Method and system for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter of the lithography apparatus

Abstract

A method, comprising: - generating (S1) N copies (304) of a neural network (306); - providing (S2) T training data records that are generated by the following steps, for i = 1 to T: a) reading (S20) parameters from an ith lithography apparatus (1, 300), b) generating (S21) an ith training data record on the basis of the read parameters; - training (S3) the N copies, comprising the following steps, for j = 1 to N: aa) training (S30) the jth copy using the ith training data record; bb) determining (S31) a jth item of change information for the trained jth copy (304); - generating (S4) an anonymized item of change information (310); - adjusting (S5) the neural network (306); - predicting (S6) an ageing effect using the adjusted neural network (312); and - producing (S7) a lithography apparatus (1, 300) or one of its components (302) on the basis of the predicted ageing effect and / or adjusting an apparatus parameter of a lithography apparatus (1, 300) on the basis of the predicted ageing effect.

Description

[0001] METHOD AND SYSTEM FOR PRODUCING A LITHOGRAPHY APPARATUS OR ONE OF ITS COMPONENTS AND / OR FOR ADJUSTING AN APPARATUS PARAMETER OF THE LITHOGRAPHY APPARATUS

[0002] The present invention relates to a method and a system for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter and / or for making a forecast regarding a failure of the lithography apparatus.

[0003] The content of priority application DE 10 2024 201 726.4 is incorporated by reference in its entirety.

[0004] Microlithography is used for producing microstructured components, such as for example integrated circuits. The microlithography process is carried out using a lithography apparatus, which has an illumination system and a projection system. The image of a mask (reticle) illuminated by way of the illumination system is projected here by way of the projection system onto a substrate, for example a silicon wafer, which is coated with a light-sensitive layer (photoresist) and arranged in the image plane of the projection system, in order to transfer the mask structure to the light-sensitive coating of the substrate.

[0005] Driven by the desire for ever smaller structures when producing integrated circuits, EUV lithography apparatuses are currently being developed that use light having a wavelength in the range of 0.1 nm to 30 nm, in particular 13.5 nm. Since most materials absorb light of this wavelength, it is necessary in such EUV lithography apparatuses to use reflective optical units, that is to say mirrors, instead of ■ as previously - refractive optical units, that is to say lens elements.

[0006] Due to high radiation loading within the lithography apparatuses, various ageing processes of the optical elements, such as for example mirrors or lens elements, occur over the apparatus lifetime. With constantly increasing laser powers and new types of exposure systems, it is additionally expected that, in the future, numerous new or unknown (ageing) effects and effect profiles will also occur, these not yet being known at present.

[0007] One example of an ageing effect that may be observed in numerous existing lithography apparatuses consists of a degradation of optical elements. By way of example, UV radiation, which is used in immersion lithography, changes the optical properties, such as refractive index, density, absorption coefficient and birefringence, of the optical materials of the optical elements that are used. Such materials are for example optical glasses, which are used predominantly in i-line (365 nm) lithography apparatuses, and synthetic quartz glass (fused silica), which is an essential component of the optical material of systems operating at a working wavelength of 248 nm (KrF) or 193 nm (ArF).

[0008] For pulsed 193-nm laser radiation, as used in immersion lithography, the change in the refractive index of the synthetic quartz glass that is often used as a lens material is known as compaction (increase in refractive index) or rarefaction (decrease in refractive index), and is described for example in US10,427,965.

[0009] Especially in the case of new, high-precision immersion lenses, the requirements on the wavefront are so high that higher-order changes of the wavefront, which occur over the service life, even in the range of a few nm, in some cases even below 1 nm, may render the lenses unusable. To extend the service life, there is the possibility of making individual optical elements exchangeable.

[0010] Since the production of such an optical exchange element takes a long time and a shutdown of the lithography apparatus results in a high loss of turnover, it is unfavourable for the customer to start the production of such an optical element only when the lithography apparatus is outside its specification, and such an optical exchange element is thus needed acutely.

[0011] For this reason, it is desirable to make changes to the optical properties of optical elements, in particular projection lenses, in a manner that is as predictable as possible, in order thereby to be able to manufacture such a required optical exchange element in advance in order to exchange it during maintenance of the lithography apparatus that is planned anyway.

[0012] In addition, a precise prediction of the optical properties of an optical element at a time of a planned exchange makes it possible to optimize the optical exchange element by applying individual aspheres such that the optical properties of the specific optical element are optimal after the exchange.

[0013] Another advantage of predicting the change in the optical properties of an optical element and knowing the effect of the settings of the lithography apparatus on such an optical element is that the user behaviour of a customer is able to be optimized. By way of example, it is thus possible to suggest, directly in the lithography apparatus, other illumination settings that maintain production but wear the components of the lithography apparatus more evenly and thus ensure a longer service life.

[0014] The loading and thus the wear of the optical elements is also influenced by their optical design and by other operating parameters of the lithography apparatus. These operating parameters include, inter alia, a wavelength, a number and / or time length of light pulses, a size of a field illuminated on a lithography mask, a distribution of an intensity of light in a pupil plane of the respective optical element, a polarization of the light, a transmission of the lithography mask and a transmission distribution. Many of these operating parameters are set actively in the lithography apparatus (for example size of the field illuminated on the mask, distribution of the intensity of light in the pupil plane, etc.) or are defined by the configuration (time length of the laser pulses).

[0015] The effects of the individual operating parameters on the optical elements of the lithography apparatus are generally known here and may be described by simulation calculations. However, since the exact usage data of the machine are known only to the customer or the user of the lithography apparatus and are highly sensitive secret business data that are not intended to leave an infrastructure of the customer, it is difficult to further investigate the effects, which are known per se, this requiring test setups that are often expensive, and to improve their prediction accuracy and to adapt to changing (operating and / or ambient) conditions.

[0016] Since the simulation -based calculation of the ageing effects or of the wear of the optical elements is also highly computationally intensive, a direct calculation of the effects of operating parameters in the lithography apparatus itself requires an excessive amount of computing capacity and computing time.

[0017] In summary, it may thus be stated that, in the current state of the art, an optimal and / or controlled prediction of ageing effects of optical elements caused by irradiation and / or effects of different operating parameters on such ageing effects is possible only to a very limited extent.

[0018] Against this background, one object of the present invention is to provide an improved method and / or system for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter of the lithography apparatus, in particular taking into account anonymization of apparatus data in order to ensure customer and / or apparatus anonymity. What is accordingly proposed is a method for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter and / or for making a forecast regarding a failure of the lithography apparatus. The method comprises the following steps: generating N copies of a neural network for predicting an ageing effect of optical elements of lithography apparatuses, wherein N > 1; providing T training data records that are generated by the following steps, for i = 1 to T, wherein T > N: a) reading parameters from an ith lithography apparatus, b) generating an ith training data record on the basis of the read parameters; training the N copies, comprising the following steps, for j = 1 to N: aa) training the jth copy using the ith training data record; bb) determining a jth item of change information for the trained jth copy; generating an anonymized item of change information on the basis of at least two of the N items of change information; adjusting the neural network on the basis of the anonymized item of change information; predicting an ageing effect using the adjusted neural network; and furthermore, optionally: producing a lithography apparatus or one of its components on the basis of the predicted ageing effect and / or adjusting an apparatus parameter and / or making a forecast regarding a failure (in particular forecasting a failure) of a lithography apparatus on the basis of the predicted ageing effect.

[0019] According to the method, multiple copies (N copies) of a neural network are generated. Preferably, a number N of the generated copies of the neural network corresponds to a number of the lithography apparatuses to each of which one of the copies is assigned. Furthermore, according to the method, at least T training data records are provided, wherein T should be at least as large as N. The training data records are preferably generated on an apparatus -specific basis for each of the lithography apparatuses or at least for a subset of the lithography apparatuses. Each of the generated training data records is then preferably used to train that copy of the neural network that is assigned to the respective lithography apparatus. In this training step, the respective copy of the neural network is thus preferably trained locally on the respective lithography apparatus. Since the neural network is preferably trained on the basis of apparatus -specific training data on each of the lithography apparatuses, this respective training does not require a large amount of computing power. An evaluation device that is preferably used in the respective lithography apparatus and that performs the training or on which the neural network is executed may thus be designed with a low computing power. This demonstrates one of the advantages of the federated learning approach used here. Specifically, on the lithography apparatus, all training data of the lithography apparatus are preferably retrievable and the respective locally executed copy of the neural network is thereby able to access these training data so as to be locally retrained based on these training data, without the training data however leaving an infrastructure of the apparatus operator. The retraining thereby does not pose any security risks.

[0020] After the respective neural network has been trained, an item of change information is determined for each lithography apparatus, this item of change information in particular recording one or more deviations between network parameters and / or hyperparameters of that copy of the neural network that is trained on the respective apparatus in relation to the incoming neural network. The respective apparatus -specific change information is processed further in a further step so as to generate an anonymized, in particular summarized item of change information. The anonymization is carried out in order to make it impossible to infer the respective lithography apparatus. This is advantageous because process-specific and / or user-specific settings and / or process parameters, which may constitute a trade secret, are often used in lithography apparatuses. In particular, if a predetermined number of items of change information in relation to a predetermined number of lithography apparatuses have been determined, the respective items of change information may be loaded onto a secure server or into a secure cloud. The server or the cloud preferably cannot be accessed externally.

[0021] The neural network copied at the outset is then trained on the basis of the anonymized item of change information in order thereby to be optimized. This makes it possible to improve prediction accuracy when predicting the ageing effect.

[0022] In other words, the global neural network is thus updated on the basis of the anonymized item of change information. A degree of change or optimization of the neural network may preferably be determined based on the anonymized item of change information. It is thus possible to prevent uncontrolled deterioration of the neural network, despite the lack of direct insight into the respective apparatus-specific item of change information.

[0023] It is particularly preferable for the neural network, trained on the basis of the anonymized item of change information, to be tested by comparison with a simulation model for simulating the ageing effect in order thereby to assess a quality and / or credibility of the optimized neural network and / or of the simulation model.

[0024] If the updated neural network is optimized in relation to the previous neural network with regard to the performance data resulting from the comparison, the updated neural network may be provided to the lithography apparatuses as a new release or software update. For this purpose, in turn, N copies of the updated neural network are preferably provided and are transmitted to the respective lithography apparatus. On the respective lithography apparatus, the copy of the updated neural network may preferably be tested again by comparison with an apparatus-specific simulation model or historical simulation data. Since merely testing or comparing the copy of the updated neural network with the respective simulation model requires comparatively little computing power compared to training a neural network, the local performance of the copy of the neural network may easily be tested on the respective lithography apparatus. Such a testing mechanism is advantageous for preventing uncontrolled deterioration (for example due to local overfitting) of the respective copy of the (updated) neural network or else for identifying local outliers.

[0025] With regard to the present method, the term “neural network” is synonymous with a machine learning model provided for predicting the ageing effect.

[0026] The prediction or diagnosis of ageing effects of an optical element using federated learning and the improved development potential of new lithography apparatuses resulting from federated learning is not limited to degradation-based ageing effects, but rather is equally applicable to other ageing effects that are possibly as yet unknown. The method described here is thus based on a federated learning approach, by way of which it is possible to comprehensively incorporate historical and / or current domain knowledge regarding ageing effects of optical elements. Federated learning makes it possible to perform or carry out the calculations of the ageing effects efficiently on the respective participating lithography apparatus. Federated learning also enables continuous optimization of the prediction accuracy of ageing effects of optical elements of the individual lithography apparatuses by using knowledge transfer and / or knowledge gain from other lithography apparatuses, without individual apparatus information being disclosed or accessible. The present method thus enables an anonymized improvement of the prediction accuracy of ageing effects of optical elements of the individual lithography apparatuses, without critical information leaving the infrastructure of the customer.

[0027] On the one hand, the present method enables apparatus -specific monitoring of the ageing effect, in particular by training and optimizing the respective copy of the neural network on the basis of apparatus parameters. In addition, retraining the neural network initially used for copying also enables a global optimization of the neural network on the basis of the anonymized item of change information. Such optimization of the neural network makes it possible to improve global system knowledge regarding the ageing effect. It may furthermore be preferable, after the neural network has been retrained on the basis of the anonymized item of change information, for a comparison with simulation data of a simulation of the ageing effect to take place. If the comparison reveals differences, systematic and / or syntactic errors of the simulation model may preferably be detected and, if necessary, corrected on the basis thereof. The prediction accuracy of the ageing effect is improved by preferably continuously optimizing the neural network on the basis of the anonymized item of change information that is respectively iteratively determined. This knowledge is then used to redesign a lithography apparatus or an apparatus component or to reset an apparatus parameter, in particular in order thereby to minimize the ageing effect. A further advantage of the present method is that, by locally training the respective copies of the neural network on each of the lithography apparatuses, it is possible to directly predict the ageing effect for each lithography apparatus. This provides a fast diagnostic path for the respective detection of the ageing effect, which leads to a better ability to act. The present federated learning approach in particular enables a reliable future prediction regarding the ageing effect, which may preferably be used to advantage in swap pool predictions. By way of example, it is thereby also possible to improve strategic decisions regarding production and / or maintenance planning for each lithography apparatus, resulting in an increase in apparatus productivity and / or a reduction in maintenance-related apparatus downtime. Furthermore, an application recommendation and / or settings recommendation may be issued for an apparatus user through the indirect analysis, provided on the basis of the anonymized item of change information, of a user behaviour and / or optimization of the prediction of the ageing effect. Such an application recommendation and / or settings recommendation may consist for example in instructing the apparatus user to make a change to an apparatus parameter in order to improve service life. Locally training the copies of the neural network on the respective lithography apparatus using federated learning enables a direct ageing effect-related “cause to result” link, thereby enabling efficient calculation of the ageing effect.

[0028] The present method also has privacy-related and security-related advantages, since an apparatus user may not have to send confidential and / or sensitive data to an apparatus manufacturer in order to receive suggestions for an optimization of apparatus parameters. This also makes it possible to prevent unwanted data leaks or data misuse. The apparatus manufacturer also does not have to provide an elaborate security architecture for data storage, since the apparatus operator preferably only receives the anonymized item of change information for retraining the neural network. This also makes it possible to reduce the amount of data traffic that arises. The information contained in the anonymized item of change information is used “blindly”, as it were, and cannot be reviewed manually. This significantly reduces the risk of data misuse.

[0029] According to one embodiment, the neural network is pretrained to predict the ageing effect, prior to generating the N copies, on the basis of historical and / or current and / or synthetic and / or simulatively generated training data regarding the ageing effect.

[0030] “Pretraining” refers to a process or a method step in which the neural network, which is in particular initially untrained, is trained on the basis of existing data and / or information regarding the ageing effect in order to learn and / or improve its ability to predict this ageing effect. The training data, in particular initial training data, which are preferably used for pretraining, may originate from various sources, including historical data (past data containing information regarding the ageing effect), current data (current information regarding the ageing effect), synthetically generated data (artificially generated data representing the ageing effect) and simulatively generated data (data created by simulations and reflecting the ageing effect). In other words, existing knowledge about the ageing effect to be predicted is used here to create a neural network. If real data from previous ageing determinations are available, the neural network may be (pre)trained on the basis of these real data. This is particularly preferable, since training on the basis of real data makes it possible to adapt a basic structure and / or hyperparameters of the neural network to a structure of the real data. As an alternative or in addition, the neural network may also be trained on the basis of artificially or synthetically generated data or on the basis of simulation data, in particular in order to achieve faster network convergence, in particular by minimizing a loss function. In the present case, the neural network initially trained in this way is furthermore improved, in particular continuously, using a federated learning approach, by using current, apparatus -specific system knowledge to retrain the neural network. On the one hand, existing real data may thereby be used to train an initial neural network. On the other hand, in the absence of real data, the apparatus -specific system knowledge may be used.

[0031] According to one embodiment, at least some of the method steps are carried out iteratively in order thereby to continually and / or continuously improve a model performance of the neural network for predicting the ageing effect.

[0032] “Carry out iteratively” means that the steps that are performed in connection with the neural network and its model performance for predicting the ageing effect are carried out repeatedly and step-by-step. Instead of training the neural network in a single step, the model is improved through a sequence of steps that are carried out repeatedly. The neural network may thereby be improved continuously. The “model performance of the neural network” preferably refers to the ability or accuracy of the neural network to predict the ageing effect. The model performance preferably describes how accurately and effectively the neural network makes these predictions. “Continuous improvement” in this case preferably means that the objective of the iterative method steps is to continuously increase the performance of the neural network over time. Each iteration is intended to achieve improvements in order to make the neural network more accurate and / or efficient in terms of predicting the ageing effect.

[0033] According to one embodiment, generating the anonymized item of change information on the basis of at least two of the N items of change information comprises^ weighting the at least two of the N items of change information! and summing the weighted at least two of the N items of change information to form the anonymized item of change information.

[0034] From the individual items of change information for the respective lithography apparatus, data processing is used to generate a single anonymized item of change information in order thereby to retrain the neural network on the basis of this anonymized item of change information. All lithography apparatuses thereby benefit from an improved neural network for predicting the ageing effect, without the need however to release or provide apparatus -specific information. The anonymized item of change information is preferably based on a weighted sum of at least two of the apparatus-specific items of change information. The weighting is preferably carried out randomly or depending on a performance or performance distribution of the copies of the neural network. When summed, it is preferable for the weights to add up to 1. By aggregating the individual weights of a large number of different lithography apparatuses and / or apparatus types, it is no longer possible to subsequently infer individual applications and / or lithography apparatuses. The anonymized item of change information thus no longer comprises any security-critical or privacy-critical knowledge. It is particularly preferable for the anonymized item of change information to be generated only when change information or related data regarding different lithography apparatuses and / or lithography apparatus types and / or apparatus manufacturers are present. This is preferable because reliable anonymization of the items of change information is able to be achieved starting only from a certain amount of different items of change information. This also ensures that neither an apparatus usage behaviour regarding individual lithography apparatuses nor an apparatus usage behaviour of certain lithography apparatus groups and / or users is able to be determined on the basis of the anonymized item of change information. By way of example, the anonymized item of change information is generated only starting from a number of 10 lithography apparatuses, in particular from different apparatus operators, and / or apparatus types, since reliable anonymization is not able to achieved before this.

[0035] An item of change information is described above only by way of example for each lithography apparatus or with the same total number as that of the lithography apparatuses (N items of change information for N lithography apparatuses). It should be understood that multiple items of change information occur for a lithography apparatus only if for example the same network is used for individual components. However, these components may be considered separately. For this reason, in other embodiments, multiple items of change information for each lithography apparatus are also possible, but are not necessary for encryption or anony miz ation .

[0036] Each item of change information preferably comprises a large number of individual parameters (in particular the weights and / or threshold values in the case of the neural network). These are preferably optimized in each training step, in particular locally, and thereby changed. The item of change information for each ith machine therefore preferably consists of a large number of individual parameters. For each of these parameters, a separate weight is preferably used for encryption or anonymization, wherein the sum of each of the parameters over all items of change information is preferably exactly one. According to one embodiment, the at least two of the N items of change information are weighted randomly or on the basis of a performance parameter distribution of the trained N copies of the neural network.

[0037] Preferably, multiple items of change information for multiple lithography apparatuses are used to generate the weighted sum. The number of apparatus -specific items of change information is preferably selected randomly in order to avoid being able to infer individual lithography apparatuses as early as in the step of selecting data for generating the anonymized item of change information. In the case of a random selection of the apparatus-specific items of change information, however, it is preferably possible to take into account a respective performance of the respective copy of the neural network in order thereby for example not to take into account any change information that comes from low-performance copies of the neural network or copies of the neural network that make poor predictions. Considering performance make be taken into account for example by way of a performance limit condition. A performance distribution based on the local items of change information may also be taken into account.

[0038] According to one embodiment, the ageing effect includes a radiation-related degradation of optical elements and / or a thermal and / or mechanical ageing effect and / or an ageing effect due to material deposition and / or corrosion and / or chemical degradation.

[0039] Radiation-related degradation describes in particular a process in which an optical element (such as a lens element or a mirror) in the lithography apparatus loses its efficiency over time due to continuous exposure to intense radiation, typically UV light or electron beams. This may happen in various ways. On the one hand, it is possible that the optical properties of the optical element change, wherein for example a transmission, a reflection and / or a refractive index of the materials of the optical element may change. Furthermore, radiation-related damage to the surface of the optical element may occur. By way of example, cracks and / or discolouration may occur on the surface of the optical elements. Thermal effects may lead to ageing of the optical element by virtue of the optical element experiencing long-term exposure to high energy, which may lead to an increase in the temperature of the optical elements. This energy input may in turn change a shape and / or alignment of the optical element. Mechanical wear of the optical element may also lead to ageing. In particular, moving parts in the lithography apparatus may be subject to wear due to regular use, which may affect the alignment and / or focus of the optical elements. Deposition of materials may likewise cause ageing of the optical elements. In an environment containing dust and / or other particles, these may accumulate on the optical elements and affect their performance. Such particles may also be produced by the outgassing of components of the lithography apparatus due to the thermal energy input, and may increase over an apparatus lifetime. Corrosion and / or chemical degradation may also influence ageing. Certain environmental conditions and / or chemical exposures may lead to corrosion of and / or other chemical damage to the optical elements.

[0040] According to one embodiment, the optical elements include lens elements and / or mirrors and / or optical sensors.

[0041] Of course, the lithography apparatus may also have other optical elements that are not mentioned explicitly here. The list given here should therefore in no way be understood as being restrictive.

[0042] According to one embodiment, producing a lithography apparatus or one of its components on the basis of the predicted ageing effect and / or adjusting an apparatus parameter of a lithography apparatus on the basis of the predicted ageing effect comprises^ retrofitting at least one component of a lithography apparatus on the basis of the predicted ageing effect in order to minimize the ageing effect; and / or adjusting an apparatus parameter of a lithography apparatus incrementally or step-by-step on the basis of the predicted ageing effect in order to minimize the ageing effect.

[0043] Retrofitting at least one component of a lithography apparatus on the basis of the predicted ageing effect preferably aims to modify or replace existing components of the lithography apparatus in order to reduce the influence of ageing. Such retrofitting may comprise updating or replacing parts such as lens elements, mirrors, mechanical parts or control elements that could be affected by ageing effects such as wear, degradation or loss of efficiency. Such retrofitting may be based on the ageing effect predicted by the neural network, and may be supplemented by further analyses, empirical data and / or simulation-based forecasts. The incremental or step-by-step adjustment of an apparatus parameter of a lithography apparatus on the basis of the predicted ageing effect comprises for example fine- tuning operating parameters of the apparatus in order to compensate for or minimize the effects of ageing. This may comprise adjusting parameters such as light intensity, focus, alignment or temperature control incrementally or step-by-step. Proactive adjustment is also possible, this preferably being carried out based on the prediction regarding the ageing effect, in order to maintain the performance of the apparatus and minimize potential downtime.

[0044] Through more accurate predictions regarding the reason and / or the type of performance losses of components, for example due to degradation, it is conceivable, in addition to setting various known parameters or settings of the components, such as for example an optimal rotation and / or mirroring, to generate new, improved illumination settings through the new findings. On the one hand, these may optimize or improve the performance of the lithography apparatus comprising already degraded components, which is synonymous with troubleshooting. On the other hand, it is thereby possible to extend the service life of the lithography apparatus and / or the components in question. In addition, it is possible to easily and accurately calculate scenarios, which may for example lead to a possible reduction in the light source strength for positive effects with regard to the service life of individual components. The more accurate prediction also makes it possible to make statements about when maintenance and / or exchange of a particular component becomes absolutely necessary in order to avoid a shutdown of the lithography apparatus until exchange parts arrive.

[0045] The predictions that are made may also be used to improve a simulation of (ageing) effects of possible new developments of components. This makes it possible to generate a better benefit assessment and / or an improved target specification. It is also possible to plan, at an early stage, which capacities are required by individual components in production planning and which components should be made exchangeable. The predictions that are made may furthermore be used to enable specifications and / or new usage settings for components. By way of example, the prediction that is made here makes it possible to rotate the illumination settings of components, which may also in particular be tested directly.

[0046] According to one embodiment, the minimization of the ageing effect is based on solving an inverse problem related to the ageing effect and / or solving a feedback problem and / or solving an optimization problem.

[0047] The wording “solving an inverse problem related to the ageing effect” is preferably understood to mean that causes of the predicted ageing effect are determined. Here, this means inferring the underlying causes or mechanisms from the predicted ageing effect. This may be achieved for example by analysing the changes in the optical properties, such as a decrease in transparency or a change in refractive index, in order to identify the specific degradation processes. The wording “solving a feedback problem” is understood to mean adjusting the lithography apparatus or one of its components or apparatus parameters so as to achieve a desired state or a desired performance. By way of example, the apparatus parameters or components of the apparatus may be adjusted so as to achieve the desired performance despite the ageing effects. The wording “solving an optimization problem” describes how to find the best solution under given limitations, in particular taking into account multiple competing factors. In the present case, this preferably means finding the optimal operating conditions or component configurations for minimizing the negative effects of ageing, in particular considering factors such as availability and performance efficiency.

[0048] Furthermore, it is conceivable to generate certain component improvements on the basis of simulation parameters that are adjusted based on ageing effects in a simulation. Furthermore, it may be possible to generate swap possibilities and / or redundancies simulatively on the basis of the ageing effects, to add these on to implementation costs and then to simulate the total cost in connection with the neural network, and thus make decisions regarding the type of future developments. In other words, it is possible to determine optical properties on the basis of predicted ageing effects in order to counteract these ageing effects. These properties may be used as input for manipulators in order thereby for example to at least partially automate simulations for the design of new components.

[0049] According to one embodiment, generating the anonymized item of change information on the basis of at least two of the N items of change information comprises the in particular respective local, unidirectional provision of the N items of change information to a server or a cloud.

[0050] The anonymized item of change information is preferably generated on a server or a cloud. Such a server or such a cloud is preferably protected from external access by appropriate security measures. By way of example, a server or a cloud may have only an intranet port, and thereby be isolated from access via the Internet.

[0051] According to one embodiment, a layer structure and / or a layer-by-layer number of neurons of the neural network is selected on the basis of the ageing effect.

[0052] In the present case, the layer structure of the neural network preferably refers to the arrangement and structure of the various layers in the neural network, including the number of layers and the type of these layers (for example convolutional layers, pooling layers, fully connected layers). The layer structure is preferably selected such that it is suitable specifically for the identification, analysis and management of ageing effects. The selection of the layer-by-layer number of neurons preferably refers to adjusting the number of neurons in each layer of the neural network. The number and distribution of the neurons in the various layers may have a significant influence on the performance and specialization of the network. The number of neurons is preferably selected on the basis of the specific requirements of the ageing effect. By way of example, a network designed to detect fine, complex ageing patterns could require a larger number of neurons in certain layers. The neural network is configured based on the specific properties and requirements of the ageing effect to be detected, both in terms of the layer structure and in terms of the number of neurons. This means that the network is trained and / or configured so as to be able to respond effectively to the specific patterns, indicators or consequences of ageing effects.

[0053] The large number of relevant measurement data of the optical unit of a lithography apparatus are in particular measurements of wavefronts describing the aberrations of optical systems or components. In order to forecast these for the future, it is preferable to incorporate the type of use of the lithography apparatus (possibly as a time series). By way of example, a convolutional neural network (CNN) having additional input layers may be integrated here into the fully connected intermediate layer of the neural network. A large number of relevant usage data of the lithography apparatus (in particular including with a data history) are thus preferably incorporated. By way of example, the following usage data are incorporated:

[0054] • (in particular illumination) settings of the lithography apparatus

[0055] • reticle transmission

[0056] • pulse data

[0057] • power of the source / laser

[0058] • material of the individual components

[0059] • machine types

[0060] • source environment

[0061] These usage data are preferably incorporated, as feature inputs, into the fully connected intermediate layer. The wavefront (Zernike) image to be considered is preferably incorporated, as input, into the convolutional neural network.

[0062] The size of the neural network is preferably flexible, as thus also is that of the layer structure. A simple convolutional network having three layers, in which the usage data are incorporated as additional input data as features after layer 1, would be conceivable. Depending on the desired complexity, a simple network having residual blocks, in which the usage conditions are also installed within the residual block, is also conceivable.

[0063] However, in any case, a suitable network would contain more than 2000 trainable parameters, which would enable efficient encryption.

[0064] According to one embodiment, the jth items of change information each contain ageing-relevant information regarding apparatus -specific operating parameters and / or regarding an apparatus configuration and / or regarding an apparatus interface and / or regarding an apparatus history and / or regarding an apparatus brand and / or regarding an apparatus type.

[0065] What is also proposed is a system for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter and / or for making a forecast regarding a failure of the lithography apparatus. The system comprises^ an in particular central computing device, i lithography apparatuses each having an evaluation device, a server or a cloud, wherein the computing device is designed to generate N copies of a neural network for predicting an ageing effect of optical elements of lithography apparatuses, wherein N > 1, and to provide an ith copy of the N copies to the evaluation device of each ith lithography apparatus, wherein the evaluation device of each of the ith lithography apparatus is designed to read parameters from the ith lithography apparatus and to generate an ith training data record on the basis of the read parameters, to train the ith copy of the N copies using the ith training data record, to determine an ith item of change information for the trained ith copy, and to provide the determined ith item of change information to the server or the cloud so as to provide a total of N items of change information, wherein the server or the cloud is designed to generate an anonymized item of change information on the basis of at least two of the N items of change information, and to provide the anonymized item of change information to the computing device, wherein the computing device and / or the evaluation device are / is designed to adjust the neural network on the basis of the anonymized item of change information, to predict an ageing effect using the adjusted neural network, and to provide a control instruction for producing a lithography apparatus or one of its components on the basis of the predicted ageing effect and / or for adjusting and / or for making a forecast regarding a failure of an apparatus parameter of a lithography apparatus on the basis of the predicted ageing effect.

[0066] The ageing effect may for example be determined locally in the evaluation device and used to output individual instructions, in particular instructions related to the individual lithography apparatuses. The ageing effect may also be determined for example in the computing device, and used to generate a global update of lithography apparatus software based on change parameters. It may also be possible to make global changes to an apparatus or product design, in particular by improving the simulation capabilities of individual components, and to compare this with other designs.

[0067] The lithography apparatus preferably comprises a projection optical unit. The lithography apparatus may also have an illumination system. The lithography apparatus or the projection exposure apparatus may be an EUV lithography apparatus. EUV stands for “extreme ultraviolet” and denotes a wavelength of the operating light of between 0.1 nm and 30 nm. The projection exposure apparatus may also be a DUV lithography apparatus. DUV stands for “deep ultraviolet” and denotes a wavelength of the operating light of between 30 nm and 250 nm.

[0068] “A(n)” should not necessarily be understood as a restriction to exactly one element in the present case. Rather, a plurality of elements, such as for example two, three or more, may also be provided. Nor should any other numeral used here be understood to the effect that there is a restriction to exactly the stated number of elements. Rather, unless indicated otherwise, numerical deviations upwards and downwards are possible.

[0069] The embodiments and features described for the method apply, mutatis mutandis, to the proposed system, and vice versa. Further possible implementations of the invention also comprise combinations, not mentioned explicitly, of features or embodiments described above or hereinafter with respect to the exemplary embodiments. In this case, a person skilled in the art will also add individual aspects as improvements or supplementations to the respective basic form of the invention.

[0070] Further advantageous configurations and aspects of the invention are the subject matter of the dependent claims and also of the exemplary embodiments of the invention that are described below. The invention is explained in detail hereinafter on the basis of preferred embodiments with reference to the accompanying figures.

[0071] Figure 1 shows a schematic meridional section of a projection exposure apparatus for EUV projection lithography!

[0072] Figure 2 shows a schematic flowchart of one exemplary embodiment of the present method for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter of the lithography apparatus! and

[0073] Figure 3 shows a schematic block diagram of one exemplary embodiment of the present system for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter of the lithography apparatus.

[0074] Unless indicated otherwise, elements that are identical or functionally identical have been provided with the same reference signs in the figures. Furthermore, it should be noted that the illustrations in the figures are not necessarily true to scale.

[0075] Figure 1 shows one embodiment of a projection exposure apparatus 1 (lithography apparatus), in particular an EUV lithography apparatus. One embodiment of an illumination system 2 of the projection exposure apparatus 1 has, in addition to a light source or radiation source 3, an illumination optical unit 4 for illuminating an object field 5 in an object plane 6. In an alternative embodiment, the light source 3 may also be provided as a module separate from the rest of the illumination system 2. In this case, the illumination system 2 does not comprise the light source 3.

[0076] A reticle 7 arranged in the object field 5 is exposed. The reticle 7 is held by a reticle holder 8. The reticle holder 8 is able to be displaced, in particular in a scanning direction, via a reticle displacement drive 9.

[0077] Figure 1 shows, for explanatory purposes, a Cartesian coordinate system with an x-direction x, a ydirection y and a z-direction z. The x-direction x runs perpendicular into the plane of the drawing. The ydirection y runs horizontally and the z- direction z runs vertically. The scanning direction in Figure 1 runs along the y direction y. The z-direction z runs perpendicular to the object plane 6.

[0078] The projection exposure apparatus 1 comprises a projection optical unit 10. The projection optical unit 10 is used to image the object field 5 into an image field 11 in an image plane 12. The image plane 12 runs parallel to the object plane 6. As an alternative, an angle other than 0° between the object plane 6 and the image plane 12 is also possible.

[0079] A structure on the reticle 7 is imaged onto a light-sensitive layer of a wafer 13 arranged in the region of the image field 11 in the image plane 12. The wafer 13 is held by a wafer holder 14. The wafer holder 14 is able to be displaced, in particular along the ydirection y, via a wafer displacement drive 15. The displacement of the reticle 7 via the reticle displacement drive 9, on the one hand, and of the wafer 13 via the wafer displacement drive 15, on the other hand, may take place in synchronicity with one another. The light source 3 is an EUV radiation source. The light source 3 emits in particular EUV radiation 16, which is also referred to hereinafter as used radiation, illumination radiation or illumination hght. The used radiation 16 in particular has a wavelength in the range between 5 nm and 30 nm. The light source 3 may be a plasma source, for example an LPP (laser-produced plasma) source, or a DPP (gas discharge-produced plasma) source. It may also be a synchrotron -based radiation source. The hght source 3 may be a free-electron laser (FEL).

[0080] The illumination radiation 16 emanating from the light source 3 is bundled by a collector 17. The collector 17 may be a collector having one or more ellipsoidal and / or hyperboloidal reflection surfaces. The at least one reflection surface of the collector 17 may be impinged upon by the illumination radiation 16 with grazing incidence (GI), that is to say with angles of incidence greater than 45°, or with normal incidence (Nl), that is to say with angles of incidence less than 45°. The collector 17 may be structured and / or coated, firstly to optimize its reflectivity for the used radiation and secondly to suppress extraneous hght.

[0081] Downstream of the collector 17, the illumination radiation 16 propagates through an intermediate focus in an intermediate focal plane 18. The intermediate focal plane 18 may constitute a separation between a radiation source module, comprising the light source 3 and the collector 17, and the illumination optical unit 4.

[0082] The illumination optical unit 4 comprises a deflection mirror 19 and a first facet mirror 20 arranged downstream thereof in the beam path. The deflection mirror 19 may be a plane deflection mirror or, alternatively, a mirror with a beam-influencing effect that goes beyond the purely deflecting effect. As an alternative or in addition, the deflection mirror 19 may be embodied as a spectral filter separating a used light wavelength of the illumination radiation 16 from extraneous hght having a wavelength that deviates therefrom. If the first facet mirror 20 is arranged in a plane of the illumination optical unit 4 that is optically conjugate to the object plane 6 as a field plane, this is also referred to as a field facet mirror. The first facet mirror 20 comprises a multiplicity of individual first facets 21, which may also be referred to as field facets. Only some of these first facets 21 are shown in Figure 1 by way of example.

[0083] The first facets 21 may be embodied as macroscopic facets, in particular as rectangular facets or as facets having an arcuate or partially circular edge contour. The first facets 21 may be embodied as plane facets or alternatively as facets with a convex or concave curvature.

[0084] As known for example from DE 10 2008 009 600 Al, the first facets 21 themselves may also be composed in each case of a multiplicity of individual mirrors, in particular a multiplicity of micromirrors. The first facet mirror 20 may in particular be designed as a microelectromechanical system (MEMS system). Reference may be made to DE 10 2008 009 600 Al for details.

[0085] The illumination radiation 16 runs horizontally, that is to say along the ydirec- tion y, between the collector 17 and the deflection mirror 19.

[0086] A second facet mirror 22 is arranged downstream of the first facet mirror 20 in the beam path of the illumination optical unit 4. If the second facet mirror 22 is arranged in a pupil plane of the illumination optical unit 4, it is also referred to as a pupil facet mirror. The second facet mirror 22 may also be arranged at a distance from a pupil plane of the illumination optical unit 4. In this case, the combination of the first facet mirror 20 and the second facet mirror 22 is also referred to as a specular reflector. Specular reflectors are known from US 2006 / 0132747 Al, EP 1 614 008 Bl and US 6,573,978. The second facet mirror 22 comprises a plurality of second facets 23. The second facets 23 are also referred to as pupil facets in the case of a pupil facet mirror.

[0087] The second facets 23 may likewise be macroscopic facets, which may for example have a round, rectangular or else hexagonal boundary, or alternatively be facets composed of micromirrors. In this respect, reference may likewise be made to DE 10 2008 009 600 Al.

[0088] The second facets 23 may have plane reflection surfaces or, alternatively, reflection surfaces with a convex or concave curvature.

[0089] The illumination optical unit 4 thus forms a double-faceted system. This basic principle is also referred to as a fly's eye integrator.

[0090] It may be advantageous to arrange the second facet mirror 22 not exactly in a plane that is optically conjugate to a pupil plane of the projection optical unit 10. In particular, the second facet mirror 22 may be arranged in a manner tilted with respect to a pupil plane of the projection optical unit 10, as is described for example in DE 10 2017 220 586 Al.

[0091] The individual first facets 21 are imaged into the object field 5 using the second facet mirror 22. The second facet mirror 22 is the last beam-shaping mirror or actually the last mirror for the illumination radiation 16 in the beam path upstream of the object field 5.

[0092] In a further embodiment (not illustrated) of the illumination optical unit 4, a transfer optical unit contributing in particular to the imaging of the first facets 21 into the object field 5 may be arranged in the beam path between the second facet mirror 22 and the object field 5. The transfer optical unit may have exactly one mirror or, alternatively, two or more mirrors, which are arranged in succession in the beam path of the illumination optical unit 4. The transfer optical unit may in particular comprise one or two normal-incidence mirrors (NI mirrors) and / or one or two grazing-incidence mirrors (GI mirrors).

[0093] In the embodiment shown in Figure 1, the illumination optical unit 4 has exactly three mirrors downstream of the collector 17, specifically the deflection mirror 19, the first facet mirror 20 and the second facet mirror 22.

[0094] In a further embodiment of the illumination optical unit 4, the deflection mirror 19 may also be omitted, and so the illumination optical unit 4 may then have exactly two mirrors downstream of the collector 17, specifically the first facet mirror 20 and the second facet mirror 22.

[0095] The imaging of the first facets 21, by way of the second facets 23 or with the second facets 23 and a transfer optical unit, into the object plane 6 is often only approximate imaging.

[0096] The projection optical unit 10 comprises a plurality of mirrors Mi, which are numbered consecutively according to their arrangement in the beam path of the projection exposure apparatus 1.

[0097] In the example illustrated in Figure 1, the projection optical unit 10 comprises six mirrors Ml to M6. Alternatives with four, eight, ten, twelve or any other number of mirrors Mi are likewise possible. The projection optical unit 10 is a doubly obscured optical unit. The penultimate mirror M5 and the last mirror M6 each have a passage opening for the illumination radiation 16. The projection optical unit 10 has an image-side numerical aperture that is larger than 0.5 and that may also be larger than 0.6 and that may for example be 0.7 or 0.75. Reflection surfaces of the mirrors Mi may be in the form of freeform surfaces without an axis of rotational symmetry. As an alternative, the reflection surfaces of the mirrors Mi may be designed as aspherical surfaces having exactly one axis of rotational symmetry of the reflection surface shape. The mirrors Mi, just like the mirrors of the illumination optical unit 4, may have highly reflective coatings for the illumination radiation 16. These coatings may be designed as multilayer coatings, in particular having alternating layers of molybdenum and silicon.

[0098] The projection optical unit 10 has a large object-image offset in the ydirection y between a ycoordinate of a centre of the object field 5 and a ycoordinate of the centre of the image field 11. This object-image offset in the ydirection y may be approximately as large as a z-distance between the object plane 6 and the image plane 12.

[0099] The projection optical unit 10 may be designed in particular to be anamorphic. In particular, it has different imaging scales Bx, By in the x-direction x and the ydirection y. The two imaging scales Bx, By of the projection optical unit 10 are preferably (Bx, By) = (+ / ■ 0.25, + / - 0.125). A positive imaging scale B means imaging without image inversion. A negative mathematical sign for the imaging scale B means imaging with image inversion.

[0100] The projection optical unit 10 thus leads, in the x-direction x, that is to say in the direction perpendicular to the scanning direction, to a reduction with a ratio of 4:1.

[0101] The projection optical unit 10 leads, in the ydirection y, that is to say in the scanning direction, to a reduction of 8A. Other imaging scales are likewise possible. Absolutely identical imaging scales, having the same mathematical sign, in the x-direction x and the ydirection y, for example with absolute values of 0.125 or 0.25, are also possible.

[0102] The number of intermediate image planes in the x-direction x and the ydirection y in the beam path between the object field 5 and the image field 11 may be the same or may, depending on the embodiment of the projection optical unit 10, be different. Examples of projection optical units with different numbers of such intermediate images in the x-direction x and ydirection y are known from US 2018 / 0074303 Al.

[0103] In each case, one of the second facets 23 is assigned to exactly one of the first facets 21 for forming in each case an illumination channel for illuminating the object field 5. This may in particular result in illumination according to Kohler's principle. The far field is broken down into a large number of object fields 5 using the first facets 21. The first facets 21 generate a plurality of images of the intermediate focus on the second facets 23 respectively assigned thereto.

[0104] The first facets 21 are each imaged by an assigned second facet 23 in superposition so as to illuminate the object field 5 onto the reticle 7. The illumination of the object field 5 is in particular as homogeneous as possible. It preferably has a uniformity error of less than 2%. Field uniformity may be attained by overlaying different illumination channels.

[0105] Arranging the second facets 23 makes it possible to geometrically define the illumination of the entrance pupil of the projection optical unit 10. Selecting the illumination channels, in particular the subset of the second facets 23 that carry light, makes it possible to set the intensity distribution in the entrance pupil of the projection optical unit 10. This intensity distribution is also referred to as an illumination setting or illumination pupil filling. A likewise preferred pupil uniformity in the region of portions of an illumination pupil of the illumination optical unit 4 that are illuminated in a defined manner may be achieved by redistributing the illumination channels.

[0106] Further aspects and details of the illumination of the object field 5 and in particular of the entrance pupil of the projection optical unit 10 are described hereinafter.

[0107] The projection optical unit 10 may have a homocentric entrance pupil in particular. It may be accessible. It may also be inaccessible.

[0108] The entrance pupil of the projection optical unit 10 often cannot be illuminated accurately with the second facet mirror 22. In the case of imaging of the projection optical unit 10 that images the centre of the second facet mirror 22 telecen- trically onto the wafer 13, the aperture rays often do not intersect at a single point. However, it is possible to find an area in which the distance between the aperture beams, determined in pairs, becomes minimal. This area represents the entrance pupil or a surface conjugate thereto in the space domain. In particular, this area exhibits a finite curvature.

[0109] It may be the case that the projection optical unit 10 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 transfer optical unit, should be provided between the second facet mirror 22 and the reticle 7. With the aid of this optical element, the different positions of the tangential entrance pupil and the sagittal entrance pupil may be taken into account.

[0110] In the arrangement of the components of the illumination optical unit 4 illustrated in Figure 1, the second facet mirror 22 is arranged in an area conjugate to the entrance pupil of the projection optical unit 10. The first facet mirror 20 is arranged in a manner tilted with respect to the object plane 6. The first facet mirror 20 is arranged in a manner tilted with respect to an arrangement plane defined by the deflection mirror 19. The first facet mirror 20 is arranged in a manner tilted with respect to an arrangement plane defined by the second facet mirror 22.

[0111] Figure 2 shows a schematic flowchart of one exemplary embodiment of the present method for producing a lithography apparatus 300 or one of its components 302 and / or for adjusting an apparatus parameter of the lithography apparatus 300. The method is preferably computer-implemented, that is to say able to be executed at least partially on a computer. The ageing effect may include a radiation-related degradation of optical elements 100 and / or a thermal and / or mechanical ageing effect and / or an ageing effect due to material deposition and / or corrosion and / or chemical degradation.

[0112] A step Si comprises generating N copies 304 of a neural network 306 for predicting an ageing effect of optical elements 100 (see for example Figure 1), in particular the mirrors M1-M6, of lithography apparatuses, wherein N is greater than 1 and preferably an integer. The optical elements may also be lens elements or sensors. The neural network 306 for predicting the ageing effect is preferably (pre)trained, prior to generating Si the N copies 304, on the basis of historical and / or current and / or synthetic and / or simulatively generated training data regarding the ageing effect. A layer structure and / or a layer-bylayer number of neurons of the neural network 306 is preferably selected and / or defined on the basis of the ageing effect to be predicted.

[0113] A step S2 comprises providing T training data records. The T training data records are generated by the following steps, each carried out for each respective lithography apparatus 300. The T training data records are generated by the following steps for i = 1 to T, wherein T is greater than or equal to N. A step S20 comprises reading parameters from an ith lithography apparatus 300. A step S21 comprises generating an ith training data record of the T training data records on the basis of the read parameters.

[0114] A step S3 comprises training each of the N copies. The training is done by carrying out the following steps for j = 1 to N. A step S30 comprises training the jth copy of the N copies 304 using the ith training data record of the T training data records. A step S31 comprises determining a jth item of change information 308 for the trained jth copy 304.

[0115] A step S4 comprises generating an anonymized item of change information 310 on the basis of at least two or any subset of the N items of change information 308. Generating S4 the anonymized item of change information 310 on the basis of at least two of the N items of change information 308 preferably comprises weighting the at least two of the N items of change information 308 and summing the weighted at least two of the N items of change information 308 to form the anonymized item of change information 310. The at least two of the N items of change information 308 are preferably weighted randomly or on the basis of a performance parameter distribution of the trained N copies 304 of the neural network 306. The jth items of change information 308 each preferably contain ageing-relevant information regarding apparatus -specific operating parameters and / or regarding an apparatus configuration and / or regarding an apparatus interface and / or regarding an apparatus history and / or regarding an apparatus brand and / or regarding an apparatus type.

[0116] A step S5 comprises adjusting the neural network 306 on the basis of the anonymized item of change information 310. A step S6 comprises predicting an ageing effect using the adjusted neural network 312.

[0117] A step S7 comprises producing a lithography apparatus or one of its components on the basis of the predicted ageing effect and / or adjusting an apparatus parameter of a lithography apparatus on the basis of the predicted ageing effect. The production S7 may comprise retrofitting at least one component 302 of a lithography apparatus 300 on the basis of the predicted ageing effect in order to minimize the ageing effect. As an alternative or in addition, the adjustment S7 may comprise adjusting an apparatus parameter of a lithography apparatus 300 incrementally or step-by-step on the basis of the predicted ageing effect in order to minimize the ageing effect. The minimization of the ageing effect is preferably based on solving an inverse problem related to the ageing effect and / or solving a feedback problem and / or solving an optimization problem.

[0118] Particularly preferably, method steps Si to S5 are carried out iteratively in order thereby to continuously improve a model performance of the neural network 306, 312 for predicting the ageing effect. This is indicated by a dashed return arrow in Figure 2.

[0119] Figure 3 shows a schematic block diagram of one exemplary embodiment of the present system 3000 for producing a lithography apparatus or one of its components and / or for adjusting an apparatus parameter of the lithography apparatus. The system is designed to carry out the present method.

[0120] The system 3000 has a preferably central computing device 3002. The system 3000 furthermore has N lithography apparatuses 1, 300, each having an evaluation device 3004. The system 3000 furthermore has a server or a cloud 3006. The computing device 3002 is designed to generate the N copies 304 of the neural network 306 for predicting an ageing effect of optical elements 100 of lithography apparatuses 1, 300, wherein N > 1, and to provide an ith copy 304 of the N copies 304 to the evaluation device 3004 of each ith lithography apparatus 1, 300.

[0121] The respective evaluation device 3004 of each of the ith lithography apparatus 1, 300 is designed to read parameters from the ith lithography apparatus 1, 300 and to generate an ith training data record of the T training data records on the basis of the read parameters, to train the ith copy 304 of the N copies 304 using the ith training data record of the T training data records, to determine the respective ith item of change information 308 for the trained ith copy 304, and to provide the determined ith item of change information 308, in particular unidirectionally, to the server or the cloud 3006 so as to provide a total of N items of change information 308.

[0122] The server or the cloud 3006 is designed to generate the anonymized item of change information 310 on the basis of at least two of the N items of change information 308, and to provide the anonymized item of change information 310 to the computing device 3002.

[0123] The computing device 3002 is designed to adjust or to retrain the neural network 306 on the basis of the anonymized item of change information 310, to predict an ageing effect using the adjusted neural network 306, and to provide a control instruction for producing a lithography apparatus 1, 300 or one of its components 302 on the basis of the predicted ageing effect and / or for adjusting an apparatus parameter of a lithography apparatus 1, 300 on the basis of the predicted ageing effect.

[0124] Although the present invention has been described on the basis of exemplary embodiments, it is modifiable in various ways. LIST OF REFERENCE SIGNS

[0125] 1 Lithography apparatus

[0126] 2 Illumination system

[0127] 3 Light source

[0128] 4 Illumination optical unit

[0129] 5 Object field

[0130] 6 Object plane

[0131] 7 Reticle

[0132] 8 Reticle holder

[0133] 9 Reticle displacement drive

[0134] 10 Projection optical unit

[0135] 11 Image field

[0136] 12 Image plane

[0137] 13 Wafer

[0138] 14 Wafer holder

[0139] 15 Wafer displacement drive

[0140] 16 Illumination radiation

[0141] 17 Collector

[0142] 18 Intermediate focal plane

[0143] 19 Deflection mirror

[0144] 20 First facet mirror

[0145] 21 First facet

[0146] 22 Second facet mirror

[0147] 23 Second facet

[0148] 100 Optical element

[0149] 300 Lithography apparatus

[0150] 302 Components

[0151] 304 Copy of a neural network

[0152] 306 Neural network 308 Item of change information

[0153] 310 Anonymized item of change information

[0154] 312 Adjusted neural network

[0155] 3000 System

[0156] 3002 Computing device

[0157] 3004 Evaluation device

[0158] 3006 Server or cloud

[0159] 51 Method step

[0160] 52 Method step

[0161] 53 Method step

[0162] 54 Method step

[0163] 55 Method step

[0164] 56 Method step

[0165] 57 Method step

[0166] 520 Method step

[0167] 521 Method step

[0168] 530 Method step

[0169] 531 Method step

[0170] Ml Mirror

[0171] M2 Mirror

[0172] M3 Mirror

[0173] M4 Mirror

[0174] M5 Mirror

[0175] M6 Mirror

Claims

PATENT CLAIMS1. Method for producing a lithography apparatus (1, 300) or one of its components (302) and / or for adjusting an apparatus parameter and / or for making a forecast regarding a failure of the lithography apparatus (1, 300), the method comprising: generating (Si) N copies (304) of a neural network (306) for predicting an ageing effect of optical elements (100) of hthography apparatuses (1, 300), wherein N > i; providing (S2) T training data records that are generated by the following steps, for i = 1 to T, wherein T > N: a) reading (S20) parameters from an ith lithography apparatus (1, 300), b) generating (S21) an ith training data record on the basis of the read parameters; training (S3) the N copies, comprising the following steps, for j = 1 to N: aa) training (S30) the jth copy using the ith training data record; bb) determining (S31) a jth item of change information for the trained jth copy (304); generating (S4) an anonymized item of change information (310) on the basis of at least two of the N items of change information (308); adjusting (S5) the neural network (306) on the basis of the anonymized item of change information; predicting (S6) an ageing effect using the adjusted neural network (312); and producing (S7) a hthography apparatus (1, 300) or one of its components (302) on the basis of the predicted ageing effect and / or adjusting an apparatus parameter and / or making a forecast regarding a failure of a hthography apparatus (1, 300) on the basis of the predicted ageing effect.

2. Method according to Claim 1, wherein the neural network (306) is pretrained to predict the ageing effect, prior to generating the N copies (304), on the basis of historical and / or current and / or synthetic and / or simulatively generated training data regarding the ageing effect.

3. Method according to Claim 1 or 2, wherein at least method steps Si to S5 are carried out iteratively in order thereby to improve a model performance of the neural network (306, 312) for predicting the ageing effect.

4. Method according to one of Claims 1 - 3, wherein generating (S4) the anonymized item of change information (310) on the basis of at least two of the N items of change information (308) comprises ■ weighting the at least two of the N items of change information (308); and summing the weighted at least two of the N items of change information (308) to form the anonymized item of change information (310).

5. Method according to Claim 4, wherein the at least two of the N items of change information (308) are weighted randomly or on the basis of a performance parameter distribution of the trained N copies (304) of the neural network (306).

6. Method according to one of Claims 1 - 5, wherein the ageing effect includes a radiation-related degradation of optical elements (100) and / or a thermal and / or mechanical ageing effect and / or an ageing effect due to material deposition and / or corrosion and / or chemical degradation.

7. Method according to one of Claims 1 - 6, wherein the optical elements (100) include lens elements and / or mirrors (M1-M6) and / or optical sensors.

8. Method according to Claim 1, wherein producing (S7) a lithography apparatus (1, 300) or one of its components (302) on the basis of the predicted ageingeffect and / or adjusting (S7) an apparatus parameter of a lithography apparatus (1,300) on the basis of the predicted ageing effect comprises: retrofitting at least one component (302) of a lithography apparatus (1, 300) on the basis of the predicted ageing effect in order to minimize the ageing effect; and / or adjusting an apparatus parameter of a lithography apparatus (1, 300) incrementally or step-or-step on the basis of the predicted ageing effect in order to minimize the ageing effect.

9. Method according to Claim 8, wherein the minimization of the ageing effect is based on solving an inverse problem related to the ageing effect and / or solving a feedback problem and / or solving an optimization problem.

10. Method according to one of Claims 1 - 9, wherein generating (S4) the anonymized item of change information (310) on the basis of at least two of the N items of change information (308) comprises in particular the respective local, unidirectional provision of the N items of change information (308) to a server or a cloud (3006).

11. Method according to one of Claims 1 - 10, wherein a layer structure and / or a layer-bylayer number of neurons of the neural network (306) is selected on the basis of the ageing effect.

12. Method according to one of Claims 1 ' 11, wherein the jth items of change information (308) each contain ageing-relevant information regarding apparatusspecific operating parameters and / or regarding an apparatus configuration and / or regarding an apparatus interface and / or regarding an apparatus history and / or regarding an apparatus brand and / or regarding an apparatus type.

13. System (3000) for producing a lithography apparatus (1, 300) or one of its components (302) and / or for adjusting an apparatus parameter of the lithography apparatus (1, 300), the system (3000) comprising: an in particular central computing device (3002), i hthography apparatuses (1, 300) each having an evaluation device (3004), a server or a cloud (3006), wherein the computing device (3002) is designed to generate N copies (304) of a neural network (306) for predicting an ageing effect of optical elements (100) of lithography apparatuses (1, 300), wherein N > 1, and to provide an ith copy (304) of the N copies (304) to the evaluation device (3004) of each ith lithography apparatus (1, 300), wherein the evaluation device (3004) of each of the ith lithography apparatus (1, 300) is designed to read parameters from the ith lithography apparatus (1, 300) and to generate an ith training data record on the basis of the read parameters, to train the ith copy (304) of the N copies (304) using the ith training data record, to determine an ith item of change information (308) for the trained ith copy (304), and to provide the determined ith item of change information (308) to the server or the cloud (3006) so as to provide a total of N items of change information (308), wherein the server or the cloud (3006) is designed to generate an anonymized item of change information (310) on the basis of at least two of the N items of change information (308), and to provide the anonymized item of change information (310) to the computing device (3002), wherein the computing device (3002) is designed to adjust the neural network (306) on the basis of the anonymized item of change information (310), to predict an ageing effect using the adjusted neural network (306), and to provide a control instruction for producing a hthography apparatus (1, 300) or one of its components (302) on the basis of the predicted ageing effect and / or for adjusting an apparatus parameter and / or for making a forecast regarding a failure of a lithography apparatus (1, 300) on the basis of the predicted ageing effect.

Images