Control apparatus, lithographic apparatus, and method of manufacturing an article

Abstract

A control device for generating a control signal for controlling a control target includes a plurality of neural networks; and a selector configured to select, from the plurality of neural networks, a neural network to be used for generating the control signal, wherein each of the plurality of neural networks is selected by the selector to be used in execution of a corresponding control mode among a plurality of control modes for controlling the control target.

Description

Technical Field

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

[0002] For control devices that control the physical quantity of the control target, typical controllers such as PID controllers, controllers based on modern control theory, or controllers using neural networks can be used. Alternatively, controllers using both those without neural networks and those with neural networks can be used. In the control device described in PTL 1, both PID controllers and controllers using neural networks are used to improve control accuracy.

[0003] Citation List

[0004] Patent documents

[0005] PTL 1: Japanese Patent Application Publication No. 2019-71405 Summary of the Invention

[0006] Technical issues

[0007] In traditional control devices that use neural networks, computational processing takes a long time, so the calculations for control may not be completed within the predetermined time, thus degrading the real-time characteristics of the control.

[0008] This invention provides a technique that is advantageous in reducing the time spent on calculations for control.

[0009] Solution to the problem

[0010] One aspect of the present invention relates to a control apparatus for generating a control signal for controlling a control objective, and the apparatus includes: a plurality of neural networks; and a selector configured to select from the plurality of neural networks a neural network to be used to generate the control signal, wherein each of the plurality of neural networks is selected by the selector to be used in the execution of a corresponding control mode among a plurality of control modes for controlling the control objective.

[0011] Beneficial effects of the invention

[0012] According to the present invention, a technique is provided that is advantageous in reducing the time spent performing calculations for control. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

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

[0015] Figure 2 This is a block diagram illustrating an example of the arrangement of the processing apparatus;

[0016] Figure 3 This is a block diagram illustrating an example arrangement of the processing apparatus;

[0017] Figure 4 This is a block diagram illustrating another example of the arrangement of the processing apparatus;

[0018] Figure 5 This is a flowchart illustrating the learning sequence in the processing system;

[0019] Figure 6 It is a timing diagram illustrating the velocity profile and acceleration profile of the test bench;

[0020] Figure 7 This is a diagram illustrating an example of a processing system being applied to a scanning exposure apparatus (photolithography apparatus);

[0021] Figure 8 This is a flowchart illustrating the exposure sequence of a scanning exposure apparatus;

[0022] Figure 9 This is a block diagram showing an example of the arrangement of a temperature controller;

[0023] Figure 10 This is a block diagram illustrating an example arrangement of a temperature controller; and

[0024] Figure 11 This is a block diagram illustrating another example of the arrangement of a temperature controller. Detailed Implementation

[0025] 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. Several features are described in the embodiments, but it is not necessary to limit the invention to all such features, and multiple such features can be suitably combined. Furthermore, in the drawings, the same reference numerals are given for the same or similar configurations, and repeated descriptions thereof are omitted.

[0026] Figure 1The configuration of a processing system PS according to an embodiment is shown. The processing system PS may include, for example, a processing device 1, a control server (high-level device) 2 that controls the processing device 1, and a learning server (learning device) 3 that performs learning to determine the parameter values ​​of a neural network included in the processing device 1. The processing device 1 is, for example, a device that performs processing on a target object, such as a manufacturing device, an inspection device, or a monitoring device. An overview of the processing may include processing, inspection, monitoring, and observation of the target object.

[0027] Processing device 1 may include a control target and control the control target using a neural network that determines parameter values ​​through reinforcement learning. Control server 2 may be configured to send control commands (e.g., drive commands) to processing device 1 and receive control results (e.g., drive results) from processing device 1. Control server 2 and learning server 3 may be formed by, for example, a general-purpose computer with programs installed or a combination of all or some of these components.

[0028] The learning server 3 can perform reinforcement learning to determine the values ​​of multiple parameters of the neural network fused in the processing device 1. More specifically, the learning server 3 can send control commands to the processing device 1 via the control server 2 and receive control results from the processing device 1. Then, the learning server 3 can calculate a reward based on the control results and update the values ​​of multiple parameters of the neural network based on the reward.

[0029] All or some of the functions of control server 2 can be integrated into learning server 3. All or some of the functions of control server 2 can be integrated into processing device 1. Processing device 1, control server 2, and learning server 3 can be physically integrated or separate. Processing device 1 as a whole can be controlled by control server 2, or it can include components controlled by control server 2 and components not controlled by control server 2. Separating control server 2 and learning server 3 may be advantageous if the computational cost of updating parameter values ​​involving the neural network is high. If there are multiple control objectives, one control server 2 and multiple learning servers 3 can be provided.

[0030] Figure 2An arrangement of processing device 1 is illustrated. Processing device 1 may include: a bench mechanism 5 including a bench (holder) ST as the control target; a sensor 6 for detecting the position or state of the bench ST; a driver 7 for driving the bench mechanism 5; and a control device 8 for giving command values ​​to the driver 7 and receiving outputs from the sensor 6. The bench ST can hold a positioning target object. The bench ST may be guided by a guide (not shown). The bench mechanism 5 may include an actuator AC for moving the bench ST. The driver 7 drives the actuator AC. More specifically, for example, the driver 7 may supply current (electrical energy) to the actuator AC corresponding to a command value given from the control device 8. The actuator AC can move the bench ST by force (mechanical energy) corresponding to the current supplied from the driver 7. The control device 8 may use a neural network that determines parameter values ​​through reinforcement learning to control the position or state of the bench ST as the control target.

[0031] Figure 3 It is shown Figure 2 The block diagram illustrates an example arrangement of the processing device 1. The processing device 1 may include a test bench mechanism 5, a sensor 6 for detecting the position or state of the test bench ST, a driver 7 for driving the test bench mechanism 5, and a control device 8 that provides manipulated variables (control signals) to the driver 7 based on control deviations and receives outputs from the sensor 6. The control device 8 may be formed, for example, by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), a general-purpose computer with a program installed, or a combination of all or some of these components.

[0032] Control device 8 generates control signals for controlling the test bench mechanism 5 (test bench ST), which is the control target. Control device 8 may include multiple neural networks 85a, 85b, and 85c, and a selector 87 that selects the neural network to be used to generate the control signal from the multiple neural networks 85a, 85b, and 85c. Based on selection information, selector 87 can select the neural network to be used to generate the control signal from the multiple neural networks 85a, 85b, and 85c. The selected neural network can be used as a compensator to generate manipulated variables based on input information (control deviation). The selection information can be generated in control device 8 or provided from another device (e.g., a higher-level device). Selection information can be generated such that the neural network to be used is selected from the multiple neural networks 85a, 85b, and 85c according to the control mode selected from multiple control modes. The multiple control modes can be control modes that can be distinguished from each other. The control mode may include, for example, a target value string (time series data of the target value).

[0033] Control device 8 may include a subtractor 86 that calculates the difference (i.e., control deviation) between a target value (e.g., target position) and a state signal indicating the state (e.g., position) of the indicator rig 5 (rig ST). Selector 87 may include multiplexer 87a and multiplexer 87b. Multiplexer 87a may supply the control deviation to a neural network specified by selection information among a plurality of neural networks 85a, 85b, and 85c. Multiplexer 87b may output the signal generated by the neural network specified by selection information among the plurality of neural networks 85a, 85b, and 85c. Note that... Figure 3 Three neural networks, 85a, 85b, and 85c, are illustrated, but the number of neural networks is arbitrary. The number of layers and neurons in each of the multiple neural networks 85a, 85b, and 85c is arbitrary. The parameter values ​​of the multiple neural networks 85a, 85b, and 85c can be determined through reinforcement learning.

[0034] Each of the plurality of neural networks 85a, 85b, and 85c may be selected by selector 87 to be used in the execution of a corresponding control mode among the control modes for controlling the control objective. Alternatively, each of the plurality of neural networks 85a, 85b, and 85c may be selected by selector 87 to be used in the execution of at least a portion of at least one of the plurality of control modes for controlling the control objective.

[0035] Figure 4 It is shown Figure 2 A block diagram illustrating another arrangement example of the processing device 1 shown in the figure. Figure 4 The arrangement example shown includes by means of Figure 3The arrangement shown is an example of an arrangement obtained by adding a main compensator 81 and an adder 82. The main compensator 81 can be, for example, a compensator including P-elements (proportional elements), I-elements (integral elements), and D-elements (differential elements) (e.g., a PID compensator), but is not limited thereto. The subtractor 86 calculates the difference between a target value (e.g., target position) and a state signal indicating the state (e.g., position) of the test bench mechanism 5 (test bench ST), and supplies the difference to both the neural network selected by the selector 87 and the main compensator 81. The adder 82 can add a first manipulated variable generated by the main compensator 81 and a second manipulated variable generated by the neural network selected by the selector 87 to generate a control signal. This control signal can be supplied to the driver 7.

[0036] Figure 5 An example of a learning sequence in the processing system PS is illustrated. This learning sequence is controlled by the learning server 3. In step S501, the learning server 3 selects one of several control modes for learning multiple neural networks 85a, 85b, and 85c. Steps S502 to S505 indicate loops to be repeatedly executed. In step S502, which is executed first in this loop, the learning server 3 initializes the parameter values ​​of the neural network to be learned. In step S502, which is executed a second time or thereafter in this loop, the learning server 3 changes the parameter values ​​of the neural network. The parameter values ​​of the neural network can be initialized and changed when the learning server 3 sends parameter values ​​to the processing device 1 via the control server 2.

[0037] In step S503, the learning server 3 sends a control mode and control commands to the processing device 1 via the control server 2, thereby operating the processing device 1. More specifically, the learning server 3 can send control information including the control mode to the control server 2, and the control server 2 can send control commands including the control mode to the processing device 1. The processing device 1 can operate the bench mechanism 5, including the bench (holder) ST as the control target, according to the control mode. The processing device 1 can monitor this operation and save the control results. The control results are data generated during the operation of the bench mechanism 5 and may include, for example, data indicating the control deviation calculated by the subtractor 86. The control results can be sent from the processing device 1 to the learning server 3 via the control server 2.

[0038] In step S505, the learning server 3 can calculate a reward based on the control result sent from the processing device 1 according to a predetermined formula. The formula can be set such that, for example, the reward value is larger when the control deviation is small during the evaluation period. In step S505, the learning server 3 determines whether to end the learning process. If the learning server 3 ends the learning process, the process proceeds to step S507; otherwise, the process returns to step S503. Whether to end the learning process can be determined based on, for example, whether the number of times the learning process (the number of times steps S502 to S505 are executed) has reached a predetermined value. In this case, if the number of times the learning process has reached the predetermined number, the learning server 3 ends the learning process; otherwise, the learning server 3 can continue learning. If the process returns to step S503, then in step S503, the parameter values ​​of the neural network can be changed according to a predetermined algorithm to increase the reward.

[0039] In step S507, the learning server 3 determines the parameter values ​​that yield the maximum reward from the rewards calculated by repeating steps S502 to S505 as the learned parameter values, and saves them as parameter values ​​for the neural network to be learned. For example, this means that if the neural network to be learned is neural network 85a, then the parameter values ​​of neural network 85a are set.

[0040] In step S508, the learning server 3 determines whether a next control mode exists (i.e., a neural network to be learned next). If a next control mode exists, the process returns to step S501. In this case, steps S501 to S507 are executed for the next control mode.

[0041] The time spent performing calculations in a control device using a neural network depends on the size of the neural network (the number of layers and neurons). The control device 8 needs to provide command values ​​(e.g., the current command value) to the driver 7 at predetermined time intervals. If the neural network is large, the calculation of the command value may not be completed within the predetermined time. To address this, in this embodiment, each of the multiple neural networks 85a, 85b, and 85c is formed by a smaller neural network, and the neural network to be used is selected from the multiple neural networks 85a, 85b, and 85c based on selection information. If the control device 8 includes only a single neural network, then that single neural network needs to handle all control modes. Therefore, the neural network is large, and the learning process takes a long time. On the other hand, if a neural network corresponding to the control mode is selected and used from the multiple neural networks 85a, 85b, and 85c, then the size of each neural network can be small. This can shorten the computation time of each neural network and also shorten the time spent performing the learning process.

[0042] Figure 6The velocity and acceleration curves of the test bench (ST) are illustrated. The velocity curve is obtained by differentiating the position curve (time-series data of the target position), and the acceleration curve is obtained by differentiating the position curve twice. The position, velocity, and acceleration curves can be understood as drive curves or control curves.

[0043] exist Figure 5 In the example shown, the drive curve includes multiple time intervals, more specifically, intervals 501 to 508. Interval 501 is the jerk interval, during which the acceleration of the controlled target increases within a positive range until the controlled target reaches uniform acceleration. Interval 502 is the uniform acceleration interval, during which the acceleration of the controlled target remains at a constant positive value (uniform acceleration). Interval 503 is the jerk interval, during which the acceleration of the controlled target decreases within a positive range until the controlled target reaches uniform velocity. Interval 504 is the uniform velocity interval, during which the controlled target moves at a uniform velocity (i.e., the acceleration of the controlled target remains zero). Interval 505 is the jerk interval, during which the absolute value of the acceleration of the controlled target increases within a negative range until the controlled target reaches negative uniform acceleration. Interval 506 is a uniform acceleration interval, during which the acceleration of the controlled target remains at a constant negative value (uniform acceleration). Interval 507 is a rapid acceleration interval, during which the absolute value of the controlled target's acceleration decreases within a negative range until the controlled target stops. Interval 508 is a stationary interval, during which the controlled target stops. Selector 87 can select the neural network to be used for control from multiple neural networks 85a, 85b, and 85c based on the current time interval among multiple time intervals. For example, information indicating the current time interval is given to selector 87 as selection information.

[0044] In one example, a first control mode can be defined from the beginning of interval 501 to the end of interval 502, a second control mode can be defined from the beginning of interval 503 to the end of interval 505, and a third control mode can be defined from the beginning of interval 506 to the end of interval 508. In this example, neural networks 85a, 85b, and 85c can be assigned to the first, second, and third control modes, respectively. In another example, a control mode can be defined by adding an offset to a time period that includes at least one of the multiple time intervals classified according to acceleration as described above. For example, a first control mode can be defined from the beginning of interval 501 to the end of interval 502 (i.e., until 10 milliseconds later), and a second control mode can be defined from the end of interval 502 (i.e., from 10 milliseconds later) to the end of interval 508.

[0045] If the time periods when multiple control modes overlap are excluded, then learning server 3 can simultaneously perform learning of multiple neural networks corresponding to the multiple control modes.

[0046] The following will refer to Figure 7 An example is described below of the application of the aforementioned processing system PS to the scanning exposure apparatus 500. The scanning exposure apparatus 500 is a stepping scanning exposure apparatus that performs scanning exposure of the substrate 14 by means of slit light formed by a slit member. The scanning exposure apparatus 500 may include an illumination optics system 23, a component stage mechanism 12, a projection optics system 13, a substrate stage mechanism 15, a first position measurement unit 17, a second position measurement unit 18, a substrate mark measurement unit 21, a substrate transfer unit 22, a temperature controller 25, drivers RD and SD, and a control unit 24.

[0047] The control unit 24 can control the illumination optics system 23, the component mounting mechanism 12, the projection optics system 13, the substrate mounting mechanism 15, the first position measurement unit 17, the second position measurement unit 18, the substrate marking measurement unit 21, the substrate transfer unit 22, and the temperature controller 25. The control unit 24 controls the process of transferring the pattern of the component 11 onto the substrate 14. The illumination optics system 23, the component mounting mechanism 12, the projection optics system 13, the substrate mounting mechanism 15, the substrate transfer unit 22, and / or the temperature controller 25 are operational units that operate to perform the process of transferring the pattern of the component 11 onto the substrate 14. The control unit 24 may be formed, for example, by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), a general-purpose computer with a program installed, or a combination of all or some of these components. The control unit 24 can be connected to... Figure 2 Or, corresponding to the control unit 8 in the processing unit 1 shown in Figure 3. The drives RD and SD can be... Figure 2 It corresponds to the driver 7 in the processing device 1 shown in 3.

[0048] The original device stage mechanism 12 and the substrate stage mechanism 15 can form a scanning mechanism for scanning the original device 11 and the substrate 14 to transfer the pattern of the original device 11 to the substrate 14. The original device stage mechanism 12 may include an original device stage RST for holding the original device 11 and a first actuator RAC for driving the original device stage RST. The first actuator RAC is driven by a first driver RD. The substrate stage mechanism 15 may include a substrate stage WST for holding the substrate 14 and a second actuator WAC for driving the substrate stage WST. The second actuator WAC is driven by a second driver SD. An illumination optical system 23 illuminates the original device 11. The illumination optical system 23 shapes the light emitted from a light source (not shown) into, for example, a long strip or arc-shaped slit light in the X direction using a light-shielding member such as a mask, and uses this slit light to illuminate a portion of the original device 11. The original 11 and the substrate 14 are held by the original stage RST and the substrate stage WST, respectively, and are arranged in a nearly optically conjugate position (on the object plane and image plane of the projection optics system 13) via the projection optics system 13.

[0049] The projection optical system 13 has a predetermined projection magnification (e.g., 1, 1 / 2, or 1 / 4) and projects the pattern of the original 11 onto the substrate 14 using slit light. The area on the substrate 14 where the pattern of the original 11 is projected (the area illuminated by the slit light) can be referred to as the illumination area. The original stage RST and the substrate stage WST are configured to be movable in a direction (Y direction) orthogonal to the optical axis (Z direction) of the projection optical system 13. The original stage RST and the substrate stage WST are scanned relative to each other synchronously at a speed ratio corresponding to the projection magnification of the projection optical system 13. This scanning of the substrate 14 relative to the illumination area in the Y direction transfers the pattern formed on the original 11 to the impact area of ​​the substrate 14. Then, by sequentially performing such scanning exposure on multiple impact areas of the substrate 14 while moving the substrate stage WST, the exposure process of one substrate 14 is completed.

[0050] The first position measurement unit 17 includes, for example, a laser interferometer and measures the position of the original component pedestal (RST). For example, the laser interferometer uses a laser beam to illuminate a reflector (not shown) provided in the original component pedestal (RST), and detects the displacement of the original component pedestal (RST) from a reference position by the interference between the laser beam reflected by the reflector and the laser beam reflected by a reference surface. The first position measurement unit 17 can obtain the current position of the original component pedestal (RST) based on the displacement. In this example, the first position measurement unit 17 can measure the position of the original component pedestal (RST) using a position measuring device (e.g., an encoder) instead of a laser interferometer.

[0051] The second position measurement unit 18 includes, for example, a laser interferometer and measures the position of the substrate stage WST. For example, the laser interferometer uses a laser beam to illuminate a reflector (not shown) provided in the substrate stage WST, and detects the displacement of the substrate stage WST (displacement from a reference position) by the interference between the laser beam reflected by the reflector and the laser beam reflected by a reference surface. The second position measurement unit 18 can obtain the current position of the substrate stage WST based on the displacement. In this example, the second position measurement unit 18 can measure the position of the substrate stage WST using a position measuring device (e.g., an encoder) instead of a laser interferometer.

[0052] The substrate marking measurement unit 21 includes, for example, an optical system and an image sensor, and can detect the position of a mark provided on the substrate 14. The substrate transfer unit 22 supplies the substrate 14 to the substrate stage WST and retrieves the substrate 14 from the substrate stage WST. The temperature controller 25 maintains a constant temperature and humidity in the chamber (not shown) of the scanning exposure apparatus 500.

[0053] Figure 8 An exposure sequence of the scanning exposure apparatus 500 is illustrated. In step S701 (substrate loading sequence), the control unit 24 controls the substrate transfer unit 22 to load (transfer) the substrate 14 onto the substrate stage WST. More specifically, in step S702 (measurement sequence), the control unit 24 performs a measurement for the alignment of the substrate 14 and the original 11. More specifically, in step S702, the control unit 24 can control the substrate stage mechanism 15 so that the mark on the substrate 14 falls within the field of view of the substrate mark measurement unit 21, and control the substrate mark measurement unit 21 to detect the position of the mark on the substrate 14. This operation can be performed on each of the plurality of marks on the substrate 14. In step S703 (exposure sequence), the control unit 24 controls the substrate stage mechanism 15, the original 11 stage mechanism 12, the illumination optical system 23, etc., so that the pattern of the original 11 is transferred to each of the plurality of impact areas on the substrate 14. In step S704 (substrate unloading sequence), the control unit 24 controls the substrate transfer unit 22 to unload (transfer) the substrate 14 on the substrate stand WST.

[0054] An example of how the processing system PS is applied to the control of the substrate stage mechanism 15 in the scanning exposure apparatus 500 will now be described. Figure 2 The control device 8, driver 7, sensor 6, and actuator AC correspond to the control unit 24, driver SD, second position measuring unit 18, and second actuator WAC, respectively. Similar to... Figure 3Alternatively, the control device 8 illustrated in Figure 4 may include a plurality of neural networks 85a, 85b, and 85c. The control unit 24 may use a neural network selected from the plurality of neural networks 85a, 85b, and 85c to control the substrate stage WST of the substrate stage mechanism 15 according to the control mode.

[0055] Each of the multiple control modes used to select the neural network to be controlled from multiple neural networks 85a, 85b, and 85c can be defined based on multiple time intervals, as shown in the reference. Figure 6 As exemplarily described above, each of the multiple neural networks 85a, 85b, and 85c can be selected during the execution of a control mode defined based on multiple time intervals.

[0056] Multiple control modes may include a control mode for step S701, a control mode for step S702, a control mode for step S703, and a control mode for step S704. The control mode for step S701 may be referred to as a loading control mode. The control mode for step S702 may be referred to as a measurement control mode. The control mode for step S703 may be referred to as an exposure control mode. The control mode for step S704 may be referred to as an unloading control mode. For example, step S701 can be controlled by neural network 85a, step S702 by neural network 85b, and step S703 by neural network 85c.

[0057] Alternatively, control for the entire time interval or a portion thereof of a sequence can be set to a control mode and assigned to a neural network. For example, neural network 85a can be applied to intervals 506 to 508 in the measurement sequence. Neural network 85b can be applied to intervals 502 and 504 in the exposure sequence.

[0058] In the measurement sequence, it is necessary to reduce the control deviation immediately following the stop of the substrate stage WST. This is done in order to... Figure 5 During interval 508 (stop interval), the position of the mark on the substrate 14 is measured by the substrate mark measurement unit 21. As an interval using a neural network in the measurement sequence, the interval can be assigned from the beginning of interval 506 during the deceleration period to the end of interval 508. The vibration component excited in interval 505 is less than the vibration component excited in interval 506. The start of the interval using a neural network in the measurement sequence is effectively set to the start of interval 506.

[0059] On the other hand, in the exposure sequence, it is necessary to reduce the deviation when the speed of WST on the substrate stage is uniform. This is done in order to... Figure 5The substrate 14 is exposed in a uniform velocity interval (uniform velocity interval) 504. As an interval in the exposure sequence using a neural network, the interval can be assigned from the beginning of interval 502 during the acceleration period to the end of interval 504. The vibrational component excited in interval 501 is less than the vibrational component excited in interval 502. Therefore, the beginning of the interval in the exposure sequence using a neural network is effectively set to the beginning of interval 502.

[0060] As described above, by providing multiple small neural networks corresponding to multiple intervals of the desired reduction in the control deviation of the substrate stage WST, and by using the neural networks while switching them, the effect of shortening the learning time and computation time is achieved.

[0061] The following will describe an example of how the processing system PS is applied to the control of the original workpiece stand mechanism 12 in the scanning exposure apparatus 500. Figure 2 The control device 8, driver 7, sensor 6, and actuator AC correspond to the control unit 24, driver RD, first position measuring unit 17, and actuator RAC, respectively. Similar to... Figure 3 Alternatively, the control device 8 illustrated in Figure 4 may include a plurality of neural networks 85a, 85b, and 85c. The control unit 24 may use a neural network selected from the plurality of neural networks 85a, 85b, and 85c to control the component rack RST of the component rack mechanism 12 according to the control mode.

[0062] Each of the multiple control modes used to select the neural network to be used from multiple neural networks 85a, 85b, and 85c can be defined based on multiple time intervals, as shown in the reference. Figure 6 As exemplarily described. Alternatively, the multiple control modes may include a control mode for the step of loading component 11, a control mode for the step of unloading component 11, and a control mode for step S703. Alternatively, control for the entire time period of a sequence or a portion thereof may be set to a single control mode and assigned a neural network.

[0063] The following will describe an example of how the processing system PS is applied to the control of the substrate transfer unit 22 in the scanning exposure apparatus 500. Figure 2 The control device 8, driver 7, sensor 6, and platform mechanism 5 correspond to the control unit 24, driver (not shown), and substrate transfer unit 22, respectively. Similar to... Figure 3 Alternatively, the control device 8 illustrated in Figure 4 may include a plurality of neural networks 85a, 85b, and 85c. The control unit 24 may use a neural network selected from the plurality of neural networks 85a, 85b, and 85c to control the substrate transfer unit 22 according to the control mode.

[0064] By applying control device 8 to the control of substrate transfer unit 22, control deviations during the driving of substrate transfer unit 22 can be suppressed, thereby improving the reproducibility of the position of substrate 14 supplied to substrate stage WST. Throughput can also be increased by suppressing control deviations while increasing acceleration and speed.

[0065] Based on reference Figure 6 The control of multiple time intervals exemplarily described can be applied to the control of the substrate transfer unit 22. That is, each of the multiple control modes for selecting the neural network to be used for control from the multiple neural networks 85a, 85b, and 85c can be defined based on multiple time intervals. In other words, each of the multiple neural networks 85a, 85b, and 85c can be selected during the execution of the control modes defined based on multiple time intervals.

[0066] Similar to the substrate stage mechanism 15, the substrate transfer unit 22 also achieves the effect of shortening the learning time and computation time by providing multiple small neural networks corresponding to multiple intervals for which control deviation is expected to be reduced, and by using the neural networks while switching them.

[0067] The following will describe Figure 1 The processing system PS shown is applied to the temperature controller 25 in the scanning exposure apparatus 500 as an example. Figure 9 An example arrangement of the temperature controller 25 is shown. The temperature controller 25 may include a temperature control unit (control device) 26, a temperature regulator 27 that adjusts the temperature of the controlled target, and a temperature sensor 28 that measures the temperature of the controlled target. The temperature control unit 26 sends command values ​​to the temperature regulator 27 at predetermined time intervals. The temperature regulator 27 adjusts the temperature in the chamber of the scanning exposure apparatus 500 via a heater and / or cooler (not shown). The temperature in the chamber of the scanning exposure apparatus 500 is measured by the temperature sensor 28, and the measurement result is sent to the temperature control unit 26.

[0068] Figure 10 It is shown Figure 9The block diagram illustrates an example arrangement of a temperature controller 25. The temperature controller 25 may include a temperature regulator 27 that regulates the temperature in a chamber as a control target, a temperature sensor 28 that measures the temperature at a predetermined location within the chamber, and a temperature control unit 26 that provides manipulated variables to the temperature regulator 27 based on control deviations and receives outputs from the temperature sensor 28. The temperature control unit 26 may be formed, for example, by a PLD (Programmable Logic Device) such as an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), a general-purpose computer with a program installed, or a combination of all or some of these components.

[0069] Temperature control unit 26 generates a control signal for controlling temperature regulator 27. Temperature control unit 26 may include multiple neural networks 85a, 85b, and 85c, and a selector 87 that selects the neural network to be used to generate the control signal from the multiple neural networks 85a, 85b, and 85c. Based on selection information, selector 87 can select the neural network to be used to generate the control signal from the multiple neural networks 85a, 85b, and 85c. The selected neural network can be used as a compensator to generate manipulated variables based on input information (control deviation). The selection information can be generated in temperature control unit 26 or provided from another device (e.g., control unit 24). Selection information can be generated such that the neural network to be used is selected from the multiple neural networks 85a, 85b, and 85c according to the control mode selected from multiple control modes. The multiple control modes can be control modes that can be distinguished from each other. The control mode can be, for example, a target value string (time series data of the target temperature).

[0070] The temperature control unit 26 may include a subtractor 86 that calculates the difference (i.e., control deviation) between the target value (target temperature) and the output (measured temperature) of the temperature sensor 28. The selector 87 may include a demultiplexer 87a and a multiplexer 87b. The demultiplexer 87a may supply the control deviation to one of the multiple neural networks 85a, 85b, and 85c specified by the selection information. The multiplexer 87b may output the signal generated by one of the multiple neural networks 85a, 85b, and 85c specified by the selection information. Note that... Figure 10 Three neural networks, 85a, 85b, and 85c, are illustrated, but the number of neural networks is arbitrary. The number of layers and neurons in each of the multiple neural networks 85a, 85b, and 85c is arbitrary. The parameter values ​​of the multiple neural networks 85a, 85b, and 85c can be determined through reinforcement learning.

[0071] Each of the plurality of neural networks 85a, 85b, and 85c may be selected by selector 87 to be used in the execution of a corresponding control mode among the control modes used to control the control target. Alternatively, each of the plurality of neural networks 85a, 85b, and 85c may be selected by selector 87 for at least a portion of at least one of the plurality of control modes used to control the control target.

[0072] Figure 11 It is shown Figure 9 A block diagram illustrating another arrangement example of the temperature controller 25. Figure 11 The arrangement example shown includes by means of Figure 10 The arrangement shown is an example of an arrangement obtained by adding a main compensator 81 and an adder 82. The main compensator 81 can be, for example, a compensator including P-elements (proportional elements), I-elements (integral elements), and D-elements (differential elements) (e.g., a PID compensator), but is not limited thereto. The subtractor 86 calculates the difference between a target value (e.g., a target temperature) and the output temperature (measured temperature) of the temperature sensor 28, and supplies the difference to both the neural network selected by the selector 87 and the main compensator 81. The adder 82 can add a first manipulated variable generated by the main compensator 81 and a second manipulated variable generated by the neural network selected by the selector 87 to generate a control signal. This control signal can be supplied to the temperature regulator 27.

[0073] The temperature within the chamber of the scanning exposure apparatus 500 can be changed, for example, by the heat generated by the original substrate stage mechanism 12 and the substrate stage mechanism 15. When the exposure sequence of the substrate begins, current flows into the actuators of the original substrate stage mechanism 12 and the substrate stage mechanism 15, thereby generating heat in the original substrate stage mechanism 12 and the substrate stage mechanism 15. Immediately after the start of the exposure sequence, the temperature change in the space within the chamber is large, then becomes gradual as the exposure sequence of the substrate continues. That is, the temperature change is correlated with the number of substrates processed.

[0074] Therefore, the selection information can be changed according to the number of substrates being processed, thereby switching between multiple neural networks 85a, 85b, and 85c to be used for temperature control. For example, neural network 30a can control the temperature during the processing time from the first substrate to the tenth substrate, during which the temperature changes significantly. Subsequently, neural network 30a can control the temperature during the processing time from the eleventh substrate to the fiftyth substrate, and neural network 30c can control the temperature during the processing time of the fifty-first substrate and subsequent substrates.

[0075] Alternatively, the driving amount of the original stage mechanism 12 and the substrate stage mechanism 15 can be varied depending on the exposure scheme (e.g., the size of the impact area), thus changing the heat value. The selection information can be changed according to the exposure scheme, thereby switching between multiple neural networks 85a, 85b, and 85c to be used for temperature control.

[0076] The example of the manufacturing system MS being applied to the scanning exposure apparatus 500 has been explained above. However, the manufacturing system MS can be applied to another type of exposure apparatus (e.g., a stepper) or another type of photolithography apparatus such as an imprinting apparatus. In this case, the photolithography apparatus is an apparatus for forming patterns on a substrate, and the concept includes exposure apparatus, imprinting apparatus, and electron beam tracing apparatus.

[0077] The following describes a method for manufacturing articles (e.g., semiconductor IC components, liquid crystal display components, or MEMS) using the aforementioned photolithography apparatus. The article manufacturing method may include a transfer step of transferring a pattern of an original onto a substrate using a photolithography apparatus, and a processing step of processing the substrate that has undergone the transfer step, thereby obtaining an article from the processed substrate.

[0078] If the photolithography apparatus is an exposure apparatus, then the article manufacturing method may include the steps of exposing a substrate (substrate, glass substrate, etc.) coated with a photosensitive agent, developing the substrate (photosensitive agent), and processing the developed substrate in other known steps. Other known steps include etching, resist removal, dicing, bonding, and encapsulation. According to this article manufacturing method, articles of higher quality than conventional articles can be manufactured. If the photolithography apparatus is an imprinting apparatus, then the article manufacturing method may include the steps of forming a pattern made of a cured product of the imprinting material by shaping an imprinting material on the substrate using a mold, and processing the substrate using the pattern.

[0079] This application claims priority to Japanese Patent Application No. 2020-131923, filed on August 3, 2020, which is hereby incorporated herein by reference.

[0080] 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 set forth to inform the public of the scope of this invention.

[0081] Reference Symbol List

[0082] 8: Control device; 81: Main compensator; 82: Adder; 85a, 85b, 85c: Neural network; 86: Subtractor; 87: Selector.

Claims

1. A control device for generating control signals for controlling a control target, the control target being included in a photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, the control device comprising: A selector, configured to select from a plurality of neural networks a neural network to be used to generate the control signal, such that the selected neural network is switched during control of the control objective. The selector is configured to select a first neural network from the plurality of neural networks for a first sequence, and to select a second neural network from the plurality of neural networks for exposing the substrate such that the pattern of the original is transferred to the substrate in a second sequence, wherein the second neural network is different from the first neural network, and The first sequence is one of the following: a measurement sequence for aligning the substrate and the component, a loading sequence for transferring the substrate to a substrate stand, and an unloading sequence for transferring the substrate on the substrate stand to a position different from the substrate stand.

2. The control device of claim 1, wherein the selector includes a multiplexer configured to output a signal generated by a neural network selected from the plurality of neural networks according to selection information during the process of transferring the pattern of the original onto the substrate.

3. A control device for generating a control signal for controlling a control target, the control target being included in a photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, the control device comprising: A selector, configured to select from a plurality of neural networks a neural network to be used to generate the control signal, such that the selected neural network is switched during control of the control objective. The acceleration curve of the control target includes multiple time intervals, including: a rapid acceleration interval where the acceleration increases rapidly in a positive range; a uniform acceleration interval where the acceleration remains at a constant positive value; a rapid acceleration interval where the acceleration decreases rapidly in a positive range; a uniform acceleration interval where the acceleration remains at zero; a rapid acceleration interval where the absolute value of the acceleration increases rapidly in a negative range; a uniform acceleration interval where the acceleration remains at a constant negative value; and a rapid acceleration interval where the absolute value of the acceleration decreases rapidly in a negative range. The selector selects the neural network to be used from the plurality of neural networks based on the current time interval among the plurality of time intervals.

4. The control device according to claim 1 or 3, wherein the selector selects the neural network to be used based on information provided from a device different from the control device.

5. The control device according to claim 1 or 3, further comprising a subtractor, a compensator, and an adder, wherein the subtractor is configured to calculate a difference between a target value and a state signal indicating the state of the control target, and the compensator is configured to calculate a manipulated variable based on the difference. The difference is fed into the neural network to be used, and the adder adds the manipulated variable and the output from the neural network to be used to generate the control signal.

6. The control device according to claim 1 or 3, wherein the control device is configured to control the position of the control target.

7. The control device according to claim 1 or 3, wherein the control device is configured to control the temperature of the control target.

8. A photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, comprising: An operation unit configured to operate for the purpose of the processing; as well as A control unit configured to generate control signals for controlling the operating unit, which is the control target. The control unit includes A selector, configured to select from a plurality of neural networks a neural network to be used to generate the control signal, such that the selected neural network is switched during control of the operating unit, and The selector is configured to select a first neural network from the plurality of neural networks for a first sequence, and to select a second neural network from the plurality of neural networks for exposing the substrate such that the pattern of the original is transferred to the substrate in a second sequence, wherein the second neural network is different from the first neural network, and The first sequence is one of the following: a measurement sequence for aligning the substrate and the component, a loading sequence for transferring the substrate to a substrate stand, and an unloading sequence for transferring the substrate on the substrate stand to a position different from the substrate stand.

9. The lithography apparatus of claim 8, wherein the selector includes a multiplexer configured to output a signal generated by a neural network selected from the plurality of neural networks according to selection information.

10. The photolithography apparatus according to claim 8 or 9, wherein The operating unit includes a scanning mechanism configured to scan the original and the substrate to transfer the pattern of the original onto the substrate. The second sequence includes exposing the substrate while scanning the substrate and the original component through the scanning mechanism.

11. A photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, comprising: An operation unit configured to operate for the purpose of the processing; as well as A control unit configured to generate control signals for controlling the operating unit, which is the control target. The control unit includes A selector, configured to select from a plurality of neural networks a neural network to be used to generate the control signal, such that the selected neural network is switched during control of the operating unit, and The acceleration curve of the operating unit includes multiple time intervals, which include: a rapid acceleration interval where acceleration increases rapidly in a positive range; a uniform acceleration interval where acceleration remains constant; a rapid acceleration interval where acceleration decreases rapidly in a positive range; a uniform acceleration interval where acceleration remains zero; a rapid acceleration interval where the absolute value of acceleration increases rapidly in a negative range; a uniform acceleration interval where acceleration remains constant; and a rapid acceleration interval where the absolute value of acceleration decreases rapidly in a negative range. The selector selects the neural network to be used from the plurality of neural networks based on the current time interval among the plurality of time intervals.

12. The lithography apparatus of claim 8 or 11, further comprising a sensor, a subtractor, a compensator, and an adder, wherein the sensor is configured to detect the state of the operating unit, the subtractor is configured to calculate a difference between a target value and an output from the sensor, and the compensator is configured to calculate a manipulated variable based on the difference. The difference is fed into the neural network to be used, and the adder adds the manipulated variable and the output from the neural network to be used to generate the control signal.

13. A method for manufacturing an article, comprising: The transfer step of transferring the pattern of the original onto the substrate using the photolithography apparatus as defined in claim 8 or 11; as well as The processing steps for the substrate that has already undergone the transfer step. The article is obtained from the substrate that has already undergone the aforementioned processing steps.

14. A control method for controlling a control target, said control target being included in a photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, said method comprising: For each of the multiple sequences, a neural network is selected from the multiple neural networks to be used to generate control signals for controlling the control objective, such that the selected neural network is switched during the control of the control objective; and The control objective is controlled using the control signals generated by the selected neural network. In the selection process, a first neural network is selected from the plurality of neural networks for a first sequence, and a second neural network is selected from the plurality of neural networks for exposing the substrate such that the pattern of the original is transferred to the substrate in a second sequence. The second neural network is different from the first neural network. The first sequence is one of the following: a measurement sequence for aligning the substrate and the component, a loading sequence for transferring the substrate to a substrate stand, and an unloading sequence for transferring the substrate on the substrate stand to a position different from the substrate stand.

15. A control method for controlling a control target, said control target being included in a photolithography apparatus for performing a process of transferring a pattern of an original onto a substrate, said method comprising: For each of the multiple sequences, a neural network is selected from the multiple neural networks to be used to generate control signals for controlling the control objective, such that the selected neural network is switched during the control of the control objective; and The control objective is controlled using the control signals generated by the selected neural network. The acceleration curve of the control target includes multiple time intervals, including: a rapid acceleration interval where the acceleration increases rapidly in a positive range; a uniform acceleration interval where the acceleration remains at a constant positive value; a rapid acceleration interval where the acceleration decreases rapidly in a positive range; a uniform acceleration interval where the acceleration remains at zero; a rapid acceleration interval where the absolute value of the acceleration increases rapidly in a negative range; a uniform acceleration interval where the acceleration remains at a constant negative value; and a rapid acceleration interval where the absolute value of the acceleration decreases rapidly in a negative range. In the selection process, the neural network to be used is selected from the plurality of neural networks based on the current time interval among the plurality of time intervals.

Images