Mass flow controller with dual-mode PID control loop

Abstract

The invention discloses a dual-mode proportional-integral-derivative (Proportional-Integral-Differentiation) method, a device and a computer readable medium. The dual-mode proportional-integral-differential (Proportional-Integral-Differentiation) method is provided. A mass flow controller (Mass Flow Controller; a mass flow controller (Mass Flow Controller, Mass Flow Controller, Mass Flow Controller, Mass Flow Controller, Mass Flow Controller; mFC). And in the training mode, the PID control loop determines the corresponding electromagnetic coil current set value according to each working state specified in the process formula, and stores the corresponding electromagnetic coil current set value for subsequent calling. During execution of a semiconductor manufacturing process, rapid flow output can be realized by using the pre-stored set values, and continuous operation of PID control is not needed. The method comprises the following steps of: performing real-time flow monitoring and statistical process control; the SPC mechanism ensures the stability of the system, and retraining can be triggered when necessary, thereby improving the speed, precision and reliability of the MFC.

Description

Cross Reference to Related Applications

[0001] This application claims priority to U.S. Patent Application No. 18 / 676,337, filed May 28, 2024. TECHNICAL FIELD

[0002] The present invention relates to mass flow controllers (MFCs) that are applied in various industries and manufacturing processes. More specifically, the present invention relates to a dual-mode PID control loop system that improves the response speed and stability of MFCs by setting a training mode for electromagnetic coil current values and an inference mode for real-time flow monitoring and stability control, thereby optimizing their overall performance. BACKGROUND

[0003] MFCs are indispensable key components in numerous industrial applications that require precise control of fluid flow, including but not limited to semiconductor manufacturing processes such as Atomic Layer Deposition (ALD) and Atomic Layer Etching (ALE). In these processes, the rapid delivery of gases is crucial for process performance and product quality. Traditional MFCs rely on the continuous operation of a PID control loop to regulate fluid flow, which is inefficient and slow to respond when there are rapid adjustment demands.

[0004] One challenge faced by existing MFCs is the delay in reaching the target flow rate due to the continuous adjustment required by the PID control loop. This delay negatively impacts the productivity and efficiency of processes that require rapid and precise flow changes. Additionally, the continuous operation of the PID control loop can exacerbate the wear and tear of system components, reducing the reliability and service life of the MFC.

[0005] Another important issue is the lack of robust mechanisms to monitor and maintain the stability of MFC operation. Fluctuations in flow and potential drift in system performance can lead to deviations from expected process conditions, affecting product quality and triggering potential process failures.

[0006] In this context, various embodiments of the present invention address the aforementioned problems, significantly improving the speed, accuracy, and stability of MFC operation. SUMMARY

[0007] The present invention relates to a mass flow controller (MFC) designed to optimize fluid flow regulation in various process systems. In some embodiments, a training mode is employed to determine set points for electromagnetic coil current when a dual mode PID control loop is active. In this training mode, the MFC controller activates the PID control loop to establish precise electromagnetic coil current values for each operating state associated with a particular fluid flow. These set points are then stored in a memory unit.

[0008] In some embodiments, these pre-stored set points can be recalled when executing a process recipe, enabling the MFC to achieve target flow without the need for continuous adjustment by the PID control loop. This significantly improves the response speed of the MFC, making it highly suitable for applications requiring fast and precise flow regulation, such as atomic layer deposition (ALD) and atomic layer etching (ALE). Rapidly reaching set points without the need to run the PID control loop contributes to improved efficiency and throughput of the process system.

[0009] Another key feature of the present invention is the ability to measure flow in inference mode during execution of a process recipe. In some embodiments, in inference mode, the flow sensor monitors flow, stores data and analyzes it through statistical process control (SPC) methods to ensure stability of the MFC operation. These data can be used to establish trend charts and monitor the stability of the flow. If an unstable or out-of-control condition is detected, a retraining event can be triggered to recalibrate the set points, ensuring continued accuracy and stability. In certain embodiments, if retraining fails to correct instability, the controller can also determine whether the MFC requires maintenance.

[0010] The present invention also includes the following feature: In some embodiments, the MFC controller is connected to a system controller of the process system, which receives a process recipe and identifies multiple operating states of the MFC. Each operating state is associated with a particular fluid flow, and the dual mode PID control loop ensures precise control and regulation of the flow.

[0011] This dual mode operation combines the training mode with the inference mode, significantly enhancing the performance and efficiency of the MFC. It provides both high-speed operation and robust process stability monitoring capabilities, which are crucial for maintaining the precision and consistency required in advanced manufacturing processes. Compared to traditional MFCs, the present invention greatly reduces the time required to reach target flow, while providing continuous monitoring and regulation functions to maintain optimal operating performance. BRIEF DESCRIPTION OF DRAWINGS

[0012] Embodiments of the present invention can be better understood through the following detailed description in conjunction with the accompanying drawings.

[0013] Figure 1A block diagram of a mass flow controller (MFC) integrating a dual-mode PID control loop and a learning engine is schematically shown.

[0014] Figure 2 An operational flow diagram of the MFC in a training mode, where the PID control loop is active, is shown.

[0015] Figure 3 An operational flow diagram of the MFC in a reasoning mode, where the PID control loop is inactive, is shown.

[0016] Figure 4 A control flow diagram of the MFC applying statistical process control (SPC) methods is depicted. DETAILED DESCRIPTION

[0017] In the following detailed description of the application, certain specific embodiments are described in order to provide a thorough understanding of the application. However, it will be clear to those skilled in the art that the application can be practiced without resorting to these specific details, or can be practiced using alternative elements or methodologies. In some cases, well-known processes, methods and components have not been described in detail in order to avoid unnecessarily obscuring aspects of the present application.

[0018] Reference is made to Figure 1 , a block diagram of a mass flow controller (MFC) 100 is schematically shown. The MFC 100 is part of a process system 101. For example, the process system 101 can be an etching or deposition apparatus. Figure 1 Further details of the process system 101 are not shown in the figure. The MFC 100 includes an inlet 102 and an outlet 104, both of which are connected to a gas flow channel 106. The inventive concept is applicable to any type of fluid, and in the present disclosure it is exemplified for a gas. Thus, the gas flow channel 106 can be considered as a flow channel in a more general sense.

[0019] In the device, a proportional valve (not shown) is used to divert a portion of the gas to a channel 108. The flow of the diverted gas is measured by a flow sensor 110. Typically, a thermal flow sensor is used, which measures the flow by detecting the temperature difference between two specific locations in a laminar flow path. Thus, the flow of the diverted gas can be considered as a representation of the flow of the gas inside the gas flow channel 106.

[0020] The structure of the MFC 100 also includes a solenoid valve. The valve contains a spring 112 that supports a plunger 114. The position of the plunger 114 determines the on / off state of the gas at the orifice 113. When the plunger 114 is in contact with the orifice 113, the gas is in the off state. In addition, the solenoid valve that cooperates with the plunger 114 is also equipped with a solenoid coil 115. When current passes through the coil 115, a magnetic force is generated. Since the plunger 114 is usually made of ferromagnetic material, the mechanical pressure exerted by the spring 112 and the magnetic force generated by the coil 115 act on the plunger 114 together. The balance of the two forces ultimately determines the position of the plunger 114.

[0021] The flow sensor 110 transmits its measurement to the MFC controller 118. The controller compares the received data with the reference value preset in the storage unit 124. If there is a deviation between the sensor reading and the reference value, the controller 118 will send an instruction to the valve driver 116. Subsequently, the valve driver 116 will adjust the current of the solenoid coil 115 according to the instruction, causing the plunger 114 to change position. After the plunger is repositioned, the flow sensor 110 will again measure the flow of the shunt gas. This calibration loop continues until the measured flow matches the reference flow. To speed up this process, the MFC controller 118 uses a PID control loop. Typically, this calibration phase lasts tens to hundreds of milliseconds, which is not ideal for ALD and ALE processes.

[0022] A significant feature of the present invention is the use of a dual-mode PID control loop 120, which enables the MFC 100 to operate in two distinct modes: training mode and inference mode.

[0023] The MFC controller 118 is connected to the system controller 126. The system controller 126 receives the process recipe 128 and converts it into a series of time-series instructions suitable for each subsystem, of which the MFC is a key component for achieving precise delivery of gases or precursors to the process chamber.

[0024] Upon receiving the process recipe, the system controller 126 analyzes it and identifies all the operating states of the MFC 100. Each operating state corresponds to a specific flow rate of a specific gas. The set value of the MFC 100 needs to be trained to correspond to different operating states. In all embodiments, the solenoid current is chosen as the set value to achieve the desired gas flow rate.

[0025] In the training mode, the dual-mode PID control loop is active. A learning engine 122 is used to establish the correspondence between the solenoid current and the flow rate of each operating state. The learning engine 122 can be implemented in software, hardware, firmware, or a combination thereof.

[0026] In the training mode, a test procedure is executed by the MFC controller 118 or the system controller 126, which includes a series of tests for the MFC 100. A typical test includes passing a gas specified in a process recipe through the MFC 100, directing it to the desired flow rate in a certain operating state, and recording the corresponding solenoid current in the storage unit 124. Each operating state requires at least one test to be performed. In the training mode, the dual-mode PID control loop 120 is active. The training mode of the MFC 100 is usually performed before processing a wafer. In mass production, the MFC 100 can be trained in various situations. In one embodiment, the test procedure is performed when a new process recipe is introduced. In another embodiment, the test can be performed after processing a preset number of wafers. It is also noted that a process system can be equipped with multiple MFCs; the training process of multiple or all MFCs can be performed simultaneously.

[0027] In the inference mode, the MFC controller 118 receives the real-time operating state from the system controller 126 and retrieves the stored solenoid current value from the storage unit 124. The valve driver 116 generates the corresponding current according to the current value retrieved by the MFC controller 118. The current passes through the solenoid 115, driving the plunger 114 to move to a predetermined position, thereby providing the flow rate required in the operating state.

[0028] Without the active dual-mode PID control loop, the speed of the MFC 100 reaching the set value can be greatly improved to the millisecond level, which will significantly improve the production efficiency of atomic layer processing processes such as ALE and ALD.

[0029] The inference mode is usually associated with the actual processing of wafers, i.e., it occurs in production events. When the dual-mode PID control loop is in the inactive state, the flow sensor 110 can be used to continuously measure the flow rate and record the results to the storage unit 124. The measurement results can be sorted by different operating states. The MFC controller 118 or the system controller 126 can use these data to establish a trend chart to monitor the operating stability of the MFC 100. SPC rules can be applied to determine whether retraining is needed. All these operations can be performed in the background and will not affect the operating speed of the MFC.

[0030] Figure 2An exemplary flowchart of a training process is shown as 200. The flow 200 begins at step 202, where the system controller 126 receives a process recipe. In step 204, the system controller 126 identifies all operating states of the MFC 100 specified in the process recipe. Each operating state corresponds to a specific gas flow rate. In step 206, the dual-mode PID control loop 120 is activated. The MFC controller 118 receives the measured flow rate from the flow sensor 110 and compares it to the stored reference value. If there is a discrepancy, the MFC controller 118 adjusts the current to the solenoid 115 through the valve driver 116. This process continues until the measured flow rate matches the reference value. A PID control mechanism known in the art is employed to speed up the adjustment. In step 208, the MFC controller 118 executes a test procedure using the learning engine 122 to determine the setpoint value of the solenoid current for each operating state. The setpoint value is measured after the flow rate reaches a steady state. In step 210, the setpoint values of the solenoid current are stored in the storage unit 124.

[0031] Figure 3 An exemplary flowchart of a reasoning process 300 is shown. The flow begins at step 302, where the system controller 126 receives a process recipe. In step 304, the operating states of the MFC 100 are identified. In step 306, the dual-mode PID control loop 120 is deactivated. The setpoint values of the solenoid current are then retrieved from the storage unit 124, which can also be preloaded into the fast cache of the MFC controller 118. The process recipe is then executed step-by-step by the system controller 128. At each operating state of the MFC 100, the MFC controller 118 reads the setpoint value for the corresponding state and instructs the valve driver 116 to provide the required solenoid current without applying the PID control loop.

[0032] A significant advantage of the present inventive concept is that the flow sensor 110 can be used to measure the steady-state flow rate corresponding to the setpoint value without being tied to the PID control loop. The measured flow rate at a selected operating state or all operating states can be used to monitor the stability of the manufacturing process.

[0033] Figure 4The illustrated flow 400 begins at step 402 where the flow of process gas is measured at a certain operating condition while the associated process step is being performed. The dual mode PID control loop 120 is inactive at this time. The measurement is stored in the storage unit 124 at step 404. In one embodiment, the system controller 126 analyzes the flow trend measured at the particular operating condition. SPC rules are applied at step 408 to assess the stability of the MFC 100. If the system controller 126 detects an out-of-control event, the controller will decide that the MFC 100 needs to be retrained by re-executing the flow 200 to reestablish the setpoint. In some cases, the system controller 126 can determine that the MFC 100 is malfunctioning and request a maintenance operation to be performed. In other embodiments, the MFC controller can perform this assessment; in further embodiments, the MFC controller 118 and the system controller 126 can cooperate to complete this assessment.

Claims

1. A mass flow controller characterized by, comprising: a fluid flow channel for guiding fluid flow; an inlet for receiving the fluid, wherein a proportional valve directs a portion of the fluid to a shunt channel connected to a flow sensor; an outlet for discharging the fluid through the fluid flow channel; a solenoid valve comprising a spring, a plunger, an orifice, and an electromagnetic coil, wherein the position of the plunger is regulated by the current applied to the electromagnetic coil, which in turn controls the flow rate of the fluid; and a MFC controller configured to adjust the flow rate of the fluid by putting a dual-mode PID control loop in an active state or an inactive state, wherein in the active state, the electromagnetic coil current value associated with a specific fluid flow rate is determined and stored in a storage unit, and in the inactive state, the current value is retrieved to position the plunger at a position to deliver the required flow rate. The MFC controller is connected to a system controller of a process system.

2. The mass flow controller of claim 1, wherein, The system controller is configured to receive a process recipe and identify a plurality of operating states of the mass flow controller.

3. The mass flow controller of claim 2, wherein, Each operating state is associated with a specific fluid flow rate.

4. The mass flow controller of claim 3, wherein, The MFC controller further comprises a learning engine, wherein the learning engine executes a test procedure.

5. The mass flow controller of claim 1, wherein, The learning procedure is executed when the dual-mode PID control loop is in the active state.

6. The mass flow controller of claim 5, wherein, A valve driver generates a current based on the retrieved current value to drive the plunger to a corresponding position to deliver the target flow rate of the specific fluid as specified in the process recipe.

7. The mass flow controller of claim 1, wherein, When the PID control loop is in the inactive state, the flow sensor measures the flow rate of the specific fluid.

8. The mass flow controller of claim 1, wherein, The measured flow rate is stored in a storage unit for trend charting and as an input for the system controller to apply SPC rules.

9. The mass flow controller of claim 8, wherein, The dual-mode PID control loop further comprises the flow sensor, the solenoid valve, and the MFC controller, wherein the position of the plunger is adjusted based on the comparison between the output of the flow sensor and a reference value provided by the MFC controller.

10. The mass flow controller of claim 1, wherein, comprising the steps of:

11. A method of controlling the flow of a fluid by a mass flow controller, characterized by, receiving a process recipe by a system controller of a process system, wherein the system controller is connected to a MFC controller; identifying a plurality of operating states of the mass flow controller by the system controller, each operating state being associated with a specific fluid flow rate; putting the dual-mode PID control loop in an active state by the MFC controller when the mass flow controller is in a learning mode; executing a test procedure by the MFC controller to determine the electromagnetic coil current value at each operating state; storing the determined electromagnetic coil current value in a storage unit connected to the MFC controller; putting the dual-mode PID control loop in an inactive state when the mass flow controller is in a reasoning mode; retrieving the stored current value at a specific operating state by the MFC controller from the storage unit; and a valve driver generates a current based on the retrieved current value to drive the solenoid valve plunger to a corresponding position to deliver the required flow rate as specified in the process recipe. ​ ​ 12. The method of claim 11, wherein, The method further includes the step of measuring flow by the flow sensor for a selected operating condition when the dual mode PID control loop is in the inactive state.

13. The method of claim 12, wherein, The method further includes the step of storing the measured flow in the memory unit.

14. The method of claim 13, wherein, The method further includes the step of establishing a trend of the measured flow at the selected operating condition by the system controller or the MFC controller.

15. The method of claim 14, wherein, The method further includes the step of applying SPC rules to the trend.

16. A method of delivering fluid to a process system by a mass flow controller, characterized by, including: measuring flow by the flow sensor for a selected operating condition when the dual mode PID control loop is in the inactive state; storing the measurement in the memory unit by the MFC controller; applying SPC rules by the system controller to analyze the flow trend at the selected operating condition; and running a test routine by the system controller or the MFC controller to determine a solenoid current value associated with the selected operating condition when the dual mode PID control loop is in the active state. The mass flow controller further includes a fluid flow passage having an inlet and an outlet.

17. The method of claim 16, wherein, The flow sensor is coupled to a shunt passage that receives a portion of the fluid flow from the fluid flow passage through a proportional valve.

18. The method of claim 17, wherein, ​

Images