ALD equipment temperature control method based on mechanism model

Abstract

The invention provides an ALD equipment temperature control method based on a mechanism model. The ALD equipment temperature control method comprises the following steps: step 1, establishing and updating the mechanism model; 2, designing an unscented Kalman estimator according to the mechanism model; 3, linearizing the mechanism model, and designing the feedback gain of the controller according to the linearized mechanism model so as to form a second controller; 4, constructing a temperature control system; and step 5, all the Inner TC acquire temperature data in real time, and the second controller outputs heating power to the corresponding heater, so that the error between the measurement value of the Inner TC and the corresponding target reference temperature is in a set range. According to the invention, the steady-state control precision is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of temperature control of semiconductor equipment reaction furnaces, and particularly relates to a temperature control method for ALD equipment based on a mechanism model. Background Art

[0002] Vertical furnace heat treatment equipment is an important process treatment equipment in the semiconductor manufacturing process, and the temperature control of the main reaction furnace has a great impact on the wafer yield.

[0003] The reaction zone of the vertical furnace usually includes N cylindrical temperature zones from top to bottom. The outer layer of each temperature zone is a heater with controllable power, which provides heat for the system; the center of the temperature zone is a wafer arranged with a certain gap up and down. There is a temperature sensor near the inner side of the heater on the central plane of the temperature zone, called Outer TC; there is a temperature sensor near the outer side of the wafer on the same horizontal plane, called Inner TC. The requirement of the process for temperature control is to make the Inner TC reach the process set temperature by adjusting the heater power.

[0004] Currently, the temperature control methods of mainstream equipment on the market generally mainly use direct PID and cascade PID.

[0005] Direct PID refers to a method in which the feedback error of each temperature zone through Inner TC is used as the input of the PID controller to directly calculate the heater power. This method is simple and easy to implement, but there are two obvious disadvantages: 1. Each temperature zone is independently controlled without considering the mutual influence between temperature zones; 2. The vertical furnace system belongs to a system with large inertia and large lag, and the attenuation of all inputs is obvious, resulting in that the PID control rate is not easy to be too fast and the system dynamics is poor; Cascade PID is divided into two-layer PID control. The feedback error of each temperature zone through Inner TC is used as the input of the upper-layer PID controller to calculate the target setting of Outer TC; then, the feedback error of Outer TC is used as the input of the lower-layer PID controller to calculate the heater power. This method has better dynamics than direct PID, but there are also two obvious disadvantages: 1. Each temperature zone is independently controlled without considering the mutual influence between temperature zones; 2. While the dynamics is improved, it causes the control quantity to fluctuate violently, increases the switching loss, and the performance improvement is limited; In addition to the commercial control methods, there are also some academic research methods. Here, they are divided into two categories: The first category mainly uses feedforward + PID or other variant adaptive PIDs. The research idea of feedforward + PID is mainly based on PID feedback control. By means of empirical feedforward, the dynamic characteristics of the control system are improved, so as to reduce overshoot and control fluctuations; parameter adaptive PID is usually used to deal with the time-varying and non-linear effects of the system. These two types of methods belong to the structural variants based on PID and do not overcome the disadvantage that PID cannot consider the temperature zone coupling.

[0006] The second category mainly uses model-based algorithms such as LQR control, LQG control, H∞ control, MPC control, etc. Most of these methods take the overall system as the research object and fully consider the coupling effect of each temperature zone. However, these algorithms all belong to linear system control algorithms. For different working conditions, it is often necessary to re-identify the system and design the controller. Moreover, the actual system is a non-linear system with non-linear characteristics such as parameter time-variation and lag. It is difficult to solve the actual system control problem by applying only one method.

[0007] Therefore, a comprehensive control method based on a model that fully considers all the characteristics of the system is needed. Summary of the Invention

[0008] The purpose of the present invention is to provide a temperature control method for an ALD device based on a mechanism model, which improves the steady-state control accuracy. The technical solution adopted is as follows: A temperature control method for an ALD device based on a mechanism model, comprising the following steps: Step 1, establish and update the mechanism model: Step 2, design an unscented Kalman estimator: select the mechanism model output in Step 1 to design an unscented Kalman estimator; Step 3, linearize the mechanism model and write it in the form of a state space equation, and design the feedback gain of the controller to form a second controller; Step 4, construct a temperature control system: Connect the input end of the unscented Kalman estimator to all Inner TCs; Connect the output end of the unscented Kalman estimator to the first input end of the second controller; Connect the second input end of the second controller to the target temperature module; the target temperature module is used to output the target reference temperature on the temperature zone side; Connect the output end of the second controller to all the heaters in the heat exchange system; Step 5, all Inner TCs obtain temperature data in real time, and the second controller outputs the heating power to the corresponding heaters, so that the error between the measured value of the Inner TC and the corresponding target reference temperature is within the set range.

[0009] Preferably, step 1 specifically includes the following steps: Step 1A: Establish a mechanism model based on the heat exchange system and select the model coefficients to be estimated, which specifically includes the following steps: Step 1A1: Establish a heat exchange system, which includes: A wafer placement area, which is located at the center of the furnace tube reaction chamber and includes several sub-areas distributed from top to bottom; a temperature control zone is formed between each sub-area, between the wafer placement area and the upper surface of the furnace tube reaction chamber, and between the wafer placement area and the lower surface of the furnace tube reaction chamber respectively; A furnace tube reaction chamber, on the outer wall of its side wall 1, heaters corresponding to the temperature control zones are provided, and a heat insulation layer is provided on the outer wall of the heaters; A first sensor Inner TC is provided near the wafer placement area, and a second sensor OuterTC is provided near the heater; And a first controller, which is located outside the furnace tube reaction chamber and is connected to Inner TC, OuterTC and each heater; Step 1A2: Establish a mechanism model: Step 1A21: Analyze the heat exchange system and select the modeling object of the mechanism model; Step 1A22: Establish a mechanism model based on the heaters and temperature zones; Step 1A3: Based on the mechanism model, select the model coefficients to be estimated; Step 1B: Design an experiment based on the heat exchange system to obtain temperature response data for estimating the model coefficients; the temperature response data includes the measured values of Inner TC and Outer TC; Step 1C: Use the non-linear mechanism model as a non-linear regression model, select the output power of the heater as the independent variable, and the measured values of Inner TC and Outer TC as the dependent variables, and use iterative least squares estimation to estimate all the model coefficients to be estimated; Step 1D: Substitute the model coefficients obtained in step 1C into the mechanism model in step 1A to update the mechanism model.

[0010] Preferably, step 1B specifically includes the following steps: Step 1B1: Replace the first controller in the heat exchange system with 5 PID controllers; connect each PID controller to the corresponding heater, Outer TC, and Inner TC; Step 1B2: Conduct experiments for all temperature zones to obtain the output power of the heaters corresponding to the temperature zones and temperature response data; the temperature response data includes: the measured value of Outer TC and the measured value of Inner TC.

[0011] Compared with the prior art, the advantages of the present invention are: By means of better temperature control combining with a mechanism model, under the requirement of a specified system stability duration, the control overshoot during system heating-up is removed, the steady-state control accuracy of the system is improved, and the power fluctuation of the heater is reduced, thereby improving the product yield and reducing the energy consumption. Description of the Drawings

[0012] Figure 1 It is a heat exchange system diagram of the ALD equipment temperature control method based on a mechanism model; Figure 2 It is a schematic diagram of forming a mechanism model; Figure 3 It is a flowchart of the mechanism model parameter identification steps; Figure 4 It is a flowchart of the ALD equipment temperature control method based on a mechanism model; Figure 5 It is a schematic diagram of target feedforward processing; Figure 6 It is a schematic diagram of lossless transformation; Figure 7 It is a control effect comparison diagram between the ALD equipment temperature control method based on a mechanism model proposed by the present invention and the cascade PID control method; Figure 8 It is a control overshoot suppression effect diagram of the feedforward processing method compared with the traditional ramp processing method; Figure 9 It is an installation schematic diagram of the No. 2 sensor OuterTC.

[0013] Wherein, 1 - side wall, 2 - glass cover. Detailed Embodiment

[0014] The ALD equipment temperature control method based on a mechanism model of the present invention will be described in more detail below with reference to the schematic diagrams, in which the preferred embodiments of the present invention are shown. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as a broad guidance for those skilled in the art and not as a limitation to the present invention.

[0015] As Figures 1 - 8 , the ALD equipment temperature control method based on a mechanism model includes the following steps: Step 1. Establish and update a mechanism model.

[0016] Step 1A. Establish a mechanism model based on the heat exchange system and select the model coefficients to be estimated.

[0017] Step 1A1. Establish a heat exchange system.

[0018] The heat exchange system includes: The wafer placement area is mounted at the center of the furnace tube reaction chamber through a quartz bracket. It includes several sub-areas distributed from top to bottom, and each sub-area holds a wafer; a temperature control zone (temperature zone) is respectively formed between each sub-area, the upper surface of the wafer placement area and the furnace tube reaction chamber, and between the wafer placement area and the lower surface of the furnace tube reaction chamber; The furnace tube reaction chamber has heaters corresponding to the temperature control zones arranged on the inner wall surface of its side wall 1, and a heat insulation layer with extremely low thermal conductivity is arranged on the outer wall of the heater; As Figure 1 shown, the segmentation represents the interval of each heater, and 5 heaters are shown.

[0019] The glass cover 2 is used to isolate the external air and the wafer placement area. It is made of a transparent material and is fixed between the wafer placement area and the outer wall of the furnace tube reaction chamber by bottom bolts. A temperature zone outside is formed between it and the wafer placement area, and a heater side is formed between it and the outer wall of the furnace tube reaction chamber; The temperature zone outside is installed with a first sensor Inner TC through a quartz capillary tube extending from the bottom of the furnace tube reaction chamber, and the heater side is installed with a second sensor OuterTC through a small hole dug out in the middle of the heater. As Figure 9 shown, the heater is annular; the inside of the glass cover is in a nearly vacuum state. As Figure 9 shown, the heater is actually a thin layer of graphite or metal sheet, on the inner side of the outer wall of the furnace body, in direct contact with the air inside the furnace, and the outer wall of the furnace body is the heat insulation layer. The hole passes through the outer wall surface and the heater and extends into the furnace body.

[0020] The first controller is located outside the furnace tube reaction chamber and is connected to Inner TC, OuterTC and each heater; The transparent glass is mounted above the wafer placement area through a quartz bracket; specifically, the transparent glass is mounted above the wafer placement area through a quartz bracket.

[0021] The temperature control flange is located at the bottom of the furnace tube reaction chamber. It is cooled by external cooling water and its temperature is always controlled to be constant; that is, the temperature control flange is a flange with cooling water temperature control, and there is a cooling control system around it to adjust its temperature.

[0022] The frosted glass is used to isolate the radiant heat transfer of the wafer and the heater to the bottom temperature control flange. It is located below between the wafer placement area and the temperature control flange and is mounted below the wafer placement area through a quartz bracket. The quartz bracket and the chassis are connected by bolts.

[0023] As Figure 1 , the reaction chamber of the furnace tube equipment applied is a cylindrical structure, and the two-dimensional sectional view of the furnace tube equipment reaction chamber is shown in the figure.

[0024] Ten temperature sensors are evenly arranged on five horizontal planes from top to bottom between the heater and the wafer placement area to reflect the temperature inside the chamber. The five temperature sensors closer to the heater side (referred to as Outer TC in the text) measure values close to the temperature of the heater body, and the five temperature sensors closer to the wafer side (referred to as Inner TC in the text) measure values close to the temperature of the wafer area.

[0025] This method requires that the reaction furnace of the ALD equipment applied has Figure 1 A similar structure as shown. By controlling the power output of the heaters set in each area, the values of the temperature sensors in each area are made to reach the target temperature.

[0026] Step 1A2: Establish a mechanism model.

[0027] Step 1A21: Analyze the heat exchange system and select the modeling object of the mechanism model.

[0028] The mechanism model is a mathematical model that describes the functional relationship from the input to the output of the controlled object and is constructed based on the first principles of the system (physical laws). Its establishment is as described below.

[0029] When conducting a mechanism analysis of the actual heat exchange system, based on the principle of focusing on the main factors and ignoring the secondary factors, the heat transfer process of the system is divided into five parts: Radiative heat transfer between each heater and the wafer, heat conduction from each heater to the thermal insulation layer, heat conduction inside the wafer, radiative heat transfer between the edge heaters and the top and bottom surfaces of the furnace tube reaction chamber, and radiative heat transfer between the wafer and the top and bottom surfaces of the furnace tube reaction chamber.

[0030] Among them, the radiative heat transfer between each heater and the wafer, that is, the radiative heat transfer between each heater and each temperature zone, that is, one heater has radiative heat transfer to 5 temperature zones.

[0031] The heat conduction inside the wafer, that is, the heat conduction between each temperature zone.

[0032] The radiative heat transfer between the edge heaters and the top and bottom surfaces of the furnace tube reaction chamber, that is, the radiative heat transfer between heater one and the top surface of the furnace tube reaction chamber, and the radiative heat transfer between heater five and the bottom surface of the furnace tube reaction chamber; The radiative heat transfer between the wafer and the top and bottom surfaces, that is, the radiative heat transfer between temperature zone one and the top surface of the furnace tube reaction chamber, and the radiative heat transfer between temperature zone five and the bottom surface of the furnace tube reaction chamber.

[0033] The heat source input of the heat exchange system is the electric power of each of the 5 heaters.

[0034] Figure 2The schematic diagram of the simplified mechanism model is shown. The system is divided into five temperature zones from top to bottom. The first temperature zone is a virtual zone containing a small amount of transparent glass and a large amount of vacuum. The second, third, and fourth temperature zones are the areas occupied by the wafers. The fifth temperature zone is a virtual zone containing frosted glass and a small amount of vacuum.

[0035] Since the radiative heat transfer between the wafer and the top and bottom surfaces is small, and the outer wall of the top surface is constantly at the ambient temperature while the bottom flange is temperature-controlled at 150 °C, the radiative heat transfer model for this part is simplified to a heat conduction model (the fourth term on the right side of Equation 1) to reduce the model complexity.

[0036] In Figure 2 the simplified model shown, the heater receives the power converted from the current , that is, the heater power .

[0037] The heat exchange between the heater and the outside world includes two parts: the radiative heat transfer between each heater and each temperature zone, and the heat conduction from each heater to the thermal insulation layer.

[0038] The heat exchange between the temperature zone and the outside world mainly includes three parts: the radiative heat transfer between each temperature zone and each heater, the heat conduction between each temperature zone, and the heat conduction from the upper and lower temperature zones (the first and fifth temperature zones) to the top and bottom surfaces.

[0039] That is, among the five parts of the above heat transfer process, the secondary factors include: the radiative heat transfer between the edge heater and the top and bottom surfaces of the furnace tube reaction chamber.

[0040] Step 1A22: Establish a mechanism model based on the heater and the temperature zone.

[0041] In this step, the relationship between the output power and temperature of the heater and the relationship between the temperatures of each temperature zone are expressed by the mechanism model constructed by analyzing the actual system.

[0042] The mechanism model includes the heat exchange model between the heater and the outside world and the heat exchange model between the temperature zone and the outside world.

[0043] The heat exchange model between the heater and the outside world is: (1)

[0044] Where: - is a vector with a dimension of 5*1, and the elements from top to bottom represent the temperatures of each heater, that is, the measured values of Outer TC. In the formula represents the vector obtained by taking the fourth power of each element in ; - A vector with a dimension of 5*1, and the elements from top to bottom represent the temperatures of each temperature zone, that is, the measured values of Inner TC. In the formula represents the vector obtained by taking the fourth power of each element in That is, in Figure 2 among the 10 temperature sensors, the 5 temperature values close to each temperature zone represent the temperatures of adjacent temperature zones respectively, and the 5 temperature values close to each heater represent the temperatures of adjacent heaters respectively.

[0045] - A scalar, the outer wall temperature of the furnace tube reaction chamber, that is, the ambient temperature, obtained through simulation; The result of - A diagonal matrix with a dimension of 5*5, and the diagonal elements from top to bottom represent the masses of each heater respectively, obtained through measurement; - A diagonal matrix with a dimension of 5*5, and the diagonal elements from top to bottom represent the specific heat capacities of each heater respectively, obtained by referring to data; - A vector with a dimension of 5*1, and the elements from top to bottom represent the temperature change values of each heater per unit time, that is the differential of - A vector with a dimension of 5*1, and the elements from top to bottom represent the powers of each heater; - A diagonal matrix with a dimension of 5*5, and the diagonal elements from top to bottom represent the emissivities of each heater respectively, obtained by referring to data; - A diagonal matrix with a dimension of 5*5, and the diagonal elements from top to bottom represent the emissivities of each temperature zone respectively, obtained by referring to data; - A scalar, the Stefan-Boltzmann constant, obtained by referring to data; - A matrix with a dimension of 5*5, the radiation angle coefficient matrix of each heater to each temperature zone. The elements in the first row of the matrix from left to right represent the radiation angle coefficients from heater 1 to heater 5 to temperature zone 1 respectively, the elements in the second row from left to right represent the radiation angle coefficients from heater 1 to heater 5 to temperature zone 2 respectively, and so on; The value of each element in

[0046] is obtained through simulation calculation or artificial setting according to the symmetry of radiative heat transfer. It can be known that the radiation angle coefficient matrix from each temperature zone to each heater is the transpose of - A diagonal matrix with a dimension of 5*5. The diagonal elements from top to bottom represent the lateral surface areas of each heater, that is, the radiation areas of each heater to each temperature zone; the lateral surface area is obtained by calculation according to the shape of the heater. - A diagonal matrix with a dimension of 5*5. The diagonal elements from top to bottom represent the lateral surface areas of each temperature zone, that is, the radiation areas of each temperature zone to each heater, and are obtained through the calculation formula for the lateral surface area of a cylinder. - A scalar, the thermal conductivity of the outer wall of the heater, obtained through simulation.

[0047] In formula (1), there are a total of 4 terms on the right side of the equal sign, among which: The meaning of the first term is the power of each heater heated by current. The meaning of the second term is the thermal radiation power of each heater to each temperature zone. The meaning of the third term is the thermal radiation power of each temperature zone to each heater. The meaning of the fourth term is the heat conduction power of each heater to the environment through the heat insulation layer on the side wall of the heater. The meaning on the left side of the equal sign is: the heating power of each heater.

[0048] The heat exchange model between the temperature zone and the outside is: (2)

[0049] Among them: - A scalar, the temperature of the outer wall of the top surface of the furnace tube reaction chamber; The result of is a vector; - A scalar, the temperature of the temperature control flange, 150 °C; The result of is a vector; Because , is the ambient temperature, so in the following description, is written as .

[0050] - A vector with a dimension of 5*1. The elements from top to bottom represent the temperature change values per unit time of each temperature zone, that is, The differential of; - A matrix with a dimension of 5*5, the heat conduction coefficient matrix between adjacent wafers (temperature zones), obtained through simulation; - A scalar, the radius of the horizontal cross-section of the wafer, that is, the radius of the cross-section, obtained through measurement; - The value of pi is 3.14; - A vector with a dimension of 5*1. The elements from top to bottom represent the heat transfer coefficients between each temperature zone and the top outer wall. The first element is obtained through simulation, and the last four elements are zero; - A vector with a dimension of 5*1. The elements from top to bottom represent the heat transfer coefficients between each temperature zone and the temperature control flange. The first four elements are zero, and the last element is obtained through simulation; - A 5*5 diagonal matrix. The diagonal elements from top to bottom represent the total mass of all solid parts (wafer, frosted glass, transparent glass) in each temperature zone, which is obtained through measurement; - A 5*5 diagonal matrix. The diagonal elements from top to bottom represent the specific heat capacities of each temperature zone, which are obtained by referring to data;

[0051] In formula (2), there are a total of 5 terms on the right side of the equal sign, among which: The meaning of the first term is the thermal radiation power from each heater to each temperature zone; The meaning of the second term is the thermal radiation power from each temperature zone to each heater; The meaning of the third term is the heat transfer power between each temperature zone; The meaning of the fourth term is the heat transfer power from each temperature zone to the top outer wall; The meaning of the fifth term is the heat transfer power from each temperature zone to the temperature control flange (bottom surface); The meaning of the left side of the equal sign is the heating power of each temperature zone.

[0052] Combining the above two energy balance equations, the final non-linear mechanism model is obtained: (3)

[0053] Step 1A3: Based on the mechanism model, select the model coefficients to be estimated.

[0054] Define , , , , as the parameters to be identified and used as the coefficients to be estimated for subsequent parameter identification.

[0055] Step 1B: Design experiments based on the heat exchange system to obtain temperature response data for estimating the model coefficients. Specifically, it includes the following steps: Step 1B1: Replace the first controller in the heat exchange system with 5 PID controllers; Each PID controller is set corresponding to a temperature zone, and each PID controller is connected to the corresponding heater, Outer TC, and Inner TC.

[0056] Step 1B2: Conduct tests simultaneously for all temperature zones to obtain the heater power and temperature response data corresponding to the temperature zones.

[0057] Control the temperature (measured value) of the Inner TC at 300 °C. After stabilizing for a period of time T, turn off the PID controller, and at the same time calculate the average value of the heater power within the previous T time at the current moment as the reference value.

[0058] During the process of turning off the PID controller, the temperature of the temperature control system may have a small fluctuation. After the system temperature stabilizes again, apply a heater power excitation with a duration of one hour, and record the temperature response data of each temperature zone and the heater power applied within this one hour. 。

[0059] The temperature response data includes: 、 。

[0060] Apply a heater power excitation with a duration of one hour, that is, the reference value plus a random signal excitation as the final heater power and continuously output for one hour.

[0061] The random signal excitation specifically refers to a random binary sequence (RBS) or a pseudo-random binary sequence (PRBS).

[0062] In this embodiment, an RBS sequence with a frequency range of 0.1 Hz to 0.001 Hz is applied to the heater through the PID controller, and the amplitude range is the current output power of the heater. When applying the excitation, the PID algorithm is not enabled.

[0063] After completing the test at 300 °C, conduct tests at 400 °C, 500 °C, and 600 °C respectively according to the same steps and record the heater output power and temperature response data.

[0064] In summary, in this step, a direct PID controller is applied to each temperature zone. From low temperature (taking 300 °C as an example) to high temperature (taking 600 °C as an example), at intervals of 100 °C. After the system temperature approaches stability for 20 minutes, while turning off the closed-loop control, calculate the average value of the control output within the previous 20 minutes at the current moment as the reference value and directly output it. At this time, the system temperature may have a small fluctuation. After the system temperature stabilizes again, apply a random signal excitation with a duration of one hour (that is, the reference value plus the signal excitation as the final control output value and continuously output for one hour), and obtain the temperature data of each temperature zone and the heater within this one hour.

[0065] Finally, a total of four one-hour identification data segments can be obtained from the experiment, which consist of the control output (heater power) of each temperature zone and the corresponding system temperature (temperature response data), and serve as the data input for the next step (offline identification of the parameters of the mechanism model).

[0066] Step 1C: Use the non-linear mechanism model as the non-linear regression model. Select the heater power ( the elements in) as the independent variable, and the measured values of Inner TC and Outer TC as the dependent variables, and use iterative least squares estimation to estimate all the model coefficients to be estimated.

[0067] That is, estimate the model coefficients of the mechanism model through the temperature response data and heater output power obtained from the identification experiment (random signal identification experiment) in Step 1B.

[0068] The model coefficient estimation algorithm usually needs to determine the parameters to be identified according to the modeling situation, and implement the least squares estimation or maximum likelihood estimation in accordance with the principle of the method of undetermined coefficients.

[0069] Step 1D: Substitute the model coefficients obtained in Step 1C into the mechanism model in Step 1A22 to update the mechanism model.

[0070] This step finally obtains a mechanism model with completely determined coefficients, which serves as the reference physical model source for the next two steps (Step 3 design of the unscented Kalman estimator based on the mechanism model and Step 4 design of the feedback controller based on the linearization of the mechanism model).

[0071] Step 2: Design an unscented Kalman estimator.

[0072] Select the mechanism model output in Step 1D to design the unscented Kalman estimator.

[0073] Designing an unscented Kalman estimator from the mechanism model belongs to the prior art.

[0074] The unscented Kalman estimator is used to convert the temperature feedback value of the system into the real-time state of the system.

[0075] Specifically, in the subsequent Step 5, in the real-time temperature control system, an excitation is applied to obtain the temperature feedback value of the system. The temperature feedback value of the system is used as a feedback signal and input into the estimator in real time in each control cycle. The estimator calculates the system state (estimated temperature) through the unscented Kalman estimation method based on the input and outputs it to the controller in real time.

[0076] Among them, the temperature feedback value includes the measured value of Inner TC.

[0077] The real-time state of the system refers to the estimated temperature of the system. The estimated temperature refers to the estimated value of the Inner TC calculated by the estimator by integrating the temperature feedback value and the mechanism model.

[0078] Unscented Kalman filter, also known as Unscented Kalman estimator, is a method for state estimation by linearizing the nonlinear function of a random variable through linear regression of n sampled points in the prior distribution and then combining this statistical linearization method (unscented transform) with the Kalman filter. Figure 6 Shows the schematic diagram of the unscented transform.

[0079] The parameters of the linearized model of the actual furnace tube under different working conditions are variable. Through the nonlinear expression of the mechanism model, the estimator can predict and update this parameter change in advance in real time without completely relying on the lagging temperature signal feedback, which enables the estimator to obtain a faster and more accurate system state estimation value.

[0080] The estimator designed through this step will be used as the pre-computation module of the controller obtained in the next step (step 4: design of the feedback controller based on the linearization of the mechanism model).

[0081] Step 3: After linearizing the mechanism model, design the feedback gain of the controller to form the second controller.

[0082] In this embodiment, the mechanism model, the estimator, and the controller are all the same, and they are all models with 5 inputs and 5 outputs.

[0083] To design the feedback gain, linearize the above-mentioned nonlinear mechanism model formula 3. The linearized system state space equation is: (4) That is, after obtaining the Jacobian matrix (the first matrix on the right side of the equal sign) of the mechanism model output in step 1D to get the linear model, a system state feedback gain controller is obtained through a state feedback gain design method based on the linear model, such as the linear quadratic optimal algorithm. The form of the controller is as follows.

[0084] (5) - A 5*5 matrix, the feedback gain matrix.

[0085] - A 5*1 vector, and the elements from top to bottom represent the target reference temperatures of each temperature zone respectively.

[0086] Figure 4 In, the linear optimal quadratic control is a method that uses the linearized model derived from the mechanism model as the reference model and combines the linear optimal quadratic control (LQR) for power control.

[0087] The linearized model generates a feedback gain matrix, which belongs to the prior art.

[0088] The optimal quadratic control (LQR) is integrated into the second controller.

[0089] Specifically, in the subsequent step 5, the second controller takes the real-time output of the estimator designed in the previous step (the unscented Kalman estimator design based on the mechanism model) as the input, and jointly calculates the final control output of the system (the real-time heater power) through the linear quadratic optimal algorithm with the reference temperature (target reference temperature) obtained by the target feedforward processing, thereby changing the power to finally achieve the effect of controlling the temperature of each temperature zone.

[0090] Figure 4 In, the "reference temperature" is the target reference temperature.

[0091] The "temperature target" includes the target value of the Inner TC.

[0092] Step 4: Construct a temperature control system.

[0093] The temperature control system includes: 1 unscented Kalman estimator and 1 second controller. That is, the target feedforward processing, the estimator, and the controller are an integral whole. The whole replaces the cascade PID controller.

[0094] Replace the first controller in the heat exchange system with 1 second controller; Connect the input end of the unscented Kalman estimator to all Inner TCs; Connect the output end of the unscented Kalman estimator to the first input end of the second controller; Connect the second input end of the second controller to the target temperature module; the target temperature module is used to output the target reference temperature outside the temperature zone.

[0095] Connect the output end of the second controller to all heaters in the heat exchange system.

[0096] In this embodiment, the target temperature module is a target feedforward processing module.

[0097] The purpose of the target feedforward processing is to eliminate control overshoot.

[0098] When designing a general controller, a ramp limit is introduced to reduce the control overshoot caused by the target step jump. However, there is a non-differentiable point of the first order at the target temperature point for the ramp limit. Although the control overshoot will be significantly improved after introducing the ramp limit, due to this non-differentiable point, especially in a temperature control system with large inertia, it is inevitable to have control overshoot.

[0099] To avoid this situation, the target feedforward processing module designs a uniformly decelerated transition interval so that the target temperature change can be guided everywhere, completely eliminating temperature control overshoot.

[0100] Among them, the expression of the transition interval of uniform deceleration is: .

[0101] Figure 5 Schematic diagram of target feedforward processing.

[0102] Step 5: All Inner TCs acquire temperature data in real time, and the second controller outputs heating power to the corresponding heater, so that the error between the measured value of the Inner TC and the corresponding target reference temperature is within the set range.

[0103] Figure 4 The whole process from given target temperature and measured temperature feedback data to calculated heater power after the implementation of controller No. 2 is demonstrated.

[0104] The No. 2 controller receives the target temperature signal of the Inner TC sent by the furnace operator through the human-computer interaction interface, and measures the temperature of the reaction chamber in real time through 10 temperature sensors. The set power of the heater is calculated according to the control method implemented in the controller, and finally the measured value of the Inner TC is maintained near the target temperature.

[0105] In addition, Figure 4 The control block diagram described in , the unscented Kalman filter method can also achieve similar effects through the extended Kalman filter EKF method; the linear quadratic optimal control method can also achieve similar effects through other feedback gain control such as model reference adaptive MRAC, H∞ control, model predictive control MPC, etc.

[0106] In summary, the simplified mechanism model proposed in this embodiment is a model simplification idea obtained by the inventor based on mechanism analysis and his own experience. The model obtained after repeated attempts is the core innovation point.

[0107] The target feedforward processing method proposed in this embodiment cleverly avoids the problem of control overshoot and achieves good experimental results in experiments, which is an innovation point that is key to protection.

[0108] Figure 7 The control effect comparison between the control method based on the mechanism model proposed in the present invention and the control method based on the cascade PID is presented. In the figure, it can be seen that the control method based on the mechanism model performs significantly better than the cascade PID in terms of control deviation.

[0109] Figure 8It is presented that the target feedforward processing method proposed by the present invention performs excellently in suppressing control overshoot compared with the traditional ramp processing method, and the new target feedforward processing method can achieve the effect of no control overshoot.

[0110] The above are only the preferred embodiments of the present invention and do not impose any limitation on the present invention. Any person skilled in the art, within the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification and other changes to the technical solution and technical content disclosed by the present invention, which are all within the content of the technical solution of the present invention and still fall within the protection scope of the present invention.

Claims

1. A temperature control method for an ALD device based on a mechanism model, characterized in that, It includes the following steps: Step 1, establish and update the mechanism model: Step 2, design an unscented Kalman estimator: Select the mechanism model output in Step 1 to design an unscented Kalman estimator; Step 3, linearize the mechanism model and write it in the form of a state-space equation, design the feedback gain of the controller to form the second controller; Step 4, construct a temperature control system: Connect the input end of the unscented Kalman estimator to all Inner TCs; Connect the output end of the unscented Kalman estimator to the first input end of the second controller; Connect the second input end of the second controller to the target temperature module; the target temperature module is used to output the target reference temperature on the temperature zone side; Connect the output end of the second controller to all heaters in the heat exchange system; Step 5, all Inner TCs obtain temperature data in real time, and the second controller outputs heating power to the corresponding heaters so that the error between the measured value of the Inner TC and the corresponding target reference temperature is within the set range.

2. The temperature control method of the ALD device based on the mechanism model according to claim 1, wherein, Step 1 specifically includes the following steps: Step 1A, establish a mechanism model based on the heat exchange system and select the model coefficients to be estimated, which specifically includes the following steps: Step 1A1, establish a heat exchange system, which includes: A wafer placement area, which is located at the center of the furnace tube reaction chamber and includes several sub-areas distributed from top to bottom; a temperature control zone is formed between each sub-area, between the wafer placement area and the upper surface of the furnace tube reaction chamber, and between the wafer placement area and the lower surface of the furnace tube reaction chamber respectively; A furnace tube reaction chamber, on the inner wall surface of its side wall (1), heaters corresponding to the temperature control zones are provided, and a heat insulation layer is provided on the outer wall of the heaters; A first sensor Inner TC arranged near the wafer placement area and a second sensor OuterTC arranged near the heater; And a first controller, which is located outside the furnace tube reaction chamber and is connected to the Inner TC, OuterTC and each heater; Step 1A2, establish a mechanism model: Step 1A21, analyze the heat exchange system and select the modeling object of the mechanism model; Step 1A22, establish a mechanism model based on the heaters and temperature zones; Step 1A3, based on the mechanism model, select the model coefficients to be estimated; Step 1B, design an experiment based on the heat exchange system to obtain temperature response data for estimating the model coefficients; the temperature response data includes the measured values of the Inner TC and the Outer TC; Step 1C, use the nonlinear mechanism model as a nonlinear regression model, select the heater output power as the independent variable, and the measured values of the Inner TC and the Outer TC as the dependent variables, and use iterative least squares estimation to estimate all the model coefficients to be estimated; Step 1D, substitute the model coefficients obtained in Step 1C into the mechanism model in Step 1A to update the mechanism model.

3. The temperature control method of the ALD device based on the mechanism model according to claim 2, wherein Step 1B specifically includes the following steps: Step 1B1, replace the first controller in the heat exchange system with 5 PID controllers; connect each PID controller to the corresponding heater, Outer TC, and Inner TC; Step 1B2: Conduct tests for all temperature zones to obtain the heater output power and temperature response data corresponding to the temperature zones; the temperature response data includes: the measured values of the Outer TC and the measured values of the Inner TC.

Images