Method for controlling the temperature of a vaporizer and a substrate processing apparatus.

Abstract

To quickly stabilize a temperature of a vaporizer at a set temperature.SOLUTION: When controlling a temperature of a vaporizer that is equipped with a heater to heat and vaporize an inflowing chemical liquid, an input power to the heater at a current time is determined based on an inflow amount of chemical liquid to the vaporizer at each time during a first prediction period from the current time to a predetermined future time, a measured temperature of the vaporizer at the current time, and a predicted temperature of the vaporizer at each time during the first prediction period.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] This disclosure relates to a method for controlling the temperature of a vaporizer and a substrate processing apparatus. [Background technology]

[0002] In film deposition equipment used as substrate processing equipment, a direct liquid injection (DLI) vaporizer is sometimes used as a means of supplying process gas (see, for example, Non-Patent Document 1). In a DLI vaporizer, the vaporizer is maintained at a target temperature (hereinafter referred to as the "set temperature") by a heater or the like, and the process gas is generated by heating and vaporizing the incoming chemical solution.

[0003] Conventionally, DLI-type vaporizers maintain a constant power input to the heater of the vaporizer, regardless of whether or not chemical solution is being added, thereby maintaining the vaporizer's temperature at a set temperature. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] "Direct vaporization method", [online], Horiba, Ltd., [Retrieved December 28, 2022], Internet <URL:https: / / www.horiba.com / jpn / semiconductor / key-technologies / element-technology / vaporization-methods / l> [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] The technology disclosed herein stabilizes the vaporizer temperature to a set temperature quickly. [Means for solving the problem]

[0006] One aspect of the technology relating to this disclosure is a temperature control method for a vaporizer equipped with a heater that heats and vaporizes an incoming chemical solution, wherein the input power to the heater at the current time is determined based on the amount of the chemical solution flowing into the vaporizer at each time in a first prediction period from the current time to a predetermined future time, the measured temperature of the vaporizer at the current time, and the predicted temperature of the vaporizer at each time in the first prediction period. [Effects of the Invention]

[0007] According to the technology disclosed herein, the vaporizer temperature can be stabilized to a set temperature quickly. [Brief explanation of the drawing]

[0008] [Figure 1] This figure illustrates a DLI-type vaporizer and its surrounding equipment to which the vaporizer temperature control method according to one embodiment of the present invention is applied. [Figure 2] This is a diagram illustrating model predictive control. [Figure 3] This graph illustrates the difference between model predictive control and PID control. [Figure 4] This is a block diagram illustrating the concept of model predictive control applied to the vaporizer temperature control method according to this embodiment. [Figure 5] This is a flowchart showing the temperature control method for a vaporizer according to this embodiment. [Modes for carrying out the invention]

[0009] Incidentally, in recent years, thin-film deposition processes using multiple process gases have become widely used, requiring thin-film deposition equipment to be equipped with multiple vaporizers. Furthermore, miniaturization of vaporizers is required in thin-film deposition equipment in order to easily accommodate the arrangement of multiple vaporizers.

[0010] As the vaporizer becomes smaller, its heat capacity decreases, making it more susceptible to the effects of the heat of vaporization of the chemical solution. If the heater input power is kept constant, the vaporizer temperature will change significantly in response to changes in the amount of chemical solution flowing into the vaporizer (hereinafter referred to as "vaporization flow rate").

[0011] Therefore, it is conceivable to control the heater input power using PID control, a common type of feedback control, to maintain the vaporizer temperature at a set temperature.

[0012] However, with PID control, the heater input power is only adjusted after the vaporizer temperature begins to deviate from the set temperature. Therefore, it takes some time for the vaporizer temperature to return to the set temperature. In particular, if the vaporization flow rate changes significantly when the chemical solution starts or stops flowing, it takes a long time for the vaporizer temperature to return to the set temperature.

[0013] If it takes time for the vaporizer temperature to return to the set temperature, for example, if the vaporizer temperature drops, the amount of chemical that cannot be vaporized increases, resulting in greater chemical loss. In addition, because the vaporizer is subjected to large temperature fluctuations for a longer period, there is a problem that the vaporizer itself deteriorates more quickly.

[0014] In contrast, the technology described herein uses model predictive control to stabilize the vaporizer temperature to a set temperature at an early stage.

[0015] Hereinafter, an embodiment of the technology relating to this disclosure will be described with reference to the drawings. Figure 1 is a diagram illustrating a DLI-type vaporizer and its surrounding equipment to which the vaporizer temperature control method according to an embodiment of the present invention is applied. In Figure 1, a DLI-type vaporizer 10 is connected to an MFC (Mass Flow Controller) 11 and a liquid MFC 12. A carrier gas is supplied to the MFC 11 from a tank (not shown). A chemical solution extruded by the pressure of a pressurized gas is supplied to the liquid MFC 12 from a chemical solution tank 13.

[0016] The vaporizer 10 includes a heater 14, a thermometer 15, and a controller 16. Power is input to the heater 14 from the outside. The thermometer 15 measures the temperature of the vaporizer 10 (hereinafter simply referred to as the "vaporizer temperature") and transmits the measurement result to the controller 16. Based on the measurement result of the thermometer 15 and the recipe information obtained from the external memory 17, the controller 16 determines the input power to the heater 14 (hereinafter simply referred to as the "heater input power") so as to maintain the vaporizer temperature at the set temperature. Here, the recipe information is, for example, the recipe information of the film forming process executed by a film forming apparatus (substrate processing apparatus) including the vaporizer 10, and the vaporization flow rate at each time is preset.

[0017] The MFC 11 controls the inflow rate of the supplied carrier gas into the vaporizer 10. The liquid MFC 12 controls the inflow rate of the supplied chemical solution into the vaporizer 10. The vaporizer 10 vaporizes by spraying the supplied chemical solution into a vaporization chamber (not shown) built therein, and mixes the vaporized chemical solution with the carrier gas to generate a process gas. At this time, the chemical solution is heated by the heat of the heater 14 to achieve stable continuous vaporization. Note that the vaporizer 10 may not include the controller 16, and an external controller may determine the heater input power based on the measurement result of the thermometer 15 and the recipe information.

[0018] By the way, in the vaporizer 10, when determining the heater input power based on the measurement result of the thermometer 15, if PID control, which is a general feedback control, is used, the heater input power starts to be adjusted after the vaporizer temperature fluctuates from the set temperature. Therefore, it takes a certain amount of time for the vaporizer temperature that has fluctuated from the set temperature to return to the set temperature. Therefore, in the present embodiment, model predictive control, which is one of the feedback controls, is used instead of PID control.

[0019] Model Predictive Control (MPC) is a control method that optimizes a system by predicting its future response at each time point in a prediction period. In MPC, a predictive model is set up that represents the dynamic characteristics of the system being controlled, and its future behavior over a finite interval from the current time in the prediction period is predicted. Possible predictive models include step response models, impulse response models, transfer function models, autoregressive moving average models (ARX models), or discrete state equations.

[0020] Next, we will explain the relationship between the output of the controlled object (hereinafter referred to as "control output") and the input to the controlled object (hereinafter referred to as "control input"). In model predictive control, if we want the control output to track a target value, we search for a time series of control inputs that minimizes the area of ​​the tracking error corresponding to the sum of the products of the differences between the target value and the control output at each time point during the prediction period from the current time to future times.

[0021] For example, consider a case in model predictive control where the control input at time t is u[t], the control output at time t is y[t], and the discrete state equation shown in equation (1) below is used as the predictive model. Note that C and D in equation (1) below are weights, which are determined through prior experiments, etc. y[t+1]=Cy[t]+Du[t] … (1)

[0022] First, using the control output of the current time and the provisionally set control input of the current time, the control output of the next time is calculated based on equation (1) above. The control output of the current time used here is an actual measured value. Next, using the calculated control output of the next time and the provisionally set control input of the next time, the control output of the next time is calculated again based on equation (1) above. This is repeated to calculate the control output of each time in the first prediction period from the current time to a predetermined future time. Then, the control input of each time in the first prediction period that minimizes the area of ​​the tracking error between the target value of each time in the first prediction period and the calculated control output of each time in the first prediction period is searched for. Here, the control input of each time is a changeable value. Then, in model predictive control, the control input of each time (see each white circle in Figure 2(B)) is adjusted to minimize the area of ​​the tracking error (the area shown by hatching in Figure 2(A)). That is, the control input of each time in the first prediction period that minimizes the area of ​​the tracking error (the time series of control inputs) is searched for. The search for the time series of control inputs here is a kind of optimization problem. Then, the first element of the time series of control inputs obtained through the search is applied to the controlled object as the actual control input at the current time.

[0023] Furthermore, at the next time after a predetermined sampling time has elapsed from the current time, the control output for the next time is calculated based on equation (1) above, using the control output for the next time and the provisionally set control input for the next time. The control output for the next time used here is an actual measured value. Next, the control output for the time after the calculated time and the provisionally set control input for the time after the next time is calculated based on equation (1) above, using the control output for the time one more than the next time. The above is repeated to calculate the control output for each time in the prediction period from the next time to a predetermined time in the future (hereinafter referred to as the "second prediction period"). Then, the control input for each time in the second prediction period that minimizes the area of ​​the tracking error between the target value for each time in the second prediction period and the calculated control output for each time in the second prediction period (the area shown by hatching in Figure 2(C)) is searched for. Then, the first element of the time series of the control inputs obtained by the search (see each white circle in Figure 2(D)) is applied to the controlled object as the actual control input for the next time.

[0024] The above process is repeated each time a predetermined sampling period has elapsed, and the control input for each time point is determined using the control output, which is the measured value at each time point. When determining the control input for the next time point after a predetermined sampling period has elapsed from the current time point, the measured value of the control output for the next time point, which is the result of applying the control input for the current time point to the controlled object, is used. Therefore, model predictive control can be considered a form of feedback control.

[0025] In model predictive control, the sampling time and prediction period are arbitrary and are set according to the dynamic characteristics of the controlled object. For example, for controlled objects whose fluctuations are difficult to converge, the sampling time and prediction period are set to be relatively long.

[0026] Figure 3 is a graph illustrating the difference between model predictive control and PID control. Here, we consider the case where the vaporization flow rate of vaporizer 10 increases at time T1. In this case, at time T1, the heat from vaporizer 10 (its heater) is absorbed by the vaporization of the chemical solution, causing the vaporizer temperature to decrease.

[0027] When controlling the vaporizer temperature using PID control, the heater input power is only adjusted after the vaporizer temperature has started to deviate from the set temperature (see dashed line in the figure) at time T1. In other words, the start of the heater input power adjustment is delayed, and it takes some time for the vaporizer temperature, which has deviated from the set temperature, to return to the set temperature.

[0028] On the other hand, when controlling the vaporizer temperature using model predictive control, once time T1 is included in the prediction period, the effect of the increase in vaporization flow rate of vaporizer 10 at time T1 is reflected in the control output, the vaporizer temperature, by the prediction model. In other words, the effect of the increase in vaporization flow rate of vaporizer 10 at time T1 can be predicted in advance of time T1, and the heater input power adjustment starts in advance of time T1 based on the predicted result. As a result, the fluctuating vaporizer temperature can be stabilized at the set temperature early. However, since the heater input power adjustment starts in advance of time T1, the vaporizer temperature starts to fluctuate from the set temperature in advance of time T1. Also, in model predictive control, the heater input power is adjusted in advance of the effect of the increase in vaporization flow rate of vaporizer 10, so the amount of fluctuation in vaporizer temperature in model predictive control is smaller than the amount of fluctuation in vaporizer temperature in PID control.

[0029] Figure 4 is a block diagram illustrating the concept of model predictive control applied to the vaporizer temperature control method according to this embodiment. In Figure 4, the controller 18 that performs model predictive control (MPC) is configured in the controller 16 and includes a predictive model 19 and an optimizer 20. The controller 18 is also input with the set temperature of the vaporizer 10 for the prediction period as a target command.

[0030] Prediction model 19 is a model that predicts the temperature of the vaporizer 10, which is the target of control. As shown in the figure, the model is input to the vaporization flow rate at each time in the prediction period, which is obtained from the recipe information. Prediction model 19 can be represented by the discrete state equation shown in equation (2) below, for example. In equation (2) below, A, B, and V are weights. x[t+1]=Ax[t]+Bu[t]+Vuf[t] … (2)

[0031] In equation (2) above, u[t] is the heater input power, which is the control input at time t, x[t] is the vaporizer temperature, which is the control output at time t, and uf[t] is the vaporization flow rate at time t. Here, x[t] is the measured value of the vaporizer temperature at time t. Furthermore, since the prediction model 19 has a term corresponding to the vaporization flow rate, the model predictive control in this embodiment can predict the vaporizer temperature while taking into account the effect of changes in the vaporization flow rate. In addition, although uf[t] is shown as a function of time t in equation (2) above, in reality it is a value predetermined in the recipe information and is therefore treated as a constant term.

[0032] The optimizer 20 is a block for solving the optimization problem. The method for solving the optimization problem used by the optimizer 20 is not particularly limited. In this embodiment, the optimizer 20 searches for a time series of heater input power that minimizes the area of ​​the tracking error of the heater 14 temperature (predicted from the prediction model 19) with respect to the set temperature of the heater 14 during the prediction period by adjusting the heater input power, which is the control input at each time step.

[0033] The controller 18 then inputs the first element of the time series of heater input power obtained through the search to the vaporizer 10 as the heater input power to be actually input. In addition, the vaporizer temperature (control output) as a result of the input heater power is measured and fed back to the controller 18.

[0034] In other words, in the model predictive control of this embodiment, the vaporization flow rate at each time point during the prediction period is obtained from the recipe information, and the actual heater input power to be input is determined using the measured value of the vaporizer temperature.

[0035] Figure 5 is a flowchart showing the temperature control method for the vaporizer according to this embodiment. The temperature control method in Figure 5 is executed, for example, by the controller 16 of the vaporizer 10 according to a predetermined program. In this embodiment, before executing the temperature control method in Figure 5, the prediction model shown in equation (2) above is identified in advance through experiments using the vaporizer 10 or a similar model thereunder. At this time, the weights A, B, and V of equation (2) above are determined.

[0036] In Figure 5, first, the vaporizer temperature at the current time is obtained (step S51), and the vaporization flow rate at each time in the first prediction period is obtained from the recipe information (step S52).

[0037] Next, using the acquired vaporizer temperature and vaporization flow rate at the current time, and the provisionally assumed heater input power at the current time, the vaporizer temperature at the next time is calculated based on equation (2) above. Then, using the calculated vaporizer temperature at the next time and the provisionally assumed control input at the next time, the vaporizer temperature at the following time is calculated again based on equation (2) above. The above is repeated to calculate the vaporizer temperature at each time in the first prediction period.

[0038] Then, by solving an optimization problem, we search for a time series of heater input power that minimizes the fluctuation in vaporizer temperature corresponding to the area of ​​the tracking error between the target temperature and the vaporizer temperature during the first prediction period (step S53).

[0039] Incidentally, considering that Vuf[t] is treated as a constant term in equation (2) above, equation (2) can also be said to be a function of the heater input power u. Here, the vaporizer temperature x(u) at each time in the calculated first prediction period is expressed as a vector X after evolving over time, and the set temperature x at each time in the first prediction period is expressed as... ref vector X ref Let us express this as follows, and also consider expressing the heater input power at each time in the first prediction period as a vector U. In this case, minimizing the fluctuation in vaporizer temperature corresponds to minimizing the quadratic evaluation function shown in equation (3) below. Note that Q and P in equation (3) below are weights. Evaluation function: (X ref - X(U)) T Q(X ref -(U)) + U T PU … (3)

[0040] Also, when an upper limit and a lower limit are set as constraint conditions for the heater input power, for example, when the maximum value of the available heater input power is set, this constraint condition is formulated as an inequality constraint. Then, while considering the constraint condition, by solving the constrained optimization problem based on the evaluation function of the above formula (3), a time series of heater input power that satisfies the constraint condition can be searched for.

[0041] Note that when the above formula (3) is represented by a scalar, the evaluation function is expressed as a quadratic formula like the following formula (4), and minimizing the evaluation function corresponds to minimizing the sum of products of the values of the following formula (4) calculated at each time in the first prediction period. Evaluation function: Q(x ref - x(u)) 2 + Pu 2 … (4)

[0042] Next, the first element of the time series of heater input power that minimizes the searched evaluation function is determined as the heater input power to be input at the current time (step S54).

[0043] Then, steps S51 to S54 are repeated every time a predetermined sampling time elapses. For example, after determining the heater input power to be input at the current time using the vaporization flow rate at each time in the first prediction period, the heater input power to be input at the next time is determined using the vaporization flow rate at each time in the second prediction period. In this embodiment, for example, 1000 msec (1 second) is assumed as the predetermined sampling time, but the sampling time may be changed according to the dynamic characteristics of the temperature of the vaporizer 10.

[0044] Furthermore, when repeating steps S51 to S54, the measured value of the vaporizer temperature at the next time is used to determine the heater input power to be input at the next time after a predetermined sampling time has elapsed from the current time. Since this measured value of the vaporizer temperature at the next time is the result of inputting the heater input power to be input, which was determined at the current time, to the vaporizer 10, the vaporizer temperature control method according to this embodiment can be called feedback control.

[0045] According to this embodiment, when predicting the vaporizer temperature at each time point during the prediction period using model predictive control, the influence of changes in vaporization flow rate obtained from recipe information can be considered in the prediction of the vaporizer temperature. That is, since the influence of changes in vaporization flow rate can be anticipated before the vaporization flow rate changes, the heater input power can be adjusted before the vaporization flow rate changes. As a result, the fluctuating vaporizer temperature can be stabilized to the set temperature early. Furthermore, because the fluctuating vaporizer temperature can be stabilized to the set temperature early, for example, the amount of chemical solution that cannot be vaporized due to a drop in the temperature of the vaporizer 10 can be reduced, thereby reducing chemical solution loss. In addition, since the time during which the vaporizer 10 is subjected to large temperature fluctuations is shortened, deterioration of the vaporizer 10 itself can be suppressed.

[0046] Furthermore, in this embodiment, the heater input power is adjusted in anticipation of the effect of the increase in vaporization flow rate of the vaporizer 10, so the amount of fluctuation in vaporizer temperature is reduced, and the input power to the heater 14 used to adjust the temperature of the vaporizer 10 can also be reduced.

[0047] Furthermore, in this embodiment, when searching for a time series of heater input power that minimizes fluctuations in vaporizer temperature, it is possible to search for a time series of heater input power considering upper and lower limits by solving a constrained optimization problem. In other words, it is possible to search for a time series of heater input power that falls between the upper and lower limits of the heater input power.

[0048] While preferred embodiments of this disclosure have been described above, this disclosure is not limited to the embodiments described above, and various modifications and changes are possible within the scope of its essence.

[0049] For example, in this embodiment, the vaporizer temperature control method is executed by the controller 16 of the vaporizer 10 according to a predetermined program, but an external controller may also execute the method according to a predetermined program.

[0050] Furthermore, the vaporizer 10 to which the vaporizer temperature control method according to this embodiment is applied may be installed not only in a film deposition apparatus, but also in other types of substrate processing apparatus that use process gases, such as an etching apparatus.

[0051] Furthermore, in this embodiment, a discrete state equation was used as the prediction model for the temperature of the vaporizer 10, but the prediction model is not limited to this, and for example, an ARX model may be used.

[0052] In this case, the prediction model is shown in equation (5) below. x[t+1]=Ax[t]+Bu[t]+Vuf[t]+A1x[t-1]+B1u[t-1]+V1uf[t-1]+A2x[t-2]+B2u[t-2]+V2uf[t-2]+... … (5)

[0053] Note that A, B, V, A1, B1, V1, A2, B2, and V2 in equation (5) above are weights.

[0054] The ARX model allows us to predict the vaporizer temperature at the next time point by considering not only the current vaporizer temperature, heater input power, and vaporization flow rate, but also the vaporizer temperature, heater input power, and vaporization flow rate at previous time points. Therefore, the ARX model is suitable when it takes time for the chemical solution to reach the vaporizer 10 from the chemical solution tank 13, and the change in vaporizer temperature lags behind the change in vaporization flow rate. [Explanation of symbols]

[0055] 10. Vaporizer 14 Heater 16 Controllers 18 Controller 19 Predictive Models

Claims

1. A method for controlling the temperature of a vaporizer equipped with a heater that heats and vaporizes the incoming chemical liquid, A method for controlling the temperature of a vaporizer, which determines the input power to the heater at the current time based on the amount of liquid chemical flowing into the vaporizer at each time in a first prediction period from the current time to a predetermined future time, the measured temperature of the vaporizer at the current time, and the predicted temperature of the vaporizer at each time in the first prediction period.

2. A method for controlling the temperature of a vaporizer according to claim 1, comprising determining a time series of input power to the heater at each time in the first prediction period to minimize the temperature fluctuation of the vaporizer during the first prediction period, and determining the first input power to the heater in the time series as the input power to the heater at the current time.

3. The vaporizer temperature control method according to claim 1, wherein the amount of the chemical solution flowing into the vaporizer at each time in the first prediction period is obtained from recipe information of a process performed by a processing apparatus equipped with the vaporizer.

4. Based on the amount of the chemical solution flowing into the vaporizer at each time in a second prediction period from the current time after a predetermined sampling time has elapsed to a predetermined future time, the measured temperature of the vaporizer at the next time, and the predicted temperature of the vaporizer at each time in the second prediction period, the input power to the heater at the next time is determined. The vaporizer temperature control method according to claim 1, wherein the measured temperature of the vaporizer at the next time is the result of the determined input power to the heater at the current time being input to the heater.

5. The input power to the heater is determined using model predictive control. The vaporizer temperature control method according to claim 1, wherein the predictive model for predicting the vaporizer temperature in the model predictive control is a discrete state equation.

6. The input power to the heater is determined using model predictive control. The vaporizer temperature control method according to claim 1, wherein the predictive model for predicting the vaporizer temperature in the aforementioned model predictive control is the ARX model.

7. A substrate processing apparatus comprising a vaporizer to which the vaporizer temperature control method according to any one of claims 1 to 6 is applied.

Images