Semiconductor process high-precision flow control system and control method thereof

Abstract

The present application is directed to the gas flow control problem of silicon nitride film deposition in semiconductor process, and provides a high-precision flow control system and method. In terms of system, three gas inlet pipes transport different gases, and the three-way pipe is used for staged mixing to reduce the risk of explosion; the second three-way pipe T-shaped design uses gravity to assist mixing; valves and sensors are arranged in each gas inlet pipe to provide comprehensive data support; the special design of the mixer realizes sufficient mixing of the gases; the specific gas flow gradient cooperates with the multi-stage mixing process to realize accurate control of the stoichiometric ratio; and the laser detector provides high-precision data at a specific angle and with advanced detection technology. The method includes the steps of gas pipeline distribution and pre-mixing, multi-parameter fusion measurement, adaptive control, dynamic calibration, and mixer optimization and output. The present application improves the quality of silicon nitride film deposition, improves the safety and flow control precision of the system, and reduces the production cost and safety risk.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to a high-precision flow control system and control method for semiconductor manufacturing processes. Background Technology

[0002] In the semiconductor manufacturing field, silicon nitride (Si3N4) thin films play a crucial role. They possess several important functions, including acting as a passivation layer to effectively protect the wafer surface from contaminants such as moisture and ions, thereby significantly improving the reliability of semiconductor devices and extending their lifespan.

[0003] Currently, chemical vapor deposition (CVD) is a commonly used method for preparing silicon nitride thin films, among which plasma-enhanced chemical vapor deposition (PECVD) and low-pressure chemical vapor deposition (LPCVD) are two main process types.

[0004] When depositing silicon nitride thin films using CVD technology, a mixture of SiH4 (20%), NH3 (30%), and N2 (50%) is typically used as the reaction source. These three gases each play an important role in the film deposition process, but also impose strict requirements on flow control.

[0005] Meanwhile, when SiH4 mixes with air and reaches a certain concentration range, it will undergo a violent oxidation reaction upon exposure to open flames, high temperatures, or strong oxidants, potentially leading to an explosion. When SiH4 is mixed with NH3, and both concentrations are within a specific range, the reactivity is significantly enhanced, the reaction rate accelerates dramatically, releasing a large amount of energy, which can then cause an explosion. Therefore, strict requirements are placed on gas flow control and delivery methods.

[0006] In summary, while the CVD process for depositing silicon nitride thin films using a mixture of SiH4, NH3, and N2 gases has significant application value in semiconductor manufacturing, it presents numerous challenges in gas flow control. These challenges directly impact the quality and performance of the silicon nitride thin films, thereby limiting the overall quality and reliability of semiconductor devices. Therefore, developing a method or device capable of effectively and precisely controlling gas flow is of great practical significance for improving the deposition quality of silicon nitride thin films and promoting the development of semiconductor manufacturing technology.

[0007] Therefore, we propose a high-precision flow control system and its control method for semiconductor manufacturing processes. Summary of the Invention

[0008] Therefore, it is necessary to address the technical problems of low flow control accuracy and uneven mixing of SiH4, NH3, and N2 mixed gases during silicon nitride thin film deposition in semiconductor manufacturing processes, as well as the limitations of traditional detection methods, which make it difficult to meet the requirements of high-precision and high-stability processes and pose an explosion risk. This paper proposes a high-precision flow control system and its control method for semiconductor manufacturing processes, which can significantly improve the quality of silicon nitride thin film deposition, enhance the overall quality and reliability of semiconductor devices, reduce the risk of explosion, and ensure production safety.

[0009] The first aspect of this invention provides a high-precision flow control system for semiconductor manufacturing processes, comprising a first inlet pipe, a second inlet pipe, a third inlet pipe, a first three-way connector, a second three-way connector, a mixer, an outlet pipe, and a laser detector. Different gases are delivered through the three inlet pipes, and the first and second three-way connectors are used to achieve staged mixing of the gases, effectively reducing the risk of explosion caused by direct mixing of SiH4 and NH3, and improving system safety. The mixer's special design, including a mixing section and a turbulence section, enables thorough mixing of the gases, ensuring uniform gas composition and guaranteeing uniform deposition of silicon nitride thin films. The laser detector allows for real-time monitoring of gas flow rate and velocity, providing precise data support for flow control and facilitating high-precision control of gas flow.

[0010] In other embodiments, the second tee is T-shaped, with two ports on the same straight line connecting to the first inlet pipe and the mixer, respectively. The first inlet pipe is located above the first tee pipe. This design utilizes the density differences between the different gases to achieve initial mixing with gravity assistance, reducing energy consumption and lessening the burden on subsequent mixing stages. Simultaneously, NH3 is injected vertically from above, creating shear force with the horizontally flowing N2-SiH4 mixture, enhancing turbulence intensity, rapidly breaking down gas stratification, reducing concentration gradients, and ensuring high homogeneity of the gas before entering the mixer, providing a foundation for subsequent precise control. Furthermore, the position of the upper inlet pipe prevents the N2-SiH4 mixture from flowing back into the NH3 pipe, ensuring unidirectional gas flow, reducing the risk of accidental mixing within the pipe, and meeting the stringent safety requirements of semiconductor processes.

[0011] In other embodiments, corresponding valves are installed on the first, second, and third air inlet pipes. Temperature sensors and flow controllers are installed on both the first and second air inlet pipes, and pressure sensors are installed on both the first air inlet pipe and the first tee pipe. These sensors enable real-time monitoring of gas parameters such as temperature, pressure, and flow rate, providing comprehensive data support for flow control. The flow controller can adjust the gas flow rate promptly based on the monitored data, ensuring the gas flow rate remains stable near the set value. The combined use of temperature and pressure sensors can also take into account the influence of temperature and pressure on gas flow rate, further improving the accuracy of flow control.

[0012] In other embodiments, the mixer includes a housing and a drive shaft axially arranged along the housing axis. Mixing blades are provided in the region where the drive shaft extends into the mixing section, and turbulence blades are provided in the region where the drive shaft extends into the turbulence section. A rotary drive device is also provided on the mixer to drive the drive shaft to rotate. The rotary drive device drives the drive shaft to rotate, causing the mixing blades and turbulence blades to stir and agitate the gas. The mixing blades can initially stir and mix the various gases entering the mixer, breaking down the stratification between gases and achieving preliminary macroscopic mixing. The turbulence blades can create strong turbulence in the turbulence section, increasing the contact area and collision frequency between gas molecules, achieving deep mixing at the microscopic level, and making the composition of the mixed gas more uniform. Simultaneously, the turbulence blades can further adjust and stabilize the gas, making the gas velocity distribution at the outlet more uniform and the pressure more stable.

[0013] In other embodiments, the gas flow rates of the second, first, and third inlet pipes decrease sequentially. This flow rate design is based on the principles of chemical reaction kinetics. SiH4, as a highly reactive gas, has its flow rate set to the lowest value to ensure sufficient dilution by N2 in the initial mixing stage, reducing the risk of spontaneous combustion. NH3, as a reactant, has a moderate flow rate to provide sufficient nitrogen for the formation of silicon nitride compounds. N2, as both a carrier gas and a diluent, has the highest flow rate, ensuring both gas delivery power and maintaining the concentration of flammable components in the mixture below the lower explosive limit (LEL). Before entering the mixer, the three gases undergo preliminary premixing in a static mixer with an internal helical blade structure, creating three-dimensional turbulence and achieving a mixing uniformity of over 95%. This flow rate gradient design, combined with a multi-stage mixing process, achieves precise control of the stoichiometric ratio (SiH4:NH3:N2 molar ratio controlled at 1:2:10), ensuring a high yield of the target product (silicon nitride) while reducing the explosion risk through dilution, significantly improving process safety.

[0014] In other embodiments, the angle between the laser detector and the gas outlet pipe is 30°-60°. This angle range is determined based on Mie scattering theory and an optical path difference optimization model. When the laser beam forms a 45° angle with the gas flow direction, it ensures sufficient optical path length (L = D / sinθ, where D is the pipe diameter) to allow gas molecules and photons to interact fully, while avoiding specular reflection interference that may occur when the incident light is perpendicular (angle > 60°). The detector employs tunable diode laser absorption spectroscopy (TDLAS) technology. Its laser emission module integrates a distributed feedback (DFB) laser, with the wavelength locked at the characteristic absorption peak of SiH4 (1693nm), a linewidth < 2MHz, and a tuning range of ±1cm. -1 This optimized design, combined with advanced detection technology, enables the system to achieve a detection limit of 0.1 ppm for SiH4 concentration and a response time of <100 ms, providing high-precision real-time data support for process control.

[0015] In other embodiments, the reactivity of the third gas, first gas, and second gas decreases sequentially, and the risk of explosion is reduced by mixing the second gas with the third gas. SiH4 (the third gas), as a typical spontaneously combustible gas, has a reactivity index as high as 9.5 (0-10 level) and a self-ignition temperature of only 25°C in air; NH3 (the first gas) has a reactivity index of 6.2 and exhibits weak reducing properties; N2 (the second gas), as an inert gas, has a reactivity index of 0. This embodiment employs an "inert dilution-gradient mixing" safety strategy. First, highly reactive SiH4 is initially mixed with a large amount of N2 (volume ratio 1:20) at the end of the third inlet pipe, utilizing the chemical inertness of N2 to inhibit the self-decomposition reaction of SiH4; subsequently, this mixture is secondary mixed with NH3 at the first three-way pipe. At this point, the SiH4 concentration has dropped to below 5% of the lower explosive limit (LEL), making it impossible to maintain a combustion chain reaction even if local temperatures rise or an ignition source is present. This dual protection system, combining gas reactivity gradient control with a physical flame arrestor, reduces the probability of explosion for the entire system.

[0016] A second aspect of this invention provides a high-precision flow control method for semiconductor processes, utilizing the aforementioned high-precision flow control system for semiconductor processes. The method includes steps such as gas pipeline distribution and premixing, multi-parameter fusion measurement, adaptive control, dynamic calibration, and mixer optimization and output. In the gas pipeline distribution and premixing step, safe mixing and initial uniform distribution of the gas are achieved through reasonable gas distribution and premixing design. N2, as an inert diluent, dilutes the SiH4 concentration to below 80% of its lower explosive limit (LEL = 1.4%) during the premixing stage; in the main mixing stage, N2 further dilutes the NH3 concentration to below 85% of its LEL (15%), effectively reducing the risk of explosion. The multi-parameter fusion measurement step collects data such as temperature, pressure, and flow rate, and uses a Kalman filter to suppress noise in the data. Then, the actual flow rate value after multi-parameter fusion is calculated, providing accurate data for flow control. The adaptive control step utilizes a deep learning-based flow prediction unit that loads historical process data, inputs current pressure and temperature data, and outputs predicted flow rate. Dynamic PID control is employed to adjust the flow controller's signal in real-time based on the difference between the target setpoint and the actual measured flow rate, bringing the actual flow rate closer to the target value and improving the system's adaptability and control accuracy. The dynamic calibration step combines the advantages of laser interferometry monitoring and quantum tunneling correction, resulting in a more accurate final corrected flow rate value. Based on the deviation between the corrected flow rate and the target value, the flow controller dynamically adjusts the flow rate in the first inlet pipe for precise control. The mixer optimization and output step uses a rotary drive to rotate the drive shaft, ensuring thorough mixing of the gas by the mixing and turbulence blades, guaranteeing uniform gas distribution. Based on the deviation between the corrected flow rate and the target value, the valve opening of the first inlet pipe is dynamically adjusted to meet the gas flow accuracy requirements in semiconductor manufacturing. This method, through the synergistic effect of multiple steps, achieves high-precision control of gas flow rate in semiconductor manufacturing, improving the quality and stability of silicon nitride thin film deposition while reducing production costs and safety risks.

[0017] In other embodiments, the formula for calculating Q1 is:

[0018] Q1=(Q MFC ×(1+k P ·ΔP)×(1+k T ·ΔT)) / (1+k P ·ΔP+k T ·ΔT)

[0019] Among them, Q MFC The initial measurement value for the mass flow controller; k P k tHere, ΔP and ΔT are pressure and temperature correction factors, respectively, and the pressure and temperature deviations are the values. This formula considers the influence of pressure and temperature on gas flow rate. By introducing correction factors and deviation values, it corrects the initial measurement value of the mass flow controller, improving the accuracy of flow measurement. In practical applications, changes in pressure and temperature affect gas density, thus affecting gas flow rate. This formula can compensate for these factors, making the measurement results closer to the true value and providing reliable data support for flow control.

[0020] In other embodiments, a mixer optimization and output step is also included. A rotary drive device rotates the drive shaft, mixing blades (located in the mixing section) perform initial mixing, and baffle blades (located in the baffle section) perform secondary agitation to ensure uniform gas distribution. The optimized mixer design can further improve the gas mixing effect, making the gas composition more uniform. The combination of initial mixing and secondary agitation can fully utilize the functions of the mixing blades and baffle blades, achieving thorough mixing of the gas at both the macroscopic and microscopic levels. Simultaneously, based on the deviation between the corrected flow rate and the target value, the flow rate of the first inlet pipe is dynamically adjusted by the flow controller, which can promptly correct flow rate deviations and ensure that the gas flow rate remains stable near the target value, meeting the stringent gas flow rate requirements of semiconductor manufacturing processes. Attached Figure Description

[0021] Figure 1 This is a three-dimensional structural diagram of the present invention.

[0022] Figure 2 This is a side view of the present invention.

[0023] Figure 3 This is a side view of the present invention.

[0024] Figure 4 for Figure 3 A sectional view of section AA in the middle.

[0025] Figure 5 This is a flowchart of the control method of the present invention.

[0026] in:

[0027] 100. First intake pipe; 101. First control valve; 200. Second intake pipe; 201. Second control valve; 300. Third intake pipe; 301. Third control valve; 400. First tee pipe; 500. Second tee pipe; 600. Mixer; 601. Mixing section; 602. Turbulence section; 603. Rotary drive device; 604. Drive shaft; 605. Mixing blades; 606. Turbulence blades; 610. Inlet; 611. Outlet; 700. Outlet pipe; 800. Laser detector; 900. Temperature sensor; 1000. Pressure sensor; 1100. Flow controller; 1200. Mounting base. Detailed Implementation

[0028] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0029] Example 1

[0030] like Figures 1-4 As shown, this embodiment discloses a flow control system for semiconductor manufacturing processes, including: a mixer 600, a rotary drive device 603, a first inlet pipe 100, a second inlet pipe 200, a third inlet pipe 300, a laser detector 800, a temperature sensor 900, a pressure sensor 1000, a flow controller 1100, a mounting base 1200, and various pipe valves. This system enables thorough mixing and precise control of the output of N2, SiH4, and NH3. Through multi-parameter fusion measurement and adaptive control technology, the system achieves high-precision control of the flow rates of the three gases, ensuring the uniformity and quality stability of the silicon nitride thin film. The special design of the mixer 600, combined with dynamic calibration using laser detection and quantum tunneling effects, further enhances the reliability and safety of the system.

[0031] like Figure 1 As shown, specifically, in this embodiment, the first air inlet pipe 100 is used to transport the first gas, which is NH3 in this embodiment; a first control valve 101 is provided on the first air inlet pipe 100.

[0032] The second air inlet pipe 200 is used to transport a second gas, which is N2 in this embodiment; a second control valve 201 is provided on the second air inlet pipe 200.

[0033] The third air inlet pipe 300 is used to transport a third gas, which is SiH4 in this embodiment; a third control valve 301 is provided on the third air inlet pipe 300.

[0034] In other embodiments, if the three gases being transported also meet the characteristics of the three gases in this embodiment, the flow control system of this embodiment can also be used for transport. At the same time, if the gas ratio changes, it can also be adjusted by adjusting the valves of the first inlet pipe 100, the second inlet pipe 200 and the third inlet pipe 300, or the flow controller 1100 on the corresponding pipeline. Therefore, this embodiment can be applied to a wide range of gas transport.

[0035] In this process, SiH4 serves as the silicon source, decomposing into silicon atoms (Si) under plasma or thermal action. The flow rate control of SiH4 is crucial, directly affecting the silicon content in the thin film. Improper SiH4 flow rate control, failing to precisely match the NH3 flow rate to maintain the stoichiometric ratio (Si:N≈3:4), will cause the thin film composition to deviate from the ideal Si3N4 composition, thus affecting the film's performance and quality. Furthermore, SiH4 is highly reactive and temperature-sensitive, easily exploding at high concentrations.

[0036] NH3 provides the nitrogen source, decomposing into nitrogen atoms (N) during the reaction to participate in the formation of silicon nitride thin films. Precise control of the NH3 flow rate is also crucial. Insufficient NH3 flow rate leads to nitrogen deficiency in the film, resulting in a silicon-rich phase (such as Si3N3). This silicon-rich phase exhibits poor insulation properties, reducing the functional effectiveness of the silicon nitride film as a dielectric and passivation layer. Excessive NH3 flow rate, on the other hand, may introduce hydrogen impurities. The presence of hydrogen impurities can adversely affect the electrical properties and stability of the film, such as increasing leakage current, and also increasing the risk of explosion.

[0037] Nitrogen (N2) constitutes 50% of the gas mixture, primarily serving as a dilution and carrier gas. It stabilizes the reaction atmosphere, preventing excessively high localized reactive gas concentrations and avoiding particle formation due to uneven concentration, thus ensuring the surface smoothness and quality of the thin film. Controlling the N2 flow rate affects the diffusion rate of the reactive gas within the chamber. An inappropriate flow rate leads to uneven distribution of the reactive gas within the chamber, causing differences in deposition rate and thickness at different locations, hindering uniform film deposition and severely impacting the performance consistency and reliability of semiconductor devices. Simultaneously, N2's inertness allows it to effectively dilute flammable gases, reducing the risk of explosion, and minimizing the impact of corrosive gases.

[0038] like Figure 1 and Figure 3 As shown, in this embodiment, the first three-way pipe 400 is connected to both the second intake pipe 200 and the third intake pipe 300, enabling the mixed transport of the second and third gases. Simultaneously, the second three-way pipe 500 is connected to both the first three-way pipe 400 and the first intake pipe 100, enabling the mixed transport of the first, second, and third gases. This design has the following advantages:

[0039] The concentration of SiH4 is diluted with N2 to prevent an explosion caused by excessively high SiH4 concentration. Furthermore, the SiH4 diluted with N2 will also avoid explosion when it combines with NH3.

[0040] As described in the background section, mixing SiH4 and NH3 can increase the risk of explosion; therefore, in related technologies, SiH4 and NH3 are transported separately.

[0041] However, in this embodiment, by premixing the second intake pipe 200 and the third intake pipe 300, unlike related technologies that use three separate pipes to transport N2, SiH4, and NH3 and then mix them simultaneously, there is still a significant risk of explosion since N2 does not dilute SiH4 and NH3. By diluting SiH4 with N2, the risk of explosion during SiH4 transport is reduced, as is the risk of explosion after SiH4 is mixed with NH3. This improves the overall safety performance. The premixing design in this embodiment significantly improves the safety of the system and simplifies the pipe layout.

[0042] like Figure 4 As shown, the mixer 600 in this embodiment is used to fully mix three gases. The mixer 600 has an air inlet 610 at one axial end, which is connected to the air outlet of the second three-way pipe 500 to facilitate the transport of mixed gas. The other end has an air outlet 611 on its radial upper surface. The mixer 600 includes a mixing section 601 and a turbulence section 602 for mixing and agitating the gases.

[0043] Specifically, the mixer 600 in this embodiment includes a housing and a drive shaft 604 axially arranged along the axis of the housing. The drive shaft 604 extends to the mixing section 601 and is provided with mixing blades 605. The drive shaft 604 extends to the turbulence section 602 and is provided with turbulence blades 606. The mixer 600 is also provided with a rotary drive device 603 that drives the drive shaft 604 to rotate.

[0044] When gas enters the mixer 600, its flow may be uneven due to factors such as pipe connections and gas pressure, resulting in uneven velocity distribution and eddies. The mixing blades 605 in the mixing section 601 provide initial stirring and mixing of the various gases entering the mixer 600. As the gas flows through the mixing blades 605, the rotation or special shape of the blades changes the flow direction and velocity, causing intense relative motion and collisions between different gases, thus breaking down the stratification and achieving initial macroscopic mixing. Under the action of the spiral mixing blades 605, the gas moves along a spiral trajectory, and different gases continuously intertwine and mix within the spiral channel, laying the foundation for subsequent deep mixing. After initial mixing in the mixing section 601, the gas enters the turbulence section 602, where the turbulence blades 606 further enhance the mixing effect. The turbulence blades 606 create strong turbulence within the turbulence section 602. In turbulent flow, gas particles undergo irregular motion, rotation, and diffusion, which greatly increases the contact area and collision frequency between gas molecules, thereby achieving deep mixing of gases at the microscopic level and making the composition of the mixed gas more uniform.

[0045] As the gas mixed in mixing section 601 flows toward outlet 611, its flow may become unstable due to factors such as energy changes and pressure fluctuations during the mixing process. The blades in turbulence section 602 can readjust and stabilize the gas, making the velocity distribution at the outlet more uniform and the pressure more stable.

[0046] like Figure 4 As shown, in this embodiment, the outlet 611 of the mixer 600 is located on the upper end face of the turbulence section 602. This design allows the gas to be smoothly discharged from the outlet 611 after passing through the turbulence section 602, reducing turbulence and noise during gas discharge. It also facilitates the normal operation of subsequent equipment and allows the small particles in the mixed gas to be deposited under the action of gravity.

[0047] The blades of the mixing section 601 and the turbulence section 602 can play a certain role in blocking and dispersing, reducing the deposition and accumulation of solid particles, droplets and other impurities carried in the gas in the mixer 600, thereby reducing the risk of equipment blockage.

[0048] like Figures 1-4 As shown, in this embodiment, the air outlet pipe 700 is connected to the air outlet 611 of the mixer 600, and a laser detector 800 is provided on the air outlet pipe 700. The laser detector 800 is fixed on the air outlet pipe 700 by the mounting base 1200.

[0049] like Figures 1-3 As shown, in this embodiment, the laser detector 800 is connected to the gas outlet pipe 700 and can emit laser into the gas outlet pipe 700. The laser emission direction is consistent with the flow direction of the mixed gas and forms an acute angle. In this embodiment, the laser detector 800 can measure the flow rate and velocity of the gas in real time by monitoring the change of optical path difference.

[0050] In this embodiment, the laser emission direction is consistent with the gas flow direction, meaning the laser emission direction is the same as the gas flow direction during gas discharge. The angle between the laser detector 800 and the outlet pipe 700 is 30°-60°. This angle range is determined based on Mie scattering theory and an optical path difference optimization model. When the laser beam forms a 45° angle with the gas flow direction, sufficient optical path length (L = D / sinθ, where D is the pipe diameter) is ensured, allowing gas molecules and photons to interact fully. Simultaneously, specular reflection interference that may occur with perpendicular incidence (angle > 60°) is avoided. The detector employs tunable diode laser absorption spectroscopy (TDLAS) technology. Its laser emission module integrates a distributed feedback (DFB) laser, with the wavelength locked at the characteristic absorption peak of SiH4 (1693nm), a linewidth < 2MHz, and a tuning range of ±1cm. -1This angle-optimized design, combined with advanced detection technology, enables the system to achieve a detection limit of 0.1 ppm for SiH4 concentration and a response time of <100 ms, providing high-precision real-time data support for process control.

[0051] In this embodiment, the second tee pipe 500 is T-shaped, with two ports on the same straight line connecting to the first air inlet pipe 100 and the mixer 600, respectively. The first air inlet pipe 100 is located above the first tee pipe 400. The density of NH3 is 0.771 kg / m³. 3 The concentration was significantly lower than that of N2 (N2, 1.25 kg / m³). 3 ) and SiH4 (SiH4, 1.44 kg / m 3 The mixture of NH3 and SiH4 gases has a lower density than air, a similar density to air, and a higher density than air. Therefore, when N2 and SiH4 mix, the overall density, being greater than air, causes it to flow downwards. NH3, entering from above, rises naturally due to its lower density, while the heavier N2-SiH4 mixture flows downwards. This stratification reduces energy consumption, achieving initial mixing with gravity assistance and reducing the burden on the subsequent mixing section 601. Furthermore, the vertical injection of NH3 from above creates shear force with the horizontally flowing N2-SiH4 mixture, enhancing turbulence. This convergence method quickly breaks up gas stratification, reduces concentration gradients, and ensures that the NH3 and N2-SiH4 mixture achieves a high degree of homogeneity before entering the mixer 600, providing a foundation for subsequent precise control.

[0052] The upper intake pipe position prevents the N2-SiH4 mixture from flowing back into the NH3 pipe, ensuring unidirectional gas flow. Especially for flammable and explosive SiH4 and corrosive NH3, preventing backflow reduces the risk of accidental mixing within the pipe, meeting the stringent safety requirements of semiconductor processes.

[0053] In this embodiment, the first intake pipe 100, the second intake pipe 200, and the third intake pipe 300 are each equipped with a corresponding valve. The first intake pipe 100 and the second intake pipe 200 are each equipped with a temperature sensor 900 and a flow controller 1100 to control the temperature and flow rate of NH3 and N2. The first intake pipe 100 and the first three-way pipe 400 are each equipped with a pressure sensor 1000 to detect the concentration of NH3 and the pressure of the N2-SiH4 mixture, respectively, to facilitate intelligent control. For details, please refer to Embodiment 2.

[0054] The gas flow rates of the second inlet pipe 200, the first inlet pipe 100, and the third inlet pipe 300 decrease sequentially. This flow rate design is based on the principle of chemical reaction kinetics: SiH4, as a highly reactive gas, has its flow rate set to the lowest value to ensure sufficient dilution by N2 in the initial stage of mixing, reducing the risk of spontaneous combustion; NH3, as a reactant, has a moderate flow rate to provide sufficient nitrogen source for the formation of silicon nitride compounds; N2, as a carrier gas and diluent, has the highest flow rate, ensuring both gas delivery power and maintaining the concentration of combustible components in the mixture below the lower explosive limit (LEL). Before entering the mixer 600, the three gases undergo preliminary premixing through a static mixer with an internal helical blade structure, creating three-dimensional turbulence and achieving a mixing uniformity of over 95%. This flow rate gradient design, combined with a multi-stage mixing process, achieves precise control of the stoichiometric ratio (SiH4:NH3:N2 molar ratio controlled at 1:2:10), ensuring a high yield of the target product (silicon nitride) while reducing the risk of explosion through the dilution effect, significantly improving process safety.

[0055] In this embodiment, the reactivity of the third, first, and second gases decreases sequentially, and the risk of explosion is reduced by mixing the second and third gases. SiH4 (the third gas), as a typical self-igniting gas, has a reactivity index as high as 9.5 (0-10 level) and a self-ignition temperature of only 25°C in air; NH3 (the first gas) has a reactivity index of 6.2 and exhibits weak reducing properties; N2 (the second gas), as an inert gas, has a reactivity index of 0. This embodiment employs an "inert dilution-gradient mixing" safety strategy: firstly, highly reactive SiH4 is initially mixed with a large amount of N2 (volume ratio 1:20) at the end of the third inlet pipe 300, utilizing the chemical inertness of N2 to inhibit the self-decomposition reaction of SiH4; subsequently, this mixture is secondary mixed with NH3 at the first three-way pipe 400, at which point the SiH4 concentration has dropped to below 5% of the lower explosive limit (LEL), making it impossible to maintain a combustion chain reaction even if local temperatures rise or an ignition source is present. This dual protection system, combining gas reactivity gradient control with a physical flame arrestor, reduces the probability of explosion for the entire system.

[0056] In summary, this embodiment details a flow control system for semiconductor manufacturing processes. Through precise design and advanced technology, this system achieves high-precision mixing and flow control of three gases: N2, SiH4, and NH3, ensuring the uniformity and quality stability of silicon nitride films.

[0057] It consists of a mixer 600, a rotary drive device 603, a first air inlet pipe 100, a second air inlet pipe 200, a third air inlet pipe 300, a laser detector 800, a temperature sensor 900, a pressure sensor 1000, a flow controller 1100, a mounting base 1200, and various pipes and valves. In terms of gas delivery, the first air inlet pipe 100, the second air inlet pipe 200, and the third air inlet pipe 300 respectively deliver three different gases, and each is equipped with a control valve: the first air inlet pipe 100 has a first control valve 101, the second air inlet pipe 200 has a second control valve 201, and the third air inlet pipe 300 has a third control valve 301.

[0058] The system employs a premixing design, achieving gas mixing and delivery via a first tee pipe 400 and a second tee pipe 500. N2 is first used to dilute SiH4, reducing the explosion risk during SiH4 delivery and subsequent mixing with NH3, improving safety, and simplifying the piping layout. The mixer 600 has an inlet 610 and an outlet 611. The inlet 610 thoroughly mixes the three gases. The blades in the mixing section 601 and the turbulence section 602 achieve preliminary and deep mixing, respectively. The mixing section 601 has mixing blades 605, and the turbulence section 602 has turbulence blades 606. The drive shaft 604 rotates the blades, stabilizing gas flow and reducing impurity deposition. The rotation drive device 603 provides power to the drive shaft 604.

[0059] The laser detector 800 measures gas flow and velocity in real time by monitoring changes in optical path difference. Its specific angle design, combined with advanced detection technology, provides high-precision real-time data for process control. The special design of the positions of the first inlet pipe 100, the second inlet pipe 200, and the third inlet pipe 300 utilizes the difference in gas density to achieve preliminary mixing with gravity assistance, reducing energy consumption, enhancing turbulence intensity, ensuring high uniformity of gas before entering the mixer 600, and preventing gas backflow to ensure safety.

[0060] The three-channel gas flow gradient design, combined with a multi-stage mixing process, precisely controls the stoichiometry, enhancing process safety. This embodiment also employs an "inert dilution-gradient mixing" safety strategy, which significantly reduces the probability of system explosion by combining gas activity gradient control with physical flame arresters.

[0061] Example 2

[0062] like Figure 5 As shown, this embodiment discloses a high-precision flow control method for semiconductor manufacturing processes, which uses a flow control system for semiconductor manufacturing processes as described in Embodiment 1, and includes the following steps:

[0063] S1: Gas pipeline distribution, delivery, and premixing

[0064] Gas distribution: The second inlet pipe 200 delivers N2, the third inlet pipe 300 delivers SiH4, and the first inlet pipe 100 delivers NH3;

[0065] The second intake pipe 200 and the third intake pipe 300 are connected by the first tee pipe 400 to achieve premixing of N2 and SiH4;

[0066] The premixed gas and the NH3 in the first inlet pipe 100 are connected through the second tee pipe 500 to form a three-gas mixed flow.

[0067] N2 is used as an inert diluent to dilute the SiH4 concentration to below 80% of its lower explosive limit (LEL = 1.4%) during the premixing stage;

[0068] During the primary mixing stage, N2 further dilutes the NH3 concentration to below 85% of its LEL (15%).

[0069] S2: Multi-parameter fusion measurement

[0070] Data acquisition: The gas temperature (T) is monitored in real time by temperature sensors 900 on the second intake pipe 200 and the first intake pipe 100. Temperature monitoring ensures that the gas is delivered within a suitable temperature range. Pressure data (P) is collected by pressure sensors 1000 on the second intake pipe 200 and the first three-way pipe 400. The initial flow rate (Q0) of each intake pipe (first intake pipe 100, second intake pipe 200, and third intake pipe 300) is obtained by flow controller 1100 (MFC).

[0071] Data fusion: Kalman filters are used to suppress noise in temperature, pressure, and flow data;

[0072] The actual flow rate after multi-parameter fusion is calculated using the following formula:

[0073] Q1=(Q MFC ×(1+k P ·ΔP)×(1+k T ·ΔT)) / (1+k P ·ΔP+k T ·ΔT)

[0074] Among them, Q MFC The initial measurement value for the mass flow controller; k P k T ΔP and ΔT are pressure and temperature correction coefficients, respectively, and are pressure and temperature deviation values.

[0075] S3: Adaptive Control

[0076] Traffic forecasting:

[0077] The deep learning-based flow prediction unit loads historical process data;

[0078] Input the current pressure (P) and temperature (T) data, and output the predicted flow rate (Q');

[0079] The prediction model uses a Long Short-Term Memory (LSTM) network with a loss function of mean squared error (MSE) and a training cycle of 500 times. The LSTM model can process time series data and improve prediction accuracy.

[0080] Dynamic PID control:

[0081] e represents the difference in flow rate between the target setpoint and the actual measured value, reflecting system deviation. By quantifying the deviation between the actual flow rate and the target flow rate, dynamic PID control is driven, ultimately achieving high-precision tracking of the actual flow rate to the target value. Dynamic PID control can adjust the flow rate in real time, ensuring stable system operation.

[0082] e = Q set -Q1

[0083] Among them, Q SET Q1 represents the target flow rate set by the user, while Q1 represents the actual flow rate output by the multi-parameter fusion measurement module.

[0084] proportionality coefficient (K) P When |e|>20%, K p =1.5; when |e|<5%, K p =0.8, the proportional coefficient can adjust the control strength according to the magnitude of the error;

[0085] Integral coefficient (K) i ): Introduce an anti-integral saturation algorithm; when the error persists, K i Gradually decays to 0.3K i0 (K i0 (as initial value);

[0086] Differential coefficients (K) d ): An incomplete differential strategy is adopted, and a first-order inertial element (τ = 0.1s) is added.

[0087] Actuator control:

[0088] The dynamic PID output control signal is sent to the flow controller to adjust the valve openings of the second intake pipe 200 and the third intake pipe 300, so that the actual flow rate (Q1) approaches the target value (Q). set ).

[0089] S4: Dynamic Calibration

[0090] Laser interferometry monitoring:

[0091] The laser detector emits a laser, forming a 30°-60° angle with the axis of the exhaust pipe;

[0092] The reflected light and the reference light superimpose to form interference fringes. The flow rate (Q2) is monitored by the laser interferometer through the change in optical path difference (ΔL). The calculation formula is as follows:

[0093] Q2=(ΔL×A×v) / sinθ

[0094] Where A is the cross-sectional area of ​​the outlet pipe, v is the average molecular velocity, and θ is the laser angle. The design of the laser angle can optimize the measurement effect of optical path difference and improve the measurement accuracy.

[0095] Quantum tunneling correction:

[0096] An ultrathin metal film (thickness <2nm) was deposited on the wall of the exhaust pipe, and the molecular collision frequency (f) was detected using the quantum tunneling effect;

[0097] Meanwhile, directly measuring f (number of molecular collisions) usually requires a high-speed counter, which is technically difficult and susceptible to noise interference. The current I generated by the quantum tunneling effect is highly correlated with f (I∝f), but through the physical properties of the metal film (such as thickness and material) and circuit design, I can be converted into a more stable flow signal Q3.

[0098] The relationship between tunneling current (I) and the quantum tunneling effect detection flux value (Q3) is: I = k × Q3 (k is the calibration coefficient);

[0099] Combining laser interferometry data (Q2) with the flow rate detected by quantum tunneling effect (Q3), the final corrected flow rate value Q after dynamic calibration is obtained. final The calculation formula is:

[0100] Q final = (Q2 + Q3) / 2

[0101] This formula can integrate the measurement results of laser interferometry and quantum tunneling effect, thereby improving the accuracy of flow calibration.

[0102] S5: Mixer Optimization and Output

[0103] Mixer driver:

[0104] The rotary drive device drives the drive shaft to rotate, and the mixing blades (located in the mixing section) perform preliminary mixing;

[0105] The turbulence-dissipating blades (located in the turbulence section) provide secondary turbulence to ensure uniform gas distribution.

[0106] Output control:

[0107] According to the corrected flow rate (Q) final ) and target value (Q)set The deviation is dynamically adjusted by the flow controller to adjust the valve opening of the first air inlet pipe 100, thereby achieving the gas flow accuracy requirements in the semiconductor manufacturing process.

[0108] In summary, this embodiment discloses a high-precision flow control method for semiconductor manufacturing processes, which, based on the flow control system of Embodiment 1, covers key aspects such as gas distribution, measurement, control, calibration, and mixed output.

[0109] In terms of gas distribution, the second inlet pipe 200 delivers N2, the third inlet pipe 300 delivers SiH4, and the first inlet pipe 100 delivers NH3. Premixing and mixing of the three gases are achieved through the first three-way pipe 400 and the second three-way pipe 500. During the premixing and main mixing stages, N2 dilutes the concentrations of SiH4 and NH3 to safe ranges, respectively.

[0110] In the multi-parameter fusion measurement stage, data is collected through temperature sensor 900, pressure sensor 1000 and flow controller 1100, noise is suppressed by Kalman filter, and the actual flow value after multi-parameter fusion is calculated by a specific formula.

[0111] In terms of adaptive control, the deep learning-based flow prediction unit uses an LSTM model to predict flow; dynamic PID regulation quantifies the deviation between the actual and target flow rates, adjusting the proportional, integral, and derivative coefficients according to the error magnitude; and actuator control adjusts the valve opening based on the dynamic PID output.

[0112] During dynamic calibration, the laser detector 800 emits a laser at a specific angle and calculates the flow rate by measuring the change in optical path difference; the quantum tunneling effect is used to detect the molecular collision frequency and convert it into a flow rate signal, and the final corrected flow rate value is obtained by combining the two data.

[0113] Finally, the mixer 600 uses a rotary drive to drive the mixing blades and turbulence blades for preliminary and secondary mixing to ensure uniform gas distribution. Then, the opening of the valve in the first inlet pipe 100 is dynamically adjusted according to the deviation between the corrected flow rate and the target value to meet the requirements of the semiconductor process for gas flow accuracy.

[0114] The above description is an explanation of the present invention and not a limitation thereof. The scope of the present invention is defined by the claims. Within the scope of protection of the present invention, any form of modification may be made.

Claims

1. A high-precision flow control system for semiconductor manufacturing processes, characterized in that, include: The first intake pipe is used to deliver the first gas; The second intake pipe is used to deliver the second gas; The third intake pipe is used to deliver the third gas; The first three-way pipe is connected to the second and third air inlet pipes respectively to realize the mixed transportation of the second gas and the third gas; The second three-way pipe is connected to the first three-way pipe and the first air inlet pipe respectively, so as to realize the mixed transportation of the first gas, the second gas and the third gas; The mixer has an air inlet at one axial end that is connected to the air outlet of the second three-way pipe, and an air outlet at the other end in a radial direction. The mixer includes a mixing section and a turbulence section for mixing and agitating the gas. The exhaust pipe is connected to the air outlet of the mixer. The laser detector is connected to the outlet pipe and can emit a laser into the outlet pipe. The laser emission direction is consistent with the flow direction of the mixed gas and forms an acute angle. The second three-way pipe is T-shaped, with two openings on the same straight line connected to the first air intake pipe and the mixer respectively, and the first air intake pipe is located above the first three-way pipe; The gas flow rates of the second intake pipe, the first intake pipe, and the third intake pipe decrease sequentially. The angle between the laser detector and the exhaust pipe is 30°-60°.

2. The high-precision flow control system for semiconductor manufacturing processes as described in claim 1, characterized in that: Each of the first, second, and third air intake pipes is equipped with a corresponding valve. Each of the first and second air intake pipes is equipped with a temperature sensor and a flow controller. Each of the first air intake pipe and the first three-way pipe is equipped with a pressure sensor.

3. The high-precision flow control system for semiconductor manufacturing processes as described in claim 1, characterized in that: The mixer includes a housing and a drive shaft axially arranged along the axis of the housing. Mixing blades are provided in the region where the drive shaft extends to the mixing section, and turbulence blades are provided in the region where the drive shaft extends to the turbulence section. A rotary drive device for driving the drive shaft to rotate is also provided on the mixer.

4. The high-precision flow control system for semiconductor manufacturing processes as described in claim 1, characterized in that: The reactivity of the third gas, the first gas, and the second gas decreases sequentially, and the risk of explosion is reduced by mixing the second gas with the third gas.

5. A high-precision flow control method for semiconductor manufacturing processes, which uses a high-precision flow control system for semiconductor manufacturing processes according to any one of claims 1-4, comprising the following steps: S1: Gas pipeline distribution, delivery, and premixing: The second intake pipe delivers N2, the third intake pipe delivers SiH4, and the first intake pipe delivers NH3; The second and third intake pipes are connected by the first tee pipe to achieve premixing of N2 and SiH4; The premixed gas and the NH3 in the first inlet pipe are connected through the second tee pipe to form a three-gas mixed flow; S2: Multi-parameter fusion measurement: The gas temperature T is monitored in real time using temperature sensors on the second and first intake pipes. Pressure data P is collected through pressure sensors on the second intake pipe and the first three-way pipe; The initial flow rate value Q0 of each intake pipe is obtained through the flow controller MFC; By collecting the above data and calculating the actual traffic value after multi-parameter fusion, we can obtain the following: ; S3: Adaptive Control The deep learning-based flow prediction unit loads historical process data; Input the current pressure P and temperature T data, and output the predicted flow rate Q'; The prediction model uses a Long Short-Term Memory (LSTM) network, with the loss function being the mean squared error (MSE), and the training cycle is 500 times. Dynamic PID control: e is the difference in flow rate between the target setpoint and the actual measured value. The formula for calculating e is: e= Q set Q1 in, Q SET Q1 is the target flow rate value set by the user, and Q1 is the actual flow rate value output by the multi-parameter fusion measurement module. The dynamic PID output control signal is sent to the flow controller to adjust the air intake volume of the second and third air intake pipes, so that the actual flow rate Q1 approaches the target value Q. set ; S4: Dynamic Calibration The laser detector emits a laser beam, and the reflected light superimposes with the reference light to form interference fringes. These fringes are observed through changes in the optical path difference. Laser interferometer monitors flow rate values The calculation formula is: Where A is the cross-sectional area of ​​the exhaust pipe, v is the average molecular velocity, and θ is the laser angle; Quantum tunneling correction: An ultrathin metal film was deposited on the wall of the exhaust pipe, and the tunneling current I was used to detect the flow rate using the quantum tunneling effect. The relationship for Q3 is: I = k × Q3, where k is the calibration coefficient; The final corrected flow rate value after dynamic calibration is obtained by combining laser interferometry data Q2 with the flow rate value Q3 detected by quantum tunneling effect. The calculation formula is: Then adjust the flow rate With the target value Q set The deviation is addressed by dynamically adjusting the flow rate in the first intake pipe using a flow controller, thus achieving precise control.

6. The high-precision flow control method for semiconductor manufacturing as described in claim 5, characterized in that: The The calculation formula is: in, These are the initial measurements for the mass flow controller; , For pressure and temperature correction factors, , These are the pressure and temperature deviation values.

7. A high-precision flow control method for semiconductor manufacturing processes as described in any one of claims 5 or 6, characterized in that: It also includes S5: Mixer optimization and output, where the rotary drive device drives the drive shaft to rotate, the mixing blades in the mixing section perform initial mixing, and the turbulence blades in the turbulence section perform secondary turbulence to ensure uniform gas distribution.

Images