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

Abstract

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

Description

[0001] The present invention relates to methods and systems for producing lithography equipment or components thereof and / or for adjusting equipment parameters and / or for predicting malfunctions of lithography equipment.

[0002] The contents of priority application DE 10 2024 201 726.4 are incorporated herein by reference in their entirety.

[0003] Microlithography is used to produce microstructured components, such as integrated circuits. The microlithography process is performed using a lithography apparatus equipped with an illumination system and a projection system. The image of the mask (etched lines) illuminated by the illumination system is projected, in this document, onto a substrate (e.g., a silicon wafer) coated with a photosensitive layer (photoresist) using the projection system, and arranged in the image plane of the projection system to transfer the mask structure to the photosensitive coating of the substrate.

[0004] Driven by the need for smaller structures in the fabrication of integrated circuits, EUV lithography equipment using light with wavelengths ranging from 0.1 nm to 30 nm (especially 13.5 nm) is currently being developed. Since most materials absorb light at this wavelength, it is necessary to use reflective optical units (i.e., mirrors) in such EUV lithography equipment instead of the previous refractive optical units (i.e., lens elements).

[0005] Due to the high radiation load within lithography equipment, optical components (such as mirrors or lenses) undergo various degradation processes during the equipment's lifespan. With the continuous increase in laser power and the emergence of new exposure systems, many new or unknown (degradation) effects and effect profiles are expected to appear in the future, which are currently unknown.

[0006] One example of degradation effects observable in many existing lithography devices is the degradation of optical components. For instance, the ultraviolet radiation used in immersion lithography alters the optical properties of the materials used in the optical components, such as refractive index, density, absorption coefficient, and birefringence. These materials include, for example, optical glass primarily used in i-line (365 nm) lithography equipment and synthetic quartz glass (fused silica), which are crucial components of the optical materials in systems operating at wavelengths of 248 nm (KrF) or 193 nm (ArF).

[0007] For 193 nm pulsed laser radiation used in immersion lithography, the refractive index change of synthetic quartz glass, which is commonly used as a lens material, is referred to as compression (increased refractive index) or sparsity (decreased refractive index), and is described, for example, in US10,427,965.

[0008] Especially with new, high-precision immersion lenses, the requirements for the wavefront are so high that high-order wavefront changes occurring during the lens's lifespan, even in the range of a few nanometers (and in some cases even below 1 nanometer), can render the lens unusable. To extend the lifespan, individual optical elements can be replaced.

[0009] Because the production of such optical replacement components is time-consuming and the shutdown of lithography equipment results in significant revenue losses, it is disadvantageous for customers to only begin production of such optical components when the lithography equipment exceeds its specifications, thus creating an urgent need for such optical replacement components.

[0010] Therefore, it is desirable to modify the optical properties of optical elements (especially projection lenses) in the most predictable way possible, so that such required optical replacement components can be manufactured in advance for replacement during lithography equipment maintenance, regardless of the schedule.

[0011] Furthermore, accurate prediction of the optical characteristics of optical components during planned replacement allows for optimization of optical replacement assemblies by applying individual aspherical surfaces, ensuring that the optical characteristics of specific optical components reach their optimal state after replacement.

[0012] Another advantage of predicting changes in the optical properties of optical components and understanding the impact of lithography equipment settings on these components is the ability to optimize customer user behavior. For example, alternative lighting settings can be directly suggested within the lithography equipment that maintain production while more evenly eroding the components, thus ensuring a longer lifespan.

[0013] The load on optical components, and consequently wear, is also affected by their optical design and other operating parameters of the photolithography equipment. These operating parameters include (but are not limited to) wavelength, number and / or duration of light pulses, magnitude of the field illuminating the photolithography mask, light intensity distribution in the pupil plane of the corresponding optical component, light polarization, and the transmittance and transmittance distribution of the photolithography mask.

[0014] Many of these operating parameters are actively set in the lithography equipment (e.g., the size of the field illuminating the mask, the light intensity distribution in the pupil plane, etc.) or defined by configuration (the duration of the laser pulse).

[0015] In this paper, the effects of various operating parameters on the optical components of the lithography equipment are generally known and can be described by simulation calculations. However, since the exact usage data of the machine is only known to the lithography equipment's customers or users, and is highly sensitive and confidential business data that is not expected to leave the customer's infrastructure, it is difficult to further study its known effects (requiring typically expensive test setups), as well as to improve its predictive accuracy and adapt to changing (operating and / or environmental) conditions.

[0016] Since simulation-based calculations of the degradation effects or wear of optical components are also highly computationally intensive, directly calculating the impact of operating parameters within the lithography equipment itself requires significant computing power and time.

[0017] In summary, it can be said that, at the current level of technology, optimal and / or controlled prediction of the degradation effects of radiation on optical components, and / or the influence of different operating parameters on such degradation effects, can be implemented to a very limited extent.

[0018] In this context, one object of the present invention is to provide an improved method and / or system for producing lithography equipment or one of its components and / or for adjusting the equipment parameters of lithography equipment, particularly taking into account the anonymization of equipment data to ensure customer and / or equipment anonymity.

[0019] Therefore, a method is proposed for manufacturing lithography equipment or one of its components, and / or for adjusting equipment parameters, and / or for predicting malfunctions of lithography equipment. This method includes the following steps: - Generate N copies (304) of the neural network to predict the degradation effects of optical elements in the lithography equipment, where N>1; T training data records are provided, which are generated through the following steps for i = 1 to T, where T ≥ N: a) Read the parameters from the i-th lithography device; b) Generate the i-th training data record based on the read parameters; - Train the N replicas by the following steps for j = 1 to N: aa) Use the i-th training data record to train the j-th copy; bb) Identify the j-th change information item of the j-th trained copy; - Generate an anonymous change information item based on at least two of the N change information items; - Adjust the neural network based on the anonymized change information item; - Predict degradation effects using a modified neural network; and alternatively: - Producing lithography equipment or one of its components based on predicted degradation effects and / or adjusting equipment parameters based on predicted degradation effects and / or making predictions about lithography equipment failures (especially predicted failures).

[0020] According to this method, multiple copies (N copies) of the neural network are generated. Preferably, the number N of generated copies of the neural network corresponds to the number of lithography devices, with one copy assigned to each lithography device. Furthermore, according to this method, at least T training data records are provided, where T should be at least as large as N. Preferably, training data records are generated on a device-specific basis for each of the lithography devices, or at least for a subset of the lithography devices. Then, preferably, each generated training data record is used to train the copy of the neural network assigned to the corresponding lithography device. Therefore, in this training step, the corresponding copy of the neural network is preferably trained locally on the corresponding lithography device. Since the neural network is preferably trained based on device-specific training data on each lithography device, this corresponding training does not require a large amount of computational power. Therefore, the evaluation device preferably used in and performing training or executing the neural network on the corresponding lithography device can be designed with low computational power. This reflects one of the many advantages of the joint learning method used herein. Specifically, on the lithography equipment, all training data is preferably retrievable, and thus the corresponding local execution copy of the neural network can access this training data for local retraining based on it, without the training data leaving the equipment operator's infrastructure. Therefore, retraining does not pose any security risks.

[0021] After training the corresponding neural network, a change information item is determined for each lithography device. This change information item specifically records one or more deviations between the network parameters and / or hyperparameters of the copy of the neural network trained on the corresponding device and the input neural network. In a further step, the device-specific change information is further processed to generate anonymized, and in particular summarized, change information items. Anonymization is performed to make it impossible to deduce the corresponding lithography device. This is advantageous because lithography devices often use process-specific and / or user-specific settings and / or process parameters that may constitute trade secrets. Specifically, if a predetermined number of change information items have been determined for a predetermined number of lithography devices, the corresponding change information items can be loaded onto a secure server or a secure cloud. The server or cloud is preferably inaccessible from the outside.

[0022] Then, the initially copied neural network is trained and optimized based on the anonymized change information. This allows for improved prediction accuracy when predicting degradation effects.

[0023] In other words, the global neural network is therefore updated based on anonymized change information items. Preferably, the degree of change or optimization of the neural network is determined based on the anonymized change information items. Therefore, uncontrolled deterioration of the neural network can be prevented even in the absence of direct insight into the device-specific change information items.

[0024] Particularly preferred is to evaluate the quality and / or reliability of the optimized neural network and / or simulation model by testing the neural network trained on anonymized change information items by comparing it with a simulation model used to simulate degradation effects.

[0025] If the updated neural network is optimized relative to the previous neural network in terms of performance data derived from comparisons, the updated neural network can be provided to the lithography equipment as a new version or software update. For this purpose, preferably, N copies of the updated neural network are provided and transmitted to the corresponding lithography equipment. On the corresponding lithography equipment, preferably, the copies of the updated neural network can be tested again by comparing them with equipment-specific simulation models or historical simulation data. Since testing or comparing the copies of the updated neural network with the corresponding simulation model requires relatively less computational power than training the neural network, the local performance of the copies of the neural network can be easily tested on the corresponding lithography equipment. This testing mechanism helps prevent uncontrolled degradation of the corresponding copies of the (updated) neural network (e.g., due to local overfitting) or helps identify local outliers.

[0026] In this context, the term "neural network" is synonymous with the machine learning model provided for predicting degradation effects.

[0027] The use of federated learning to predict or diagnose degradation effects on optical components, and the potential for improving new lithography equipment through federated learning, are not limited to degradation effects based on degradation itself, but are equally applicable to other degradation effects that may not yet be discovered. Therefore, the method described in this paper is based on federated learning, which comprehensively integrates historical and / or current domain knowledge regarding degradation effects on optical components. Federated learning enables the efficient execution or calculation of degradation effects on the relevant participating lithography equipment. Federated learning also enables the continuous optimization of the prediction accuracy of degradation effects on optical components of a given lithography equipment through knowledge transfer and / or knowledge acquisition from other lithography equipment, without disclosing or accessing the relevant equipment information. Therefore, this method can anonymize and improve the prediction accuracy of degradation effects on optical components of a given lithography equipment without removing critical information from the customer's infrastructure.

[0028] On the one hand, this method enables device-specific degradation effect monitoring, particularly by training and optimizing corresponding copies of the neural network based on device parameters. Furthermore, retraining the initially used neural network allows for global optimization of the neural network based on anonymized change information terms. This optimization enhances global system knowledge regarding degradation effects. Moreover, after retraining the neural network based on anonymized change information terms, it is preferably compared with simulation data from a degradation effect simulation. If the comparison reveals differences, system errors and / or syntax errors in the simulation model can preferably be detected and corrected as necessary. By continuously optimizing the neural network, preferably based on separately iteratively determined anonymized change information terms, the prediction accuracy of degradation effects can be improved. This knowledge can then be used to redesign the lithography equipment or equipment components, or to reset device parameters, particularly to minimize degradation effects. Another advantage of this method is that by locally training the corresponding copy of the neural network on each lithography equipment, the degradation effects of each lithography equipment can be directly predicted. This provides a rapid diagnostic path for the detection of degradation effects, leading to better actionability. This joint learning method is particularly capable of achieving reliable future predictions of degradation effects, and is preferably used in exchange pool prediction. For example, it can also improve strategic decisions regarding production and / or maintenance plans for each lithography unit, thereby increasing unit production rates and / or reducing maintenance-related downtime. Furthermore, application and / or setting recommendations can be provided to unit users through indirect analysis of user behavior and / or optimization of degradation effect predictions based on anonymized change information items. These application and / or setting recommendations can, for example, include instructing unit users to change unit parameters to improve lifespan. Using joint learning to locally train a copy of the neural network on the corresponding lithography unit enables the realization of direct “causal” links related to degradation effects, thereby efficiently calculating degradation effects.

[0029] This method also offers privacy and security advantages, as device users may not need to send confidential and / or sensitive data to device manufacturers to receive recommendations for device parameter optimization. This also helps prevent unnecessary data leaks or misuse. Device manufacturers also do not need to provide complex security architectures for data storage, as device operators preferably only receive anonymized change information items used to retrain neural networks. This also reduces the amount of data traffic generated. The information contained in the anonymized change information items is used "blindly" and cannot be manually viewed. This significantly reduces the risk of data misuse.

[0030] According to one embodiment, before generating N copies, the neural network is pre-trained based on training data generated from historical and / or current and / or synthetic and / or simulated data regarding degradation effects to predict degradation effects.

[0031] "Pre-training" refers to a process or methodological step in which a neural network (particularly an initially untrained neural network) is trained based on existing data and / or information about degradation effects to learn and / or improve its ability to predict those degradation effects. Training data (particularly initial training data, preferably used for pre-training) can be derived from various sources, including historical data (past data containing information about degradation effects), current data (current information about degradation effects), synthetic data (artificially generated data representing degradation effects), and simulation-generated data (data created through simulation and reflecting degradation effects). In other words, this paper uses existing knowledge about the degradation effect to be predicted to create the neural network. If real data from previous degradation measurements is available, the neural network can be (pre)trained based on this real data. This is particularly preferred because training based on real data allows the basic structure and / or hyperparameters of the neural network to adapt to the structure of the real data. Alternatively or additionally, the neural network can also be trained based on artificially or synthetically generated data or on simulation data, particularly to achieve faster network fusion, especially by reducing the loss function. In this example, a joint learning approach is used to retrain the neural network using current device-specific system knowledge, resulting in further improvements (especially continuous improvements) to the neural network initially trained in this way. On the one hand, existing real-world data can thus be used to train the initial neural network. On the other hand, device-specific system knowledge can be used even when real-world data is unavailable.

[0032] According to one embodiment, at least some of the steps of the method are performed iteratively to continuously and / or persistently improve the model performance of the neural network used to predict degradation effects.

[0033] "Iterative execution" means that the steps related to the performance of the neural network and its model used to predict degradation effects are repeated and progressively executed. Instead of training the neural network in a single step, the model is improved through a series of repeatedly executed steps. This allows for continuous improvement of the neural network. "Model performance of the neural network" preferably refers to the ability or accuracy of the neural network in predicting degradation effects. Model performance preferably describes the accuracy and effectiveness with which the neural network makes these predictions. In this context, "continuous improvement" preferably means that the goal of the iterative method steps is to continuously improve the performance of the neural network over time. Each iteration aims to achieve improvement so that the neural network is more accurate and / or efficient in predicting degradation effects.

[0034] According to one embodiment, generating anonymized change information items based on at least two of N change information items includes: Weight at least two of the N change information items; and At least two of the weighted N change information items are summed to form the anonymized change information item.

[0035] Based on the various change information items of the corresponding lithography equipment, a single anonymized change information item is generated using data processing, and the neural network is then retrained based on this anonymized change information item. All lithography equipment thus benefit from the improved neural network for predicting degradation effects without releasing or providing equipment-specific information. The anonymized change information item is preferably based on a weighted sum of at least two of the equipment-specific change information items. Preferably, the weighting is performed randomly or depending on the performance or performance distribution of the copies of the neural network. When summing, the weights preferably add up to 1. By aggregating the various weights of a large number of different lithography equipment and / or equipment types, it becomes impossible to subsequently infer the individual applications and / or lithography equipment. Therefore, the anonymized change information item no longer includes any security-critical or privacy-critical knowledge. Particularly preferred is that the anonymized change information item is generated only when change information or relevant data about different lithography equipment and / or lithography equipment types and / or equipment manufacturers exists. This is preferred because reliable anonymization of the change information items can be achieved by starting only with a specific number of different change information items. This also ensures that anonymized change information items cannot determine the equipment usage behavior of individual lithography devices, nor can they determine the equipment usage behavior of certain lithography device groups and / or users. For example, anonymized change information items should only be generated starting with a quantity of 10 lithography devices, especially starting with different device operators and / or device types, because reliable anonymization cannot be achieved before this.

[0036] The above description of change information items is provided by way of example, referring only to each lithography device or the total number of lithography devices (N change information items for N lithography devices). It should be understood that multiple change information items for a lithography device only occur when, for example, the components of the same network terminology. However, these components can also be considered separately. Therefore, multiple change information items for each lithography device are possible in other embodiments, but are not necessary for encryption or anonymization.

[0037] Each change information item preferably includes a large number of individual parameters (especially weighted sums and / or thresholds in the case of neural networks). These parameters are preferably optimized at each training step, particularly locally and thereby changed. Therefore, the change information item for each i-th machine preferably includes a large number of individual parameters. For each of these parameters, it is preferably encrypted or anonymized using individual weights, wherein the sum of each of the parameters across all change information items is preferably exactly 1.

[0038] According to one embodiment, the performance parameter distribution of N randomly or neurally-based trained replicas is used to weight at least two of the N change information items.

[0039] Preferably, a weighted sum is generated using multiple change information items from multiple lithography devices. Preferably, the number of device-specific change information items is randomly selected to avoid inferring the individual lithography devices early in the step of selecting the data used to generate the anonymized change information items. However, in the case of randomly selecting device-specific change information items, it is preferable to consider the corresponding performance of the corresponding copies of the neural network, thereby, for example, ignoring any change information from low-performance copies of the neural network or copies of the neural network that make poor predictions. For example, performance constraints can be considered. The performance distribution based on local change information items can also be considered.

[0040] According to one embodiment, the degradation effect includes radiation-related degradation and / or thermal and / or mechanical degradation effects of the optical element and / or degradation effects due to material deposition and / or corrosion and / or chemical degradation.

[0041] Radiation-related degradation specifically describes the process by which optical components (such as lenses or mirrors) in a lithography apparatus lose their efficiency over time due to continuous exposure to strong radiation (typically ultraviolet light or electron beams). This can occur in several ways. Firstly, the optical properties of the components may change, such as variations in the transmittance, reflectance, and / or refractive index of the component material. Secondly, radiation-related damage may occur on the surface of the components. For example, cracks and / or discoloration may appear on the surface. Thermal effects can also cause degradation due to prolonged exposure to high energy, which can lead to increased temperatures. This energy input can then alter the shape and / or alignment of the components. Mechanical wear on the components can also contribute to degradation. Specifically, moving parts in the lithography apparatus may wear down due to daily use, which can affect the alignment and / or focusing of the components. Material deposition can also cause degradation. In environments containing dust and / or other particles, these can accumulate on the components and affect their performance. These particles can also be generated by exhaust from components of the lithography apparatus due to thermal input and may increase over the lifespan of the apparatus. Corrosion and / or chemical degradation can also contribute to deterioration. Certain environmental conditions and / or chemical exposures can cause corrosion and / or other chemical damage to optical components.

[0042] According to one embodiment, the optical element includes a lens element and / or a mirror and / or an optical sensor.

[0043] Of course, lithography equipment may also have other optical elements not explicitly mentioned in this article. Therefore, the list provided herein should not be construed as limiting.

[0044] According to one embodiment, producing a lithography apparatus or one of its components based on predicted degradation effects and / or adjusting the apparatus parameters based on predicted degradation effects includes: Modify at least one component (302) of the lithography equipment based on the predicted degradation effect to minimize the degradation effect; and / or Based on the predicted degradation effect, the equipment parameters of the lithography equipment are adjusted incrementally or gradually to minimize the degradation effect.

[0045] Modifying at least one component of a lithography apparatus based on predicted degradation effects preferably aims to modify or replace existing components of the lithography apparatus to reduce the impact of degradation. Such modification may include updating or replacing components that may be affected by degradation effects, such as wear, degradation, or efficiency reduction, such as lens elements, mirrors, mechanical parts, or control components. This modification may be based on degradation effects predicted by neural networks and may be supplemented by further analysis, empirical data, and / or simulation-based predictions. Adjusting the apparatus parameters of the lithography apparatus incrementally or gradually based on predicted degradation effects includes, for example, fine-tuning the operating parameters of the apparatus to compensate for or minimize the effects of degradation. This may include incrementally or gradually adjusting parameters such as light intensity, focal length, alignment, or temperature control. Active adjustments may also be made, preferably based on predictions of degradation effects, to maintain apparatus performance and minimize potential downtime.

[0046] By more accurately predicting the causes and / or types of component performance loss (e.g., due to degradation), in addition to setting various known parameters or settings of the components (e.g., optimal rotation and / or mirroring), it is conceivable that new and improved illumination settings can be generated through these new findings. On the one hand, these can optimize or improve the performance of lithography equipment including degraded components, which is synonymous with troubleshooting. On the other hand, this can extend the lifespan of the lithography equipment and / or components in question. Furthermore, scenarios that may lead to a reduction in light source intensity, for example, can be easily and accurately calculated, positively impacting the lifespan of individual components. More accurate predictions can also indicate when specific components absolutely require maintenance and / or replacement, thus avoiding downtime of the lithography equipment before replacement parts arrive.

[0047] The predictions made can also be used to simulate potential new (degradation) effects of components. This allows for the generation of better benefit assessments and / or improved target specifications. It also enables early-stage planning of the required capacity for individual components in the production schedule and which components should be replaceable. Furthermore, the predictions can be used to enable component specifications and / or new usage settings. For example, the predictions made in this paper can rotate the lighting settings of components, which can also be directly tested.

[0048] According to one embodiment, minimizing the degradation effect is based on solving the inverse problem associated with the degradation effect and / or solving the feedback problem and / or solving the optimization problem.

[0049] The phrase “solving the inverse problem related to degradation effects” is preferably understood to mean identifying the cause of the predicted degradation effect. Here, this means inferring the underlying cause or mechanism from the predicted degradation effect. For example, this can be achieved by analyzing changes in optical properties (such as a decrease in transparency or a change in refractive index) to identify specific degradation processes. The phrase “solving the feedback problem” is understood to mean adjusting the lithography equipment or one of its components, or equipment parameters, to achieve a desired state or performance. For example, equipment parameters or components can be adjusted to achieve the desired performance even in the presence of degradation effects. The phrase “solving the optimization problem” describes how to find the optimal solution under given constraints (especially considering multiple competing factors). In the present context, this preferably means finding the optimal operating conditions or component configuration to minimize the negative effects of degradation, particularly considering factors such as availability and performance efficiency.

[0050] Furthermore, it is conceivable to generate certain component improvements based on simulation parameters adjusted for degradation effects observed in the simulation. Additionally, exchange possibilities and / or redundancies can be simulated based on degradation effects, added to implementation costs, and then the total cost associated with the neural network can be simulated, thus making decisions about the type of future development. In other words, optical properties can be determined based on predicted degradation effects to counteract them. These properties can be used as input to a manipulator to, for example, at least partially automate the simulation design of new components.

[0051] According to one embodiment, generating anonymized change information items based on at least two of N change information items specifically includes providing the N change information items locally and unidirectionally to a server or the cloud.

[0052] Anonymized change information items are preferably generated on a server or in the cloud. Preferably, such a server or cloud is protected from external access by appropriate security measures. For example, the server or cloud may have only an intranet port, thus isolating it from access via the Internet.

[0053] According to one embodiment, the layer structure and / or the number of neurons per layer of the neural network are selected based on the degradation effect.

[0054] In this example, the layer structure of the neural network preferably refers to the arrangement and structure of the layers in the neural network, including the number of layers and the types of these layers (e.g., convolutional layers, pooling layers, fully connected layers). The layer structure is preferably chosen to be particularly suitable for identifying, analyzing, and managing degradation effects. The selection of the number of neurons layer by layer preferably refers to adjusting the number of neurons in each layer of the neural network. The number and distribution of neurons in each layer can have a significant impact on the network's performance and specifications. The number of neurons is preferably selected based on the specific requirements of the degradation effect. For example, a network designed to detect fine, complex degradation patterns may require a large number of neurons in some layers. In terms of both layer structure and the number of neurons, the neural network is configured based on the specific characteristics and requirements of the degradation effect to be detected. This means that the network is trained and / or configured to effectively respond to specific patterns, pointers, or results of degradation effects.

[0055] A large amount of relevant measurement data from the optical units of the lithography equipment, particularly wavefront measurements describing the aberrations of the optical system or components, is required. To predict future data, it is preferable to incorporate the usage type of the lithography equipment (which may be in time-series form). For example, a convolutional neural network (CNN) with additional input layers can be integrated into the fully connected intermediate layers of the neural network. Therefore, it is preferable to incorporate a large amount of relevant usage data from the lithography equipment (especially including historical data). For example, the following usage data can be incorporated: • Setup of the lithography equipment (especially the lighting); • Line engraving transmission; • Pulse Design • Source / laser power • Materials of each component • Machine type • Source Environment These data are preferably used as feature inputs and incorporated into a fully connected intermediate layer. Preferably, the wavefront (Zernike) image to be considered is incorporated as input into the convolutional neural network.

[0056] Preferably, the size of the neural network is flexible, as is the size of the layer structure. A simple convolutional network with three layers is conceivable, where the data is used as additional input data and is incorporated as features after layer 1. Depending on the desired complexity, a simple network with remaining blocks is also conceivable, where the conditions are also installed within the remaining blocks.

[0057] However, in any case, a suitable network will contain more than 2,000 trainable parameters, which will enable efficient encryption.

[0058] According to one embodiment, each of the j-th change information items includes deterioration-related information about device-specific operating parameters, and / or device configuration, and / or device interface, and / or device history and / or device brand and / or device type.

[0059] A system for producing lithography equipment or one of its components and / or for adjusting equipment parameters and / or for predicting malfunctions of lithography equipment is also proposed. The system includes: Especially the central computing device; i lithography devices, each with its own evaluation unit; Server or cloud; The central computing unit is designed to generate N copies of a neural network for predicting the degradation effects of optical elements in a lithography device, where N>1; and to provide the i-th copy of the N copies to an evaluation unit for each i-th lithography device. The evaluation device for each of the i-th lithography equipment is designed to read parameters from the i-th lithography equipment; generate the i-th training data record based on the read parameters; use the i-th training data record to train the i-th copy among N copies; determine the i-th change information item of the trained i-th copy; and provide the determined i-th change information item to a server or cloud to provide a total of N change information items. The server or cloud is designed to generate anonymized change information items based on at least two of N change information items, and to provide the anonymized change information items to the computing device. The computing device is designed to adjust a neural network based on anonymized change information items; predict degradation effects using the adjusted neural network; and provide control instructions for producing one of the lithography equipment or its components based on the predicted degradation effects and / or adjusting equipment parameters based on the predicted degradation effects and / or making predictions about lithography equipment failures.

[0060] For example, degradation effects can be determined locally in the evaluation apparatus and used to output individual instructions, particularly those concerning individual lithography devices. Degradation effects can also be determined in a computing device and used to generate global updates to the lithography device software based on changing parameters. Global changes can also be made to the device or product design, particularly by improving the simulation capabilities of individual components and comparing them with other designs.

[0061] The lithography apparatus preferably includes a projection optics unit. The lithography apparatus may also have an illumination system. The lithography apparatus or projection exposure apparatus may be an EUV lithography apparatus. EUV stands for "Extreme Ultraviolet" and indicates that the operating light wavelength is between 0.1 nm and 30 nm. The projection exposure apparatus may also be a DUV lithography apparatus. DUV stands for "Deep Ultraviolet" and indicates that the operating light wavelength is between 30 nm and 250 nm.

[0062] In this context, "one" should not be construed as limited to a single element. Rather, multiple elements may be provided, such as two, three, or more. Any other numerical values ​​used herein should also not be construed as limited to the exact number of the components. Rather, unless otherwise stated, upward and downward numerical deviations are possible.

[0063] The embodiments and features described in relation to this method, with necessary modifications, are applicable to the proposed system, and vice versa.

[0064] Other possible embodiments of the present invention also include combinations of features or embodiments not explicitly mentioned in the foregoing or following description of exemplary embodiments. In such cases, those skilled in the art will also add various aspects as improvements or supplements to the corresponding basic form of the invention.

[0065] Further advantageous configurations and aspects of the invention are the subject of the dependent claims and also the subject of exemplary embodiments of the invention described below. The invention will now be described in detail based on preferred embodiments with reference to the accompanying drawings.

[0066] Figure 1 A schematic meridional section of the projection exposure apparatus for EUV projection lithography is shown; Figure 2 A schematic flowchart illustrating an exemplary embodiment of a method for producing a lithography apparatus or a component thereof and / or for adjusting apparatus parameters of a lithography apparatus; and Figure 3 A schematic diagram of an exemplary embodiment of a system for producing a lithography apparatus or one of its components and / or for adjusting the apparatus parameters of a lithography apparatus is shown.

[0067] Unless otherwise specified, identical or functionally equivalent elements have the same reference numerals in the drawings. Furthermore, it should be noted that the drawings are not necessarily to scale.

[0068] Figure 1An embodiment of a projection exposure apparatus 1 (lithography apparatus), particularly an EUV lithography apparatus, is shown. One embodiment of the illumination system 2 of the projection exposure apparatus 1 includes, in addition to a light source or radiation source 3, an illumination optics 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 include the light source 3.

[0069] The lithography 7, arranged in the object field 5, is exposed. The lithography 7 is carried by the lithography carrier 8. The lithography carrier 8 can be displaced by the lithography displacement driver 9, specifically displaced along the scanning direction.

[0070] For ease of explanation, Figure 1 A Cartesian coordinate system with x-direction x, y-direction y, and z-direction z is shown. The x-direction x is perpendicular to the drawing plane. The y-direction y extends horizontally, and the z-direction z extends vertically. Figure 1 The scanning direction extends along the y-direction (y). The z-direction (z) is perpendicular to the object plane (6).

[0071] The projection exposure apparatus 1 includes a projection optics unit 10. The projection optics unit 10 is used to image the object field 5 onto an image field 11 in an image plane 12. The image plane 12 extends parallel to the object plane 6. Alternatively, the angle between the object plane 6 and the image plane 12 may be an angle other than 0°.

[0072] The structure on the photolithography 7 is imaged onto the photosensitive layer of wafer 13, which is positioned within the region of image field 11 in image plane 12. Wafer 13 is supported by wafer carrier 14. Wafer carrier 14 can be displaced via wafer displacement driver 15, specifically along the y-direction. The displacement of photolithography 7 via photolithography displacement driver 9 and the displacement of wafer 13 via wafer displacement driver 15 can occur synchronously with each other.

[0073] Light source 3 is an EUV radiation source. Light source 3 emits, specifically, EUV radiation 16, hereinafter also referred to as working radiation, illumination radiation, or illuminating light. Specifically, the wavelength of working radiation 16 is in the range of 5 nm to 30 nm. Light source 3 can be a plasma source, such as an LPP (laser-generated plasma) source or a DPP (dual-phase plasma) source. The light source can also be a synchrotron radiation source. Light source 3 can be a free-electron laser (FEL).

[0074] Illumination radiation 16 emitted from light source 3 is focused by concentrator 17. Concentrator 17 may be a concentrator having one or more elliptical and / or hyperboloidal reflective surfaces. At least one reflective surface of concentrator 17 may be illuminated by illumination radiation 16 with grazing incidence (GI) (i.e., an incident angle greater than 45°) or normal incidence (NI) (i.e., an incident angle less than 45°). Concentrator 17 may be structured and / or coated, primarily to optimize its reflectivity to the radiation used, and secondarily to suppress intrusive light.

[0075] Downstream of the concentrator 17, the illumination radiation 16 propagates through the intermediate focal point on the intermediate focal plane 18. The intermediate focal plane 18 can constitute a separation between the radiation source module (including the light source 3 and the concentrator 17) and the illumination optical unit 4.

[0076] The illumination optics unit 4 includes a deflector 19 and a first faceted mirror 20 disposed downstream of it in the beam path. The deflector 19 may be a planar deflector, or alternatively, a mirror with beam-affecting effects (beyond pure deflection). Alternatively or additionally, the deflector 19 may be embodied as a spectral filter that separates the wavelength of the illumination radiation 16 used from external light with wavelength deviations. If the first faceted mirror 20 is disposed in the plane of the illumination optics unit 4 that is optically conjugate to the object plane 6, which serves as the field plane, this is also referred to as a field faceted mirror. The first faceted mirror 20 includes a plurality of individual first facets 21, which may also be referred to as field facets. Figure 1 Some of these first facets 21 are shown only as examples.

[0077] The first facet 21 can be represented as a macroscopic facet, particularly a rectangular facet, or a facet with an arc-shaped or partially circular edge profile. The first facet 21 can be represented as a planar facet, or alternatively as a facet with convex or concave curvature.

[0078] For example, as can be seen from DE 10 2008 009 600 A1, the first facet 21 itself can also be composed of multiple individual mirrors (especially multiple micromirrors) in each case. The first facet mirror 20 can be specifically designed as a microelectromechanical system (MEMS system). For details, please refer to the reference DE 10 2008 009 600 A1.

[0079] The illumination radiation 16 extends horizontally between the condenser 17 and the deflector 19, that is, it extends along the y-direction.

[0080] The second faceted mirror 22 is arranged downstream of the first faceted mirror 20 in the beam path of the illumination optical unit 4. If the second faceted mirror 22 is arranged in the pupil plane of the illumination optical unit 4, it is also called a pupil faceted mirror. The second faceted mirror 22 can also be arranged at a certain distance from the pupil plane of the illumination optical unit 4. In this case, the combination of the first faceted mirror 20 and the second faceted mirror 22 is also called a specular mirror. Specular mirrors are known from US2006 / 0132747 A1, EP 1 614 008 B1, and US 6,573,978.

[0081] The second faceted reflector 22 includes a plurality of second facets 23. In the case of a pupil faceted reflector, the second facets 23 are also referred to as pupil facets.

[0082] The second facet 23 can also be a macroscopic facet, for example, it can have circular, rectangular or hexagonal boundaries, or alternatively it can be a facet composed of multiple micromirrors. In this regard, see also DE 10 2008 009 600A1.

[0083] The second facet 23 may have a planar reflective surface, or alternatively, a reflective surface with convex or concave curvature.

[0084] Therefore, the illumination optical unit 4 forms a biplane system. This basic principle is also known as a fly-eye integrator.

[0085] It may be advantageous to arrange the second faceted mirror 22 imprecisely on a plane that forms an optical conjugate with the pupil plane of the projection optics unit 10. Specifically, the second faceted mirror 22 may be arranged in an oblique manner relative to the pupil plane of the projection optics unit 10, for example as described in DE 10 2017 220 586 A1.

[0086] The second faceted mirror 22 is used to image each of the first facets 21 into the object field 5. The second faceted mirror 22 is the last beam-shaping mirror in the upstream beam path of the object field 5, or in fact the last mirror of the illumination radiation 16.

[0087] In another embodiment (not shown) of the illumination optics unit 4, a transmission optics unit, particularly helpful in imaging the first facet 21 into the object field 5, can be arranged in the beam path between the second facet mirror 22 and the object field 5. The transmission optics unit may have exactly one mirror, or alternatively, two or more mirrors arranged consecutively in the beam path of the illumination optics unit 4. The transmission optics unit may specifically include one or two normal incident mirrors (NI mirrors) and / or one or two grazing incident mirrors (GI mirrors).

[0088] exist Figure 1 In the embodiment shown, the illumination optical unit 4 has three mirrors downstream of the condenser 17, specifically a deflector 19, a first faceted mirror 20, and a second faceted mirror 22.

[0089] In another embodiment of the illumination optical unit 4, the deflector 19 may also be omitted, and thus the illumination optical unit 4 may have exactly two mirrors downstream of the condenser 17, specifically a first faceted mirror 20 and a second faceted mirror 22.

[0090] The first plane 21 is typically only an approximate image when it is imaged onto the object plane 6 via the second plane 23 or by using the second plane 23 and the transmission optical unit.

[0091] The projection optics unit 10 includes a plurality of mirrors Mi, which are sequentially numbered according to their arrangement in the beam path of the projection exposure device 1.

[0092] exist Figure 1 In the example shown, the projection optics unit 10 includes six mirrors M1 to M6. Alternatives with four, eight, ten, twelve, or any other number of mirrors M1 are also possible. The projection optics unit 10 is a double-shielded optics unit. The penultimate mirror M5 and the last mirror M6 each have a channel opening for illumination radiation 16. The image-side numerical aperture of the projection optics unit 10 is greater than 0.5, and may also be greater than 0.6, and may, for example, be 0.7 or 0.75.

[0093] The reflecting surface of mirror Mi can be a freeform surface without an axis of rotational symmetry. Alternatively, the reflecting surface of mirror Mi can be designed as an aspherical surface with exactly one axis of rotational symmetry of its shape. Mirror Mi (like the mirror of illumination optics unit 4) can have a high-reflectivity coating for illumination radiation 16. These coatings can be designed as multilayer coatings, particularly with alternating layers of molybdenum and silicon.

[0094] The projection optical unit 10 has a large object-image offset y along the y-direction y between the y-coordinate of the center of the object field 5 and the y-coordinate of the center of the image field 11. The object-image offset along the y-direction y can be approximately as large as the z-distance between the object plane 6 and the image plane 12.

[0095] The projection optics unit 10 can be specifically designed to be deformable. Specifically, it has different imaging ratios βx and βy along the x-direction and the y-direction. The two imaging ratios βx and βy of the projection optics unit 10 are preferably (βx, βy) = (+ / - 0.25, + / - 0.125). A positive imaging ratio β means imaging without image inversion. A negative imaging ratio β means imaging with image inversion.

[0096] Therefore, the projection optical unit 10 results in a reduction of 4:1 in the x-direction (that is, in the direction perpendicular to the scanning direction).

[0097] The projection optical unit 10 results in a reduction ratio of 8:1 in the y-direction (that is, in the scanning direction).

[0098] Other imaging scales are also possible. Absolutely identical imaging scales with the same mathematical sign are also possible in the x-direction (x) and y-direction (y), for example, with absolute values ​​of 0.125 or 0.25.

[0099] The number of intermediate image planes in the x-direction (x) and y-direction (y) of the beam path between the object field 5 and the image field 11 may be the same or different depending on the embodiment of the projection optics unit 10. Examples of projection optics units with different numbers of such intermediate images in the x-direction (x) and y-direction (y) are known from US 2018 / 0074303 A1.

[0100] In each case, one of the second facets 23 is assigned to exactly one of the first facets 21 to form an illumination channel for illuminating the object field 5 in each case. This may specifically result in illumination according to Köhler's principle. The far field is decomposed into a large number of object fields 5 using the first facets 21. The first facets 21 generate multiple images with intermediate focal points on the corresponding second facets 23 assigned to them.

[0101] The first facet 21 is freely assigned to the second facet 23, which are then superimposed to form an image to illuminate the object field 5 onto the photolithography 7. The illumination of the object field 5 should be as uniform as possible, preferably with a uniformity error of less than 2%. Field uniformity can be obtained by superimposing different illumination channels.

[0102] The second facet 23 is arranged to geometrically define the illumination of the entrance pupil of the projection optical unit 10. An illumination channel (specifically a subset of the light-carrying second facet 23) is selected such that the intensity distribution of the entrance pupil of the projection optical unit 10 can be configured. This intensity distribution is also referred to as the illumination setting or illumination pupil filling.

[0103] The same preferred pupil uniformity in the region of the illumination pupil of the illumination optical unit 4 illuminated in a defined manner can be achieved by reallocating the illumination channel.

[0104] The following describes further aspects and details of the illumination of the object field 5, particularly the illumination of the entrance pupil of the projection optical unit 10.

[0105] The projection optics unit 10 may specifically have a concentric entrance pupil. This concentric entrance pupil can be accessible. This concentric entrance pupil can also be inaccessible.

[0106] The entrance pupil of the projection optics unit 10 is typically not accurately illuminated by the second faceted mirror 22. In the case of imaging the projection optics unit 10 onto the wafer 13 with the center of the second faceted mirror 22 telecentrically projected onto it, the aperture rays typically do not intersect at a single point. However, a region can be found where the distance between paired, defined aperture beams becomes minimal. This region represents the entrance pupil of the spatial domain or its conjugate surface. Specifically, this region exhibits a finite curvature.

[0107] It is possible that the projection optics unit 10 has different positions for the tangential and sagittal beam paths. In this case, the imaging element (especially the optical element of the transmission optics unit) should be positioned between the second faceted mirror 22 and the photolithography 7. With the aid of this optical element, the different positions of the tangential and sagittal entrance pupils can be taken into account.

[0108] exist Figure 1 In the arrangement of the components of the illumination optical unit 4 shown, the second faceted mirror 22 is arranged in the region conjugate with the entrance pupil of the projection optical unit 10. The first faceted mirror 20 is arranged in an inclined manner relative to the object plane 6. The first faceted mirror 20 is arranged in an inclined manner relative to the arrangement plane defined by the deflection mirror 19. The first faceted mirror 20 is arranged in an inclined manner relative to the arrangement plane defined by the second faceted mirror 22.

[0109] Figure 2 A schematic flowchart illustrating an exemplary embodiment of a method for producing a lithography apparatus 300 or one of its components 302 and / or for adjusting apparatus parameters of the lithography apparatus 300. The method is preferably computer-implemented, that is, capable of being executed at least partially on a computer. Degradation effects may include radiation-related degradation and / or thermal degradation and / or mechanical degradation effects of the optical element 100 and / or degradation effects due to material deposition and / or corrosion and / or chemical degradation.

[0110] Step S1 includes generating N copies 304 of the neural network 306 for predicting the optical elements 100 of the lithography apparatus (e.g., see...). Figure 1 The degradation effect (especially of mirrors M1-M6) is considered, where N is greater than 1 and preferably an integer. Optical elements may also be lens elements or sensors. Preferably, before generating (S1)N copies 304, a neural network 306 for predicting the degradation effect is (pre)trained based on historical and / or current and / or synthetic and / or simulated training data regarding the degradation effect. The layer structure and / or the number of neurons per layer of the neural network 306 are preferably selected and / or defined based on the degradation effect to be predicted.

[0111] Step S2 includes providing T training data records. The T training data records are generated through the following steps, each step performed for a corresponding lithography device 300. The T training data records are generated by the following steps, where i = 1 to T, and T is greater than or equal to N. Step S20 includes reading parameters from the i-th lithography device 300. Step S21 includes generating the i-th training data record from the T training data records based on the read parameters.

[0112] Step S3 includes training each of the N replicas. Training is accomplished by performing the following steps for j = 1 to N. Step S30 includes training the j-th replica among the N replicas 304 using the i-th training data record from the T training data records. Step S31 includes determining the j-th change information item 308 of the trained j-th replica 304.

[0113] Step S4 includes generating an anonymized change information item 310 based on at least two or any subset of N change information items 308. Generating the anonymized change information item 310 based on at least two of the N change information items 308 preferably includes weighting at least two of the N change information items 308 and summing the weighted at least two of the N change information items 308 to form the anonymized change information item 310. The at least two of the N change information items 308 are weighted randomly or based on a distribution of performance parameters of N trained replicas 304 of a neural network 306. Each of the j-th change information items 308 preferably contains deterioration-related information regarding device-specific operating parameters, and / or device configuration, and / or device interface, and / or device history, and / or device brand, and / or device type.

[0114] Step S5 includes adjusting the neural network 306 based on the anonymized change information item 310.

[0115] Step S6 includes using a modified neural network 312 to predict degradation effects.

[0116] Step S7 includes manufacturing the lithography equipment or one of its components based on the predicted degradation effect and / or adjusting the equipment parameters of the lithography equipment based on the predicted degradation effect. Manufacturing S7 may include modifying at least one component of the lithography equipment 300 based on the predicted degradation effect to minimize the degradation effect. Alternatively or additionally, adjusting S7 may include incrementally or gradually adjusting the equipment parameters of the lithography equipment 300 based on the predicted degradation effect to minimize the degradation effect. Minimizing the degradation effect is based on solving inverse problems and / or feedback problems and / or optimization problems related to the degradation effect.

[0117] Particularly preferably, method steps S1 to S5 are performed iteratively to continuously improve the model performance of neural networks 306 and 312 used to predict degradation effects. This is achieved by... Figure 2 The dashed line indicates the return arrow.

[0118] Figure 3 This is a schematic block diagram illustrating an exemplary embodiment of a system 3000 for producing a lithography apparatus or one of its components and / or for adjusting apparatus parameters of a lithography apparatus. The system is designed to perform the methods shown.

[0119] System 3000 includes, in particular, a central computing unit 3002. System 3000 also includes N lithography devices 1, 300, each lithography device having an evaluation device 3004. System 3000 also includes a server or cloud 3006.

[0120] The computing device 3002 is designed to generate N copies 304 of the neural network 306 for predicting the degradation effect of the optical element 100 of the lithography equipment 1, 300, where N>1, and to provide the i-th copy 304 of the N copies 304 to the evaluation device 3004 of each i-th lithography equipment 1, 300.

[0121] The corresponding evaluation device 3004 of each of the i-th lithography equipment 1 and 300 is designed to read parameters from the i-th lithography equipment 1 and 300; generate the i-th training data record among T training data records based on the read parameters; use the i-th training data record among the T training data records to train the i-th copy 304 among N copies 304; determine the corresponding i-th change information item 308 of the trained i-th copy 304; and provide the determined i-th change information item 308 (particularly unidirectionally) to a server or cloud 3006 to provide a total of N change information items 308.

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

[0123] The computing device 3002 is designed to adjust or retrain the neural network 306 based on anonymized change information 310, use the adjusted neural network 306 to predict degradation effects; and provide control instructions for producing one of the lithography apparatus 1, 300 or its components 302 based on the predicted degradation effects and / or for adjusting the device parameters of the lithography apparatus 1, 300 based on the predicted degradation effects.

[0124] Although the invention has been described based on exemplary embodiments, modifications can be made in various ways.

[0125] List of reference numerals 1: Photolithography equipment 2: Lighting System 3: Light source 4: Illumination Optical Unit 5: Object Field 6: Object plane 7: Photolithography 8: Photolithography carrier 9: Photolithography displacement driver 10: Projection Optical Unit 11: Image Field 12: Image plane 13: Wafer 14: Wafer Carrier 15: Wafer displacement driver 16: Lighting radiation 17: Concentrator 18: Intermediate focal plane 19: Deflecting Mirror 20: First faceted mirror 21: First facet 22: Second faceted mirror 23: Second facet 100: Optical components 300: Photolithography equipment 302: Component 304: A copy of a neural network 306: Neural Networks 308: Change Information Item 310: Anonymization of change information items 312: Adjusted neural network 3000: System 3002: Computing device 3004: Evaluation device 3006: Server or cloud S1: Method and Steps S2: Method and Steps S3: Methods and Steps S4: Methods and Steps S5: Methods and Steps S6: Methods and Steps S7: Methods and Steps M1: Reflector M2: Reflector M3: Reflector M4: Reflector M5: Reflector M6: Reflector

Claims

1. A method for producing a lithography apparatus (1, 300) or one of its components (302), and / or a method for adjusting apparatus parameters, and / or for predicting failures of said lithography apparatus (1, 300), said method comprising the steps of: N copies (304) of a (S1) neural network (306) are generated to predict the degradation effect of the optical element (100) of the lithography device (1, 300), where N>1; Provide (S2)T training data records, which are generated by the following steps for i = 1 to T, where T ≥ N: a) Read (S20) the parameters from the i-th lithography device (1, 300); b) Generate the i-th training data record based on the read parameters (S21); Training (S3) the N replicas includes the following steps for j = 1 to N: aa) Use the i-th training data record to train (S30) the j-th copy; (bb) Determine (S31) the j-th change information item of the j-th trained copy (304); Anonymized change information item (310) is generated (S4) based on at least two of the N change information items (308); The neural network (306) is adjusted (S5) based on the anonymized change information item. The degradation effect (S6) is predicted using a modified neural network (312); and (S7) Produce lithography equipment (1, 300) or one of its components (302) based on the predicted degradation effect and / or adjust equipment parameters based on the predicted degradation effect and / or make predictions about failures of lithography equipment (1, 300).

2. The method of claim 1, wherein, prior to generating the N copies (304), the neural network (306) is pre-trained based on historical and / or current and / or synthetic and / or simulated training data regarding the degradation effect to predict the degradation effect.

3. The method of claim 1 or 2, wherein at least method steps S1 to S5 are performed iteratively to improve the model performance of the neural network (306, 312) for predicting the degradation effect.

4. The method of any one of claims 1-3, wherein generating (S4) the anonymized change information item (310) based on at least two of the N change information items (308) comprises: Weight at least two of the N change information items (308); as well as At least two of the weighted N change information items (308) are summed to form the anonymized change information item (310).

5. The method of claim 4, wherein at least two of the N change information items (308) are weighted randomly or based on the performance parameter distribution of the N trained replicas (304) of the neural network (306).

6. The method of any one of claims 1-5, wherein the degradation effect includes radiation-related degradation and / or thermal and / or mechanical degradation effect of the optical element (100) and / or degradation effect due to material deposition and / or corrosion and / or chemical degradation.

7. The method according to any one of claims 1-6, wherein the optical element (100) comprises a lens element and / or a reflector (M1-M6) and / or an optical sensor.

8. The method of claim 1, wherein producing (S7) the lithography apparatus (1, 300) or one of its components (302) based on the predicted degradation effect and / or adjusting the apparatus parameters of the (S7) lithography apparatus (1, 300) based on the predicted degradation effect includes: Modify at least one component (302) of the lithography equipment (1, 300) based on the predicted degradation effect to minimize the degradation effect; and / or Based on the predicted degradation effect, the equipment parameters of the lithography equipment (1, 300) are adjusted incrementally or gradually to minimize the degradation effect.

9. The method of claim 8, wherein minimizing the degradation effect is based on solving the inverse problem and / or the feedback problem and / or the optimization problem associated with the degradation effect.

10. The method of any one of claims 1-9, wherein generating (S4) the anonymized change information item (310) based on at least two of the N change information items (308) specifically includes providing the N change information items (308) locally and unidirectionally to a server or cloud (3006).

11. The method of any one of claims 1-10, wherein the layer structure and / or the number of neurons per layer of the neural network (306) are selected based on the degradation effect.

12. The method of any one of claims 1-11, wherein each of the j-th change information items includes deterioration-related information about device-specific operating parameters and / or device configuration and / or device interface and / or device history and / or device brand and / or device type.

13. A system (3000) for producing a lithography apparatus (1, 300) or one of its components (302) and / or for adjusting apparatus parameters of said lithography apparatus (1, 300), said system (3000) comprising: In particular, the central computing unit (3002); i lithography devices (1, 300), each having an evaluation device (3004); Server or cloud (3006); The central computing device (3002) is configured to generate N copies (304) of a neural network (306) for predicting the degradation effect of the optical element (100) of the lithography apparatus (1, 300), where N > 1; and to provide the evaluation device (3004) of the N copies (304) to each of the i-th lithography apparatus (1, 300). The evaluation device (3004) of each of the i-th lithography devices (1, 300) is designed to read parameters from the i-th lithography device (1, 300); generate an i-th training data record based on the read parameters; use the i-th training data record to train the i-th copy (304) among the N copies (304); determine the i-th change information item (308) of the trained i-th copy (304); and provide the determined i-th change information item (308) to the server or the cloud (3006) to provide a total of N change information items (308). The server or cloud (3006) is designed to generate an anonymized change information item (310) based on at least two of the N change information items (308), and to provide the anonymized change information item (310) to the computing device (3002). The computing device (3002) is designed to adjust the neural network (306) based on the anonymized change information item (310); predict degradation effects using the adjusted neural network (306); and provide control instructions for producing a lithography apparatus (1, 300) or one of its components (302) based on the predicted degradation effects and / or adjusting apparatus parameters and / or making predictions about failures of the lithography apparatus (1, 300) based on the predicted degradation effects.

Images