CONTROL DEVICE, IMPRINT DEVICE, AND ARTICLE MANUFACTURING METHOD
The control device addresses disturbance forces in manufacturing equipment by using a compensator and disturbance observer to enhance control accuracy and precision in semiconductor and MEMS manufacturing.
Patent Information
- Application Number
- JP2020144872
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-08-28
- Publication Date
- 2024-12-04
- Estimated Expiration
- 2040-08-28
AI Technical Summary
Existing control devices for manufacturing equipment, such as those used in semiconductor devices and MEMS, face challenges in maintaining control accuracy due to disturbance forces applied by mounting parts and guide components, which affect the precision of force application on controlled objects.
A control device incorporating a compensator that calculates an operation amount based on a force target value and a target driving force, a drive unit to apply force, an estimation unit to estimate disturbance forces, and a correction unit to generate command values, utilizing a disturbance observer to account for environmental changes and improve control accuracy.
The solution enhances control accuracy by effectively compensating for disturbance forces, ensuring precise force application and maintaining control precision in manufacturing processes.
Smart Images

Figure 0007594383000022 
Figure 0007594383000023 
Figure 0007594383000024
Abstract
Description
[Technical field]
[0001] The present invention relates to a control device, an imprint apparatus, and a method for manufacturing an article. [Background technology]
[0002] A manufacturing apparatus for manufacturing an article such as a semiconductor device or MEMS is required to perform high-precision control in order to improve the performance and / or yield of the article. To achieve this, it is necessary to control the controlled object in the manufacturing apparatus with high precision, and it is necessary to accurately apply a desired control force to the controlled object. However, since the controlled object is often accompanied by a mounting section and / or a guide for supplying power and / or fluid, a disturbance force is applied to the controlled object, and there is a problem that the desired control force cannot be applied to the controlled object. As a method for solving this problem, Patent Document 1 describes an active vibration isolation apparatus having a feedback correction section (disturbance observer) that estimates a disturbance input to a system based on an output from the system. The feedback correction section includes a nominal model that estimates an input based on an output from the system. [Prior art documents] [Patent documents]
[0003] [Patent Document 1] JP 2003-108236 A Summary of the Invention [Problem to be solved by the invention]
[0004] However, the vibration isolation apparatus of Patent Document 1 does not have a means for determining a nominal model, and therefore, when the state of the vibration isolation apparatus or the environment in which the vibration isolation apparatus is placed changes, the nominal model cannot be made to follow the changes.
[0005] An object of the present invention is to provide a technique that is advantageous for improving or maintaining control accuracy. [Means for solving the problem]
[0006] One aspect of the present invention relates to a control device for processing an object to be processed by pressing an object against the object, the control device comprising: a compensator that determines a manipulation amount based on a sum of a force target value that is a target value of a force required to process the object to be processed and a target driving force that is a target value of a driving force required to drive the object; a driving unit that applies a force to the object; an estimation unit that estimates an estimate of a disturbance force applied to the object based on information about a state of the object; a correction unit that generates a command value to be given to the driving unit by correcting the manipulation amount based on the estimated value so that the effect of the disturbance force is reduced; and a determination unit that executes a process to determine a plurality of parameter values to be used by the estimator to obtain the estimated value based on the information. Effect of the Invention
[0007] The present invention provides a technique that is advantageous for improving or maintaining control accuracy. [Brief description of the drawings]
[0008] [Figure 1] FIG. 2 is a schematic diagram showing the configuration of a control device in the first embodiment. [Diagram 2] FIG. 2 is a block diagram showing the configuration of a force feedback control system in the first embodiment. [Diagram 3] FIG. 2 is a block diagram showing the configuration of a force feedback control system that performs disturbance force compensation in the first embodiment. [Figure 4] FIG. 1 is a schematic diagram showing a disturbance estimation model using a three-element model. [Diagram 5] Schematic diagram showing state space. [Figure 6] Schematic diagram showing a disturbance estimation model using three 3-element models. [Figure 7] FIG. 2 is a block diagram showing the configuration of a position feedback control system in the first embodiment. [Figure 8] 3 is a flowchart showing a machine learning method according to the first embodiment. [Figure 9] FIG. 11 is a block diagram showing the configuration of a force feedback control system that performs disturbance force correction in a second embodiment. [Figure 10] Schematic diagram showing a neural network. [Figure 11] FIG. 13 is a schematic view showing the configuration of an imprint apparatus according to a fifth embodiment. [Figure 12] 13 is a flowchart showing an imprint process in an imprint apparatus according to a fifth embodiment. [Figure 13] FIG. 3 is explained using a specific device. [Figure 14] FIG. 8 is an explanatory diagram of FIG. 7 using a specific device. [Figure 15] FIG. 13 is a diagram for explaining an optimization method according to the fourth embodiment. [Figure 16] FIG. 13 is a diagram for explaining an optimization method according to the fourth embodiment. [Figure 17] 1A to 1C are diagrams illustrating a method for manufacturing an article. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] Hereinafter, the embodiments will be described in detail with reference to the attached drawings. Note that the following embodiments do not limit the invention according to the claims. Although the embodiments describe a number of features, not all of these features are essential to the invention, and the features may be combined in any manner. Furthermore, in the attached drawings, the same reference numbers are used for the same or similar configurations, and duplicated descriptions are omitted.
[0010] FIG. 1 shows the configuration of the control device 1100 of the first embodiment. Here, FIG. 1(a) shows a state before the processing unit 1010 and the workpiece 1001 come into contact, and FIG. 1(b) shows a state in which the processing unit 1010 and the workpiece 1001 come into contact. The control device 1100 can be configured to process (treat) the workpiece 1001, for example, by pressing the processing unit (object) 1010 against the workpiece (object to be processed) 1001 with a desired force. The control device 1100 controls the force applied to the processing unit (object) 1010. In the following description, the driving direction of the processing unit 1010 is defined as the X-axis.
[0011] The control device 1100 may include, for example, a processing unit 1010, a driving unit 1015 for driving the processing unit 1010 in a predetermined direction (X-axis direction), a measuring unit 1025 for measuring the displacement of the processing unit 1010, and a control unit 1035 for outputting an operation amount (command value) to the driving unit 1015. Furthermore, the control device 1100 may include a mounting unit (member) 1050 for supplying energy (e.g., power) and fluid (gas and / or liquid) required for processing to the processing unit 1010. The mounting unit 1050 is directly or indirectly connected to the processing unit 1010, and a disturbance force may be applied to the processing unit 1010 via the mounting unit 1050. The driving unit 1015 may be configured to apply a force in the X-axis direction to the processing unit 1010. The driving unit 1015 may include, for example, an actuator such as a linear actuator. The measuring unit 1025 may be, for example, a linear encoder or the like. The control unit 1035 may be configured, for example, by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array), or an ASIC (Application Specific Integrated Circuit), or a general-purpose or dedicated computer with a program embedded therein, or a combination of all or part of these.
[0012] The control unit 1035 can have a mode in which position feedback control is performed with the processing unit 1010 as the control target, and a mode in which force feedback control is performed with the processing unit 1010 as the control target. The control unit 1035 can perform position feedback control of the processing unit 1010 when the processing unit 1010 is moved to the vicinity of the workpiece 1001, and can perform force feedback control of the processing unit 1010 when the workpiece 1001 is processed by the processing unit 1010.
[0013] FIG. 2 shows a block diagram illustrating the configuration and operation of the control device 1100 when performing force feedback control. Here, FIG. 2(a) shows a case where no disturbance force Fd is applied to the processing unit 1010 by the mounting unit 1050, and FIG. 2(b) shows a case where the disturbance force Fd is applied to the processing unit 1010 by the mounting unit 1050. First, FIG. 2(a) will be explained. Fr indicates a target value of the force applied to the processing unit 1010 (hereinafter, the force target value), Gc indicates a control compensator, Ki indicates a current gain, and Kt indicates a thrust constant of the driving unit 1015. G0 indicates a transfer function from force to displacement of the processing unit 1010, and X0 indicates a displacement of the processing unit 1010 (a position measurement value by the measuring unit 1025). The output F output by the drive unit 1015 is the sum of a driving force F 0 required to drive the processing unit 1010 and a processing force Fa required to process the workpiece 1001 .
[0014] In order to control the machining force Fa according to the force target value Fr, it is necessary to feed back the machining force Fa. However, in the control device 1100 of the first embodiment, due to the convenience of machining, the machining force Fa cannot be directly measured. Therefore, in FIG. 2(a), the output F obtained by multiplying the operation voltage Vc by the thrust constant Kt and the current gain Ki is fed back. The operation voltage Vc can be understood as an operation amount or a command value. Since the output F is the sum of the machining force Fa and the driving force F0, it is necessary to add the target value of the driving force F0 to the force target value Fr. Therefore, in FIG. 2(a), the target driving force F0r is added to the force target value Fr. The target driving force F0r may be obtained by multiplying the displacement X0 of the machining unit 1010 by 1 / G0 as shown in FIG. 2(a), or may be prepared in advance as a table according to the displacement X0. This allows the machining force Fa to be controlled according to the force target value Fr.
[0015] However, in reality, due to the influence of the mounting portion 1050 shown in Fig. 1, a disturbance force Fd is applied to the processing portion 1010 as shown in Fig. 2(b), so the processing force Fa becomes Fa' = Fa + Fd, and the workpiece 1001 cannot be processed with the processing force Fa equal to the force target value Fr. In the case of the first embodiment, the disturbance force Fd depends on the displacement and speed of the processing portion 1010, so as shown in Fig. 2(b), if the transfer function from the displacement of the processing portion 1010 to the disturbance force Fd is Gd and the displacement of the processing portion 1010 is X0, the disturbance force Fd is expressed as Fd = Gd x X0.
[0016] Therefore, in the first embodiment, as shown in FIG. 3, a disturbance estimation unit (estimation unit) Gde is used as a disturbance observer. The disturbance estimation unit Gde estimates an estimated disturbance force Fde, which is an estimated value of the disturbance force Fd applied to the machining unit 1010, based on information on the state of the machining unit 1010. The disturbance estimation unit Gde can estimate the estimated disturbance force Fde applied to the machining unit 1010, for example, based on the displacement X0 of the machining unit 1010 measured by the measurement unit 1025. By subtracting a voltage equivalent to the estimated disturbance force Fde from the operation voltage Vc, the machining unit 1010 can output a machining force Fa that is equal to the force target value Fr. The disturbance estimation unit 1305 uses a plurality of parameter values determined in advance to obtain the estimated disturbance value Fde based on information on the state of the machining unit 1010. The plurality of parameter values are determined by the determination unit 1350. Note that reference numerals 301 to 307 in FIG. 3 indicate the correspondence with the components shown in FIG. 13 which will be described below.
[0017] FIG. 13 shows an example of the physical configuration of the control device 1100 corresponding to FIG. 3. The control unit 1035 may include a target value generating unit 1301, a compensator 1302, a correction unit 1303, a dynamic stiffness multiplier 1304, and a disturbance estimator 1305. The target value generating unit 1301 generates a force target value Fr. The compensator 1302 obtains an operation voltage Vc from the sum of the force target value Fr and the target driving force F0r. The compensator 1302 corresponds to a block composed of the control compensator Gc (301), the current gain Ki (302), and the thrust constant Kt (303) in FIG. 3, and the gain is Gc / (1+GcKtKi). The correction unit 1303 corresponds to a block composed of 1 / Ki (304), 1 / Kt (305), and the calculator (subtractor) 306 in FIG. 3. The dynamic stiffness multiplier 1304 corresponds to 1 / G0 (307) in Fig. 3. The corrector 1303 corrects the operation amount (command value) to be given to the driver 1015 based on the estimated disturbance force Fde so as to reduce the influence of the disturbance force Fd. The disturbance estimator 1305 corresponds to the disturbance estimator Gde (306) in Fig. 3.
[0018] Next, the disturbance estimation unit Gde (306) will be described. The disturbance estimation unit Gde is a model of the disturbance transfer function Gd shown in FIG. 2(b) and FIG. 3. The disturbance transfer function Gd represents the disturbance caused by the mounting unit 1050 shown in FIG. 1. The disturbance caused by the mounting unit 1050 is generated by the displacement and speed of the processing unit 1010, and can be modeled by a three-element model using a spring (rigidity) and a damper (damping). Therefore, the disturbance estimation unit Gde can also be modeled by a three-element model. FIG. 4 shows a disturbance estimation model using one three-element model. When the displacement of the processing unit 1010 is X0 and the displacement of the spring K2 is X1, the estimated disturbance force Fde is expressed by Equation 1 and Equation 2.
[0019]
number
[0020]
number
[0021] Numbers 1 and 2 can be transformed into numbers 3 and 4, respectively.
[0022]
number
[0023]
number
[0024] Since Equation 3 represents a state equation and Equation 4 represents an output equation, the disturbance estimation model shown in Fig. 4 can be expressed as a state space shown in Fig. 5. The state matrix A, input matrix B, output matrix C, and direct feedthrough matrix D in Fig. 5 are expressed by Equations 5 to 8.
[0025]
number
[0026]
number
[0027]
number
[0028]
number
[0029] Therefore, the disturbance estimation unit Gde can be expressed by the state space shown in FIG. 5, and when the displacement X0 of the machining portion 1010 is input to the disturbance estimation unit Gde, an estimated disturbance force Fde can be obtained.
[0030] Moreover, by solving the differential equations expressed by Equation 1 and Equation 2, the estimated disturbance force Fde can be expressed as a function of time t as shown in Equation 9. T in Equation 9 is the time constant of the disturbance estimation model and disturbance estimation unit Gde shown in FIG. 4, and is expressed by Equation 10.
[0031]
number
[0032]
number
[0033] When the mounting unit 1050 is composed of a plurality of elements having different time constants, it is necessary to combine a plurality of three-element models. In the first embodiment, a disturbance estimation model in which three three-element models are combined as shown in Fig. 6 is used. In the case of the disturbance estimation model shown in Fig. 6, the state matrix A, input matrix B, output matrix C, and direct feedthrough matrix D of the disturbance estimation unit Gde are expressed by equations 11 to 14.
[0034]
number
[0035]
number
[0036]
number
[0037]
number
[0038] Further, the disturbance force Fd applied to the processing unit 1010 can be calculated from the displacement X0 of the processing unit 1010 and the output F output by the driving unit 1015 at that time when the processing unit 1010 is subjected to position feedback control. FIG. 7 shows a block diagram showing the configuration and operation of the control device 1100 when the processing unit 1010 is subjected to position feedback control. The control device 1100 shown in FIG. 7 includes an acquisition unit 1450 that acquires the disturbance force Fd. Xr represents the displacement target value (position command value) of the processing unit 1010. When the processing unit 1010 is subjected to position feedback control, the output F output by the driving unit 1015 is F=F0-Fd, and even if the disturbance Fd is applied, the driving force F0 is applied to the processing unit 1010, so that the processing unit 1010 can be positioned according to the displacement target value Xr. Therefore, the disturbance force Fd can be calculated by subtracting the driving force F0 from the output F (=F0-Fd) and multiplying the result by -1. The driving force F0 can be calculated by multiplying the displacement X0 of the processing portion 1010 by 1 / G0.
[0039] When acquiring the disturbance force Fd, the acquiring unit 1450 performs position feedback control to operate the driving unit 1015 based on the difference between the displacement target amount Xr and the displacement X0 so that the displacement X0 (measurement value) of the processing unit 1010 coincides with the displacement target amount Xr that commands the position of the processing unit 1010. In this position feedback control, the acquiring unit 1450 acquires the disturbance force Fd based on the difference between the displacement target amount Xr and the displacement X0 and the displacement X0 (measurement value). In the example of FIG. 7, F is calculated based on the difference between the displacement target amount Xr and the displacement X0, F0 is calculated based on the displacement X0 (measurement value), and Fd is calculated from F and F0. Note that the reference numerals 701 to 708 in FIG. 7 indicate the correspondence with the components shown in FIG. 14 described below.
[0040] FIG. 14 shows an example of the physical configuration of the control device 1100 corresponding to FIG. 7. The control unit 1035 can include a target value generating unit 1401, a compensator 1402, a dynamic stiffness multiplier 1403, and calculators 1404, 1405, and 1406. The target value generating unit 1401 generates a displacement target value Xr (position command value). The calculator 1406 calculates the difference (displacement deviation) between the displacement target value Xr and the displacement X0. The compensator 1402 obtains a driving force F (=F0-Fd) from the difference (displacement deviation) between the displacement target value Xr and the displacement X0. The compensator 1402 corresponds to a block composed of the control compensator Gc (701), the current gain Ki (702), and the thrust constant Kt (703) in FIG. 7, and the gain is GcKtKi. The calculator 1404 corresponds to the calculator 706 in FIG. 7. In FIG. 7, a driving force F (=F0-Fd) calculated by a block consisting of a control compensator Gc (701), a current gain Ki (704), and a thrust constant Kt (705) is input to a calculator 706. A calculator 1405 corresponds to the calculator 707 in FIG. 7. A dynamic stiffness multiplier 1403 corresponds to 1 / G0 (708) in FIG. 7. The compensator 1402, the dynamic stiffness multiplier 1403, and calculators 1404, 1405, and 1406 constitute an acquisition unit 1450 that acquires a disturbance force Fd. A determination unit 1350 shown in FIG. 13 determines a plurality of parameter values used by the disturbance estimation unit 1305 through machine learning based on the disturbance force Fd acquired by the acquisition unit 1450. Here, the plurality of parameter values are used to obtain an estimated disturbance value Fde based on information on the state of the machining unit 1010.
[0041] Incidentally, in order for the estimated disturbance force Fde generated by the disturbance estimation unit 1305 to be a value that correctly estimates the disturbance force Fd, it is necessary to correctly determine multiple parameter values of the disturbance estimation unit Gde. When the disturbance estimation model is the three three-element model shown in FIG. 6, the total number of spring and damper elements is nine, and it is very difficult to determine the nine parameter values by desk calculation. Therefore, in the first embodiment, the determination unit 1350 determines multiple parameter values of the disturbance estimation unit Gde by machine learning. As a method of machine learning, for example, unsupervised learning can be used. For example, the function of performing machine learning can be incorporated into the control unit 1035.
[0042] First, the control unit 1035 acquires data to be used for training in order to perform machine learning. As data, the processing unit 1010 is subjected to position feedback control, and disturbance forces Fd at various displacements X0 are collected together with the displacement X0 of the processing unit 1010. The collected data of the displacement X0 is set as displacement data_X0, and the collected data of the disturbance force Fd is set as disturbance force data data_Fd. The control unit 1035 inputs the displacement X0 to the disturbance estimation unit Gde, and collects the estimated disturbance force Fde output from the disturbance estimation unit Gde as estimated disturbance force data data_Fde. The control unit 1035 uses the data collected in this manner as training data.
[0043] In unsupervised learning, it is desired to make the estimated disturbance force Fde match the disturbance force Fd as closely as possible. Therefore, the control unit 1035 can be configured to optimize multiple parameter values of the disturbance estimation unit Gde so as to minimize the loss function L, with the function for calculating the square sum error between the estimated disturbance force Fde and the disturbance force Fd being a loss function L. The loss function L is shown in Equation 15. In Equation 15, i is the data number of the training data, and n is the total number of training data.
[0044]
number
[0045] As the loss function, in addition to the sum of squares error, for example, a function that calculates the mean sum of squares error, the mean absolute error, or the mean squared logarithmic error may be adopted.
[0046] Next, an optimization method in machine learning will be described by way of example. As an optimization method in machine learning, for example, a gradient method can be used. FIG. 8(a) illustrates the procedure of the gradient method. First, in step SS1, the control unit 1035 selects training data for learning from the collected training data. The training data for learning may be randomly selected from the collected training data, or training data in which the feature to be learned appears well may be selected. Next, in step SSS2, the control unit 1035 obtains the gradient of each parameter value in order to reduce the value of the loss function. When the value of the loss function is L, the gradient of each parameter value can be expressed, for example, by Equation 16.
[0047]
number
[0048] Next, in step SS3, the control unit 1035 updates each parameter value by a small amount in the gradient direction. The parameter value can be updated, for example, according to Equation 17. η in Equation 17 represents a learning rate, which is a parameter that determines how much the parameter value is updated in one learning session. In general, the learning rate η can be about 0.001 to 0.1.
[0049]
number
[0050] In step SS4, the control unit 1035 calculates the loss function L for the updated parameter values. In step SS5, the control unit 1035 determines whether or not a termination condition is satisfied, and terminates the learning if the termination condition is satisfied, and returns to step SS1m to continue the learning if the termination condition is not satisfied.
[0051] FIG. 8(b) shows another example of a procedure that is a partial modification of the procedure of the gradient method shown in FIG. 8(a). In the procedure of the gradient method shown in FIG. 8(b), as shown in Equation 18, a change amount ΔL of the loss function L for the stored loss function Lold is calculated, and a step SS6 is added to determine whether the change amount ΔL is equal to or less than a threshold. If the change amount ΔL is not equal to or less than the threshold, the process returns to step SS3 to update the parameter value, and the loss function L is calculated in step SS4. The procedure of FIG. 8(b) is a method in which, once a gradient is obtained and a direction to move in is determined, the parameter value is changed in the direction of the gradient, and when the decrease in the loss function L stops, a new gradient is obtained and the parameter value is updated. This method can obtain an optimal solution that minimizes the loss function L more quickly than the procedure shown in FIG. 8(a).
[0052]
number
[0053] The learning rate η may be gradually decreased according to the progress of learning, so that when the parameter value is far from the optimal solution, the parameter value is changed largely, and when the parameter value is approaching the optimal solution, the parameter value is changed finely, so that the optimal solution can be reached more quickly and accurately.
[0054] In the first embodiment, the gradient method is used as the optimization method in machine learning, but other optimization methods may be adopted.
[0055] The process for determining the parameter value of the disturbance estimation unit Gde may be executed while the control device 1100 is in operation (while the workpiece 1001 is being processed), or may be executed by driving the processing unit 1010 according to a dedicated drive profile during non-operational time. When the parameter value of the disturbance estimation unit Gde is determined during non-operational time, the process for determining the parameter value of the disturbance estimation unit Gde may be executed, for example, periodically, at an arbitrary timing, or when a predetermined condition is satisfied. For example, when the error of the estimated disturbance force Fde with respect to the disturbance force Fd exceeds a certain ratio (for example, 5%), the parameter value of the disturbance estimation unit Gde may be updated or re-determined.
[0056] FIG. 9 shows a block diagram showing the configuration of a force feedback control system in which disturbance force correction is performed in the second embodiment. In the second embodiment, a disturbance estimation unit Gden configured with a neural network may be used instead of the disturbance estimation unit Gde in the first embodiment. The disturbance estimation unit Gden may include an input layer, a plurality of intermediate layers, and an output layer, as exemplified in FIG. 10. The total number of intermediate layers and the number of nodes in each layer may be appropriately set. The parameter values of the disturbance estimation unit Gden may be optimized, for example, by a method similar to the method of determining the parameter values of the disturbance estimation unit Gde in the first embodiment. By using the disturbance estimation unit Gden configured with a neural network, it is possible to estimate disturbances with higher accuracy than the disturbance estimation unit Gde using a three-element model.
[0057] In the third embodiment, the parameter values of the disturbance estimation unit Gde in the first embodiment are determined using reinforcement learning. In reinforcement learning, various types of values can be used as a reward that is a goal during learning. As an example, the "action evaluation function Q π Here is an example of learning using the action evaluation function Q π (s, a) can be expressed by number 19.
[0058]
number
[0059] In Equation 19, t is time, s is state, a is action, π is policy, E π {} represents the expected value under policy π, r represents the reward, γ represents the discount rate for the future reward, and k represents the time until the future reward.
[0060] The acting subject (agent), here the control unit 1035, is determined by the action evaluation function Q π The robot acts to maximize (s, a). In so doing, it selects between the "optimal action based on past experience" and the "search for new actions" aimed at obtaining further rewards, according to a given policy π. Since the formula is an expectation value that takes future rewards into account, it can handle situations where rewards decrease in the short term, but large rewards can be obtained in the long term. This allows the action evaluation function Q π It is possible to learn the states and actions that maximize (s, a).
[0061] The action evaluation function Q shown in Equation 19 π The reward at (s, a) can be defined as the inverse of the evaluation function in Equation 15, as shown in Equation 20.
[0062]
number
[0063] The formula 20 is an equation that expects the reward r to be large if the disturbance force Fde estimated by the disturbance estimation unit Gde is close to the disturbance force Fd of the actual machine.
[0064] Action a can be defined, for example, as follows: The number of actions a is the number of parameter values 9×2=18.
[0065]
number
[0066] Here, j represents the number of times of learning. Action a is an action that increases or decreases the spring constant and damping coefficient of the disturbance estimation model by minute amounts ΔK and ΔC in order to search for optimal parameter values of the disturbance estimation unit Gde. The magnitudes of ΔK and ΔC can be obtained by multiplying the spring constant and damping coefficient of the initial state by a certain ratio, and the ratio can be within a range of, for example, about 0.1% to 10%. Note that the action is not limited to action a, and other actions may be defined as long as they can search for optimal parameter values of the disturbance estimation unit Gde.
[0067] The strategy π can be defined as follows:
[0068] Policy π: Randomly act a i Select the action evaluation function Q π If (s, a) increases, the parameter values of the disturbance estimation unit Gde are updated.
[0069] Here, future rewards are not taken into consideration, the time until the future reward is set to 0, the discount rate γ of the future reward is set to 0.01, and the behavior evaluation function Q π (s, a) may be equivalent to reward r.
[0070] Randomly selected action a jh (h is between 1 and 18) to change the parameter values of the disturbance estimation unit Gde. The action evaluation function Q π (s, a) is the action evaluation function Q for the initial state s0 π If it is increasing with respect to (s, a), the parameter values of the disturbance estimator Gde in state s1 at t=1 are updated. On the other hand, the action evaluation function Q π (s, a) is the action evaluation function Q for the initial state s0 π If the parameter value of the disturbance estimator Gde in state s1 at t=1 is not increased with respect to (s, a), the parameter value of the disturbance estimator Gde in state s1 at t=1 is not updated. π (s, a) is the previous action evaluation function Q π If the parameter of the disturbance estimation unit Gde is better than (s, a), the action evaluation function Q πThe parameter values of the disturbance estimation unit Gde can be optimized so that (s, a) increases. Here, since time t is equivalent to the number of learning epochs, one advance in time t can be said to be "one learning".
[0071] The definition of the policy π is not limited to the above example. Any condition may be added, such as not selecting an action that transitions to a state that has already been explored. The time k until the future reward is set to a value of 1 or more, and the action evaluation function Q π (s, a) may be maximized. In this case, the number of learning times required until optimization increases, but it becomes possible to determine the parameter values of the disturbance estimation unit Gde without getting stuck in a local solution.
[0072] In the fourth embodiment, the parameter value of the disturbance estimation unit Gde in the first embodiment is determined using an optimization method, not machine learning. Fig. 15 shows an example of optimizing K11 using the optimization method. The control unit 1035 calculates a loss function Lp when K11 is increased by ΔK11 and a loss function Lm when K11 is decreased by ΔK11. The control unit 1035 then compares Lp and Lm, and changes K11 in a direction that can reduce the loss function L. Here, an example will be described in which Lp is smaller.
[0073] First, the control unit 1035 substitutes Lp into L_old. Next, the loss function L when K11 is further increased by ΔK11 is calculated. If L is compared with L_old and L is smaller, L is substituted into L_old, K11 is further increased by ΔK11, the loss function L is calculated, and L is compared with L_old. This loop is continued until the loss function L does not become smaller. If L is larger than L_old, the control unit 1035 exits the loop, substitutes L_old into L_K11, and substitutes K11_old into K11. Finally, K11 and L_K11 are output.
[0074] Next, a method for optimizing parameter values of a plurality of parameters will be described with reference to Fig. 16. In the example of Fig. 16, the parameter values of parameters K11, K12, and C12 are optimized, but the parameter values of more parameters may be optimized simultaneously, or the parameter values of several parameters may be optimized as in the fourth embodiment.
[0075] In the example of FIG. 16, the control unit 1035 first assigns the current K11, K12, and C12 to K11_old, K12_old, and C12_old, respectively. Next, the control unit 1035 optimizes K11, K12, and C12, respectively, and outputs the optimized K11, K12, and C12 and the respective loss functions L_K11, L_K12, and L_C12. Furthermore, the control unit 1035 compares L_K11, L_K12, and L_C12, and changes the parameter value of the parameter that can minimize the loss function L. For example, if L_K11 is the smallest, the control unit 1035 changes K11, and further assigns L_K11 to L. Next, the control unit 1035 compares L and L_old, and if L is smaller than L_old, assigns L to L_old, and further optimizes K11, K12, and C12, respectively. The loop of comparing L_K11, L_K12, and L_C12, changing the parameter value of the parameter that can make the loss function L the smallest, and updating L is continued until the loss function L is no longer small. If L is larger than L_old, the control unit 1035 exits the loop and substitutes K11_old for K11, K12_old for K12, and C12_old for C12. Then, the control unit 1035 finally outputs K11, K12, and C12. In this manner, the parameter values of the parameters of the disturbance estimation unit Gde may be determined using an optimization method instead of machine learning.
[0076] The fifth embodiment provides an example in which the disturbance estimation unit Gde or the control device 1100 of the first embodiment is applied to an imprint apparatus. Fig. 11 shows a schematic configuration of an imprint apparatus 100 in the fifth embodiment. Fig. 11(a) shows a state before the mold 10 and the imprint material 60 on the substrate 1 come into contact, and Fig. 11(b) shows a state in which the mold 10 and the imprint material 60 on the substrate 1 come into contact. In the following description, two axes that are orthogonal to each other in a plane parallel to the surface of the substrate 1 are defined as the X-axis and the Y-axis, and an axis perpendicular to the X-axis and the Y-axis is defined as the Z-axis.
[0077] The imprint apparatus 100 may include a substrate operation unit 23 that holds the substrate 1, a supply unit 18 that supplies the imprint material 60, a mold operation unit 24 that holds the mold 10, a light source 16, an alignment scope 21, and a control unit 35. The imprint apparatus 100 brings the imprint material 60 supplied onto the substrate 1 into contact with the mold 10 and applies energy for curing to the imprint material 60, thereby forming a pattern of a cured material to which the concave-convex pattern of the mold 10 has been transferred. The imprint apparatus 100 in FIG. 11 may be used to manufacture articles such as semiconductor devices.
[0078] The substrate operation unit 23 may include a substrate chuck 2, a θ stage 3 (rotation drive unit), and an XY stage 4 (XY drive unit). The substrate chuck 2 holds the substrate 1 by vacuum or electrostatic adsorption. In FIG. 11, the substrate 1 is held by the substrate chuck 2. The θ stage 3 is disposed on an XY stage 4 for correcting the position of the substrate 1 in the θ direction (rotation direction around the Z axis) and positioning the substrate 1 in the X and Y directions. The XY stage 4 may be driven in the X and Y directions by a linear motor 19. The θ stage 3 and the XY stage 4 hold the substrate chuck 2 and move the substrate 1 held by the substrate chuck 2. The XY stage 4 is placed on a base 5. A linear encoder 6 is attached to the base 5 in the X and Y directions to measure the position of the XY stage 4. A support 8 stands on the base 5 and supports a top plate 9.
[0079] As the substrate 1, for example, a single crystal silicon substrate or an SOI (Silicon On Insulator) substrate is used. As the substrate 1, glass, ceramics, metal, semiconductor, resin, etc. may be used, and a member or layer made of a material different from the base material may be formed on the surface of the base material as necessary. Specifically, the substrate may include a silicon wafer, a compound semiconductor wafer, quartz glass, etc. The substrate 1 may have a plurality of shot regions, and an imprint material 60 may be supplied to the shot regions by a supply unit 18. The imprint apparatus 100 can form a pattern on the entire surface of the substrate 1 by repeating an imprint process for forming a pattern of the imprint material for each shot region.
[0080] As the imprint material 60, a curable composition (sometimes called an uncured resin) that is cured by applying energy for curing is used. As the energy for curing, electromagnetic waves or the like can be used. The electromagnetic waves can be, for example, light having a wavelength selected from the range of 10 nm to 1 mm, such as ultraviolet light. The curable composition can be a composition that is cured by irradiation with light. Among these, a photocurable composition that is cured by irradiation with light contains at least a polymerizable compound and a photopolymerization initiator, and may further contain a non-polymerizable compound or a solvent as necessary. The non-polymerizable compound is at least one selected from the group consisting of a sensitizer, a hydrogen donor, an internal mold release agent, a surfactant, an antioxidant, and a polymer component. In this embodiment, as an example, a photocurable composition that has a property of being cured by ultraviolet light is used as the imprint material 60. The imprint material 60 can be arranged on the substrate 1 in the form of droplets, or in the form of islands or films formed by connecting a plurality of droplets, by the supply unit 18. The viscosity of the imprint material (at 25° C.) may be, for example, 1 mPa·s or more and 100 mPa·s or less. The imprint material 60 may be applied in the form of a film onto the substrate by a spin coater or a slit coater.
[0081] The supply unit 18 (dispenser) supplies the imprint material 60 onto the substrate 1. The supply unit 18 has, for example, a discharge nozzle (not shown) and supplies the imprint material 60 onto the substrate 1 from the discharge nozzle. Note that in this embodiment, the supply unit 18 supplies the imprint material onto the substrate 1 by, for example, dropping the liquid imprint material 60 onto the surface of the substrate 1. The amount of the imprint material supplied by the supply unit 18 may be determined based on the required thickness of the imprint material, the pattern density to be formed, and the like. Furthermore, the supply unit 18 does not necessarily have to be provided in the imprint apparatus 100, and the imprint material may be supplied onto the substrate 1 by a supply unit provided outside the imprint apparatus 100.
[0082] The mold 10 is a mold for forming the imprint material on the substrate. The mold may also be called a mold, a template, or an original plate. The mold 10 has, for example, a rectangular outer periphery, and is provided with a pattern region P on the surface facing the substrate 1, in which a concave-convex pattern to be transferred to the imprint material 60 supplied onto the substrate 1 is formed in a three-dimensional shape. The pattern region P is also called a mesa portion. The pattern region P is formed in a convex portion of several tens of μm to several hundreds of μm so that the portion other than the pattern region P of the mold 10 (the region surrounding the pattern region P) does not come into contact with the substrate 1. The mold 10 is made of a material that transmits light (ultraviolet rays) for curing the imprint material on the substrate, such as quartz.
[0083] The mold operation unit 24 may include a mold chuck (mold holding unit) 11, a mold stage 22, and a linear actuator 15 (mold driving unit). The mold chuck 11 holds the mold 10 by vacuum suction force or electrostatic suction force. The mold chuck 11 is held by the mold stage 22. The mold stage 22 has a function of adjusting the Z position of the mold 10 and a tilt function for correcting the inclination of the mold 10. The linear actuator 15 drives the mold 10 held by the mold chuck 11 in the Z axis direction to bring the mold 10 into contact with the imprint material 60 on the substrate 1 and separate it. The linear actuator 15 is, for example, an air cylinder or a linear motor. The mold chuck 11 and the mold stage 22 each have an opening (not shown) that allows the light irradiated from the first light source 16 to pass through to the mold 10.
[0084] In a process of curing the imprint material 60 on the substrate 1, the light source 16 irradiates the substrate 1 with light (ultraviolet light) for curing the imprint material 60 via a collimator lens 17a. The light source 16 may be a light source that generates, for example, i-line (365 nm), but may generate light of other wavelengths. The beam splitter 20 is disposed in the optical path between the light source 16 and the mold operation unit 24, and separates the light used by the alignment scope 21 to observe the contact state of the mold 10 from the light for curing the imprint material 60. The alignment scope 21 captures an image of the pattern area P of the mold 10 via the beam splitter 20.
[0085] The control unit 35 controls the operation and adjustment of each part constituting the imprint apparatus 100. The control unit 35 is composed of, for example, a computer, is connected to each part constituting the imprint apparatus 100 via a communication path, and can execute control of each part in accordance with a program, etc. The control unit 35 may be provided within the imprint apparatus 100, or may be provided in a location separate from the imprint apparatus 100 and controlled remotely.
[0086] Next, the imprint process by the imprint apparatus 100 configured as described above will be described. FIG. 12 is a flowchart showing the imprint process in the imprint apparatus 100 according to the fifth embodiment. Each step can be executed by the control of each part of the imprint apparatus 100 by the control unit 35. First, various parameters required for imprinting are set (step S0). Then, the XY stage 4 of the imprint apparatus 100 is driven to move the substrate chuck 2 on which the substrate 1 is placed in the X and Y directions, and a shot area to be the target of the imprint process (hereinafter, the target shot area) is positioned under the supply unit 18 (step S1). Then, a predetermined amount of uncured imprint material 60 is supplied onto the substrate 1 (step S2).
[0087] Next, the XY stage 4 is driven again to move the substrate chuck 2 so that the target shot area is positioned opposite the pattern area P of the mold 10, and the θ stage 3 is driven to correct the θ direction position of the substrate 1 (step S3). Next, the linear actuator 15 is driven to move the mold stage 22 in the -Z direction, and the mold 10 is brought into contact with the uncured imprint material 60 on the substrate 1 (contact step, step S4). In step S4, instead of moving the mold stage 22, the substrate chuck 2 may be moved in the Z direction, or the mold stage 22 and the substrate chuck 2 may be moved separately. The control unit 35 determines whether the contact force generated when the mold 10 is brought into contact with the uncured imprint material 60 on the substrate 1 is optimal (step S5).
[0088] If the contact force is not optimal (step S5, NO), the mold stage 22 changes the inclination of the mold chuck 11 so that the contact force of the mold 10 with the imprint material 60 becomes a predetermined value. In addition, the force pressing the mold 10 against the imprint material is adjusted by changing the pressing amount using the linear actuator 15 (step S6).
[0089] If the contact force is optimal (step S5, YES), the alignment marks AM formed on the mold 10 and the substrate 1 are detected by the alignment scope 21, and alignment is performed based on the detected measurement results. The relative deviation between the mold 10 and the substrate 1 is obtained from the measurement results, and the XY stage 4 and the θ stage 3 are driven to align the mold 10 with the substrate 1 (step S7).
[0090] After the mold 10 and the substrate 1 are aligned, the imprint material 60 on the substrate 1 is irradiated with light (ultraviolet rays) by the light source 16 (main exposure) to harden the imprint material 60 (step S8). The irradiation area for the main exposure is the entire surface of the shot area. After the irradiation of light (ultraviolet rays) for a predetermined period of time is completed, the linear actuator 15 is driven to raise the mold stage 22 in the +Z direction, and a mold release process (step S9) is performed in which the mold 10 is separated from the hardened imprint material 60 on the substrate 1. In step S9, instead of moving the mold stage 22, the substrate chuck 2 may be moved in the Z direction, or the mold stage 22 and the substrate chuck 2 may be moved separately.
[0091] Thereafter, it is determined whether pattern formation has been completed for all shot areas on the substrate 1 (step S102). If there are any shot areas remaining in which to form a pattern of the imprint material, the XY stage 4 is driven to move the substrate 1 in order to supply the imprint material 60 to the next target shot area (step S1). This series of processes is repeated until pattern formation has been completed for all shot areas on the substrate 1. When pattern formation has been completed for all shot areas, the XY stage 4 is driven to move the substrate 1 to a predetermined position (step S113), and the imprint process for one substrate 1 is completed.
[0092] In step S6, the force pressing the mold 10 against the imprint material is adjusted, and the mounting unit 50 for supplying power and fluid is connected to the mold operation unit 24, and the mounting unit 50 applies a disturbance force to the mold operation unit 24. In the fifth embodiment, the control unit 35 has a disturbance estimation unit Gde, and supplies the displacement Z0 of the mold stage 22 measured by a position measuring device (not shown) of the mold operation unit 24 to the disturbance estimation unit Gde. Then, the disturbance estimation unit Gde calculates an estimated disturbance Fde, and the operation voltage Vc can be corrected based on the estimated disturbance force Gde. The disturbance estimation unit Gde is a plurality of three-element models, and unsupervised machine learning can be used to determine the parameter values of the disturbance estimation unit Gde, as in the first embodiment.
[0093] In the sixth embodiment, the disturbance estimation unit Gden of the second embodiment is applied to an imprint apparatus. The disturbance estimation unit Gden is configured with a neural network, and the parameter values of the disturbance estimation unit Gden can be determined by unsupervised machine learning, as in the second embodiment.
[0094] In the seventh embodiment, the disturbance estimation unit Gde of the third embodiment is applied to an imprint apparatus. The disturbance estimation unit Gde is a plurality of three-element models, and the parameter values of the disturbance estimation unit Gde can be determined using reinforcement learning, as in the third embodiment.
[0095] In the first to seventh embodiments, the disturbance applied to the processing unit 1010 or the mold operation unit 24 is estimated by the disturbance estimation unit, and the operation amount can be corrected based on this. However, in the case of an imprint apparatus, various disturbance factors affect each other in a complex manner, so that the overlay accuracy may not be improved only by the correction based on the disturbance force by the mounting unit 50. Therefore, in the eighth embodiment, in order to estimate the disturbance force in the contact process in which the imprint material on the substrate and the mold are brought into contact with each other, the parameter value of the disturbance estimation unit is determined using disturbance force data_Fdimp calculated based on the measurement result of the overlay inspection device. For example, a neural network can be used as the disturbance estimation unit. In addition to the displacement Z0 of the mold chuck 11 during the imprint process, the operation voltage Vcimp to the linear actuator 15 that drives the mold chuck 11 can be used as an input to the disturbance estimation unit. The disturbance estimation unit can output an estimated disturbance force Fdeimp based on the displacement Z0 and the operation voltage Vcimp. Since the disturbance force data data_Fdimp is a value output for each shot area, the estimated disturbance force Fdeimp during the contact process can be collected for each shot area and used as the estimated disturbance force data data_Fdeimp. Using the disturbance force data data_Fdimp during the contact process and the estimated disturbance force data data_Fdeimp as training data, the disturbance estimator parameter values can be determined using unsupervised machine learning, as in the second embodiment. This allows the disturbance estimator to estimate the disturbance during imprinting that affects the overlay accuracy, and correct the operation amount based on the disturbance, thereby improving the overlay accuracy.
[0096] In the ninth embodiment, reinforcement learning is used to determine the parameter values of the disturbance estimator in the eighth embodiment, as in the third embodiment.
[0097] Hereinafter, an article manufacturing method will be described as a tenth embodiment. A pattern of a cured product formed using an imprint apparatus is used permanently on at least a part of various articles, or temporarily when manufacturing various articles. The article is an electric circuit element, an optical element, a MEMS, a recording element, a sensor, a mold, or the like. Examples of the electric circuit element include volatile or non-volatile semiconductor memories such as DRAM, SRAM, flash memory, and MRAM, and semiconductor elements such as LSI, CCD, image sensor, and FPGA. Examples of the mold include a mold for imprinting.
[0098] The pattern of the cured product is used as it is as at least a part of a component of the article, or is used temporarily as a resist mask, which is removed after etching or ion implantation in a substrate processing step.
[0099] Next, a method for manufacturing an article will be described, in which a pattern is formed on a substrate by the above-mentioned imprinting apparatus, the substrate on which the pattern is formed is processed, and an article is manufactured from the processed substrate. As shown in Fig. 17(a), a substrate 1z such as a silicon wafer having a workpiece 2z such as an insulator formed on its surface is prepared, and then an imprinting material 3z is applied to the surface of the workpiece 2z by an inkjet method or the like. Here, a state in which the imprinting material 3z in the form of multiple droplets is applied onto the substrate is shown.
[0100] As shown in Fig. 17(b), the imprinting mold 4z is placed so that the side on which the concave-convex pattern is formed faces the imprinting material 3z on the substrate. As shown in Fig. 17(c), the substrate 1z to which the imprinting material 3z has been applied is brought into contact with the mold 4z, and pressure is applied. The imprinting material 3z fills the gap between the mold 4z and the workpiece 2z. When light is irradiated through the mold 4z in this state as energy for hardening, the imprinting material 3z hardens.
[0101] 17(d), after the imprint material 3z is cured, the mold 4z and the substrate 1z are separated, and a pattern of the cured product of the imprint material 3z is formed on the substrate 1z. In this cured product pattern, the recesses of the mold correspond to the protrusions of the cured product, and the protrusions of the mold correspond to the recesses of the cured product, i.e., the recessed and protruding pattern of the mold 4z is transferred to the imprint material 3z.
[0102] As shown in Fig. 17(e), when etching is performed using the pattern of the cured material as an etching-resistant mask, the portion of the surface of the workpiece 2z where there is no cured material or where only a thin layer remains is removed, forming a groove 5z. As shown in Fig. 17(f), when the pattern of the cured material is removed, an article having grooves 5z formed on the surface of the workpiece 2z can be obtained. Here, the pattern of the cured material is removed, but it may be used as an interlayer insulating film included in a semiconductor element or the like, that is, a component of an article, without being removed after processing.
[0103] The invention is not limited to the above-described embodiments, and various modifications and variations are possible without departing from the spirit and scope of the invention. Accordingly, the following claims are appended to apprise the public of the scope of the invention. [Explanation of symbols]
[0104] 1010: Processing section, 1001: Workpiece, 1015: Drive section, 1303: Correction section, 1305: Disturbance estimation section
Claims
1. A control device for processing an object to be processed by pressing an object against the object, comprising: a compensator that determines an operation amount based on a sum of a force target value that is a target value of a force required to process the processing object and a target driving force that is a target value of a driving force required to drive the object; A drive unit that applies a force to the object; an estimation unit that estimates an estimate of a disturbance force applied to the object based on information about a state of the object; a correction unit that generates a command value to be provided to the drive unit by correcting the manipulated variable based on the estimated value so that an effect of the disturbance force is reduced; a determination unit that performs a process of determining a number of parameter values used by the estimator to obtain the estimated value based on the information; A control device comprising:
2. The disturbance force is applied to the object via a member directly or indirectly connected to the object. The control device according to claim 1 .
3. the plurality of parameter values includes parameter values indicative of stiffness and damping of the member; 3. The control device according to claim 2.
4. Further comprising a measurement unit for measuring the information.
4. The control device according to claim 1, wherein the control device is a control unit.
5. An acquisition unit for acquiring the disturbance force is further provided, The determination unit determines the plurality of parameter values through machine learning based on the disturbance force obtained by the acquisition unit.
5. The control device according to claim 1, wherein the control device is a control unit.
6. the acquisition unit acquires the disturbance force based on the difference and the measurement value in a position feedback control in which the drive unit is operated based on a difference between a position command value that commands a position of the object and the measurement value so that the measurement value of the position of the object coincides with the position command value, 6. The control device according to claim 5.
7. The machine learning is unsupervised learning.
7. The control device according to claim 5 or 6.
8. In the unsupervised learning, one of the following loss functions is used: sum of squares error, mean sum of squares error, mean absolute error, and mean squared logarithmic error. The control device according to claim 7 .
9. The machine learning is reinforcement learning.
7. The control device according to claim 5 or 6.
10. In the reinforcement learning, the reciprocal of one of the sum of squares error, the mean sum of squares error, the mean absolute error, and the mean squared logarithmic error is used as a reward. The control device according to claim 9 .
11. The mold is incorporated in an imprinting device that contacts an imprinting material on a substrate with a mold and hardens the imprinting material to form a pattern made of a hardened product of the imprinting material.
11. The control device according to claim 1 ,
12. A force applied to a mold holding part that holds the mold is controlled as a force applied to the object. The control device according to claim 11 .
13. A control unit is further provided for controlling a first mode in which position feedback control is performed with the object as a control target, and a second mode in which force feedback control is performed with the object as a control target, the control unit brings the object close to the processing target in the first mode, and then controls the object in the second mode; The compensator generates the manipulated variable in the second mode.
13. The control device according to any one of claims 1 to 12.
14. An imprinting apparatus for forming a pattern made of a cured product of an imprinting material by pressing a mold against an imprinting material on a substrate and curing the imprinting material, comprising: A mold holding unit that holds the mold; a compensator that determines an operation amount based on a sum of a force target value that is a target value of a force required to process the imprint material and a target driving force that is a target value of a driving force required to drive the mold holding unit; A drive unit that applies a force to the mold holding unit; an estimation unit that estimates an estimate of a disturbance force applied to the mold holding unit based on information about a state of the mold holding unit; a correction unit that generates a command value to be provided to the drive unit by correcting the manipulated variable based on the estimated value so that an effect of the disturbance force is reduced; a determination unit that performs a process of determining a number of parameter values used by the estimator to obtain the estimated value based on the information; An imprint apparatus comprising:
15. The disturbance force is applied to the mold holding part via a member directly or indirectly connected to the mold holding part. The imprint apparatus according to claim 14 .
16. the plurality of parameter values includes parameter values indicative of stiffness and damping of the member; The imprint apparatus according to claim 15 .
17. Further comprising a measuring device for measuring the information. The imprint apparatus according to any one of claims 14 to 16.
18. An acquisition unit for acquiring the disturbance force is further provided, The determination unit determines the plurality of parameter values through machine learning based on the disturbance force obtained by the acquisition unit.
18. The imprint apparatus according to claim 14,
19. the acquisition unit acquires the disturbance force based on the difference and the measurement value in a position feedback control in which the drive unit is operated based on a difference between a position command value that commands a position of the mold holding unit so that a measurement value of the position of the mold holding unit coincides with the position command value, the difference being between the position command value and the measurement value.
20. The imprint apparatus according to claim 18 .
20. A control unit is further provided for controlling a first mode in which position feedback control is performed with the mold holding unit as a control object, and a second mode in which force feedback control is performed with the mold holding unit as a control object, the control unit brings the mold holding unit closer to the imprint material in the first mode, and then controls the mold holding unit in the second mode; The compensator generates the manipulated variable in the second mode.
20. The imprint apparatus according to claim 14,
21. forming a pattern on a substrate by the imprint apparatus according to any one of claims 14 to 20; processing the substrate on which the pattern has been formed in the process; A method for manufacturing an article, comprising:
Citation Information
Patent Citations
Feeder and laser beam recorder
JP2002367239A
Active vibration damping method
JP2003108236A
Drive control method of moving body, exposure method, robot control method, drive control device, exposure device, and robot device
JP2010123957A
Imprint apparatus and method of manufacturing product using the same
JP2013021155A
Control device, lithography device, measuring device, machining device, planarization device, and method for manufacturing article
JP2019071405A
