Control device, method of adjusting a control device, lithographic apparatus, and method of manufacturing an article

Abstract

A control device for generating control signals for controlling a controlled object includes a first compensator, a corrector, a second compensator, and a computing device. The first compensator is configured to generate a first signal based on a control deviation of the controlled object. The corrector is configured to generate a correction signal by correcting the control deviation according to a computational expression having adjustable coefficients. The second compensator is configured to generate a second signal from a neural network based on the correction signal. The computing device is configured to generate the control signal based on the first signal and the second signal.

Description

Technical Field

[0001] This invention relates to a control device, a method for adjusting the control device, a photolithography apparatus, and a method for manufacturing articles. Background Technology

[0002] Recently, the requirements for improving control accuracy have become more stringent, and in some cases, the required accuracy cannot be achieved solely through traditional feedback control. Therefore, efforts are underway to use neural network controllers (PTL 1) in parallel with traditional controllers. The parameters of a neural network controller are tuned through machine learning; however, this controller presents reliability issues. For example, a controller generated through machine learning may produce abnormal outputs under conditions significantly different from those given during learning (i.e., when the state of the controlled object changes or the disturbance environment changes). To address this issue, a technique has been proposed that incorporate limiting units in the output stage of the neural network controller to restrict the output (PTL 2).

[0003] Citation List

[0004] Patent documents

[0005] PTL 1: Japanese Patent Publication No. 7-603563

[0006] PTL 2: Japanese Patent Publication No. 2019-71505 Summary of the Invention

[0007] Technical issues

[0008] When the state of the controlled object changes and / or the disturbance environment changes, the predetermined neural network parameter values ​​of a traditional control device using neural networks may no longer be optimal, and control accuracy deteriorates. In this case, control accuracy can be improved by relearning to redetermine the neural network parameter values. However, relearning takes a considerable amount of time. Moreover, the predetermined learning sequence is executed during relearning, so the device cannot produce anything. Therefore, the execution of relearning reduces the device's productivity.

[0009] The purpose of this invention is to provide a technique that is advantageous in improving tolerance to changes in the state of the controlled object and / or changes in the disturbance environment.

[0010] Solution to the problem

[0011] One aspect of the present invention relates to a control apparatus for generating a control signal for controlling a controlled object, the control apparatus comprising a first compensator, a corrector, a second compensator, and a computing device, the first compensator being configured to generate a first signal based on a control deviation of the controlled object, the corrector being configured to generate a corrected signal by correcting the control deviation according to a computational expression having adjustable coefficients, the second compensator being configured to generate a second signal from a neural network based on the corrected signal, and the computing device being configured to generate the control signal based on the first signal and the second signal.

[0012] Advantages of the invention

[0013] This invention provides a technique that is advantageous in improving the tolerance to changes in the state of the controlled object and / or changes in the disturbance environment. Attached Figure Description

[0014] Figure 1 This is a diagram illustrating a configuration example of the system according to the first embodiment;

[0015] Figure 2 This is a diagram illustrating a configuration example of the system according to the first embodiment;

[0016] Figure 3 This is a block diagram illustrating an example configuration of the controller in the system of the first embodiment;

[0017] Figure 4 This is a block diagram illustrating an example configuration of the controller in the system of the first embodiment;

[0018] Figure 5 This is a diagram illustrating a configuration example of the system according to the first embodiment;

[0019] Figure 6 This is a flowchart illustrating an example of system operation when the system of the second embodiment is applied to a production apparatus;

[0020] Figure 7A This is a flowchart illustrating an example of the process of adjusting (or readjusting) the parameter values ​​of the corrector;

[0021] Figure 7B This is a flowchart illustrating an example of the process of adjusting (or readjusting) the parameter values ​​of the corrector;

[0022] Figure 8 This is a diagram illustrating an example of interference suppression characteristics;

[0023] Figure 9 This is a diagram showing the configuration of the bench control device according to the second embodiment;

[0024] Figure 10This is a block diagram illustrating an example configuration of the controller of the bench control device according to the second embodiment;

[0025] Figure 11 This is a diagram illustrating an example configuration of the exposure apparatus according to the third embodiment;

[0026] Figure 12 This is a diagram illustrating an example of position control deviation in the third embodiment; and

[0027] Figure 13 This is a diagram illustrating an example of the results of frequency analysis in the third embodiment. Detailed Implementation

[0028] In the following, embodiments will be described in detail with reference to the accompanying drawings. Note that the following embodiments are not intended to limit the scope of the claimed invention. Multiple features are described in the embodiments, but the invention is not limited to claiming all such features, and multiple such features may be suitably combined. Furthermore, in the drawings, similar or analogous configurations are given the same reference numerals, and redundant descriptions thereof are omitted.

[0029] Figure 1 The configuration of the system SS according to the first embodiment is shown. The system SS can be applied, for example, to a manufacturing apparatus for manufacturing articles. This manufacturing apparatus may include, for example, a processing apparatus for processing materials or components that form parts of an article or a part of an article. The processing apparatus may be, for example, a photolithography apparatus for transferring a pattern onto a material or component, a film forming apparatus for forming a film on a material or component, an etching apparatus for etching a material or component, and a heating apparatus for heating a material or component.

[0030] System SS may include, for example, sequence unit 101, control device 100, and controlled object 103. Control device 100 may include controller 102. Control device 100 or controller 102 may generate a manipulated variable signal MV as a control signal for controlling controlled object 103. When system SS is applied to a production system, a production sequence may be provided to sequence unit 101. The production sequence may define the production process. Based on this production sequence, sequence unit 101 may generate a target value R for controlling controlled object 103 and provide the target value R to control device 100 or controller 102.

[0031] Control device 100 or controller 102 can perform feedback control on controlled object 103. More specifically, based on the control deviation, which is the difference between a target value R provided from sequence unit 101 and a controlled variable CV provided from controlled object 103, control device 100 or controller 102 can control controlled object 103 such that the control variable CV of controlled object 103 follows the target value R. Controlled object 103 may have a sensor for detecting the control variable CV, and the control variable CV detected by the sensor can be provided to controller 102. Each of the target value R, manipulated variable MV, and control variable CV can be time-series data whose value changes according to the elapsed time.

[0032] like Figure 2 As shown, the system SS may include a learning unit 201. The learning unit 201 can be configured as part of the control device 100, or it can be configured as an external device of the control device 100. When the learning unit 201 is configured as an external device of the control device 100, the learning unit 201 can be detached from the control device 100 after learning is complete. The learning unit 201 can be configured to send a prepared learning sequence to a sequence unit 101. The sequence unit 101 can generate a target value R based on the learning sequence and provide the target value R to the controller 102.

[0033] Controller 102 can generate a manipulation quantity MV based on a control deviation, which is the difference between a target value R generated by sequence unit 101 according to a learning sequence and provided by it, and a control quantity CV provided by the controlled object 103. Controller 102 has a neural network and can generate the manipulation quantity MV by using the neural network. The manipulation quantity MV generated by controller 102 can be provided to the controlled object 103, and the controlled object 103 can operate according to the manipulation quantity MV. The control quantity CV, as a result of this operation, can be provided to controller 102. Controller 102 can provide learning unit 201 with a historical operation log indicating the operation of controller 102 based on the target value R. Based on this operation log, learning unit 201 can determine the parameter values ​​of the neural network and set the parameter values ​​in the neural network of controller 102. The parameter values ​​can be determined by machine learning such as reinforcement learning.

[0034] Figure 3An example configuration of controller 102 is shown. Controller 102 may include a first compensator 301 for generating a first signal S1 based on the control deviation E of the controlled object 103, and a corrector 303 for generating a correction signal CS by correcting the control deviation E according to an operational expression with adjustable coefficients. Controller 102 may also include a second compensator 302 for generating a second signal S2 from a neural network based on the correction signal CS, and an arithmetic unit 306 for generating a manipulated quantity MV as a control signal based on the first signal S1 and the second signal S2. The manipulated quantity MV is the sum of the first and second signals S1 and S2, therefore the arithmetic unit 306 may be configured as an adder. From another perspective, the manipulated quantity MV is a signal corrected based on the first and second signals S1 and S2. Controller 102 may include a subtractor 305 for generating the control deviation E as the difference between the target value R and the control quantity CV.

[0035] The controller 102 may further include an operation log recorder 304. The learning unit 201 may be configured to determine the parameter values ​​of the neural network of the second compensator 302. For this learning process of the learning unit 201, the operation log recorder 304 may record the operation logs required for the learning process of the learning unit 201 and provide the recorded operation logs to the learning unit 201. The operation logs may be a correction signal CS as input data to the second compensator 302 and a second signal S2 as output data of the second compensator 302, but may also be other data.

[0036] The following will explain some configuration examples of the corrector 303. The first to fifth configuration examples provide examples of operational expressions to be used by the corrector 303 to generate a correction signal CS based on the control deviation E. The operational expressions can be monomials or polynomials.

[0037] In the first configuration example, the corrector 303 has control characteristics represented by the operational expression in Equation 1. In this operational expression, x is the input (E) of the corrector 303, y is the output (CS) of the corrector 303, and K p It is an arbitrary coefficient (constant).

[0038] y = K p x ...(1)

[0039] In the second configuration example, the corrector 303 has control characteristics represented by the operational expression in Equation 2. In this operational expression, x is the input (E) of the corrector 303, y is the output (CS) of the corrector 303, t is time, and K... i The coefficients are arbitrary (constant). Note that integration can be performed multiple times. The integration can be a definite integral or an indefinite integral over a given time interval.

[0040] y = K i ∫x dt...(2)

[0041] In the third configuration example, the corrector 303 has control characteristics represented by the operational expression in Equation 3. In this operational expression, x is the input (E) of the corrector 303, y is the output (CS) of the corrector 303, t is time, and K d The coefficients are arbitrary (constant). Note that differentiation can be performed multiple times.

[0042]

[0043] In the fourth configuration example, the corrector 303 has control characteristics represented by the operational expression in Equation 4. In this operational expression, x is the input (E) of the corrector 303, y is the output (CS) of the corrector 303, and K p K i and K d The coefficients are arbitrary (constant). Note that integration and differentiation can be performed multiple times.

[0044]

[0045] In the fifth configuration example, the corrector 303 has control characteristics represented by the operational expression in Equation 5. In this operational expression, x is the input (E) of the corrector 303, y is the output (CS) of the corrector 303, n is the integral order of the multiple integral, m is the derivative order, and K... p K is an arbitrary coefficient (constant). i_n K are arbitrary coefficients (constants) of the nth multiple integral, and K d_m These are arbitrary coefficients (constants) of the m-th order differential.

[0046]

[0047] Each of the first to fifth configuration examples can be understood as the following example: in this example, the operational expression to be used by the corrector 303 to generate the correction signal CS includes at least one of the following: a term proportional to the control deviation E, a term that is integrated once or more with respect to the control deviation E, and a term that is differentiated once or more with respect to the control deviation E.

[0048] The coefficients (constants) K of the operational expressions in the first to fifth configuration examples p K i K d K i_n and K d_mThis is an example of the adjustable parameters of the corrector 303. When the state of the controlled object 103 changes and / or the disturbance environment changes while the system SS is operating, this change or these changes can be controlled by adjusting the values ​​(parameter values) of the operational expressions (coefficients) exemplified as examples of the first to fifth configurations. The time required to adjust the values ​​of the operational expressions (coefficients) of the corrector 303 is shorter than the time required for the neural network to relearn. Therefore, control accuracy can be maintained without reducing the productivity of the system SS. That is, tolerance to changes in the state of the controlled object and / or changes in the disturbance environment can be improved by using the corrector 303.

[0049] Figure 4 Another configuration example of controller 102 is shown. For example... Figure 4 As shown, the controller 102 may have multiple (two or more) neural networks 302. A correction signal CS generated by the corrector 303 can be provided to the multiple neural networks 302. Alternatively, the correction signal CS generated by the corrector 303 can be provided to a selected neural network 302 among the multiple neural networks 302. One of the multiple neural networks 302 can be selected by the selector 401 based on the operating mode of the controlled object 103, and the output of the neural network 302 selected by the selector 401 is provided as a second signal S2 to the arithmetic unit 306. Information indicating the operating mode to be used by the selector 401 to select the neural network can be provided from the sequence unit 101.

[0050] The parameter values ​​of multiple neural networks 302 can be determined according to the operating mode of the controlled object 103. For example, when the controlled object 103 includes a stage and the position of the stage needs to be controlled, different neural networks can be used for the operating mode in the acceleration range where the stage is accelerated and for other operating modes. Alternatively, two systems SS can be incorporated into the exposure apparatus, with one system SS controlling the board stage (substrate stage) and the other system SS controlling the mask stage (original stage). In this case, different neural networks can be used in the system SS for the operating mode in the synchronization range where the board stage and the mask stage are driven synchronously and for other operating modes.

[0051] As described above, when multiple neural networks 302 are used, the use of the corrector 303 can improve the tolerance to changes in the state of the controlled object and / or changes in the disturbance environment.

[0052] like Figure 5As shown, the control device 100 may include a setting unit 202 for setting parameter values ​​of the corrector 303. The setting unit 202 can perform an adjustment process to adjust the parameter values ​​of the corrector 303, and determine the parameter values ​​of the corrector 303 through this adjustment process, or it can determine the parameter values ​​of the corrector 303 based on a command from a user. In the former case, the setting unit 202 can send a check sequence to the sequence unit 101 for checking the operation of the controller 102, and based on this check sequence, cause the sequence unit 101 to generate a target value R. Then, the setting unit 202 can obtain an operation log (e.g., control deviation) from the controller 102 operating based on the target value R, and determine the parameter values ​​of the corrector 303 based on the operation log. The setting unit 202 with the above functions can be understood as an adjustment unit for adjusting the parameter values ​​of the corrector 303.

[0053] During the production process where sequence unit 101 generates the target value R based on the production sequence, setting unit 202 can obtain an operation log (e.g., control deviation) from controller 102 and determine whether to adjust the parameter value of corrector 303 based on this operation log. Alternatively, a determining unit for determining whether to adjust the parameter value of corrector 303 by using setting unit 202 can be installed separately from setting unit 202.

[0054] Figure 6An operational example of the system SS in a first embodiment when the system SS is applied to a production unit is shown. In step S501, the sequence unit 101 can generate a target value R based on a given production sequence and provide the target value R to the control device 100 or the controller 102. The control device 100 or the controller 102 can control the controlled object 103 based on the target value R. In step S502, the setting unit 202 can obtain the operation log (e.g., control deviation) of the controller 102 in step S501. In step S503, the setting unit 202 can determine whether to perform an adjustment (or readjustment) of the parameter value of the corrector 303 based on the operation log obtained in step S502. For example, when the operation log meets a predetermined condition, the setting unit 202 can determine to perform an adjustment (or readjustment) of the parameter value of the corrector 303. The predetermined condition is a condition for stopping production, for example, a condition where the control deviation obtained from the operation log exceeds a predetermined value. If it is determined that the setting unit 202 will adjust (or readjust) the parameter values ​​of the corrector 303, then the process proceeds to step S504; otherwise, the process proceeds to step S505. In step S504, the setting unit 202 performs the adjustment (or readjustment) of the parameter values ​​of the corrector 303. This adjustment is performed while the parameter values ​​of the second compensator 302 remain in their previous state, and the parameter values ​​(coefficients) of the corrector 303 are reset through the adjustment.

[0055] In step S505, sequence unit 101 determines whether production should be terminated according to the production sequence. If the result in step S505 is "No", then the process returns to step S501. If the result in step S501 is "Yes", then production is terminated. In the above process, even when production is about to be stopped, the parameter value of corrector 303 can be adjusted immediately and production can be restarted while minimizing production interruption.

[0056] In step S504, the setting unit 202 can send a check sequence to the sequence unit 101, causing the sequence unit 101 to execute the check sequence and obtain the operation log (e.g., control deviation) from the check sequence from the controller 102. Then, the setting unit 202 can perform frequency analysis on the operation log, determine the frequency to be improved based on the analysis results, and determine the parameter value of the corrector 303 such that the control deviation at that frequency does not exceed a predetermined value. A more practical example of step S504 will be explained in the second embodiment.

[0057] Figure 8 An example of measurement results for interference suppression characteristics is shown. Interference suppression characteristics are the frequency response of the control deviation E when a sine wave is given as the manipulation variable MV. Figure 8In the graph, the horizontal axis represents frequency, and the vertical axis represents the gain of the interference suppression characteristic. The interference suppression characteristic represents the frequency response of the control deviation E when interference is added to the manipulated input; therefore, a large gain indicates low interference suppression effectiveness. Conversely, a small gain indicates high interference suppression effectiveness. Figure 8 In the image, the broken line indicates the interference suppression characteristics before adjustment.

[0058] When determined Figure 8 When performing step S504, the frequencies indicated by the alternating long and short dashed lines are the frequencies at which the interference suppression characteristics are to be improved. For example, interference suppression characteristics indicated by solid lines can be obtained. The gain of the interference suppression characteristics decreases at the frequencies to be improved, indicating that the interference suppression characteristics have been improved.

[0059] The second embodiment will now be explained. Matters not mentioned in the second embodiment follow the first embodiment. Figure 9 An example is shown of the control system SS or control device 100 of the first embodiment being applied to a bench control device 800. The bench control device 800 is configured to control a bench 804. The bench control device 800 may include, for example, a control board 801, a current driver 802, a motor 803, a bench 804, and a sensor 805. The control board 801 corresponds to the control device 100 or controller 102 in the system SS of the first embodiment. The current driver 802, motor 803, bench 804, and sensor 805 correspond to the controlled object 103 in the system SS of the first embodiment. However, the current driver 802 may also be incorporated into the control board 801. Although in Figure 9 Although not shown, the bench control device 800 may include a sequence unit 101, a learning unit 201, and a setting unit 202.

[0060] The target position value can be supplied from the sequence unit 101 to the control board 801. Based on this target position value supplied from the sequence unit 101 and the position information supplied from the sensor 805, the control board 801 can generate a current command as a control signal or as a manipulated quantity (manipulation command), and supply this current command to the current driver 802. Furthermore, the control board 801 can supply an operation log to the sequence unit 101.

[0061] The current driver 802 can supply current to the motor 803 in accordance with current commands. The motor 803 can be an actuator that converts the current supplied from the current driver 802 into driving force and uses this driving force to drive the stage 804. The stage 804 can hold an object such as a plate or mask. The sensor 805 can detect the position of the stage 804 and supply the obtained position information to the control board 801.

[0062] Figure 10An example configuration of the control board 801 is shown as a block diagram. The control board 801 may include a first compensator 301 for generating a first signal S1 based on the position control deviation E of the test bench 804, which is the controlled object, and a corrector 303 for generating a correction signal CS by correcting the control deviation E according to an arithmetic expression with adjustable coefficients. The control board 801 may also include a second compensator 302 for generating a second signal S2 from the correction signal CS using a neural network, and an arithmetic unit 306 for generating a current command as a control signal or as a manipulation signal based on the first signal S1 and the second signal S2. The control board 801 may further include a subtractor 305 for generating the control deviation E as the difference between the position target value PR and the position information.

[0063] The bench control device 100 of the second embodiment may include a learning unit 201, which may be configured to perform learning for determining the parameter values ​​of the neural network of the second compensator 302. For this learning by the learning unit 201, an operation log recorder 304 may record the operation log required for the learning by the learning unit 201 and provide the recorded operation log to the learning unit 201. The operation log may be, for example, a correction signal CS as input data of the second compensator 302 and a second signal S2 as output data of the second compensator 302, but may also be other data.

[0064] The bench control device 100 of the second embodiment may include a setting unit 202. The setting unit 202 may perform an adjustment process to adjust the parameter values ​​of the calibrator 303, determine the parameter values ​​of the calibrator 303 through this adjustment process, and set the determined parameter values. It may also set the parameter values ​​of the calibrator 303 based on commands from the user.

[0065] Will refer again Figure 6This section explains an example of the operation of the bench control device 800 in the second embodiment when it is applied to a production apparatus. In step S501, the sequence unit 101 can generate a position target value PR based on a given production sequence and provide the position target value PR to the bench control device 800. The bench control device 800 can control the position of the bench 804 based on the position target value PR. In step S502, the setting unit 202 can acquire the operation log (e.g., control deviation) of the control board 801 from step S501. In step S503, the setting unit 202 can determine, based on the operation log acquired in step S502, whether to perform an adjustment (or readjustment) of the parameter value of the corrector 303. For example, if the operation log meets predetermined conditions, then the setting unit 202 can determine to perform an adjustment (or readjustment) of the parameter value of the corrector 303. The predetermined conditions are conditions used to stop production, such as the condition that the maximum value of the position control deviation exceeds a predetermined value when the bench 804 is driven at a constant speed. If it is determined that the setting unit 202 will adjust (or readjust) the parameter values ​​of the corrector 303, the process proceeds to step S504; otherwise, the process proceeds to step S505. In step S504, the setting unit 202 may adjust (or readjust) the parameter values ​​of the corrector 303. In step S505, the sequence unit 101 determines whether to terminate production according to the production sequence. If the result in step S505 is "No", the process returns to step S501. If the result in step S505 is "Yes", production is terminated.

[0066] Figure 7A and 7B The illustration shows a practical example of the process of adjusting (or readjusting) the parameter values ​​of the corrector 303 in step S504. In step S601, the setting unit 202 may send a check sequence for checking the operation of the bench control device 800 to the sequence unit 101, and based on this check sequence, the sequence unit 101 generates a position target value PR. In step S602, the setting unit 202 may obtain the position control deviation E as an operation log from the controller 102 operating based on the position target value PR. Figure 12 An example of position control deviation is shown. Figure 12 In the graph, the horizontal axis indicates time, and the vertical axis indicates the position control deviation E. The curve indicated by the dotted line represents the position control deviation E before the parameter values ​​of the corrector 303 were adjusted, and shows that the position control accuracy has deteriorated.

[0067] In step S603, the setting unit 202 can perform frequency analysis on the position control deviation E obtained in step S602. Figure 13 An example of the results of the frequency analysis in step S603 is shown. Figure 13 In the diagram, the horizontal axis represents frequency, and the vertical axis represents the power spectrum. The dotted line indicates the frequency showing the maximum spectrum before adjustment. In step S604, the setting unit 202 can determine the frequency in the power spectrum that shows the maximum spectrum and is the frequency to be improved.

[0068] Steps S605 to S610 indicate a practical example of the adjustment process for the parameter values ​​of the corrector 303. This example will be explained using the steepest descent method as the parameter value adjustment method, but another method can also be used. In step S605, the setting unit 202 initializes n to 1. For example, when the operational expression of the corrector 303 is configured with three terms (i.e., a first-order integral term, a proportional term, and a first-order differential term), the parameters with the values ​​to be adjusted are the three parameters K. i K p and K d Equation 6 below shows the parameter value p in the nth adjustment. n .

[0069]

[0070] In step S606, the setting unit 202 can set the parameter value p n In the first adjustment, the parameter value p1 is set to an arbitrary initial value. In the nth adjustment, the parameter value p can be set to be indicated by Equation 8 (described later). n .

[0071] Used to adjust parameter value p n The objective function J(p) n The gain could be, for example, the gain of the interference suppression characteristic at the frequency determined in step S604. In step S607, the setting unit 202 can measure the target function J(p) n The gradient vector gradJ(p) n Gradient vector grad J(p) n The gradient vector grad J(p) can be given by Equation 7. n This can be achieved by configuring the parameter value p. n element K i-n K p-n and K d-n To measure by changing a very small amount.

[0072]

[0073] In step S608, setting unit 202 as the convergence determination unit of the steepest descent method can determine the gradient vector gradJ(p) n Check whether the value of each element of ) is equal to or less than a specified value. If the gradient vector grad J(p) is... nIf the value of each element of ) is equal to or less than a specified value, then the setting unit 202 can terminate the adjustment of the parameter values ​​of the corrector 303. On the other hand, the gradient vector grad J(p n If the value of each element exceeds a specified value, the setting unit 202 can calculate the parameter value p in step S609. n+1 In this step, the parameter value p n+1 It can be calculated using, for example, an arbitrary constant α greater than 0 according to Equation 8. In step S610, setting unit 202 increments the value of n by 1 and returns to step S606.

[0074] p n+1 =p n -αgrad J(p n ...(8)

[0075] In step S611, the setting unit 202 can send a check sequence for checking the operation of the bench control device 800 to the sequence unit 101, and the sequence unit 101 generates a position target value PR based on the check sequence. In step S612, the setting unit 202 can obtain the position control deviation E as an operation log from the controller 102 operating based on the position target value PR.

[0076] In step S613, the setting unit 202 determines whether the position control deviation E obtained in step S612 is equal to or less than a predetermined value. If the position control deviation E exceeds the predetermined value, the setting unit 202 can return to step S601 and perform the adjustment again. If the position control deviation E is equal to or less than the predetermined value, the setting unit 202 can terminate the adjustment.

[0077] In the second embodiment, when the state of the controlled object, including bench 804, changes and / or the disturbance changes, this change can be controlled by adjusting the parameter values ​​of the corrector 303. Figure 12 In the example shown, the position control deviation indicated by the dotted line is reduced to the position control deviation indicated by the solid line, thus improving control accuracy.

[0078] In the example indicated by Equation 6, the corrector 303 has only three parameters, which is much smaller than the number of parameters in a typical neural network. For example, when using a deep neural network, if the input layer has an order of 5, the hidden layers have an order of 32×2, and the output layer has an order of 8, then the number of parameters is 1545. The time required to adjust the parameter values ​​of the corrector 303 is less than the time required to determine 1545 parameter values ​​through relearning. This allows control accuracy to be maintained without sacrificing the productivity of the bench control device 800.

[0079] Figure 11An example configuration of the exposure apparatus EXP according to the third embodiment is illustrated schematically. The exposure apparatus EXP can be configured as a scanning exposure apparatus. The exposure apparatus EXP may include, for example, an illumination source 1000, an illumination optics system 1001, a mask stage 1003, a projection optics system 1004, and a plate stage 1006. The illumination source 1000 may include, but is not limited to, a mercury lamp, an excimer laser source, or an EUV source. The exposure light 1010 from the illumination source 1000 is shaped by the illumination optics system 1001 into the shape of the irradiated area of ​​the projection optics system 1004 with uniform illumination. In one example, the exposure light 1010 may be shaped into a rectangle extending in the X direction, which is an axis perpendicular to the plane formed with the Y-axis and Z-axis. The exposure light 1010 may be shaped into an arc shape depending on the type of projection optics system 1004. The shaped exposure light 1010 is emitted onto the pattern of the mask (original) 1002, and the exposure light 1010 that has passed through the pattern of the mask 1002 forms an image of the pattern of the mask 1002 on the surface of the plate (substrate) 1005 via the projection optics system 1004.

[0080] Mask 1002 is held by mask stand 1003 by, for example, vacuum adsorption. Plate 1005 is held by chuck 1007 of plate stand 1006 by, for example, vacuum adsorption. The positions of mask stand 1003 and plate stand 1006 can be controlled by a multi-axis position control device including a position sensor 1030 such as a laser interferometer or laser ruler, a drive system 1031 such as a linear motor, and a controller 1032. The position measurement value output from the position sensor 1030 can be provided to the controller 1032. The controller 1032 generates a control signal (manipulation signal) based on the position control deviation, which is the difference between the position target value and the position measurement value, and provides the control signal to the drive system 1031, thereby driving the mask stand 1003 and plate stand 1006. By simultaneously driving the mask stage 1003 and the plate stage 1006 in the Y direction to perform scanning exposure on the plate 1005, the pattern of the mask 1002 is transferred to the photosensitive material on the plate 1005.

[0081] The following will explain how the second embodiment is applied to the control of the board platform 1006. (Refer to...) Figure 9The control board 801 corresponds to the controller 1032, the current driver 802 and the motor 803 correspond to the drive system 1031, the stage 804 corresponds to the board stage 1006, and the sensor 805 corresponds to the position sensor 1030. By applying a controller with a neural network to the control of the board stage 1006, the position control deviation of the board stage 1006 can be reduced. Therefore, overlap accuracy can be improved. The parameter values ​​of the neural network can be determined by a predetermined learning sequence. However, if the state of the controlled object has changed and / or the interference environment has changed since the learning time, the control accuracy of the board stage 1006 decreases. Even in such cases, the parameter values ​​of the corrector can be adjusted within a shorter time period than the time required to relearn the neural network. Therefore, control accuracy can be maintained without sacrificing the productivity of the exposure apparatus.

[0082] The following will explain how the second embodiment is applied to the control of the mask stage 1003. (Refer to...) Figure 9 The control board 801 corresponds to the controller 1032, the current driver 802 and the motor 803 correspond to the drive system 1031, the stage 804 corresponds to the mask stage 1003, and the sensor 805 corresponds to the position sensor 1030.

[0083] Even when the second embodiment is applied to the mask stage 1003, the position control deviation of the mask stage 1003 can be reduced. Therefore, overlap accuracy can be improved. The parameter values ​​of the neural network can be determined through a predetermined learning sequence. However, if the state of the controlled object changes and / or the interfering environment changes since the learning time, the control accuracy of the mask stage 1003 decreases. Even in such cases, the parameter values ​​of the corrector can be adjusted within a shorter time period than required to relearn the neural network. Therefore, control accuracy can be maintained without sacrificing the productivity of the exposure apparatus.

[0084] The second embodiment can be applied not only to the control of the stage in an exposure apparatus, but also to the control of the stage in other lithography apparatuses such as imprinting apparatuses and electron beam lithography apparatuses. Furthermore, the first or second embodiment can be applied to the control of moving parts of a conveyor mechanism for transporting articles, such as the moving parts of a hand for holding the article.

[0085] The photolithography apparatus described above can be used to perform article manufacturing methods for manufacturing various articles (e.g., semiconductor IC devices, liquid crystal display devices, and MEMS). The article manufacturing method includes a transfer step of transferring a pattern of a workpiece onto a substrate using the aforementioned photolithography apparatus, a processing step of processing the substrate that has undergone the transfer step, and obtaining an article from the processed substrate. When the photolithography apparatus is an exposure apparatus, the transfer step may include an exposure step of exposing the substrate to the workpiece, and a development step of developing the substrate that has undergone the exposure step.

[0086] This invention can be implemented by supplying a program for implementing one or more functions of the above embodiments to a system or device via a network or storage medium, and by causing one or more processors in the computer of the system or device to read and execute the program. This invention can also be implemented by circuitry (e.g., an ASIC) for implementing one or more functions.

[0087] This invention is not limited to the embodiments described above, and various changes and modifications can be made within the spirit and scope of this invention. Therefore, the appended claims are presented to inform the public of the scope of this invention.

[0088] This application claims priority to Japanese Patent Application No. 2020-152293, filed on September 10, 2020, which is hereby incorporated herein by reference.

[0089] Reference Symbol List

[0090] 100: Control device; 102: Controller; 103: Controlled object; 301: First compensator; 302: Second compensator; 303: Corrector; 305: Subtractor; 306: Computing device.

Claims

1. A control device for generating control signals for controlling a controlled object, comprising: A first compensator is configured to generate a first signal based on the control deviation of the controlled object; A corrector configured to generate a correction signal by correcting the control deviation according to an operational expression having adjustable coefficients; A second compensator is configured to generate a second signal from a neural network based on the correction signal; A computing device configured to generate the control signal based on the first signal and the second signal; as well as The setting unit is configured to perform frequency analysis on the control deviation, determine the frequency to be improved based on the analysis results, and determine the parameter values ​​of the coefficients of the operational expression such that the control deviation at the frequency does not exceed a specified value.

2. The control device according to claim 1, wherein, The computational expression includes a term proportional to the control deviation.

3. The control device according to claim 1, wherein, The operational expression includes a term that integrates the control deviation at least once.

4. The control device according to claim 1, wherein, The operational expression includes a term that is differentiated at least once over the control deviation.

5. The control device according to claim 1, wherein, The operational expression includes at least one of the following: a term proportional to the control deviation, a term that integrates the control deviation at least once, and a term that differentiates the control deviation at least once.

6. The control device according to claim 1, wherein, The setting unit determines the parameter values ​​of the coefficients of the operational expression while the parameter values ​​of the neural network remain in the previous state.

7. The control device according to claim 1, wherein, The setting unit determines the parameter value of the operation expression based on the interference suppression characteristic, wherein the interference suppression characteristic represents the frequency response of the control deviation when interference is added to the control signal.

8. The control device according to any one of claims 1 to 7, further comprising a learning unit configured to determine the parameter values ​​of the neural network through machine learning.

9. A photolithography apparatus for transferring a pattern of an original onto a substrate, comprising a control device defined in any one of claims 1 to 8 and configured to control the position of the substrate or the original.

10. A method for manufacturing an article, comprising: The transfer step of transferring the pattern of the original to the substrate by using the photolithography apparatus as defined in claim 9; as well as The processing steps involve processing the substrate that has already undergone the transfer steps. The article is obtained from a substrate that has already undergone the aforementioned processing steps.

11. A method for adjusting a control device, the control device comprising a first compensator, a corrector, a second compensator, and a computing device, wherein the first compensator is configured to generate a first signal based on a control deviation of a controlled object, the corrector is configured to generate a correction signal by correcting the control deviation according to a computational expression having adjustable coefficients, the second compensator is configured to generate a second signal from a neural network based on the correction signal, and the computing device is configured to generate a control signal based on the first signal and the second signal, the method comprising: The characteristics of the corrector are adjusted while the parameters of the neural network remain in their previous state. Perform frequency analysis on the control deviation; The frequency to be improved is determined based on the analysis results; and The parameter values ​​of the coefficients in the operational expression are determined such that the control deviation at the frequency does not exceed a specified value.

Images