Optimization method and optimization system of wafer cleaning and cooling system

Abstract

The invention discloses an optimization method and an optimization system of a wafer cleaning and cooling system in the technical field of information, and the optimization method comprises the steps: manufacturing different fan blade structures and refrigerant nozzle samples according to a numerical simulation result, building a refrigeration system performance test platform, adopting a high-speed thermocouple array and an infrared thermal imager, and carrying out the testing of the performance of the refrigeration system. Wafer surface temperature distribution characteristics at different wafer processing technology temperatures are tested, the optimally designed wafer cooling uniformity control effect is verified, and the numerical model is corrected and perfected; numerical simulation and experimental test results are integrated, a quantitative correlation model among fan structure parameters, refrigerant flow, wafer temperature uniformity and refrigeration system energy efficiency is established, a rapid evaluation tool for wafer cooling system design and model selection is formed, and the design and model selection of the wafer cooling system can be evaluated according to different wafer material attributes and processing temperature requirements. And a fan layout scheme and a refrigerant temperature and flow regulation and control strategy are formulated.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an optimization method and an optimization system for a wafer cleaning and cooling system. Background Art

[0002] Cooling cleaning is a crucial step in the production of semiconductor wafers. This process mainly involves spraying cooling liquid onto the surface of the wafer through a nozzle to achieve rapid cooling and effective cleaning.

[0003] However, there is a significant technical contradiction in the current nozzle structure design, namely the trade-off between cooling uniformity and cooling efficiency. In order to achieve highly uniform cooling of the wafer surface, it is necessary to reasonably design the nozzle structure so that the coolant is ideally distributed on the wafer surface, which requires precise control of the nozzle's spray angle. Improving cooling efficiency requires that the coolant can quickly remove heat from the wafer surface, which places higher demands on the spray speed and droplet size. High-speed and fine droplets are conducive to cooling, but may affect the uniform distribution of droplets. How to balance and optimize the nozzle structure design, improve cooling efficiency while meeting uniformity requirements, and find a suitable combination of spray angle, speed and droplet size parameters is a challenging technical problem. This requires in-depth research on the intrinsic relationship between nozzle structure and spray performance, and through a large number of experiments and simulation optimization, to obtain the best nozzle design that takes into account both cooling uniformity and efficiency. Summary of the invention

[0004] The present invention provides an optimization method for a wafer cleaning cooling system, which mainly comprises the following steps:

[0005] According to the wafer size specifications and material properties, the wafer thermal conductivity coefficient and specific heat capacity values ​​are determined, the wafer thickness uniformity and surface roughness data are obtained, and the wafer temperature field distribution model is established according to the wafer defect density distribution. Based on the temperature field distribution model, the heat flux density homogenization principle is adopted to divide the ideal coolant injection area of ​​the wafer, and the coolant flow required for each area is calculated to obtain the coolant flow distribution plan;

[0006] Using computational fluid dynamics methods, a three-dimensional air duct model including a fan, a guide plate, and a nozzle is constructed based on the coolant injection area and flow distribution scheme obtained in the previous step. The flow and heat transfer characteristics of the refrigerant are simulated in the air duct model to obtain the refrigerant temperature adjustment range and flow control threshold. Through parametric design and multi-objective optimization algorithm, the fan speed control strategy and refrigerant flow control scheme with the best energy efficiency of the refrigeration system are searched under the constraint of wafer temperature uniformity.

[0007] Based on the optimized fan blade geometry and refrigerant control scheme, the large eddy simulation method is used to perform fine numerical simulation of fan airflow pulsation, analyze the influence of airflow pulsation frequency and amplitude on the dynamic temperature distribution of the wafer surface, and evaluate the wafer cooling uniformity control effect of the fan structure design. According to the evaluation results, the refrigeration system operation mode is optimized by adjusting the fan speed control strategy and refrigerant injection angle to obtain the ideal transient temperature distribution on the wafer surface.

[0008] According to the numerical simulation results, different fan blade structures and refrigerant nozzle samples were made, and a refrigeration system performance test platform was built. A high-speed thermocouple array and infrared thermal imager were used to test the wafer surface temperature distribution characteristics under different wafer processing temperatures, verify the wafer cooling uniformity control effect of the optimized design, and modify and improve the numerical model.

[0009] By integrating numerical simulation and experimental test results, a quantitative correlation model between fan structural parameters and refrigerant flow rate and wafer temperature uniformity and refrigeration system energy efficiency is established, forming a rapid evaluation tool for wafer cooling system design and selection. According to different wafer material properties and processing temperature requirements, fan layout plans and refrigerant temperature flow control strategies are formulated;

[0010] Verify the rapid evaluation tool, compare it with experimental data, evaluate the accuracy and scope of application of the tool, optimize and modify the evaluation tool based on the verification results to improve its reliability, use the optimized rapid evaluation tool to make a preliminary selection of the refrigeration system, determine the fan structure and refrigerant control scheme that best suits the wafer cooling needs, build a wafer cooling system prototype based on the selection results, integrate the optimized fan and refrigerant nozzle, conduct wafer cooling performance evaluation tests, test the wafer temperature uniformity control effect of the refrigeration system under different wafer defect density distributions and surface roughness conditions, and optimize the fan noise level and vibration characteristics;

[0011] The prototype test results are compared and analyzed with the previous model predictions, the design method is corrected and improved, and the wafer cooling system design and operation optimization specifications are formed for application in actual wafer cleaning equipment. The wafer surface temperature distribution and the refrigeration system energy efficiency are monitored in real time, and the operation data is collected. According to the data analysis results, the fan speed and refrigerant control mode are adjusted at fixed time intervals (such as every hour or each processing batch). The adjustment is based on changes in wafer surface temperature uniformity deviation, refrigeration system energy consumption indicators and processing technology requirements. Through this regular adjustment mechanism, the refrigeration system performance is continuously improved to ensure long-term stability of wafer cooling uniformity and refrigeration efficiency.

[0012] The present invention provides an optimization system for a wafer cleaning and cooling system. The optimization system is established by an optimization method for a wafer cleaning and cooling system, and mainly includes:

[0013] Wafer temperature field distribution model building module, used to analyze wafer temperature distribution and cooling requirements;

[0014] Refrigerant flow and heat transfer characteristics simulation module, used to optimize coolant injection area and flow distribution;

[0015] Fan structure and refrigerant control scheme optimization module to improve wafer cooling uniformity and system energy efficiency;

[0016] Wafer cooling system performance test and evaluation module, used to verify and correct numerical models;

[0017] A rapid evaluation tool development module for wafer cooling system design and selection to develop the optimal cooling strategy;

[0018] Practical application and continuous optimization module of wafer cooling system to ensure long-term stable cooling effect.

[0019] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0020] The optimization method of the wafer cleaning cooling system of the present invention first establishes a wafer temperature field distribution model, divides the coolant injection area and calculates the flow distribution. Then a three-dimensional air duct model is constructed to simulate the refrigerant flow and heat transfer characteristics, and optimize the fan speed and refrigerant flow control scheme. Large eddy simulation is used to analyze the influence of air flow pulsation on the wafer temperature distribution, and the fan and refrigerant parameters are adjusted to obtain an ideal temperature distribution. The numerical model is verified by experiment, and a correlation model between fan parameters, refrigerant flow and wafer temperature uniformity is established to form a rapid evaluation tool. The tool is used to perform system selection and prototype testing to optimize noise and vibration characteristics. In practical applications, the operating parameters are regularly adjusted according to real-time monitoring data to continuously improve performance. The present invention realizes precise and uniform control of the wafer surface temperature, improves the energy efficiency of the refrigeration system, provides customized cooling solutions for different wafer materials and processing technologies, and effectively guarantees the quality and yield of wafer manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The present invention is a flowchart of the method for optimizing the wafer cleaning cooling system.

[0022] Figure 2 It is a structural schematic diagram of the optimization system of the wafer cleaning and cooling system of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] like Figure 1-Figure 2The optimization system of the wafer cleaning and cooling system of this embodiment may specifically include:

[0025] Step S101, according to the wafer size specifications and material properties, determine the wafer thermal conductivity and specific heat capacity values, obtain the wafer thickness uniformity and surface roughness data, and establish a wafer temperature field distribution model based on the wafer defect density distribution. Based on the temperature field distribution model, the heat flux density homogenization principle is adopted to divide the ideal coolant injection area of ​​the wafer, and the coolant flow required for each area is calculated to obtain the coolant flow distribution plan.

[0026] Step S101 also includes the following steps: obtaining the thermal conductivity and specific heat capacity values ​​of the wafer, and using the measurement data of the wafer thickness uniformity and surface roughness as input parameters of the temperature field distribution model; using the finite element analysis method to establish the wafer temperature field distribution numerical model based on the wafer defect density distribution data; dividing the wafer into a number of ideal coolant injection areas according to the principle of heat flux density uniformity; for each injection area, calculating the required coolant flow rate according to its area and heat flux density; comprehensively considering the coolant flow requirements of each injection area to obtain a coolant flow distribution plan for the entire wafer; inputting the flow distribution plan into the cooling system control module to achieve precise temperature control of the wafer.

[0027] Specifically, when manufacturing semiconductor wafers, thermal management of the wafer is a critical link. First, the thermal conductivity and specific heat capacity of the wafer are the basic physical parameters that determine its thermal properties. For example, the thermal conductivity of a silicon wafer is about 148W / m·K, while the specific heat capacity is about 0.7J / g·K. These parameters can be obtained by consulting the Handbook of Materials Science, and they are crucial to understanding how heat energy propagates during the heating process of the wafer. Next, the measurement data of the thickness uniformity and surface roughness of the wafer are equally important for simulating heat propagation. Assume that there is a slight difference in the thickness of a wafer between the center and the edge, which will affect the distribution of heat in the wafer. For example, if the center is thicker than the edge, it will be more difficult for the heat in the center to dissipate, resulting in uneven temperature. Surface roughness also affects the efficiency of heat exchange, and a wafer with a rougher surface may have better heat dissipation performance due to the increased surface area. After understanding these basic data, the finite element analysis method can be used to establish a numerical model of the temperature field distribution of the wafer. This model can help predict the temperature distribution of the wafer under different process conditions, thereby optimizing the heating and cooling process. For example, simulations can reveal that certain areas of the wafer may overheat during the heating process, which requires adjustments to the heater layout or power output. Based on the principle of heat flux uniformity, the wafer can be divided into several ideal coolant injection areas. The cooling requirements of each area may be different, depending on the heat flux and area of ​​the area. For example, if the heat flux in an area is high, more coolant is required in this area. Through calculations, the required coolant flow rate for each area can be obtained, and then the coolant flow distribution plan for the entire wafer can be obtained. Finally, this flow distribution plan is input into the cooling system control module to achieve precise control of the wafer temperature. This precise control can not only prevent the wafer from being damaged due to excessive temperature, but also ensure the quality and performance of the wafer during the production process. For example, by accurately controlling the coolant flow rate, it is possible to ensure that the wafer has a uniform temperature throughout the production process, thereby improving the consistency and reliability of the product. Through this series of steps, from the acquisition of basic physical parameters to the precise control of the coolant flow rate, the temperature of the wafer during the production process can be effectively managed, which is crucial to improving the performance and reliability of semiconductor equipment.

[0028] Step S102, using computational fluid dynamics methods, a three-dimensional air duct model including a fan, a guide plate, and a nozzle is constructed based on the coolant injection area and flow distribution scheme obtained in the previous step. The flow and heat transfer characteristics of the refrigerant are simulated in the air duct model to obtain the refrigerant temperature adjustment range and flow control threshold. Through parametric design and multi-objective optimization algorithm, the fan speed control strategy and refrigerant flow control scheme with the best energy efficiency of the refrigeration system are searched under the constraint of wafer temperature uniformity.

[0029] According to the physical parameters of the refrigerant and the structural dimensions of the duct, a three-dimensional duct model is established and a grid is divided; the initial values ​​of the fan speed and the refrigerant flow rate are set as boundary conditions for the simulation calculation; the computational fluid dynamics method is used to solve the fluid flow and heat transfer equations inside the duct to obtain the velocity field, pressure field and temperature field distribution inside the duct; according to the temperature distribution on the wafer surface, it is judged whether the wafer temperature meets the uniformity requirement, if not, the fan speed and the refrigerant flow rate are adjusted, and the simulation calculation is repeated; through parametric design, multiple groups of design schemes are automatically generated within the value range of the fan speed and the refrigerant flow rate; a multi-objective optimization algorithm is used, with the energy efficiency of the refrigeration system and the wafer temperature uniformity as the optimization goals, the optimal solution is searched in the design scheme, and the fan speed control strategy and the refrigerant flow control scheme are obtained; the optimized fan speed and refrigerant flow rate are set as the control parameters of the refrigeration system to realize the automatic optimization of the refrigerant temperature adjustment range and the flow control threshold.

[0030] Specifically, during the wafer cooling process, precise temperature control is crucial. In order to achieve precise temperature control of the wafer, the refrigeration system needs to be designed and optimized. First, a three-dimensional duct model needs to be established based on the physical properties of the refrigerant, such as the density, specific heat capacity, thermal conductivity of the refrigerant, and the structural dimensions of the duct. For example, SolidWorks, CATIA, and CAD software can be used to establish a three-dimensional model of the duct. Then, the duct model is imported into the meshing software, such as ICEM CFD and ANSYS Mesh, for meshing. The density of the mesh will affect the accuracy of the calculation results and needs to be selected according to the actual situation. Next, the initial values ​​of the fan speed and refrigerant flow rate are set as the boundary conditions for the simulation calculation. For example, the fan speed can be set to 3000RPM and the refrigerant flow rate can be set to 0.5kg / s. These initial values ​​can be determined based on experience or preliminary calculations. Computational fluid dynamics (CFD) methods, such as CFD software such as Fluent and CFX, are used to solve the fluid flow and heat transfer equations inside the duct to obtain the velocity field, pressure field, and temperature field distribution inside the duct. For example, the velocity at the center of the air duct can be simulated to be 10m / s, the pressure is 101325Pa, and the temperature is 20℃. Then, according to the temperature distribution on the wafer surface, it is judged whether the wafer temperature meets the uniformity requirements. For example, if the standard deviation of the wafer surface temperature is less than 1℃, it is considered to meet the uniformity requirements. If not, it is necessary to adjust the fan speed and refrigerant flow rate and re-simulate the calculation. For example, if the wafer edge temperature is too high, the fan speed can be increased to increase the cooling effect; if the wafer center temperature is too low, the refrigerant flow rate can be reduced to reduce the cooling intensity. In order to find the best fan speed and refrigerant flow combination, a parametric design method can be used. Within the range of fan speed and refrigerant flow rate, multiple groups of design schemes are automatically generated. For example, the fan speed range can be set to 2000RPM to 4000RPM, the refrigerant flow rate range can be set to 0.3kg / s to 0.7kg / s, and multiple groups of combinations can be generated with a certain step size. A multi-objective optimization algorithm is used to search for the optimal solution in the design scheme with the energy efficiency of the refrigeration system and the uniformity of the wafer temperature as the optimization goals. For example, you can use optimization algorithms such as genetic algorithms and particle swarm algorithms to find the fan speed and refrigerant flow combination that makes the refrigeration system most energy efficient and the wafer temperature uniformity best. Assume that the optimization result is a fan speed of 3500RPM and a refrigerant flow of 0.6kg / s. Finally, set the optimized fan speed and refrigerant flow as the control parameters of the refrigeration system to achieve automatic optimization of the refrigerant temperature adjustment range and flow control threshold. Doing so can improve the efficiency of the refrigeration system, reduce energy consumption, ensure the uniformity of wafer temperature, and improve product quality.For example, the optimization results can be written into the control program of the refrigeration system so that the refrigeration system can adjust the fan speed and refrigerant flow rate in real time according to the temperature of the wafer, thereby achieving precise temperature control.

[0031] Step S103, for the optimized fan blade geometry and refrigerant control scheme, a large eddy simulation method is used to perform a fine numerical simulation of the fan airflow pulsation. The influence of the airflow pulsation frequency and amplitude on the dynamic temperature distribution on the wafer surface is analyzed, and the wafer cooling uniformity control effect of the fan structure design is evaluated. According to the evaluation results, the refrigeration system operation mode is optimized by adjusting the fan speed control strategy and the refrigerant injection angle to obtain the ideal transient temperature distribution on the wafer surface.

[0032] According to the optimized fan blade geometry and refrigerant control scheme, a numerical model of fan airflow pulsation is established by using a large eddy simulation method; dynamic temperature distribution data of the wafer surface under different airflow pulsation frequencies and amplitudes are obtained through numerical simulation; based on the temperature distribution data, the temperature uniformity index of the wafer surface is calculated, and the control effect of the fan structure design on the wafer cooling uniformity is evaluated; if the temperature uniformity index does not reach the preset threshold, the fan speed control strategy and the refrigerant injection angle parameter are optimized by using a machine learning algorithm according to the correlation between the airflow pulsation characteristics and the wafer surface temperature distribution; the optimized fan speed control strategy and the refrigerant injection angle parameter are input into the numerical model, and the optimized transient temperature distribution of the wafer surface is obtained through simulation; it is determined whether the optimized wafer surface temperature uniformity index meets the preset conditions, and if so, the optimized refrigeration system operation mode is determined, otherwise the optimization step of the machine learning algorithm is returned to continue the optimization; according to the determined optimal operation mode of the refrigeration system, a dynamic control curve of the fan speed and the refrigerant injection is generated to guide the fan structure manufacturing and the refrigeration system control parameter setting to obtain the ideal cooling uniformity of the wafer surface.

[0033] Specifically, in the wafer manufacturing process, temperature uniformity is critical to the quality of the final product. To achieve the ideal cooling uniformity, the fan and refrigerant injection need to be precisely controlled. The following is a detailed analysis and example of the optimization scheme. First, the large eddy simulation (LES) method is used to model the fan airflow pulsation. Compared with the traditional Reynolds-averaged Navier-Stokes (RANS) method, LES can more accurately capture the transient characteristics of turbulence, especially in the case of high Reynolds numbers, and can more accurately simulate the complex airflow pulsation generated by the fan. For example, assuming that the number of fan blades is 12 and the speed is 5000RPM, LES can simulate the vortex structure and airflow separation phenomenon generated by the rotation of the blades, as well as airflow pulsations of different frequencies and amplitudes. Through LES numerical simulation, the dynamic temperature distribution data of the wafer surface can be obtained. For example, when the time step is 0.001 seconds, the temperature values ​​of each point on the wafer surface at each time step can be recorded to form a time series of temperature field data. Assuming that the wafer diameter is 300 mm, the wafer surface can be divided into a 100x100 grid, and the temperature data is recorded at each grid point. Based on the acquired temperature distribution data, the temperature uniformity index of the wafer surface can be calculated. Commonly used indicators include maximum temperature difference, standard deviation, and root mean square error. For example, the difference between the highest and lowest temperatures on the wafer surface can be calculated as the maximum temperature difference, or the standard deviation of all temperature values ​​can be calculated to evaluate the degree of temperature dispersion. Assuming that the average value of the wafer surface temperature is 25°C, the maximum temperature difference is 2°C, and the standard deviation is 0.5°C. The calculated temperature uniformity index is compared with the preset threshold. If the index does not reach the preset threshold, for example, the maximum temperature difference exceeds 1°C, or the standard deviation exceeds 0.3°C, further optimization is required. The preset threshold is determined according to the specific process requirements. For example, some processes have very high requirements for temperature uniformity, and the maximum temperature difference cannot exceed 0.5°C. In order to optimize the fan speed control strategy and the refrigerant injection angle, a machine learning algorithm can be used. For example, data of different fan speeds, refrigerant injection angles, and wafer surface temperature uniformity indicators can be collected to construct a data set. Then, the model is trained using machine learning algorithms such as support vector machines and neural networks to learn the correlation between airflow pulsation characteristics and wafer surface temperature distribution. Through the trained model, the temperature uniformity of the wafer surface under different parameter combinations can be predicted, and the best parameter combination can be found. For example, the machine learning model may find that the best temperature uniformity can be obtained when the fan speed is adjusted to 4500RPM and the refrigerant injection angle is adjusted to 30 degrees. The optimized parameters are input into the numerical model, and the LES simulation is performed again to obtain the optimized transient temperature distribution of the wafer surface. For example, the optimized fan speed and refrigerant injection angle are used for simulation to obtain new temperature distribution data. The optimized wafer surface temperature uniformity index is calculated again, and it is determined whether it meets the preset conditions.If satisfied, the optimized refrigeration system operation mode is determined. For example, if the optimized maximum temperature difference is less than 1°C and the standard deviation is less than 0.3°C, the preset conditions are considered to be met. Based on the determined optimal operation mode, a dynamic control curve for fan speed and refrigerant injection is generated. For example, the fan speed and refrigerant injection angle can be adjusted in real time according to the change in wafer temperature to maintain the temperature uniformity of the wafer surface. These control curves can be used to guide fan structure manufacturing and refrigeration system control parameter settings. For example, a frequency converter can be designed to control the fan speed, or a servo motor can be designed to control the angle of the refrigerant nozzle, based on the control curve.

[0034] According to the correlation between the airflow pulsation characteristics and the wafer surface temperature distribution, the support vector machine algorithm is used to optimize the fan speed control strategy and the refrigerant injection angle parameters. The optimized transient temperature distribution of the wafer surface is obtained through simulation, and it is judged whether the optimized wafer surface temperature uniformity index meets the preset conditions. If so, the optimized refrigeration system operation mode is determined. Otherwise, the machine learning algorithm optimization step is returned to continue the optimization to obtain the dynamic control curve of the fan speed and refrigerant injection.

[0035] The airflow pulsation characteristic data and the wafer surface temperature distribution data are obtained; the airflow pulsation characteristic data include airflow velocity and pressure parameters; the wafer surface temperature distribution data is obtained by an infrared thermal imager. The airflow pulsation characteristic data and the wafer surface temperature distribution data are analyzed by a multivariate linear regression algorithm, and a correlation model between the airflow pulsation characteristic and the wafer surface temperature distribution is established. The initial values ​​of the fan speed parameter and the refrigerant injection angle parameter are set. The support vector machine algorithm is used to predict the wafer surface temperature distribution based on the established correlation model, the initial value of the fan speed parameter and the initial value of the refrigerant injection angle parameter, and the fan speed parameter and the refrigerant injection angle parameter are adjusted according to the prediction result to obtain a first fan speed parameter and a first refrigerant injection angle parameter. According to the first fan speed parameter and the first refrigerant injection angle parameter, a simulation is performed using computational fluid dynamics software to obtain the transient temperature distribution data of the wafer surface. The wafer surface temperature uniformity index is calculated. A Gaussian process regression algorithm is used to fit the dynamic control curve of the fan speed and the dynamic control curve of the refrigerant injection angle to generate a smooth dynamic control curve.

[0036] Specifically, there are many types of sensors used to obtain data on airflow pulsation characteristics and wafer surface temperature distribution. For example, a hot wire anemometer or a pitot tube can measure airflow velocity, and a pressure sensor can measure airflow pressure. The data from these sensors can be used to construct an airflow pulsation characteristic data set. At the same time, an infrared thermal imager can capture the temperature distribution on the wafer surface and construct a temperature distribution data set. Assuming a wafer with a diameter of 300 mm, the infrared thermal imager can divide its surface into a 100x100 grid, and each grid point records temperature data, such as 25°C, 26°C, etc. The multivariate linear regression algorithm can be used to analyze the association between airflow pulsation characteristics and wafer surface temperature distribution. For example, the higher the fan speed, the faster the airflow speed, and the lower the average temperature of the wafer surface may be. Through multivariate linear regression analysis, a model can be established, such as: temperature = a*speed + b*pressure + c, where a, b, and c are regression coefficients. This model can be used to predict the wafer surface temperature distribution under different airflow pulsation characteristics. Before using the support vector machine algorithm to optimize the fan speed and refrigerant injection angle, the initial value needs to be set. For example, the initial value of the fan speed can be set to 5000RPM, and the initial value of the refrigerant injection angle can be set to 45 degrees. These initial values ​​can be determined based on experience or pre-experimental results. The support vector machine algorithm can predict the temperature distribution on the wafer surface based on the established association model and initial parameters, and adjust the fan speed and refrigerant injection angle parameters. For example, if the predicted wafer edge temperature is too high, you can try to increase the refrigerant injection angle, or increase the fan speed to enhance the cooling effect of the edge. The goal of the support vector machine algorithm is to find the best parameter combination so that the temperature uniformity of the wafer surface reaches the best state. Computational fluid dynamics (CFD) software can be used to simulate the transient temperature distribution on the wafer surface. For example, the optimized fan speed and refrigerant injection angle parameters can be input into the CFD software to simulate the temperature distribution on the wafer surface at different time points. Assuming that the time step is 0.001 seconds, the temperature values ​​of each point on the wafer surface can be recorded at each time step. For example, at 0.001 seconds, the temperature of a point is 25.5℃, and at 0.002 seconds, the temperature of the point is 25.6℃. The temperature uniformity index of the wafer surface can be calculated in many ways, such as the maximum temperature difference and the standard deviation. For example, the difference between the highest and lowest temperatures on the wafer surface can be calculated as the maximum temperature difference. Assuming that the highest temperature on the wafer surface is 26°C and the lowest temperature is 24°C, the maximum temperature difference is 2°C. The calculated temperature uniformity index is compared with the preset threshold. If the index value is less than the preset threshold, for example, the maximum temperature difference is less than 1°C, it is considered that the preset condition is met. The Gaussian process regression algorithm can further optimize the fan speed and refrigerant injection angle parameters optimized by the support vector machine algorithm. For example, assume that the fan speed obtained by the support vector machine algorithm is 4800RPM and the refrigerant injection angle is 48 degrees.The Gaussian process regression algorithm can fit these parameters to generate a smoother dynamic control curve. For example, the fan speed can increase smoothly from 4750RPM to 4850RPM over time instead of abruptly. This can improve control accuracy and stability and avoid oscillation or overshoot in the system.

[0037] Step S104: According to the numerical simulation results, different fan blade structures and refrigerant nozzle samples are manufactured, and a refrigeration system performance test platform is built. High-speed thermocouple arrays and infrared thermal imagers are used to test the wafer surface temperature distribution characteristics at different wafer processing temperatures, verify the wafer cooling uniformity control effect of the optimized design, and correct and improve the numerical model.

[0038] Step S104 also includes the following steps: according to the fan blade structural parameters and the nozzle size, the fan blade and the nozzle are numerically simulated using CFD software to obtain flow field and temperature field distribution data. According to the obtained wafer surface temperature distribution data, the wafer surface temperature uniformity is calculated. If the wafer cooling uniformity does not reach a preset threshold, the parameters of the numerical model are corrected according to the test results, and the parameters include the fan speed and the nozzle flow rate.

[0039] Step S105, combining numerical simulation and experimental test results, establish a quantitative correlation model between fan structural parameters and refrigerant flow rate and wafer temperature uniformity and refrigeration system energy efficiency. Form a rapid evaluation tool for wafer cooling system design and selection, and formulate fan layout plans and refrigerant temperature flow control strategies based on different wafer material properties and processing temperature requirements.

[0040] According to the preprocessed standardized data set, a quantitative relationship model between the fan structural parameters, the refrigerant flow rate, the wafer temperature uniformity, and the refrigeration system energy efficiency is constructed. Based on the quantitative relationship model, a rapid evaluation tool for the wafer cooling system design selection is developed. The wafer material properties and the processing temperature requirements are obtained. A genetic algorithm or a particle swarm algorithm is used, with the wafer temperature uniformity and the refrigeration system energy efficiency as the objective function, to obtain the optimized fan layout scheme and the refrigerant flow control strategy.

[0041] Specifically, in the design process of the wafer cooling system, it is first necessary to collect experimental test data and simulation data, which cover the wafer temperature distribution and system energy consumption under different fan structure parameters and refrigerant flow conditions. For example, by setting different fan speeds and nozzle sizes in the laboratory, recording the temperature changes of the wafer during processing, and using software to simulate the temperature distribution under the same conditions, the difference between reality and theory can be compared. Data preprocessing is a key step, and outliers and erroneous data caused by equipment failure or operational errors need to be eliminated. For example, if the temperature recorded in a certain experiment suddenly jumps to an unreasonable high value, this is likely caused by a thermocouple failure, and such data should be removed from the data set. Next, the data is cleaned and normalized to ensure that data from different sources and batches are compared under the same standard, thereby forming a standardized data set. The construction of a quantitative relationship model is carried out by analyzing the processed data. Multiple regression algorithms or support vector machine algorithms can be used to predict wafer temperature uniformity and system energy efficiency based on fan structure parameters and refrigerant flow. For example, by inputting a specific fan speed and nozzle flow, the model can output the expected wafer temperature distribution uniformity and system energy consumption. Based on these models, a rapid evaluation tool for wafer cooling system design selection can be developed. The tool allows users to input wafer material properties and processing temperature requirements, as well as fan structure parameters and refrigerant flow rates, to quickly obtain predicted wafer temperature uniformity and system energy efficiency. The development of such a tool can greatly improve design efficiency and reduce the need for experimental testing. In actual application scenarios, obtaining wafer material properties and processing temperature requirements is based on specific production requirements. For example, for silicon wafers that need to be processed at high temperatures, their material properties and temperature requirements will be very different from those of glass wafers processed at room temperature. The optimization of fan layout schemes and refrigerant flow control strategies is achieved through genetic algorithms or particle swarm algorithms. These algorithms can simulate the principle of survival of the fittest in the natural selection process and find the optimal fan layout and refrigerant flow configuration through iteration. For example, if the wafer temperature uniformity of the initial design does not meet the preset threshold, the algorithm will automatically adjust the position and number of fans until a new solution that meets the temperature uniformity requirements is found. Finally, the output optimization results will directly affect the performance of the wafer cooling system. The obtained fan layout scheme and refrigerant flow control strategy will be used in actual system design and manufacturing to ensure that the wafers can be maintained at optimal temperature conditions during processing, thereby ensuring product quality and production efficiency.

[0042] Step S106, verify the rapid evaluation tool, and evaluate the accuracy and scope of application of the tool by comparing it with the experimental data. Based on the verification results, optimize and modify the evaluation tool to improve its reliability. Use the optimized rapid evaluation tool to make a preliminary selection of the refrigeration system to determine the fan structure and refrigerant control scheme that best suits the wafer cooling needs. Based on the selection results, build a wafer cooling system prototype and integrate the optimized fan and refrigerant nozzle. Carry out wafer cooling performance evaluation tests, test the wafer temperature uniformity control effect of the refrigeration system under different wafer defect density distributions and surface roughness conditions, and optimize the fan noise level and vibration characteristics.

[0043] Step S107, compare and analyze the prototype test results with the previous model predictions, correct and improve the design method, and form the wafer cooling system design and operation optimization specifications. Applied in actual wafer cleaning equipment, the operation data is collected by real-time monitoring of the wafer surface temperature distribution and the refrigeration system energy efficiency. According to the data analysis results, the fan speed and refrigerant control mode are adjusted at fixed time intervals (such as every hour or each processing batch). The adjustment basis includes the deviation of wafer surface temperature uniformity, the energy consumption index of the refrigeration system, and the changes in the processing technology requirements. Through this regular adjustment mechanism, the performance of the refrigeration system is continuously improved to ensure the long-term stability of the wafer cooling uniformity and refrigeration efficiency.

[0044] Acquire the wafer surface temperature distribution data; calculate the temperature uniformity deviation value according to the wafer surface temperature distribution data; acquire the refrigeration system energy consumption index data and the processing technology requirement parameters; if the temperature uniformity deviation value or the refrigeration system energy consumption index exceeds a preset threshold, then according to the temperature uniformity deviation value, the refrigeration system energy consumption index and the processing technology requirement parameters, adopt a fuzzy control algorithm to obtain the fan speed adjustment amount and the refrigerant control mode adjustment amount; according to the fan speed adjustment amount, adopt a PID control algorithm to obtain the fan speed control signal; according to historical operation data and current process parameters, adopt a machine learning algorithm to establish the wafer temperature prediction model; according to the prediction result of the wafer temperature prediction model, adjust the fan speed and the refrigerant control mode.

[0045] Specifically, temperature control is crucial in the wafer manufacturing process. The uniformity of temperature directly affects the yield and performance of the chip. The following describes how to build an accurate wafer temperature control system. First, it is necessary to obtain the distribution data of the wafer surface temperature. A high-precision platinum resistance temperature sensor array can be used, evenly distributed at different locations on the wafer surface. For example, on a wafer with a diameter of 300mm, 100 sensors can be arranged to form a 10x10 matrix. Each sensor collects temperature data in real time and transmits it to the central control system to form a real-time temperature distribution map. The platinum resistance sensor is chosen because of its high accuracy, good stability, and suitability for high temperature environments. Then, based on the collected temperature data, the wafer temperature uniformity deviation is calculated. For example, the highest temperature measured is 205℃, the lowest temperature is 200℃, and the average temperature is 202℃. The uniformity deviation is (205-200) / 202=0.025, that is, 2.5%. The smaller the uniformity deviation, the more uniform the temperature distribution. At the same time, it is also necessary to monitor the energy consumption indicators of the refrigeration system. Use a power meter to monitor the power consumption of the refrigeration system in real time. For example, the current power consumption is 10kW. This indicator is used to evaluate the energy efficiency of the system and provide a basis for subsequent optimization. Reducing energy consumption can not only save costs, but also reduce the impact on the environment. In addition, it is necessary to obtain the process requirement parameters of the current processing batch from the process database. For example, the target temperature of the current batch is 200℃, and the temperature uniformity tolerance is ±1%. These parameters are set according to the specific process flow and chip type, and are the key to ensuring product quality. Next, adjust the fan speed and refrigerant control mode based on the data analysis results. If the wafer temperature uniformity deviation exceeds the preset threshold (for example, 2%) or the refrigeration system energy consumption index exceeds the preset threshold (for example, 12kW), the fan speed and refrigerant control mode need to be adjusted. The control strategy can use a fuzzy control algorithm. For example, when the uniformity deviation is large, the fan speed can be increased to enhance airflow circulation, thereby improving temperature uniformity. When the energy consumption is too high, the refrigerant flow rate can be reduced to reduce energy consumption. The fuzzy control algorithm can flexibly adjust the control parameters according to the actual situation to achieve more refined control. If the fan speed needs to be adjusted, the PID control algorithm is used for adjustment. The PID control algorithm can automatically adjust the fan speed according to the deviation between the set target speed and the actual speed to stabilize it at the target value. For example, if the fan speed needs to be increased from 1000rpm to 1200rpm, the PID controller will calculate the appropriate control signal based on the difference between the current speed and 1200rpm to drive the fan motor to the target speed. Finally, in order to achieve more advanced control, a predictive model can be used to predict the wafer temperature. Using machine learning algorithms, such as neural networks, the wafer temperature in the future can be predicted based on historical operating data and current process parameters.For example, the wafer temperature in the next hour can be predicted based on the temperature data, fan speed, refrigerant flow rate and other data from the past week, as well as the current process parameters. Based on the prediction results, the fan speed and refrigerant control mode can be adjusted in advance to achieve predictive control, avoid temperature fluctuations, and further improve the accuracy and efficiency of temperature control. Regularly training and updating the model with the latest operating data can improve prediction accuracy and adapt to process changes.

[0046] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for optimizing a wafer cleaning cooling system, characterized in that: The method comprises the following steps: Step S101, according to the wafer size specifications and material properties, determine the wafer thermal conductivity and specific heat capacity values, obtain the wafer thickness uniformity and surface roughness data, establish a wafer temperature field distribution model for the wafer defect density distribution, and divide the wafer into ideal coolant injection areas based on the temperature field distribution model and the heat flux density uniformization principle, and calculate the coolant flow required for each area to obtain a coolant flow distribution plan; Step S102, using computational fluid dynamics methods, based on the coolant injection area and flow distribution scheme obtained in the previous step, construct a three-dimensional air duct model including a fan, a guide plate, and a nozzle, simulate the flow and heat transfer characteristics of the refrigerant in the air duct model, obtain the refrigerant temperature adjustment range and the flow control threshold, and search for the fan speed control strategy and refrigerant flow control scheme with the best energy efficiency of the refrigeration system under the condition of satisfying the wafer temperature uniformity constraint through parameterized design and multi-objective optimization algorithm; Step S103, for the fan blade geometry and refrigerant control scheme obtained by optimization, a large eddy simulation method is used to perform a fine numerical simulation of the fan airflow pulsation, the influence of the airflow pulsation frequency and amplitude on the dynamic temperature distribution on the wafer surface is analyzed, and the wafer cooling uniformity control effect of the fan structure design is evaluated. According to the evaluation results, the refrigeration system operation mode is optimized by adjusting the fan speed control strategy and the refrigerant injection angle to obtain an ideal transient temperature distribution on the wafer surface; Step S104, according to the numerical simulation results, different fan blade structures and refrigerant nozzle samples are manufactured, a refrigeration system performance test platform is built, and a high-speed thermocouple array and an infrared thermal imager are used to test the wafer surface temperature distribution characteristics under different wafer processing temperatures, verify the wafer cooling uniformity control effect of the optimized design, and correct and improve the numerical model; Step S105, combining numerical simulation and experimental test results, establishing a quantitative correlation model between fan structural parameters and refrigerant flow rate and wafer temperature uniformity and refrigeration system energy efficiency, forming a rapid evaluation tool for wafer cooling system design and selection, and formulating fan layout plans and refrigerant temperature flow rate control strategies according to different wafer material properties and processing temperature requirements; Step S106, verifying the rapid evaluation tool, comparing it with the experimental data, evaluating the accuracy and scope of application of the tool, optimizing and correcting the evaluation tool based on the verification results to improve its reliability, using the optimized rapid evaluation tool to preliminarily select the refrigeration system, determining the fan structure and refrigerant control scheme that best suits the wafer cooling requirements, building a wafer cooling system prototype based on the selection results, integrating the optimized fan and refrigerant nozzle, conducting a wafer cooling performance evaluation test, testing the wafer temperature uniformity control effect of the refrigeration system under different wafer defect density distributions and surface roughness conditions, and optimizing the fan noise level and vibration characteristics; Step S107, compare and analyze the prototype test results with the previous model predictions, correct and improve the design method, form the wafer cooling system design and operation optimization specifications, and apply them in actual wafer cleaning equipment. By real-time monitoring of the wafer surface temperature distribution and the refrigeration system energy efficiency, collect operation data, and adjust the fan speed and refrigerant control mode at fixed time intervals based on the data analysis results. The adjustment basis includes changes in wafer surface temperature uniformity deviation, refrigeration system energy consumption indicators and processing technology requirements. Through this regular adjustment mechanism, the refrigeration system performance is continuously improved to ensure long-term stability of wafer cooling uniformity and refrigeration efficiency.

2. The optimization method of the wafer cleaning cooling system according to claim 1, characterized in that: The step S101 also includes the following steps: Obtain the thermal conductivity and specific heat capacity values ​​of the wafer, and use the measurement data of wafer thickness uniformity and surface roughness as input parameters of the temperature field distribution model; According to the wafer defect density distribution data, a finite element analysis method is used to establish a numerical model of the wafer temperature field distribution; According to the heat flux uniformity principle, the wafer is divided into a plurality of ideal coolant spraying areas; For each of the injection regions, the required coolant flow rate is calculated according to the area and heat flux density thereof; Based on the coolant flow requirements of each of the injection areas, a coolant flow distribution plan for the entire wafer is obtained; The flow distribution scheme is input into a cooling system control module to achieve precise temperature control of the wafer.

3. The method for optimizing the wafer cleaning cooling system according to claim 1, characterized in that: The step S102 also includes: According to the physical parameters of the refrigerant and the structural dimensions of the duct, a three-dimensional duct model is established and the grid is divided; Set the initial values ​​of fan speed and refrigerant flow rate as boundary conditions for simulation calculation; Using computational fluid dynamics methods to solve the fluid flow and heat transfer equations inside the air duct, and obtain the velocity field, pressure field and temperature field distribution inside the air duct; According to the temperature distribution on the wafer surface, determine whether the wafer temperature meets the uniformity requirement. If not, adjust the fan speed and refrigerant flow rate and re-perform simulation calculation; Through parametric design, multiple sets of design solutions are automatically generated within the value range of the fan speed and the refrigerant flow rate; A multi-objective optimization algorithm is used to optimize the cooling system energy efficiency and wafer temperature uniformity, and the optimal solution is searched in the design scheme to obtain the fan speed control strategy and the refrigerant flow control scheme; The fan speed and refrigerant flow rate obtained by optimization are set as control parameters of the refrigeration system to achieve automatic optimization of the refrigerant temperature adjustment range and the flow rate control threshold.

4. The method for optimizing the wafer cleaning cooling system according to claim 1, characterized in that: The step S103 also includes: Based on the optimized fan blade geometry and refrigerant control scheme, a numerical model of fan airflow pulsation is established using large eddy simulation method. The dynamic temperature distribution data of the wafer surface under different airflow pulsation frequencies and amplitudes are obtained through numerical simulation; Calculate the temperature uniformity index of the wafer surface based on the temperature distribution data, and evaluate the control effect of the fan structure design on the wafer cooling uniformity; If the temperature uniformity index does not reach the preset threshold, a machine learning algorithm is used to optimize the fan speed control strategy and the refrigerant injection angle parameters according to the correlation between the airflow pulsation characteristics and the wafer surface temperature distribution; The optimized fan speed control strategy and refrigerant injection angle parameters are input into the numerical model, and the optimized transient temperature distribution of the wafer surface is obtained through simulation; Determine whether the optimized wafer surface temperature uniformity index meets the preset conditions, and if so, determine the optimized refrigeration system operation mode, otherwise return to the machine learning algorithm optimization step to continue optimization; According to the determined optimal operation mode of the refrigeration system, a dynamic control curve of the fan speed and refrigerant injection is generated to guide the fan structure manufacturing and the setting of the refrigeration system control parameters to obtain the ideal cooling uniformity on the wafer surface; It also includes: according to the correlation between the airflow pulsation characteristics and the wafer surface temperature distribution, the support vector machine algorithm is used to optimize the fan speed control strategy and the refrigerant injection angle parameters, and the optimized wafer surface transient temperature distribution is obtained through simulation, and it is judged whether the optimized wafer surface temperature uniformity index meets the preset conditions. If so, the optimized refrigeration system operation mode is determined, otherwise the machine learning algorithm optimization step is returned to continue the optimization, and the dynamic control curve of the fan speed and refrigerant injection is obtained.

5. The method for optimizing the wafer cleaning cooling system according to claim 4, characterized in that: According to the correlation law between the airflow pulsation characteristics and the wafer surface temperature distribution, the support vector machine algorithm is used to optimize the fan speed control strategy and the refrigerant injection angle parameter, and the optimized wafer surface transient temperature distribution is obtained through simulation, and it is judged whether the optimized wafer surface temperature uniformity index meets the preset conditions. If it meets the preset conditions, the optimized refrigeration system operation mode is determined, otherwise the machine learning algorithm optimization step is returned to continue the optimization, and the dynamic control curve of the fan speed and the refrigerant injection is obtained, including the following steps: Acquiring the airflow pulsation characteristic data and the wafer surface temperature distribution data; The airflow pulsation characteristic data includes airflow velocity and pressure parameters; The wafer surface temperature distribution data is obtained by an infrared thermal imager; The airflow pulsation characteristic data and the wafer surface temperature distribution data are analyzed by using a multivariate linear regression algorithm to establish a correlation model between the airflow pulsation characteristic and the wafer surface temperature distribution; Setting initial values ​​of the fan speed parameter and the refrigerant injection angle parameter; Using a support vector machine algorithm, based on the established association model, the initial value of the fan speed parameter, and the initial value of the refrigerant injection angle parameter, predict the wafer surface temperature distribution, and adjust the fan speed parameter and the refrigerant injection angle parameter according to the prediction result to obtain a first fan speed parameter and a first refrigerant injection angle parameter; According to the first fan speed parameter and the first refrigerant injection angle parameter, a simulation is performed using computational fluid dynamics software to obtain transient temperature distribution data on the wafer surface; Calculating the wafer surface temperature uniformity index; A Gaussian process regression algorithm is used to fit the dynamic control curve of the fan speed and the dynamic control curve of the refrigerant injection angle to generate a smooth dynamic control curve.

6. The method for optimizing a wafer cleaning cooling system according to any one of claims 1 to 5, characterized in that: The step S104 also includes the following steps: According to the fan blade structural parameters and the nozzle size, CFD software is used to perform numerical simulation on the fan blade and the nozzle to obtain flow field and temperature field distribution data; Calculating the temperature uniformity of the wafer surface according to the acquired wafer surface temperature distribution data; If the wafer cooling uniformity does not reach a preset threshold, the parameters of the numerical model are corrected according to the test results, and the parameters include the fan speed and the nozzle flow rate.

7. An optimization system for a wafer cleaning cooling system, characterized in that: The optimization system is established by the optimization method of the wafer cleaning and cooling system according to any one of claims 1 to 6, and the system comprises: Wafer temperature field distribution model building module, used to analyze wafer temperature distribution and cooling requirements; Refrigerant flow and heat transfer characteristics simulation module, used to optimize coolant injection area and flow distribution; Fan structure and refrigerant control scheme optimization module to improve wafer cooling uniformity and system energy efficiency; Wafer cooling system performance test and evaluation module, used to verify and correct numerical models; A rapid evaluation tool development module for wafer cooling system design and selection to develop the optimal cooling strategy; Practical application and continuous optimization module of wafer cooling system to ensure long-term stable cooling effect.

Images