WO2023104658A1

WO2023104658A1 - Projection exposure apparatus comprising a correction determination module

Info

Publication number: WO2023104658A1
Application number: PCT/EP2022/084227
Authority: WO
Inventors: Joachim Hartjes; Alexander Wolf; Toralf Gruner
Original assignee: Carl Zeiss Smt Gmbh
Priority date: 2021-12-10
Filing date: 2022-12-02
Publication date: 2023-06-15
Also published as: DE102021214139A1

Abstract

The invention relates to a microlithography projection exposure apparatus (10) for imaging mask structures (21) on a substrate (30), comprising a correction determination module (62) configured to generate an execution variant (64a) of a correction command (64) which specifies at least one corrective measure (KM1-KM12) in respect of the projection exposure apparatus. The correction determination module (62) is configured to use an optimisation algorithm (66) to generate the execution variant of the correction command from measured values (MW1-MW12) of parameters (P1-P12) that influence an imaging quality of the projection exposure apparatus, wherein the measured values are determined by means of at least two different measuring sensors (S1-S12). The optimisation algorithm (66) is configured as a machine learning method which is designed to be trained on the basis of a plurality of training datasets (70) which each comprise training values (AW1-AW12) of the parameters (P1-P12) that influence the imaging quality and information about a training variant (64l) of the correction command (64) associated with the respective training values.

Description

Projection exposure system with a correction determination module

The present application claims the priority of German patent application 10 2021 214 139.0 of December 10, 2021. The entire disclosure of this patent application is incorporated herein by reference.

Background of the Invention

The invention relates to a projection exposure system for microlithography for imaging mask structures onto a substrate with a correction determination module and a method for operating such a projection exposure system. A projection exposure system for microlithography is used in the production of semiconductor components to produce structures on a substrate in the form of a semiconductor wafer. For this purpose, the projection exposure system includes a projection lens having a plurality of optical elements for imaging the mask structures onto the wafer during an exposure process.

To ensure the most precise possible imaging of the mask structures on the wafer, a projection objective with the lowest possible wavefront aberrations is required. Projection lenses are therefore typically equipped with manipulators that make it possible to minimize wavefront errors by corrective measures in the form of changes in the state of individual optical elements of the projection lens. Examples of such state changes include: a change in position in one or more of the six degrees of freedom of the rigid body of the relevant optical element, an application of heat and/or cold to the optical element and a deformation of the optical element. To this end, the aberration characteristic of the projection lens is usually measured regularly and changes are made if necessary in the aberration characteristic between each measurement is determined by simulation. For example, lens heating effects can be taken into account in the calculation. The calculation of the manipulator changes to be carried out to correct the aberration characteristic takes place in a correction determination module using an optimization algorithm that generates a travel distance, which is also referred to as a “manipulator change model”. In this case, a correction command specifying the manipulator changes to be carried out is generated on the basis of measured values from a wavefront sensor. Such optimization algorithms are described, for example, in WO 2010/034674 A1.

However, correction commands which are generated by means of conventional optimization algorithms are often very imprecise and often no longer completely meet increasing demands on the correction accuracy of the imaging behavior of the projection exposure system.

Underlying Task

It is an object of the invention to provide a projection exposure system and a method of the type mentioned at the outset, with which the aforementioned problems are solved and, in particular, an improved correction accuracy of the imaging behavior of the projection exposure system can be achieved.

Solution according to the invention

The aforementioned object can be achieved according to the invention, for example, with a projection exposure system for microlithography for imaging mask structures on a substrate with a correction determination module, which is configured to an embodiment variant of a correction command, which min at least one corrective measure with regard to the projection exposure system. The correction determination module is configured to generate the execution variant of the correction command using an optimization algorithm from measured values of parameters influencing an imaging quality of the projection exposure system, the measured values being determined using at least two different measuring sensors. The optimization algorithm is configured as a machine learning method, which is set up to be learned using a large number of learning data sets, each of which includes learning values of the parameters influencing the imaging quality and an indication of a learning variant of the correction command assigned to the respective learning values.

According to different embodiments, the measured values are determined using at least three, at least four, at least five or at least ten different measuring sensors. Optical elements in the aforementioned sense include mirrors and/or lenses of the illumination system and of the projection objective of a projection exposure system. Furthermore, optical elements are also to be understood, for example, as individual facet elements of a facet mirror of the lighting system. The correction command is used to correct an imaging property of the projection lens. Since the optimization algorithm is configured as a machine learning method, it is based on artificial intelligence.

The at least one corrective measure of the determined correction command can be, for example, at least one travel specification for one or more manipulators of optical elements of the projection exposure system. In the event that multiple manipulators are provided, the travel command can include multiple travel specifications, specifically one travel specification for some or all of the manipulators. A manipulator can be configured, for example, to carry out a rigid-body movement on one of the optical elements or to apply heat, cold, pressure, light of a specific wavelength or currents to the optical element locally or over a large area. So can one Be configured manipulator for applying a pressure distribution to or for generating a temperature distribution in such an optical element. A manipulator for generating a temperature distribution can be configured in particular as a so-called preheater, with which areas of the optical element which are heated less by the exposure radiation than other areas when a reticle is exposed, are heated to set a homogeneous temperature profile.

A preheater can also be used to reduce a temporal variation in the temperature of the optical element, which occurs when using reticles of different brightness in the course of the exposure process over time. This can be done in that the preheater fills a reduced irradiation intensity, which occurs when a comparatively dark reticle is exposed, to a reference value, in particular a maximum value for the irradiation intensity.

According to one variant, this means that the preheater irradiates the optical element with radiation of suitable intensity in such a way that the heating power, which is generated in the optical element by the exposure radiation coming from the reticle and by the radiation of the preheater, is as high as the heating power an exposure radiation having the reference value. According to a further variant, “filling up” the irradiation intensity to a reference value means that the preheater generates the difference in heating power between the exposure radiation set to the reference value and the exposure radiation coming from the reticle, for example by means of heating elements embedded in the optical element. The temperature variation in the optical element caused by the use of different reticles over the course of the exposure process over time is minimized by filling up the irradiation intensity. The influence of inhomogeneities regularly contained in the material of the optical element, which would otherwise bring about a temperature-induced temporal variation in the optical image errors, is fixed by minimizing the temporal variation in the temperature of the optical element by means of the preheater. In this way, a more precise compensation of the optical image errors is made possible.

The solution according to the invention is based on the knowledge that measured values that are based solely on the measurement of a wavefront sensor are often not suitable for adequately describing an error state of the relevant projection exposure system. When a correction command is determined solely on the basis of the measured values of a measurement sensor, typically a wavefront sensor, it is often not possible to identify the status parameters on which the error status is based with sufficient accuracy. As a result, the corrective measures specified by the corrective command often do not correct the state parameter deviations that are actually responsible for the error state with sufficient precision. As described below with reference to FIG. 4, it is possible, for example, that thermally induced image errors and image errors caused by rigid-body movements of optical elements of the projection lens produce essentially the same error distributions in the Zernike coefficients Z2 to Z7 measured by means of a wavefront sensor. The use according to the invention of at least two different measuring sensors, such as a wavefront sensor and at least one position sensor on one or more optical elements, now makes it possible to more precisely identify the status parameters on which the error status is based.

Furthermore, the solution according to the invention with the machine learning method provides a concept with which the measured values determined by means of at least two different measuring sensors can be flexibly and efficiently converted into a correction command can be implemented. The conventional approach, in which a target function based on a sensitivity matrix is minimized, proves to be not sufficiently practicable due to the difficult to model interactions between the measured values determined with the various measuring sensors. The configuration of the optimization algorithm according to the invention as a machine learning method, on the other hand, enables a targeted and efficient conversion of the measured values into a correction command without having to do a great deal of modeling work for the optical system.

According to one specific embodiment, the projection exposure system also includes a plurality of optical elements for illuminating and imaging the mask structures onto the substrate, with the corrective measure relating to at least one of the optical elements. According to one embodiment variant, the projection exposure system includes at least one manipulator, which is configured to change an optical effect of at least one of the optical elements by manipulating a property of the optical element along a travel, and the correction command includes a travel specification for the manipulator. In this case, the travel specification can order a rigid-body movement of the relevant optical element to be carried out. Alternatively, the travel specification can also define a preheating power to be provided on an optical element.

According to a further embodiment, the projection exposure system includes a reticle displacement stage for displacing a reticle having the mask structures to be imaged and a substrate displacement stage for displacing the substrate during an exposure process, and the correction command includes an instruction for modifying control signals to the reticle displacement stage and/or the substrate displacement stage. The control signals are used to control the reticle transfer stage and the substrate transfer stage during the exposure process. The instruction for modifying the control signals thus specifies an actuation of the reticle displacement stage and/or the substrate displacement stage that deviates from the standard actuation. This deviating actuation can include a modification of the focus position or decentering, a modification of the tilt, speed or acceleration of at least one of the transfer platforms or a higher spatial derivation of the movement profile of the transfer platforms over time.

According to a further embodiment, the correction command includes repair instructions to be carried out manually in the case in which a correction that can be carried out automatically is not considered appropriate. Such a correction that can be carried out automatically comprises, for example, a correction that can be carried out by means of the at least one manipulator or a correction of the control signals of the reticle and/or the substrate shifting stage. A correction that can be carried out automatically is not to be regarded as appropriate if the limits of the correction that can be carried out automatically are exceeded, or if the measured values leave a normal range and dangerous trends in particular can be identified.

The repair instruction to be carried out manually is to be understood as an instruction that a suitable person, such as a service technician, has to carry out on the projection exposure system, in particular on the projection lens, in order to improve the imaging behavior of the projection exposure system. For example, such a repair instruction can order the replacement of one of the optical elements of the projection exposure system. For example, a mirror of the illumination system or of the projection objective or also other optical elements, such as an illumination CGH that defines the angular distribution of the mask illumination, can be determined for replacement. The repair instruction can support the on-site service engineer in troubleshooting, or serve as a suggestion for preventive maintenance directed to a remote machine monitor.

A repair instruction to be carried out manually can also include an early announcement of a repair that is becoming necessary. For example, service campaigns based on trends can be requested in good time, so that too a longer logistical chain with the delivery of spare parts is minimally invasive for the end customer. This increases the availability of the projection exposure system and thus the utility value for the semiconductor manufacturer.

According to a further embodiment, the information about the learning variant of the correction command in at least one of the learning data records includes the relevant learning variant itself. This means that the relevant learning data record includes the learning values and the learning correction command assigned to the learning values. The teach-in correction command can be determined in particular by means of a simulation method.

According to a further specific embodiment, the information about the training variant of the correction command includes feedback on the suitability of a training variant of the correction command determined by the optimization algorithm on the basis of the training values. This means that in this case the training values of one of the training data sets are first transmitted to the optimization algorithm, the optimization algorithm uses these training values to determine a training correction command and reports this back. The optimization algorithm is then given feedback as to whether the correction command determined is suitable or not suitable for error correction of the simulated system state. This acknowledgment represents the aforementioned information and is therefore part of the relevant teach-in data record in accordance with the above definition. In the case in which the feedback is negative, the optimization algorithm can, if necessary, determine an improved teach-in correction command, for which purpose feedback is in turn provided. Alternatively, a suitable teach-in correction command can then also be specified for the optimization algorithm.

According to a further embodiment, the measurement sensors include a wavefront sensor and a temperature sensor. In this case, the temperature sensor can be configured, for example, as an infrared camera or as a thermistor. According to a further embodiment, the measuring sensors comprise at least two elements from the following group: a wavefront sensor, a temperature sensor, a pressure sensor, at least one intensity sensor arranged in the beam path of the projection exposure system, a sensor for determining a position and/or an orientation of one of the optical elements and a Sensor for determining the gas composition.

The sensor for determining the position of the optical element of the projection exposure system can be configured as an encoder or as an interferometric distance measuring device. The optical element whose position is determined can be a mirror that fills the entire cross section of the exposure beam path or a corresponding lens of the illumination optics or of the projection objective. Furthermore, the optical element can be, for example, a single facet of a facet mirror in the illumination optics.

According to one embodiment variant, the at least one intensity sensor includes one or more elements from the following group: an image position sensor, a uniformity sensor, a scattered light sensor and a sensor for measuring a reflectivity of an optical element of the projection exposure system.

The image position sensor is configured to measure a position of a projected aerial image of a reticle mark in the image plane of the projection lens.

The projected aerial image can be, for example, a line structure with a line width that is of the order of magnitude of the wavelength of the exposure radiation.

The uniformity sensor is configured to measure the uniformity of the intensity of the exposure radiation in a plane of the projection exposure system, eg the mask plane or the image plane. The scattered light sensor can be based on the principle of the so-called Kirk test. In this test, a test structure with an isolated dark area, which is arranged in a much larger bright field, is placed in the object plane of the projection exposure system. The scattered light sensor scans the area of the test structure in the image plane and calculates the scattered light component from the intensity measured in the dark area of the image.

According to a further embodiment, the optimization algorithm is based on an artificial neural network with at least five layers, in particular with at least ten layers or with at least 15 layers.

According to a further embodiment, the training data records include a development of the training values over time, and the optimization algorithm is configured to take into account a development of the measured values over time when generating the correction command.

According to a further embodiment, the projection exposure system also includes a simulation module which is configured to determine the training data sets by simulating various system states of the projection exposure system.

According to one embodiment variant, the simulation module is configured to simulate system states which are caused by different points in time in the exposure process of the projection exposure system, by different error events in relation to the projection exposure system, by different values for the parameters influencing an imaging quality of the projection exposure system and/or by different configurations of the parameters to be imaged Mask structures are defined.

The aforementioned object can also be achieved, for example, with a method for operating a projection exposure system for microlithography, which is configured for imaging mask structures on a substrate. The method according to the invention comprises the steps: Training an optimization algorithm configured as a machine learning method using a large number of training data sets, each of which includes training values of parameters influencing an imaging quality of the projection exposure system and information on a training variant of a correction command assigned to the respective training values, the Correction command specifies at least one corrective measure with regard to the projection exposure system, determining measured values of the parameters influencing the imaging quality using at least two different measuring sensors, and generating an execution variant of the correction command from the measured values using the trained optimization algorithm. In particular, the method includes executing the generated execution variant of the correction command.

According to one embodiment, a learning check takes place after the optimization algorithm has been learned. According to one embodiment variant, when the optimization algorithm is trained, at least one of the training data sets is retained and during the learning control, a control variant of the correction command is generated from the training values of the retained training data set using the trained optimization algorithm and compared with the training variant of the correction command. In the case of extensive agreement, the learning process is regarded as successful and the learned optimization algorithm is released for use in the exposure mode of the projection exposure system. In the event that there is insufficient agreement, further training data sets are generated by the simulation module and the training process is continued with these.

According to one embodiment, the training data sets are determined by simulating different system states of the projection exposure system. According to one embodiment variant, the different system states include different lighting settings. Under an illumination setting, a specific angle-resolved intensity distribution of the mask structures is in operation of the projection exposure system. Examples of lighting settings include annular lighting, quadrupole lighting, or other more complex lighting configurations. Furthermore, the various system states can include different states in the maintenance cycle, for example a state of the projection exposure system after a maintenance operation in which a mirror was re-attached to a holding device by means of an adhesive. In the period immediately following the maintenance procedure, adhesive drift may occur and affect the optical properties of the mirror. The influence of the adhesive drift can be taken into account when simulating the system state.

The features specified with regard to the above-mentioned embodiments, exemplary embodiments or embodiment variants, etc. of the projection exposure system according to the invention can be correspondingly transferred to the method according to the invention and vice versa. These and other features of the embodiments according to the invention are explained in the description of the figures and the claims. The individual features can be realized either separately or in combination as embodiments of the invention. Furthermore, they can describe advantageous embodiments which are independently protectable and whose protection may only be claimed during or after the application is pending.

Brief description of the drawings

The above and other advantageous features of the invention are illustrated in the following detailed description of exemplary embodiments according to the invention with reference to the attached schematic drawings. It shows:

1 shows an embodiment according to the invention of a projection exposure system for microlithography with a correction determination module which is used for this purpose is configured to generate a correction command using an optimization algorithm,

2 shows a simulation module for determining training data sets for training the optimization algorithm, a first embodiment of a mode of operation for training the optimization algorithm and the mode of operation of the correction determination module in an exposure mode of the projection exposure system,

3 shows a further embodiment of the mode of operation when learning the optimization algorithm, and

4 Error distributions of Zernike coefficients for exemplary thermally caused image errors and for exemplary image errors caused by rigid-body movements of optical elements of the projection exposure system.

Detailed description of exemplary embodiments according to the invention

In the exemplary embodiments or embodiments or design variants described below, elements that are functionally or structurally similar to one another are provided with the same or similar reference symbols as far as possible. Therefore, for an understanding of the features of each element of a particular embodiment, reference should be made to the description of other embodiments or the general description of the invention.

A Cartesian xyz coordinate system is shown in FIG. 1 for explanation. The y-direction runs perpendicularly into the plane of the drawing. The x-direction is horizontal and the z-direction is vertical. 1 shows an embodiment of a projection exposure system 10 for microlithography according to the invention. The description of the basic structure of the projection exposure system 10 and its components should not be understood as limiting here. The projection exposure apparatus 10 described here is an embodiment for EUV lithography. Due to this operating wavelength, all optical elements are designed as mirrors. However, the invention is not limited to projection exposure systems in the EUV wavelength range. Further embodiments according to the invention are designed, for example, for operating wavelengths in the UV range, such as 365 nm, 248 nm or 193 nm. In this case at least some of the optical elements are configured as conventional transmission lenses.

An illumination system 12 of the projection exposure system 10 includes, in addition to a radiation source 14, illumination optics 16 for illuminating an object field in an object plane 18. The object plane 18 can also be referred to as a mask plane. In this case, a reticle 20 arranged in the object field with mask structures 21 arranged thereon is exposed. The reticle 20 is held by a reticle holder 22 . The reticle holder 22 can be displaced in particular in a scanning direction 25 via a reticle displacement stage 24 . The scanning direction 25 runs along the x-direction in FIG. 1 . The z-direction runs perpendicular to the object plane 18.

The projection exposure system 10 also includes a projection lens 26.

The projection lens 26 is used to image the object field in an image field in an image plane 28. The image plane 28, also referred to as the wafer plane, runs parallel to the object plane 18. Alternatively, an angle different from 0° between the object plane 18 and the image plane 28 is also possible.

The mask structures 21 are imaged on the reticle 20 on a light-sensitive layer of a substrate arranged in the region of the image field in the image plane 28 in the form of a wafer 30 . The wafer 30 is held by a wafer holder 32 . The wafer holder 32 is by means of a substrate transfer stage or a wafer displacement platform 34 along the x-direction, in particular in a wafer scanning direction 36 opposite to the scanning direction 25. The displacement of the reticle 20 on the one hand via the reticle displacement platform 24 and on the other hand the wafer 30 via the wafer displacement platform 34 can be synchronized with one another.

During an exposure process, the reticle displacement stage 24 and the wafer displacement stage 34 are controlled by a central exposure control unit 58 with corresponding control signals C1 and C2. The control signals C1 and C2 specify the respective speed of the traversing platforms 24 and 34 in the timing of the exposure process.

In the present exemplary embodiment, the radiation source 14 is an EUV radiation source. The radiation source 14 emits exposure radiation 38 in the form of EUV radiation, which is also referred to below as useful radiation. In particular, the useful radiation has a wavelength in the range between 5 nm and 30 nm, in particular a wavelength of approximately 13.5 nm or approximately 6.8 nm. The radiation source 14 can be a plasma source, for example an LPP (laser Produced Plasma (laser generated plasma) source or a DPP (Gas Discharged Produced Plasma) source. It can also be a synchrotron-based radiation source. The radiation source 14 can also be a free-electron laser (free-electron laser, FEL).

The exposure radiation 38 emanating from the radiation source 14 is bundled by a collector 40 . The collector 40 can be a collector with one, as shown in FIG. 1, or with several ellipsoidal and/or hyperboloidal reflection surfaces. The at least one reflection surface of the collector 40 can be used in grazing incidence (Grazing Incidence, Gl), ie with angles of incidence greater than 45°, or in normal incidence (Normal Incidence, NI), ie with Angles of incidence smaller than 45°, as shown in FIG. 1, are exposed to the exposure radiation 38. The collector 40 can be structured and/or coated on the one hand to optimize its reflection for the useful radiation and on the other hand to suppress stray light.

After the collector 40, the exposure radiation 38 propagates through an intermediate focus in an intermediate focal plane 42. The intermediate focal plane 42 can represent a separation between a radiation source module, comprising the radiation source 14 and the collector 40, and the illumination optics 16. The course of the exposure radiation 38 through the illumination optics 16 and the projection lens 26 is referred to as the useful beam path 44 .

The illumination optics 16 includes a deflection mirror 46 and, downstream of this in the beam path, a first facet mirror 48. The deflection mirror 46 can be a planar deflection mirror or alternatively a mirror with an effect that influences the bundle beyond the pure deflection effect. Alternatively or additionally, the deflection mirror 46 can be designed as a spectral filter, which separates a useful light wavelength of the exposure radiation 38 from stray light of a different wavelength. If the first facet mirror 48 is arranged in a plane of the illumination optics 16 which is optically conjugate to the object plane 18 as the field plane, it is also referred to as a field facet mirror. The first facet mirror 48 includes a multiplicity of individual first facets 50, which are also referred to below as field facets. Some of these facets 50 are shown in FIG. 1 only by way of example.

The first facets 50 can be embodied as macroscopic facets, in particular as rectangular facets or as facets with an arcuate or part-circular edge contour. The first facets 50 can be embodied as planar facets or alternatively as convexly or concavely curved facets. Each of the facets 50 is mounted in an individually adjustable manner by means of a respective actuator. In particular, a tilting of the respective facet 50 about two mutually orthogonal tilting axes is possible.

As is known, for example, from DE 10 2008 009 600 A1, the first facets 50 themselves can each also be composed of a multiplicity of individual mirrors, in particular a multiplicity of micromirrors. The first facet mirror 48 can be embodied in particular as a microelectromechanical system (MEMS system). Reference is made to DE 10 2008 009 600 A1 for details.

A second facet mirror 52 is arranged after the first facet mirror 48 in the useful beam path 44 of the illumination optics 16 . If the second facet mirror 52 is arranged in a pupil plane of the illumination optics 16, it is also referred to as a pupil facet mirror. The second facet mirror 52 can also be arranged at a distance from a pupil plane of the illumination optics 16 . In this case, the combination of the first facet mirror 48 and the second facet mirror 52 is also referred to as a specular reflector. Specular reflectors are known from US2006/0132747 A1, EP 1 614 008 B1 and US Pat. No. 6,573,978.

The second facet mirror 52 includes a plurality of second facets 54. In the case of a pupil facet mirror, the second facets 54 are also referred to as pupil facets.

The second facets 54 can also be macroscopic facets, which can have round, rectangular or hexagonal borders, for example, or alternatively facets composed of micromirrors. In this regard, reference is also made to DE 10 2008 009 600 A1. The second facets 54 can have plane, or alternatively convexly or concavely curved, reflection surfaces. Each of the facets 54 is mounted in an individually adjustable manner by means of a respective actuator. In particular, a tilting of the respective facet 54 about two mutually orthogonal tilting axes is possible. The illumination optics 16 thus forms a double-faceted system. This basic principle is also known as the Fly's Eye Integrator.

It can be advantageous not to arrange the second facet mirror 52 exactly in a plane which is optically conjugate to a pupil plane of the projection objective 26 .

The individual first facets 50 are imaged in the object field with the aid of the second facet mirror 52 . The second facet mirror 52 is the last beam-forming mirror or actually the last mirror for the exposure radiation 38 in the useful beam path 44 in front of the object plane 18.

In a further embodiment of the illumination optics 16, not shown, transmission optics can be arranged in the useful beam path 44 between the second facet mirror 52 and the object plane 18, which contributes in particular to the imaging of the first facets 50 in the object field. The transmission optics can have exactly one mirror, but alternatively also have two or more mirrors, which are arranged one behind the other in the useful beam path of the illumination optics 16 . The transmission optics can in particular include one or two mirrors for perpendicular incidence (NI mirror, normal incidence mirror) and/or one or two mirrors for grazing incidence (GI mirror, gracing incidence mirror).

The illumination optics 16, which begins after the collector 40, comprises exactly three mirrors in the embodiment shown in Fig. 1, namely the deflection mirror 46, the first facet mirror 48 designed as a field facet mirror and the second facet mirror 52 designed as a pupil facet mirror.

In a further embodiment of the illumination optics 16, the deflection mirror 46 can also be omitted, so that the illumination optics 16 then has exactly two mirrors may have, namely the first facet mirror 48 and the second facet mirror 52.

The imaging of the first facets 50 by means of the second facets 54 or with the second facets 54 and transmission optics in the object plane 18 is generally only an approximate imaging.

The projection lens 26 includes a plurality of optical elements in the form of mirrors Ei, which are numbered according to their arrangement in the useful beam path 44 of the projection exposure system 10 .

In the example shown in FIG. 1, the projection lens 26 includes four mirrors E1 to E4. Alternatives with six, eight, ten, twelve or another number of mirrors are also possible. The projection objective 26 has an image-side numerical aperture which is greater than 0.5 and which can also be greater than 0.6 and which can be 0.7 or 0.75, for example.

Reflective surfaces of the mirrors Ei can be designed as free-form surfaces without an axis of rotational symmetry. Alternatively, the reflection surfaces of at least some of the mirrors Ei can be designed as aspherical surfaces with exactly one axis of rotational symmetry of the shape of the reflection surfaces. Just like the mirrors of the illumination optics 16, the mirrors Ei can have highly reflective coatings for the exposure radiation 38. These coatings can be designed as multilayer coatings, in particular with alternating layers of molybdenum and silicon.

The projection lens 26 has a large object-image offset in the y-direction between a y-coordinate of a center of the object field and a y-coordinate of the center of the image field. This object-image offset in the y-direction can be approximately as large as a z-distance between the object plane 18 and the image plane 28. The projection lens 26 can in particular be anamorphic. In particular, it has different image scales βx, βy in the x and y directions. The two image scales βx, βy of the projection objective 26 are preferably at (βx, βy)=(+/−0.25, +/-0.125). A positive image scale ß means an image without image reversal. A negative sign for the imaging scale ß means imaging with image reversal.

The projection objective 26 thus leads to a reduction in the ratio 4:1 in the y-direction, that is to say in the direction perpendicular to the scanning direction 25 . The projection objective 26 leads to a reduction of 8:1 in the x-direction, ie in the scanning direction 25 . Other imaging scales are also possible. Image scales with the same sign and absolutely the same in the x and y directions, for example with absolute values of 0.125 or 0.25, are also possible.

The number of intermediate image planes in the x and y directions in the useful beam path 44 between the object field 18 and the image field 28 can be the same or, depending on the design of the projection lens 26, can be different. Examples of projection lenses with different numbers of such intermediate images in the x and y direction are known from US 2018/0074303 A1.

In each case one of the second facets 54 embodied as pupil facets is assigned to exactly one of the first facets 48 embodied as field facets to form an illumination channel for illuminating the object field. In this way, in particular, lighting can result according to Köhler's principle. The far field is broken down into a large number of object fields with the aid of the first facets 50 . The first facets 50 generate a plurality of images of the intermediate focus on the second facets 54 assigned to them.

The first facets 50 are each imaged onto the reticle 20 by an associated second facet 54 in a superimposed manner for illuminating the object field. In particular, the illumination of the object field is as homogeneous as possible. It preferably has a uniformity error of less than 2%. The Field uniformity can be achieved by superimposing different illumination channels.

The illumination of the entrance pupil of the projection objective 26 can be geometrically defined by an arrangement of the second facets 54 . The intensity distribution in the entrance pupil of the projection lens 26 can be set by selecting the illumination channels, in particular the subset of the pupil facets that guide light. This intensity distribution is also referred to as an illumination setting.

A likewise preferred pupil uniformity in the area of defined illuminated sections of an illumination pupil of the illumination optics 16 can be achieved by redistributing the illumination channels.

The arrangement of the facets 50 and 54 of the facet mirrors 48 and 52 that is desired during an exposure process is set by means of the above-mentioned actuators assigned to the respective facets 50 and 54 . For this purpose, respective control signals C3 and C4 are transmitted to the facet mirrors 48 and 52 .

Further aspects and details of the illumination of the object field and, in particular, the entrance pupil of the projection lens 26 are described below.

The projection objective 26 can in particular have a homocentric entrance pupil. This can be accessible. It can also be inaccessible.

The entrance pupil of the projection objective 26 cannot regularly be illuminated exactly with the second facet mirror 52 . When the projection lens 26 is imaged, which images the center of the second facet mirror 52 telecentrically onto the wafer 30, the aperture rays often do not intersect at a single point. However, an area can be found in which the distance between the aperture rays, which is determined in pairs, is minimal. This area represents the entrance pupil or a surface conjugate to it in position space. In particular, this surface shows a finite curvature.

It may be that the projection objective 26 has different positions of the entrance pupil for the tangential and for the sagittal beam path. In this case, an imaging element, in particular an optical component of the transmission optics, should be provided between the second facet mirror 52 and the reticle 20 . With the help of this optical element, the different positions of the tangential entrance pupil and the sagittal entrance pupil can be taken into account.

In the arrangement of the components of the illumination optics 26 shown in FIG. 1, the second facet mirror 52 is arranged in a surface conjugate to the entrance pupil of the projection objective 26 . The first facet mirror 48 is arranged tilted to the object plane 18 . The first facet mirror 48 is also arranged tilted relative to an arrangement plane that is defined by the deflection mirror 46 . The first facet mirror 48 is also arranged tilted relative to an arrangement plane that is defined by the second facet mirror 52 .

The mirrors E1 to E4 of the projection lens 26 are each movably mounted. For this purpose, the mirrors E1 to E4 are assigned respective manipulators M1 to M4 for performing rigid-body movements. According to one embodiment, each of the four mirrors E1 to E4 of the projection objective 26 can be manipulated in all six rigid body degrees of freedom (x, y and z displacement and tilting about the x, y and z axis), in which case the The number of rigid-body degrees of freedom that can be controlled by means of the manipulators M1 to M4 is twenty-four.

In FIG. 1 four such rigid body degrees of freedom are illustrated by way of example using the manipulators M1 to M4 designed as movement manipulators. The manipulators M1, M2 and M3 each allow a shift Exercise of the associated optical elements E1, E2 and E3 in the x-direction and thus essentially parallel to the plane in which the respective reflecting surface of the mirror lies. The manipulator M4 is configured to tilt the mirror M4 by rotating it about a tilting axis 56 arranged parallel to the y-axis. This changes the angle of the reflecting surface of E4 with respect to the incident radiation. Further degrees of freedom for the manipulators M1 to M4 are conceivable. For example, a displacement of a relevant optical element transverse to its optical surface and/or rotation about a reference axis perpendicular to the reflecting surface can be provided.

Generally speaking, each of the manipulators M1 to M4 shown here is intended to bring about a displacement of the associated optical element E1 to E4 by performing a rigid-body movement along a predetermined adjustment path. Such an adjustment path can, for example, combine translations in different directions, tilting and/or rotations in any way.

Alternatively or additionally, manipulators can also be provided, which are configured to undertake a different type of change in a state variable of the associated optical element by appropriate actuation of the manipulator. In this regard, an actuation can take place, for example, by subjecting the optical element to a specific temperature distribution or a specific force distribution. In this case, the travel can be given by changing the temperature distribution on the optical element or by applying a local stress to an optical element designed as a deformable lens or as a deformable mirror.

The mirror E3 is configured, for example, with such a manipulator M5 for applying specific temperature distributions. The manipulator M5 is configured as a heating device, which includes a plurality of electric heating elements 60 for respectively providing heating energy. In other words, the manipulator M5 serves to provide a preheating power at the mirror E3. The manipulator M5 can therefore also be called a preheater. The heating elements 60 are arranged below a reflective coating 61 in a substrate of the mirror E3 and are configured to generate heat from electrical energy. Electrical lines are also routed through the mirror substrate to each heating element 60 for the purpose of power supply. In this case, the manipulator M5 can then be used to homogenize a local temperature distribution on the mirror E3 and/or to reduce a temporal variation in the temperature of the mirror E3, which can occur, for example, when using reticle 20 of different brightness.

The projection exposure system 10 also includes a large number of measuring sensors S1 to S12, which are configured to generate measured values MW1 to MW12 of parameters P1 to P12 influencing an imaging quality of the projection exposure system 10 and to transmit these to a correction determination module 62. A first of these measuring sensors is a wavefront sensor S1 integrated into the wafer displacement stage 34, which is configured to determine measured values MW1 of first parameters P1 influencing the imaging quality in the form of aberration parameters, such as Zernike coefficients, of the projection lens 26.

Another of the measurement sensors is an image position sensor S2, which is also integrated into the wafer displacement platform 34, for measuring the parameter P2 in the form of an image position. This is configured to measure a position of an aerial image of a marking on the reticle 20 projected by the projection objective 26 in the image plane 28 and to transmit it to the correction determination module 62 as measured value MW2. The projected aerial image can, for example, have a line width which is typically of the order of the wavelength of the exposure radiation 38 . Another of the measurement sensors is a uniformity sensor S3 that is also integrated into the wafer displacement platform 34 . This is configured to measure the parameter P3 in the form of a uniformity of the exposure radiation 38 in the image plane 28 and to transmit it to the correction determination module 62 as measured value MW3. Further uniformity sensors for measuring the uniformity of the exposure radiation 38 in a further plane, such as the object plane 18, can also be provided. A further uniformity sensor S12 is also integrated in the reticle displacement stage 24 . This sensor is used to measure the parameter P12 in the form of a homogeneity of the illumination of the object field in the object plane 19.

Another of the measurement sensors is a scattered light sensor S4 that is also integrated into the wafer displacement platform 34 . This is configured to measure the parameter P4 in the form of a scattered light component of the exposure radiation 38 in the image plane 28 and to transmit it to the correction determination module 62 as a measured value MW4. The measurement using the scattered light sensor S4 can be based on the principle of the so-called Kirk test. In this test, a test structure with an isolated dark area, which is arranged in a much larger lighter field, is placed in the object plane 18. This is done, for example, by arranging a suitable test reticle in the object plane 18. The scattered light sensor S4 scans the area of the test structure in the image plane 28 and calculates the scattered light component from the intensity measured in the area of the image of the dark area.

Another of the measuring sensors is a temperature sensor S5, which is configured as an infrared camera in the embodiment shown in FIG. 1 and is used to measure the parameter P5 in the form of a temperature distribution on the reflective coating 61 of the mirror E3. The measured temperature distribution is transmitted to correction determination module 62 as measured value MW5. As an alternative or in addition, further temperature sensors can be provided for measuring the respective temperature distribution on the surfaces of other mirrors in the projection lens 26 and/or in the illumination optics 16 . About it In addition, one or more temperature sensors can be provided for the respective measurement of a time profile of a temperature value at a specific location. Such a temperature sensor can be configured as a thermistor, for example.

Another of the measurement sensors is a pressure sensor S6 arranged inside the projection lens 26 . Furthermore or alternatively, pressure sensors can also be arranged at other locations in the projection exposure system 10 . The pressure sensor S6 measures the parameter P6 in the form of the atmospheric gas pressure in the projection lens 26 and transmits this as a measured value MW5 to the correction determination module 62. In addition, a gas analysis sensor S1 1 can be used to determine the parameter P1 1 in the form of the gas composition within the projection exposure system 10, for example in Projection lens 26, be provided in particular for residual gas analysis.

Other measuring sensors are intensity sensors S7 and S8, which can be inserted into useful beam path 44 directly in front of mirror E1 or in the area of intermediate focal plane 42 to measure parameter P7 or P8 in the form of the intensity of exposure radiation 38. The respective measurement result, either in the form of an individual intensity value or in the form of a local intensity distribution, such as a two-dimensional intensity distribution, is transmitted to correction determination module 62 as measured value MW7 or MW8. The positioning of the intensity sensors S7 and S8 shown in FIG. 1 should only be understood as an example. Intensity sensors S7 and S8 and possibly other intensity sensors can also be arranged at other suitable positions in useful beam path 44 .

Another of the measuring sensors is designed as a position sensor S9 for determining the parameter P9 in the form of the position of the mirror E1. The measured position is transmitted to correction determination module 62 as measured value MW9. The measured position can be location coordinates and/or angle information include tilting of the mirror E1. The position sensor S9 can be an encoder or an interferometric distance measuring device, for example. The position sensor S9 is only shown as an example in FIG. Additional position sensors of this type can be provided to determine the respective position of other mirrors of the projection exposure system 10 .

Another of the measurement sensors is designed as a position sensor S10 for determining the parameter P10 in the form of the position of a facet 50 of the facet mirror 48 . The measured position is transmitted to correction determination module 62 as measured value MW 10 . The measured position can include location coordinates and/or angle information for the tilt position of the facet 50. Analogous to the position sensor S9, the position sensor S10 can be an encoder or an interferometric distance measuring device, for example. The position sensor S10 is only shown as an example in FIG. Further position sensors of this type can be provided for determining the respective position of further facets 50 of facet mirror 48 or of facets of facet mirror 52 .

The measurement using the sensors S1 to S12 can take place regularly after each exposure of a wafer or after exposure of a complete set of wafers. The measuring frequency of some sensors can also be higher than the measuring frequency of other sensors.

The correction determination module 62 determines an execution variant 64a of a correction command 64 from the transmitted measured values MW1 to MW12, in any case from at least two transmitted measured values is explained. The correction command 64a comprises a multiplicity of corrective measures, in the embodiment illustrated in FIG. 1 the corrective measures KM1 to KM12. The corrective measures KM1 to KM5 are travel specifications for the manipulators M1 to M5. As discussed above, manipulators M1-M4 are configured to perform rigid body motions on mirrors E1-E4. The travel specifications of the corrective measures KM1 to KM4 thus specify the rigid body movements to be carried out by the manipulators M1 to M4. The travel specification of the corrective measure KM5 specifies how the heating elements 60 are to be operated and thus which heating load distribution is to be set at the mirror E5.

The corrective measures KM6 to KM9 are instructions to the exposure control unit 58 to modify the control signals C1 to C4 emanating from it. The corrective measures KM6 and KM7 indicate how the movement sequences of the reticle displacement stage 24 and of the wafer displacement stage 34 controlled by the control signals C1 and C2 are to be corrected during an exposure process. The corrective measures KM8 and KM9 indicate how the settings of the facet mirrors 48 and 52 specified by the control signals C3 and C4 are to be corrected, resulting in a correction of the illumination setting.

The corrective measure KM10 represents a repair instruction to be carried out manually. This is to be understood as an instruction that a suitable person, such as a service technician, has to carry out on the projection exposure system 10 in order to improve the imaging behavior. Such a repair instruction can, for example, order the replacement of an optical element of the projection exposure system 10, for example the deflection mirror 46, as illustrated in FIG. In addition to the replacement of other optical elements, such a repair instruction can, for example, also order a repair of mechanical modules, such as the reticle transfer stage 24 or the wafer transfer stage 34 . In an embodiment of the projection exposure system 10 that is not shown in the drawing, the illumination optics 16 have a so-called illumination CGH. The replacement of such a lighting CGH can also be part of the repair instructions mentioned. 2 illustrates the mode of operation of the correction determination module 62. This is configured to use the optimization algorithm 66 to generate the execution variant 68a of the correction command 68 from the measured values of a measurement vector 68m. The measurement vector 68m includes measured values MW1 to MWm of the parameters P1 to Pm; in the exemplary embodiment according to FIG. 1 these are the measured values MW1 to MW12 of the parameters P1 to P12. The correction command includes information about the corrective measures KM1 to KMm; in the exemplary embodiment according to FIG. 1, these are the corrective measures KM1 to KM12.

The optimization algorithm 66 is a machine learning method, i.e. a learning method based on artificial intelligence, based on an artificial neural network with at least five layers. The optimization algorithm 66 is set up to be learned or trained using a large number of learning data sets 70 . The training data sets 70 are generated by a simulation module 72 and made available to the optimization algorithm 66 before the exposure operation of the projection exposure system 10 is started. According to one embodiment, in addition to the originally provided training data sets, further training data sets 70 can also be generated by the simulation module 72 during the exposure operation and made available 66 to the optimization algorithm.

According to one embodiment, at least one of the training data sets 70 generated by the simulation module 72 is retained during the training of the optimization algorithm 66 . After completion of the learning process, a learning check is carried out, in which the learning vector 68I of the at least one retained learning data record 70 is made available to the learned optimization algorithm 66 . From this, the optimization algorithm 66 determines a control variant of the correction command 64 which is then compared with the training variant 64I of the correction command 64 contained in the training data record 70 in question. In the case of extensive agreement, the learning process is regarded as successful and the learned optimization algorithm 66 is released for use in the exposure mode of the projection exposure system 10 . In the event that there is insufficient agreement, further training data records are generated by the simulation module 72 and the training process is continued with these.

The simulation module 72 can be part of the projection exposure system 10 or be independent of it. The simulation module 72 simulates various system states SZj of the projection exposure system 10, provided with the reference symbol 74, and determines a training data record 70 for each of the system states SZj Learning vector 68a comprises learning values AW1 to AWm of parameters P1 to Pm, in the exemplary embodiment according to FIG. 1 these are learning values AW1 to AW12 of parameters P1 to P12. The learning values AW1 to AWm of the learning vector 68I correspond to the measured values MW1 to MWm of the measuring vector 68m in that they each relate to values for the parameters P1 to Pm. The training values AW1 to AWm differ from the measured values MW1 to MWm only in that the training values are determined by simulating the relevant system states 74 using the simulation module 72 for the relevant parameters, while the measured values are determined by measuring the parameters on the real projection exposure system 10 .

The various system states 74, which are simulated in the simulation module 72, can be defined by one or more of the state parameters 76 listed below: First, various points in time ti in the exposure process of the projection exposure system 10 can serve as state parameters 76-1. This means that the respective system state SZj of the projection exposure system 10 is calculated at different points in time t 1 , t 2 , ta etc. in the time course of the exposure of a wafer 30 or a set of wafers 30 . For example, the heating that takes place over the course of exposure over time, and possibly cooling in exposure interruptions, of mirror elements E1 to E4 of the projection lens 26 and possibly different before mechanical conditions of the reticle transfer stage 24 and the wafer transfer stage 34 are taken into account. In other words, the training data records 70 can include a time development of the training values AW1 to AWm. The optimization algorithm 66 is then configured in particular to take into account a temporal development of the measured values MW1 to MWm when generating the correction command 64a.

Furthermore, different structure types of mask structures 21 to be imaged can serve as further state parameters 76-2 for defining the system states SZj. For example, imaging may be simulated using the following texture types: dense vertical lines, dense horizontal lines, isolated vertical lines, isolated horizontal lines, dense dot structures, and/or isolated dot structures.

Furthermore, various error events FE1, FE2, etc. in relation to the projection exposure system 10 can serve as further state parameters 76-3 for defining the system states SZj. Such an error event can be given, for example, by the entry of gas into the useful beam path 44, whereupon it is possible that interferometric position sensors, for example, deliver erroneous measured values, which can lead to incorrect positioning of mirrors in the useful beam path 44.

Furthermore, one or more of the parameters P1 to Pm can serve as status parameters 76-4, to which specific values for simulating the system statuses SZj are then assigned. In addition, settings relating to the control signals C1 to C4 can also serve as status parameters 76-5. By choosing different settings for the control signals C1 and C2, different operating modes of the reticle transfer stage 24 and the wafer transfer stage 34 can be simulated. Different lighting settings can be simulated by selecting the settings for the control signals C3 and C4. FIG. 3 illustrates an embodiment with an alternative mode of operation of the correction determination module 62 and of the simulation module 72. This embodiment differs from the embodiment according to FIG.

2 that the simulation module 72 initially only transmits the training vector 681 to the correction determination module 62 as the result of the simulation of the respective system state 74 . This then uses the optimization algorithm 66 to determine a training variant 641 of the correction command, also referred to as a training correction command 641, and sends this back to the simulation module 72 with the question of whether the training correction command 641 is suitable for error correction in the simulated system state 74 or not . In response to this question, the simulation module 72 sends corresponding feedback 80 to the correction determination module 62. This feedback 80 contains either the answer “Yes” (Y) or “No” (N) to the above question. Alternatively or additionally, the feedback 80 can also include a quantitative evaluation in the form of a scalar value of a target function. The training data set 70 transmitted in this embodiment for the relevant system state 74 to the correction determination module 62 for training the optimization algorithm 66 thus effectively includes the training vector 681 and the associated feedback 80, which in this text is also used as an indication of the training variant determined by the correction determination module 62 641 of the correction command 64 is designated.

In the case in which the feedback 80 has the answer “no”, the optimization algorithm 66 can, if necessary, determine an improved training correction command 641, for which purpose a feedback 80 from the simulation module 72 in turn takes place. This process is repeated until the feedback 80 provided by the simulation module 72 contains a "yes" answer.

In FIG. 4, measurement results of the wavefront sensor S1 are illustrated by way of example using error distributions of the Zernike coefficients Z2 to Z7 in the area of the exposure slit of the projection exposure system 10. The left-hand section of FIG. 4 relates to measurement results of thermally induced image errors 78t, ie image errors which are due to inhomogeneous temperature distributions on one or more of the mirrors E1 to E4, and in the right-hand section of FIG. 4 to measurement results of image errors 78s caused by rigid-body movements of the mirrors E1 to E4. As can be seen from the illustration in FIG. 4, the error distributions of the Zernike coefficients Z2 to Z6 for the image errors 78t and 78s are almost identical in the exemplary embodiments chosen here, only the error distributions of the Zernike coefficient Z7 show significant differences. This exemplary embodiment illustrates the advantage of the measure according to the invention of measuring measured values (cf. MW1 to MW12) of several parameters (cf. P1 to P1) influencing the imaging quality of the projection exposure system 10 and making them available to the optimization algorithm 66 for determining the correction command 64a.

If only the measured values MW1 of the wavefront sensor S1 were made available to the optimization algorithm in the form of error distributions of the Zernike coefficients, it would not be able to clearly identify the cause of the image errors, at least in the exemplary embodiment illustrated in FIG. 4, namely whether the cause is thermal type is caused by rigid body displacements of the mirrors E1 to E2 or whether a combination of both causes is present. As a result, the optimization algorithm 66 might not determine the optimum correction command. Due to the use according to the invention of measured values of several parameters, possibly also of measured values that specify mirror positions (cf. position sensors S9 and S10), the optimization algorithm 66 can implicitly take into account the actual causes of the image errors with a high success rate and thus issue a correspondingly effective correction command 64a determine.

The above description of exemplary embodiments, embodiments or embodiment variants is to be understood as an example. The disclosure thus made will enable those skilled in the art to understand the present invention and the advantages attendant thereto, while also encompassing variations and modifications to the described structures and methods that would become apparent to those skilled in the art. Therefore, all such Variations and modifications insofar as they come within the scope of the invention as defined in the appended claims, and equivalents, are to be covered by the protection of the claims.

Reference List

10 projection exposure system for microlithography

12 lighting system

14 radiation source

16 illumination optics

18 object level

20 reticles

22 reticle holders

24 reticle transfer stage

25 scan direction

26 projection lens

28 image plane

30 wafers

32 wafer holders

34 wafer transfer stage

36 wafer scan direction

38 exposure radiation

40 collector

42 intermediate focal plane

44 useful beam path

46 deflection mirrors

48 first facet mirror

50 facet

52 second facet mirror

54 facet

56 tilt axis

58 exposure control unit

60 heating element

61 reflective coating

62 correction determination module

64 correction command 64a Execution variant of a correction command

641 Learning variant of a correction command

66 optimization algorithm

68m measurement vector

681 training vector

70 training records

72 simulation module

74 system states

76 state parameters

78t thermal caused image errors

78s Image artifacts caused by rigid body motion

80 feedback

C1-C4 control signals

E1 -E4 mirror of the projection lens

KM1 -KM12 corrective actions

MW1 - MW12 measured values

M1 -M5 manipulators

P1-P12 parameters

51 wavefront sensor

52 image position sensor

53 uniformity sensor

54 scattered light sensor

55 temperature sensor

56 pressure sensor

57 intensity sensor

58 intensity sensor

59 position sensor

510 position sensor

511 gas analysis sensor

512 uniformity sensor

Claims

37 claims

1. Projection exposure system for microlithography for imaging mask structures on a substrate with a correction determination module which is configured to generate an embodiment variant of a correction command which specifies at least one corrective measure with regard to the projection exposure system, the correction determination module being configured to determine the execution variant of the correction command using an optimization algorithm to generate measured values from parameters influencing an imaging quality of the projection exposure system, the measured values being determined using at least two different measuring sensors, and the optimization algorithm being configured as a machine learning method which is set up to use a large number of training data sets which each include training values of the parameters influencing the imaging quality and an indication of a training variant of the correction command assigned to the respective training values.

2. Projection exposure system according to claim 1, which further comprises a plurality of optical elements for illuminating and imaging the mask structures onto the substrate, wherein the corrective measure relates to at least one of the optical elements.

3. The projection exposure system as claimed in claim 2, wherein the projection exposure system comprises at least one manipulator which is configured to change an optical effect of at least one of the optical elements by manipulating a property of the optical element along a travel, and the correction command comprises a travel specification for the manipulator .

4. Projection exposure system according to one of the preceding claims, 38 wherein the correction command comprises repair instructions to be carried out manually in the event that a correction that can be carried out automatically is not considered appropriate.

5. Projection exposure system according to one of the preceding claims, wherein the information on the training variant of the correction command in at least one of the training data records includes the relevant training variant itself.

6. Projection exposure system according to one of the preceding claims, wherein the information on the training variant of the correction command includes feedback on the suitability of a training variant of the correction command determined by the optimization algorithm using the training values.

7. Projection exposure system according to one of the preceding claims, in which the measuring sensors comprise a wavefront sensor and a temperature sensor.

8. Projection exposure system according to one of the preceding claims, wherein the measurement sensors comprise at least two elements from a wavefront sensor, a temperature sensor, a pressure sensor, at least one intensity sensor arranged in the beam path of the projection exposure system, a sensor for determining a position and/or an orientation of one of the optical elements and a sensor for determining the group having the gas composition.

9. The projection exposure system as claimed in claim 8, wherein the at least one intensity sensor comprises one or more elements from the group comprising an image position sensor, a uniformity sensor, a scattered light sensor and a sensor for measuring a reflectivity of an optical element of the projection exposure system.

10. Projection exposure system according to one of the preceding claims, where the optimization algorithm is based on an artificial neural network with at least five layers.

11 . Projection exposure system according to one of the preceding claims, wherein the training data records include a time development of the training values and the optimization algorithm is configured to take into account a time development of the measured values when generating the correction command.

12. Projection exposure system according to one of the preceding claims, which further comprises a simulation module, which is configured to determine the training data sets by simulating different system states of the projection exposure system.

13. The projection exposure system as claimed in claim 12, in which the simulation module is configured to simulate system states which are caused by different points in time in the exposure process of the projection exposure system, by different error events in relation to the projection exposure system, by different values for the parameters influencing an imaging quality of the projection exposure system and/or or are defined by different configurations of the mask structures to be imaged.

14. A method for operating a projection exposure system for microlithography for imaging mask structures on a substrate, with the steps:

- Learning an optimization algorithm configured as a machine learning method using a large number of learning data sets, each of which includes learning values of parameters influencing an imaging quality of the projection exposure system and information on a learning variant of a correction command assigned to the respective learning values, the correction command including at least one corrective measure with regard to the projection exposure system specifies, - Determination of measured values of the parameters influencing the imaging quality by means of at least two different measuring sensors, and

- Generation of an execution variant of the correction command from the measured values using the learned optimization algorithm.

15. The method as claimed in claim 14, in which a learning check takes place after the learning of the optimization algorithm.

16. The method as claimed in claim 14 or 15, in which the training data sets are determined by simulating various system states of the projection exposure system.