it is necessary to continuously monitor the changes in the silicon melt level to provide a basis for production process analysis and adjustment. However, due to the high temperature, enclosed environment, and strong thermal radiation inside the single-crystal furnace, the silicon melt cannot be directly detected using conventional contact level measurement methods. Therefore, it is necessary to indirectly estimate the silicon melt level state by combining equipment operating parameters and multi-source detection information to meet the requirements of continuous and stable production of single-crystal silicon rods. As the semiconductor industry continues to demand higher dimensional accuracy and production stability for monocrystalline silicon materials, establishing a measurement method that can adapt to complex thermal environments and achieve dynamic sensing of silicon melt level is of great significance for improving the stability of the constant diameter pulling process of monocrystalline silicon rods.
[0003] Currently, in the constant-diameter pulling process of monocrystalline silicon rods, the silicon melt level is usually difficult to measure directly. It often relies on operator observation, experience, or inference of the melt state through indirect parameters such as crystal diameter, pulling speed, and heating power. This results in the inability to continuously and accurately obtain the level changes. At the same time, the single-parameter feedback method is easily affected by factors such as thermal field fluctuations, thermal expansion of silicon melt, changes in feeding, and equipment differences, which can lead to deviations in the level estimation results and make it difficult to reflect the true melt supply and demand status in a timely manner. In addition, there is a lack of adaptive correction mechanisms for the operating characteristics of different monocrystalline furnaces, which cannot effectively eliminate the measurement drift and environmental disturbances during long-term operation. This results in insufficient crystal diameter stability during the constant-diameter pulling process, affecting the production quality and process control accuracy of monocrystalline silicon rods. Summary of the Invention
[0004] This invention provides a method for controlling the constant-diameter pulling of monocrystalline silicon rods based on silicon melt level estimation. This addresses the challenges of directly measuring the silicon melt level during the constant-diameter pulling process, which often relies on operator observation, experience, or inference of the melt state using indirect parameters such as crystal diameter, pulling speed, and heating power. This results in inconsistent and inaccurate acquisition of melt level changes. Furthermore, the single-parameter feedback method is susceptible to fluctuations in the thermal field, thermal expansion of the silicon melt, changes in feed rate, and equipment differences, leading to deviations in the level estimation results and making it difficult to reflect the true melt supply and demand status in a timely manner. Additionally, the lack of an adaptive correction mechanism for different monocrystalline furnace operating characteristics prevents the effective elimination of measurement drift and environmental disturbances during long-term operation, resulting in insufficient crystal diameter stability during constant-diameter pulling and impacting the production quality and process control accuracy of monocrystalline silicon rods.
[0005] The present invention provides a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation, which specifically includes the following steps:
[0006] S1. Simultaneously collect multi-source data during the operation of the single crystal furnace, including silicon material input mass, crucible sidewall temperature, free liquid surface position, single crystal silicon rod pulling speed, and crystal diameter; based on the multi-source data, calculate the mass output of the single crystal silicon rod and obtain the mass change of the silicon melt; convert the mass change of the silicon melt into the volume change of the silicon melt, and combine it with the final silicon melt liquid level estimate at the previous sampling time to obtain the initial estimate of the silicon melt liquid level;
[0007] S2. Introduce the historical final silicon melt level estimate and the historical silicon melt free surface position to calculate the level estimation error in the historical stable constant diameter drawing process; by analyzing the influence of temperature change and free surface position deviation on the level estimation error, calculate the temperature correction coefficient and the free surface position correction coefficient.
[0008] S3. Based on the temperature correction coefficient and the free surface position correction coefficient, and combined with the liquid level estimation error, the initial estimated value of the silicon melt liquid level is corrected to obtain the final estimated value of the silicon melt liquid level; based on the final estimated value of the silicon melt liquid level, the change in silicon melt liquid level is calculated, and the stable state value of the silicon melt liquid level is obtained; based on the stable state value of the silicon melt liquid level, the stable state of the liquid level at the current sampling time is determined.
[0009] Preferably, a sampling time interval is introduced, and the volume of the single-crystal silicon rod is calculated by combining the pulling speed and crystal diameter of the single-crystal silicon rod; the volume of the single-crystal silicon rod is converted into a mass output quantity by combining the single-crystal silicon density, and the mass change of the silicon melt is calculated; based on the mass change of the silicon melt, the silicon melt density is introduced to calculate the volume change of the silicon melt.
[0010] Preferably, the inner diameter of the crucible is introduced, the liquid level change value corresponding to the change in the volume of the silicon melt is calculated, and combined with the final estimated value of the silicon melt liquid level at the previous sampling time, the initial estimated value of the silicon melt liquid level at the current sampling time is obtained.
[0011] Preferably, a historical stable constant diameter drawing window is introduced, and a reference temperature is constructed based on the crucible sidewall temperature within the historical stable constant diameter drawing window; the temperature normalization change is obtained based on the crucible sidewall temperature at the current sampling time and the reference temperature; and the free surface position normalization deviation is obtained based on the free surface position of the silicon melt at the current sampling time and the initial estimated value of the silicon melt level.
[0012] Preferably, based on the liquid level estimation error, historical temperature normalization change, and historical free surface position normalization deviation corresponding to historical sampling points within the historical stable constant diameter drawing window, a liquid level correction model is constructed by introducing a temperature correction coefficient and a free surface position correction coefficient to be fitted, and the least squares method is used for fitting to obtain the temperature correction coefficient and the free surface position correction coefficient.
[0013] Preferably, based on the normalized temperature change and the normalized deviation of the free liquid surface position at the current sampling time, and combined with the temperature correction coefficient and the free liquid surface position correction coefficient, the liquid level correction amount at the current sampling time is calculated; based on the liquid level correction amount, the initial estimate of the silicon melt liquid level is corrected to obtain the final estimate of the silicon melt liquid level.
[0014] Preferably, based on the final estimated silicon melt level, the real-time silicon melt level change corresponding to the current sampling time and the historical silicon melt level change corresponding to the historical sampling time within the historical stable constant diameter drawing window are calculated respectively; based on the historical silicon melt level change within the historical stable constant diameter drawing window, the standard deviation of the silicon melt level change is calculated.
[0015] Preferably, based on the standard deviation of silicon melt level change, and combined with the real-time silicon melt level change and the historical silicon melt level change, the current silicon melt level stability value and the historical silicon melt level stability value are constructed respectively; based on the historical silicon melt level stability value, a level stability threshold is set and compared with the current silicon melt level stability value to determine the level stability at the current sampling time.
[0016] The beneficial effects of the technical solution of the present invention are:
[0017] 1. By utilizing the mass conservation relationship between the input mass of silicon material and the mass consumed in the growth of single-crystal silicon rods, the liquid level change that cannot be directly measured is converted into a calculable volume change. Combined with the geometry of the crucible, the initial liquid level estimation is completed, avoiding the problem of delayed state judgment caused by the difficulty in liquid level measurement in traditional methods.
[0018] 2. By introducing temperature correction and free liquid surface position constraints, the impact of temperature fluctuations, thermal expansion, and detection errors on the liquid level estimation results can be reduced, making the liquid level estimation results more consistent with the actual operating state of the single crystal furnace.
[0019] 3. By introducing historical stable constant diameter drawing data, the temperature correction coefficient and free liquid surface position correction coefficient are obtained by fitting using the least squares method, realizing the adaptive update of the liquid level correction model coefficients; at the same time, the standard deviation of silicon melt liquid level change and the liquid level stability threshold are established by statistically analyzing the historical silicon melt liquid level changes, which serve as liquid level stability criteria, thus improving the adaptability and engineering application scope of the method.
[0020] 4. By comparing the current stable value of the silicon melt level with the stable threshold value, abnormal changes in the silicon melt level can be identified in a timely manner, providing reliable data for the stable control of the diameter of monocrystalline silicon rods, reducing crystal diameter fluctuations caused by changes in melt supply, and improving the stability of the monocrystalline silicon rod production process and product quality. Attached Figure Description
[0021] Figure 1 This is a flowchart of a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation, as described in this invention. Detailed Implementation
[0022] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0024] The following description, in conjunction with the accompanying drawings, details a specific scheme for a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation, provided by the present invention.
[0025] See attached document Figure 1 The diagram illustrates a flowchart of a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation, according to an embodiment of the present invention. The method includes the following steps:
[0026] S1. Simultaneously collect multi-source data during the operation of the single crystal furnace, including silicon material input mass, crucible sidewall temperature, free liquid surface position, single crystal silicon rod pulling speed, and crystal diameter; based on the multi-source data, calculate the mass output of the single crystal silicon rod and obtain the mass change of the silicon melt; convert the mass change of the silicon melt into the volume change of the silicon melt, and combine it with the final silicon melt liquid level estimate at the previous sampling time to obtain the initial estimate of the silicon melt liquid level;
[0027] First, a data acquisition system for measuring the silicon melt level is established. Using the height of the free surface of the silicon melt within the single-crystal furnace as the measured level, multi-source data is collected during the furnace's operation. An indirect measurement model for the silicon melt level, suitable for high-temperature environments where direct contact is not permitted, is then established to achieve continuous measurement and estimation of the free surface height of the silicon melt. The multi-source data includes: silicon input mass, crucible sidewall temperature, free surface position, single-crystal silicon rod pulling speed, and crystal diameter.
[0028] In practical implementation, a mass acquisition unit, a temperature acquisition unit, a free liquid surface position acquisition unit, and a process parameter acquisition unit are deployed in the single crystal furnace equipment. The mass acquisition unit includes a weighing sensor installed on the feeding mechanism to acquire the mass of silicon material entering the single crystal furnace, i.e., the input mass of silicon material; the temperature acquisition unit includes a temperature sensor installed around the crucible to acquire the temperature of the crucible sidewall; the free liquid surface position acquisition unit includes a visual inspection device, such as an infrared camera, installed at the furnace observation window to acquire the spatial position of the free liquid surface of the silicon melt in the acquired image of the silicon melt surface, i.e., the vertical height of the free liquid surface of the silicon melt relative to the bottom of the crucible; the process parameter acquisition unit includes a lifting mechanism position encoder and a crystal diameter detection device, used to acquire the single crystal silicon rod lifting speed based on displacement differential and to acquire the crystal diameter based on image edge recognition, respectively.
[0029] All the above-mentioned acquisition units are connected to a unified time base to synchronously acquire data. Let the time interval between two adjacent sampling times be . , It is obtained directly from the timestamp difference of the acquisition system. Let the sampling sequence number in the current liquid level estimation process be . , ,in, Let the number of sampling points in the current liquid level estimation process be denoted as the number of sampling points in the current process. Each sampling time is , take two adjacent sampling times and The time period between is defined as the first Each sampling period is denoted as . ,in The process quantities accumulated during this sampling period are denoted by the following subscript: Mark, and at time Complete the calculation and storage; store the multi-source data collected by each acquisition unit into the silicon melt level database according to a unified time base;
[0030] Furthermore, based on the mass change relationship of the silicon melt within the single crystal furnace, the mass change of the silicon melt during the current sampling period is calculated. The specific calculation process is as follows:
[0031] Let the first The mass of silicon material entering the single crystal furnace at each sampling time was: The unit is kg. The data is directly collected by the weighing sensor installed on the feeding mechanism; let the first... The mass output of the single-crystal silicon rod formed within each sampling period is The unit is kg. It is calculated based on the geometric relationship of crystal growth. Specifically, let the first... The crystal diameter at each sampling time is The unit is m, in the first... The pulling speed of the single-crystal silicon rod at each sampling time is The unit is m / s, and the density of single-crystal silicon is... The unit is kg / m³. The physical property parameters are obtained from the material property parameter table of the silicon material currently used in production. The material property parameter table is a material specification document provided by the silicon single crystal manufacturer.
[0032] In the At each sampling moment, based on the cross-sectional area of the single-crystal silicon rod, the pulling speed, and the sampling time interval... Calculate the volume of the monocrystalline silicon rod formed within the corresponding time period, and then combine it with the monocrystalline silicon density to convert it into the mass output of the monocrystalline silicon rod: , in, Pi; Indicates the first Cross-sectional area of a single-crystal silicon rod at each sampling time; Indicates by the first From the sampling time to the first sampling time The axial growth length of the single-crystal silicon rod during each sampling time; Indicates by the first From the sampling time to the first sampling time The volume of the single-crystal silicon rod formed during each sampling time.
[0033] According to the The input mass of silicon material and the output mass of the single crystal silicon rod at the sampling time are calculated. Change in the mass of the silicon melt at each sampling time : , in, The unit is kg.
[0034] Furthermore, the change in mass of the silicon melt is converted into a change in volume of the silicon melt based on the silicon melt density. Let the first... The density of the silicon melt at each sampling time is The unit is kg / m³, and the density value of silicon melt is based on the first... Crucible sidewall temperature obtained at each sampling time The temperature-density relationship of silicon melt was determined by consulting a pre-defined table of silicon melt temperature-density parameters, which is a well-known table of silicon material properties in the field. Based on the relationship between mass and volume, the first... Change in the volume of the silicon melt at each sampling time : , in, The unit is m³.
[0035] In obtaining the volume change of the silicon melt Then, based on the geometric constraints of the crucible space, the initial estimate of the silicon melt level is calculated. The specific implementation process is as follows: the inner diameter of the crucible is obtained in advance through the structural parameter table of the single crystal furnace equipment provided by the equipment manufacturer. and crucible height All units are in meters (m). Since the effective cross-sectional area of the crucible changes little during constant-diameter drawing, the change in the volume of the silicon melt can be directly divided by the area of the circle corresponding to the inner diameter of the crucible top. The change in liquid level in the cylindrical space is calculated using a linear approximation.
[0036] Specifically, starting from the estimation of the silicon melt level, i.e. First, calculate the estimated value of the silicon melt level at the initial sampling time; assume the initial charge mass is... The unit is kg. The density of the silicon melt is obtained from a weighing sensor after the loading is completed; let the initial density of the silicon melt be... The unit is kg / m³, which is the initial density of the silicon melt. Using numerical values known in the art, as a specific embodiment, the melt density of silicon at its melting point (approximately 1410°C) can be taken here, specifically 2520 kg / m³, which can be found in the "Handbook of Physicochemical Properties of Silicon Materials"; the specific calculation formula for the estimated silicon melt level at the initial sampling time is as follows: , in, This represents the estimated level of the molten silicon at the initial sampling time, in meters (m).
[0037] In the At each sampling time, based on the final estimated value of the silicon melt level obtained at the previous sampling time... and the change in the volume of the silicon melt during the current sampling period. Calculate the first Initial estimate of the silicon melt level at each sampling time : , in, Indicates the first The initial estimated value of the silicon melt level at each sampling time is obtained based on mass conservation and crucible geometry, in meters. It indicates that it has already reached the 1st The final estimated value of the silicon melt level calculated at each sampling time; Indicates the first The liquid level change value corresponding to the change in volume of the silicon melt within each sampling period.
[0038] S2. Introduce the historical final silicon melt level estimate and the historical silicon melt free surface position to calculate the level estimation error in the historical stable constant diameter drawing process; by analyzing the influence of temperature change and free surface position deviation on the level estimation error, calculate the temperature correction coefficient and the free surface position correction coefficient.
[0039] During actual single-crystal furnace operation, temperature changes in the silicon melt lead to variations in liquid density and volume. Furthermore, visual inspection, mass measurement, and geometric conversion processes all introduce liquid level measurement errors. Therefore, if calculations are performed solely based on mass conservation and geometric relationships, the resulting initial liquid level estimate will inevitably deviate from the true liquid level, failing to yield a completely accurate measurement result. To address this, this invention introduces a multi-factor error correction process after obtaining the initial estimate of the silicon melt level to compensate for the impact of the aforementioned liquid measurement errors on measurement accuracy.
[0040] First, obtain the number Crucible sidewall temperature at each sampling time The unit is ℃. Temperature is directly obtained from the temperature acquisition unit. A historical stable constant diameter drawing window is introduced, and the average value of the crucible sidewall temperature within the historical stable constant diameter drawing window is set as the reference temperature. The unit is °C. The historical stable constant diameter pulling window is defined as follows: during continuous operation of the single crystal furnace, the absolute value of the deviation of the measured crystal diameter from the target diameter (e.g., 218-220 mm) does not exceed ±1 mm, and no active adjustment of major process parameters such as heating power, pulling speed, or crucible lifting is made for at least 30 consecutive minutes. The start time of this window is the moment when all the above conditions are met simultaneously for the first time, and the end time is the moment when any one of the conditions is no longer met for the first time.
[0041] Calculate the change between the current crucible sidewall temperature and the reference temperature, and normalize the temperature change to obtain the normalized temperature change: , in, For the first Normalized temperature change at each sampling time; To prevent the denominator from being zero, a small constant value can be taken here. .
[0042] Simultaneously, the free liquid surface position acquisition unit of the silicon melt acquires the first... Position of the free surface of the silicon melt at each sampling time The unit is meters (m), representing the position of the free surface of the molten silicon. This indicates the vertical height of the free surface of the silicon melt relative to the bottom of the crucible, as obtained by the visual inspection device, and does not indicate the position of the solid-liquid interface.
[0043] Based on the position of the free liquid surface of the silicon melt Initial estimate of silicon melt level Calculate the normalized deviation of the free surface position based on the deviation between them: , in, Indicates the first The normalized deviation of the free surface position at each sampling time. Normalization is performed so that the deviation of the free liquid surface position in single crystal furnaces of different sizes can participate in the subsequent liquid level correction calculation on a uniform scale.
[0044] The calculated normalized temperature change and the normalized deviation of the free liquid surface position are stored in the silicon melt level database.
[0045] Furthermore, the historical stable constant diameter drawing window was extracted from the silicon melt level database. The historical final silicon melt level estimate and the historical silicon melt free surface position corresponding to each historical sampling time are denoted as follows: and The unit is meters. This represents the historical sampling sequence number within the historical stable isopathic pull window, where , This indicates the number of sampling points included within the historical stable isochronous drawing window. It should be noted that... Real-time sampling index in the current liquid level estimation process These belong to two different data indexing systems. Used to indicate the real-time sampling time during the current operation of the single crystal furnace. Used to represent the historical sampling time in a historical stable constant diameter drawing window.
[0046] Furthermore, based on the historical final silicon melt level estimate and the historical silicon melt free surface position, the level estimation error within the historical stable constant diameter drawing window is calculated: , in, Indicates the first isochronous pull within the historical stable isochronous pull window The liquid level estimation error corresponding to each historical sampling time is expressed in meters (m).
[0047] Furthermore, based on the liquid level estimation error, the normalized change in historical temperature, and the normalized deviation in historical free surface position corresponding to all historical sampling points within the historical stable constant diameter drawing window, a liquid level correction model is established: ,in, , These represent the temperature correction factor and the free surface position correction factor, respectively. and They represent the first The normalized variation of historical temperature and the normalized deviation of historical free surface position at each historical sampling time are calculated. Specifically, the least squares fitting method is used to minimize the deviation between the output of the liquid level correction model and the liquid level estimation error, in order to calculate the temperature correction coefficient. and free surface position correction factor : , in, This represents the liquid level correction factor corresponding to the temperature change, i.e., the temperature correction factor, with the unit being meters (m). This represents the liquid level correction factor corresponding to the deviation of the free liquid surface position, i.e., the free liquid surface position correction factor, with the unit being meters (m). Indicates the first isochronous pull within the historical stable isochronous pull window The normalized change in historical temperature at each historical sampling moment is a dimensionless quantity. Indicates the first isochronous pull within the historical stable isochronous pull window The normalized deviation of the historical free liquid surface position at each historical sampling moment; This indicates the number of historical sampling points within the historical stable isochronous pull window.
[0048] When the single crystal furnace equipment is replaced or the thermal field structure changes, it is necessary to re-accumulate a period of historical stable constant diameter drawing window data after the new thermal field structure has been operating stably, and extract this data from the silicon melt level database to recalculate. and This allows for the parameter update of the liquid level correction model.
[0049] S3. Based on the temperature correction coefficient and the free surface position correction coefficient, and combined with the liquid level estimation error, the initial estimated value of the silicon melt liquid level is corrected to obtain the final estimated value of the silicon melt liquid level; based on the final estimated value of the silicon melt liquid level, the change in silicon melt liquid level is calculated, and the stable state value of the silicon melt liquid level is obtained; based on the stable state value of the silicon melt liquid level, the stable state of the liquid level at the current sampling time is determined.
[0050] During the current operation of the single crystal furnace, the temperature correction coefficient and free liquid surface position correction coefficient obtained in step S2 are used to adjust the first... Normalized temperature change at each sampling time and the normalization deviation of the free liquid surface position Substitute into the liquid level correction model, calculate the first Liquid level correction at each sampling time: , in, Indicates the first The liquid level correction amount corresponding to each sampling time is expressed in meters (m).
[0051] Subsequently, according to the The initial estimate of the silicon melt level calculated at each sampling time. and liquid level correction amount The final estimated value of the silicon melt level is calculated and stored in the silicon melt level database. , in, Indicates the first The final estimated value of the silicon melt level at each sampling time, in meters.
[0052] The calculation obtained in this step This will serve as the reference value for the recursive calculation of the liquid level in step S1 at the next sampling time, i.e., the value in step S1. With this step These are the final estimated values of the silicon melt level at consecutive moments in the same recursive sequence, and the two constitute a complete recursive update relationship for level estimation.
[0053] After continuously obtaining the final estimated silicon melt level at multiple sampling moments during the current operation, the stability of the current silicon melt level change is further analyzed. First, based on the current... The sampling time and the first The change in silicon melt level is calculated based on the estimated final silicon melt level at each sampling time. , in, Indicates the first The change in silicon melt level at each sampling time is expressed in meters (m).
[0054] To obtain the range of silicon melt level variation under normal operating conditions, the historical silicon melt level variation is calculated based on the estimated historical silicon melt level within the historical stable constant diameter drawing window: , in, Indicates the first isochronous pull within the historical stable isochronous pull window The change in silicon melt level at each historical sampling moment is expressed in meters. Indicates the first isochronous pull within the historical stable isochronous pull window The estimated value of the silicon melt level at each historical sampling time.
[0055] The standard deviation of the silicon melt level change is calculated based on all historical changes in silicon melt level within the historical stable constant diameter drawing window: , in, This represents the standard deviation of the silicon melt level variation within a historically stable constant-diameter drawing window, expressed in meters (m). This represents the average change in the silicon melt level within a historically stable constant-diameter drawing window, expressed in meters (m).
[0056] Through the above statistical process, the normal fluctuation range of silicon melt level changes under historically stable constant diameter drawing conditions was obtained.
[0057] Furthermore, based on the relationship between the current change in silicon molten liquid level during operation and the fluctuation range of silicon molten liquid level within the historical stable constant diameter drawing window, the current [number]th [stage] is calculated. Steady state values of silicon melt level at each sampling time: , in, Indicates the first The steady-state value of the silicon melt level at each sampling time is a dimensionless quantity.
[0058] Simultaneously, based on the historical stable constant-diameter drawing window's silicon melt level variation, the historical stable state value of the silicon melt level is calculated: , in, Indicates the first isochronous pull within the historical stable isochronous pull window The stable state value of the silicon melt level at each historical sampling moment.
[0059] Furthermore, the historical stable silicon melt level stability values within the historical stable constant diameter drawing window were analyzed. Sort the values in ascending order and select the 95th percentile as the threshold for stable liquid level. .
[0060] When the current number Each sampling time satisfies At that time, it was considered that the current change in the silicon melt level was within the historical stable measurement range, the level was stable, and the constant diameter drawing process was operating normally; when If the current change in the silicon melt level deviates from the historical stable measurement range, the level is abnormal, and the constant diameter drawing process may be unstable, requiring intervention.
[0061] In summary, a method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation has been completed.
[0062] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0063] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
[0064] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation, characterized in that, Specifically, the following steps are included: S1. Simultaneously collect multi-source data during the operation of the single crystal furnace, including silicon material input mass, crucible sidewall temperature, free liquid surface position, single crystal silicon rod pulling speed, and crystal diameter; based on the multi-source data, calculate the mass output of the single crystal silicon rod and obtain the mass change of the silicon melt; convert the mass change of the silicon melt into the volume change of the silicon melt, and combine it with the final silicon melt liquid level estimate at the previous sampling time to obtain the initial estimate of the silicon melt liquid level; S2. Introduce the historical final silicon melt level estimate and the historical silicon melt free surface position to calculate the level estimation error in the historical stable constant diameter drawing process; by analyzing the influence of temperature change and free surface position deviation on the level estimation error, calculate the temperature correction coefficient and the free surface position correction coefficient. S3. Based on the temperature correction coefficient and the free surface position correction coefficient, and combined with the liquid level estimation error, the initial estimated value of the silicon melt liquid level is corrected to obtain the final estimated value of the silicon melt liquid level; based on the final estimated value of the silicon melt liquid level, the change in silicon melt liquid level is calculated, and the stable state value of the silicon melt liquid level is obtained; based on the stable state value of the silicon melt liquid level, the stable state of the liquid level at the current sampling time is determined.
2. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 1, characterized in that, By introducing a sampling time interval and combining the pulling speed and crystal diameter of the single-crystal silicon rod, the volume of the single-crystal silicon rod is calculated; by combining the single-crystal silicon density, the volume of the single-crystal silicon rod is converted into a mass output quantity, and the mass change of the silicon melt is calculated; based on the mass change of the silicon melt, the silicon melt density is introduced to calculate the volume change of the silicon melt.
3. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 2, characterized in that, By introducing the inner diameter of the crucible, the liquid level change value corresponding to the change in silicon melt volume is calculated. Combined with the final estimated value of silicon melt liquid level at the previous sampling time, the initial estimated value of silicon melt liquid level at the current sampling time is obtained.
4. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 3, characterized in that, A historically stable constant diameter drawing window is introduced, and a reference temperature is constructed based on the crucible sidewall temperature within the historically stable constant diameter drawing window. Based on the crucible sidewall temperature at the current sampling time, combined with the reference temperature, the normalized temperature change is obtained. Based on the free surface position of the silicon melt at the current sampling time and the initial estimated value of the silicon melt level, the normalized deviation of the free surface position is obtained.
5. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 4, characterized in that, Based on the liquid level estimation error, historical temperature normalization change, and historical free surface position normalization deviation corresponding to historical sampling points within the historical stable constant diameter drawing window, a liquid level correction model is constructed by introducing the temperature correction coefficient and free surface position correction coefficient to be fitted, and the least squares method is used for fitting to obtain the temperature correction coefficient and free surface position correction coefficient.
6. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 5, characterized in that, Based on the normalized temperature change and the normalized deviation of the free liquid surface position at the current sampling time, and combined with the temperature correction coefficient and the free liquid surface position correction coefficient, the liquid level correction amount at the current sampling time is calculated. Based on the liquid level correction amount, the initial estimate of the silicon melt liquid level is corrected to obtain the final estimate of the silicon melt liquid level.
7. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 6, characterized in that, Based on the final estimated silicon melt level, calculate the real-time silicon melt level change at the current sampling time and the historical silicon melt level change at the historical sampling time within the historical stable constant diameter drawing window; and calculate the standard deviation of the silicon melt level change based on the historical silicon melt level change within the historical stable constant diameter drawing window.
8. The method for controlling the constant diameter pulling of single-crystal silicon rods based on silicon melt level estimation according to claim 7, characterized in that, Based on the standard deviation of silicon melt level change, and combined with the real-time silicon melt level change and the historical silicon melt level change, the current silicon melt level stability value and the historical silicon melt level stability value are constructed respectively. Based on the historical silicon melt level stability value, a level stability threshold is set and compared with the current silicon melt level stability value to determine the level stability at the current sampling time.