Method for determining corrections for controlling lithographic and / or metrology process and associated device

By using trained artificial intelligence models, especially regression tree or neural network models in lithography devices, to predict and compensate for perturbation parameters caused by the second component, the problem of strong responsiveness of feedback controllers in the prior art is solved. Feedforward correction is achieved, which significantly improves the control accuracy and production efficiency of lithography equipment.

CN118974664APending Publication Date: 2024-11-15ASML NETHERLANDS BV

Patent Information

Application Number
CN202380032059.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-04-14
Filing Date
2023-03-03
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The feedback controller in existing lithography equipment is highly reactive and requires errors to be used to correct it. There is a lack of feedforward control methods to reduce the impact of disturbing power caused by the second component, especially in the process of lithography or measurement. Accurate feedforward correction.

Method used

Using trained artificial intelligence models, especially based on regression tree or neural network models, the feed-forward correction of lithography equipment or measurement equipment is achieved to reduce electromagnetic disturbances by input data. The impact of

Benefits of technology

Through the prediction and compensation correction of artificial intelligence models, the impact of disturbed power on lithography equipment or measurement equipment is significantly reduced, the control accuracy and production efficiency are improved, and the impact of disturbed power on lithography equipment is reduced, such as in EUV light The disturbance power in the engraving machine can be reduced by 96% or more.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118974664A_ABST
    Figure CN118974664A_ABST
Patent Text Reader

Abstract

A method of determining a correction for controlling at least one first component of a lithographic or metrology apparatus and an associated apparatus are disclosed. The method comprises: obtaining a trained artificial intelligence model, the trained artificial intelligence model having been trained to map input data related to at least one second component of the lithographic apparatus or metrology apparatus to perturbation parameters related to perturbation of the at least one first component, the disturbance has been caused by the at least one second component; obtaining input data related to the at least one second component; and using the trained artificial intelligence model to determine a compensation correction for the disturbance parameter according to the input data.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to European application 22168402.0 filed on April 14, 2022, and the entire contents of said European application are incorporated herein by reference. Technical Field

[0003] The present invention relates to a method for manufacturing a device using photolithography. The present invention also relates to a computer program product for implementing such a method. Background Art

[0004] A lithographic apparatus is a machine configured to apply a desired pattern to a substrate. For example, a lithographic apparatus may be used in the manufacture of integrated circuits (ICs). A lithographic apparatus may, for example, project a pattern at a patterning device (e.g., a mask) onto a layer of radiation-sensitive material (resist) disposed on a substrate.

[0005] To project a pattern onto a substrate, a lithographic apparatus may use electromagnetic radiation. The wavelength of this radiation determines the minimum size of a feature that can be formed on the substrate. A lithographic apparatus using extreme ultraviolet (EUV) radiation (having a wavelength in the range of 4 nm to 20 nm, such as 6.7 nm or 13.5 nm) may be used to form smaller features on a substrate than a lithographic apparatus using, for example, radiation having a wavelength of 193 nm.

[0006] In order to achieve accuracy in positioning of the desired features, and thus minimize parameters such as overlay and edge placement errors, accurate and precise control of the stages (e.g., a reticle stage for holding a reticle or mask, and at least one wafer stage for holding a wafer or substrate) and / or beam actuators (e.g., beam steering mirrors) is very important. In many cases, such corrections are determined and applied in a feedback loop. However, feedback corrections rely on the presence of an error to be corrected in the first place; that is, a correction cannot be determined until an error has occurred.

[0007] It is desirable to improve methods for determining corrections for control within a lithography or metrology process, and in particular methods for providing feed-forward corrections. Summary of the invention

[0008] In a first aspect of the present invention, a method for determining a correction for controlling at least one first component of a lithography device or a measurement device is provided; the method comprises: obtaining a trained artificial intelligence model, which has been trained to map input data related to at least one second component of the lithography device or the measurement device to disturbance parameters related to a disturbance of the at least one first component, the disturbance having been caused by the at least one second component; obtaining input data related to the at least one second component; and using the trained artificial intelligence model to determine a compensating correction for the disturbance parameters based on the input data.

[0009] In a second aspect of the invention there is provided a computer program comprising program instructions operable when run on a suitable apparatus to perform the method of the first aspect, as well as a non-transitory computer program carrier and processing means comprising such a computer program.

[0010] In a third aspect of the present invention, a lithography tool for patterning a substrate or measuring a patterned structure on a substrate is provided; the lithography tool comprises: at least one first component; at least one second component; a processor; and a non-transitory computer program carrier, the non-transitory computer program carrier comprising a trained artificial intelligence model, the trained artificial intelligence model has been trained to map input data related to the at least one second component to disturbance parameters related to disturbances of the at least one first component, the disturbances having been caused by the at least one second component.

[0011] In the context of the third embodiment, a "lithography tool" may include any tool used in a lithography process. This may include a lithography exposure tool or scanner used to perform the exposure in the lithography process, or a metrology tool (such as a scatterometer or scanning electron microscope) used to monitor and / or control such a lithography process. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Embodiments of the invention will now be described, by way of example only, with reference to the accompanying schematic drawings, in which:

[0013] - Figure 1 A schematic diagram of a first lithographic apparatus is depicted;

[0014] - Figure 2 A schematic overview of a lithography unit is depicted;

[0015] - Figure 3 An EUV lithography system including a lithography apparatus and a radiation source is depicted;

[0016] - Figure 4depicts a schematic representation of a scatterometry device;

[0017] - Figure 5 is an exemplary flow chart of a method for determining perturbation parameters using a regression tree model;

[0018] - Figure 6 is a flow chart of a component control loop illustrating how electromagnetic perturbation forces may be obtained for training purposes;

[0019] - Figure 7 is an exemplary flow chart of a method for determining a perturbation parameter using an ensemble of regression tree models;

[0020] - Figure 8 is a flow chart of a component control loop that applies a corrective or compensatory force as may be determined using a regression tree according to an embodiment of the present invention; and

[0021] - Fig. 9 is an exemplary flow chart of a method for determining disturbance parameters using an artificial neural network model. DETAILED DESCRIPTION

[0022] In this document, the terms "radiation" and "beam" are used to cover all types of electromagnetic radiation, including ultraviolet radiation (e.g., having a wavelength of 365 nm, 248 nm, 193 nm, 157 nm or 126 nm) and EUV (extreme ultraviolet radiation, e.g., having a wavelength in the range of about 5 to about 100 nm).

[0023] In this document, the terms "radiation" and "beam" are used to cover all types of electromagnetic radiation, including ultraviolet radiation (e.g., having a wavelength of 365nm, 248nm, 193nm, 157nm or 126nm) and EUV radiation (extreme ultraviolet radiation, e.g., having a wavelength in the range of about 5nm to 100nm).

[0024] The terms "reticle", "mask" or "patterning device" as used in the present invention may be broadly interpreted to refer to a general patterning device that can be used to impart a patterned cross-section to an incident radiation beam that corresponds to the pattern to be produced in a target portion of the substrate. In this context, the term "light valve" may also be used. In addition to classical masks (transmissive or reflective, binary, phase-shifting, hybrid, etc.), other examples of such patterning devices include programmable mirror arrays and programmable LCD arrays.

[0025] Figure 1A lithographic apparatus LA is schematically depicted. The lithographic apparatus LA comprises an illumination system (also referred to as an illuminator) IL configured to condition a radiation beam B (e.g., UV radiation, DUV radiation, or EUV radiation), a mask support (e.g., a mask stage) MT configured to support a patterning device (e.g., a mask) MA and connected to a first positioner PM configured to accurately position the patterning device MA according to certain parameters, a substrate support (e.g., a wafer stage) WT configured to hold a substrate (e.g., a wafer coated with resist) W, connected to a second positioner PW configured to accurately position the substrate support according to certain parameters, and a projection system (e.g., a refractive projection lens system) PS configured to project a pattern imparted to the radiation beam B by the patterning device MA onto a target portion C of the substrate W (e.g., a portion comprising one or more dies).

[0026] In operation, the illumination system IL receives a radiation beam from a radiation source SO, for example via a beam delivery system BD. The illumination system IL may include various types of optical components for directing, shaping and / or controlling the radiation, such as refractive, reflective, magnetic, electromagnetic, electrostatic and / or other types of optical components, or any combination thereof. The illuminator IL may be used to condition the radiation beam B to have a desired spatial and angular intensity distribution in its cross-section at the plane of the patterning device MA.

[0027] The term "projection system" PS as used in the present invention should be interpreted broadly to cover various types of projection systems suitable for the exposure radiation used and / or for other factors such as the use of immersion liquid or the use of vacuum, including refractive, reflective, catadioptric, anamorphic, magnetic, electromagnetic and / or electrostatic optical systems or any combination thereof. Any term "projection lens" used herein may be considered synonymous with the more general term "projection system" PS.

[0028] The lithographic apparatus LA may be of a type in which at least a portion of the substrate may be covered by an immersion liquid having a relatively high refractive index, such as water, so as to fill the space between the projection system PS and the substrate W - this is also known as immersion lithography. More information on immersion techniques is given in US 6952253, which is incorporated herein by reference.

[0029] The lithographic apparatus LA may also be of a type having two or more substrate supports WT (also referred to as a "dual stage"). In such a "multi-stage" machine, the substrate supports WT may be used in parallel and / or a substrate W on one of the substrate supports WT may be used to expose a pattern on another substrate W while a step of preparing the substrate W for subsequent exposure is being performed on the substrate W on another substrate support WT.

[0030] In addition to the substrate support WT, the lithographic apparatus LA may also comprise a measurement platform. The measurement platform is arranged to hold sensors and / or cleaning means. The sensors may be arranged to measure properties of the projection system PS or properties of the radiation beam B. The measurement platform may hold a plurality of sensors. The cleaning means may be arranged to clean a part of the lithographic apparatus, for example a part of the projection system PS or a part of the system providing immersion liquid. The measurement platform may be moved under the projection system PS when the substrate support WT is away from the projection system PS.

[0031] In operation, the radiation beam B is incident on the patterning device (e.g. mask) MA held on the mask support MT and is patterned by a pattern (design layout) present on the patterning device MA. Having traversed the mask MA, the radiation beam B passes through the projection system PS which focuses the beam onto a target portion C of the substrate W. With the aid of the second positioner PW and the position measurement system IF, the substrate support WT can be accurately moved, for example in order to position different target portions C in the path of the radiation beam B in a focused and aligned position. Similarly, the first positioner PM and possibly another position sensor (in Figure 1 The patterning device MA (not explicitly depicted in the figure) can be used to accurately position the patterning device MA relative to the path of the radiation beam B. The patterning device MA and the substrate W can be aligned using mask alignment marks M1, M2 and substrate alignment marks P1, P2. Although the substrate alignment marks P1, P2 occupy dedicated target portions as illustrated, the marks can be located in the space between target portions. When the substrate alignment marks P1, P2 are located between the target portions C, these substrate alignment marks are called scribe line alignment marks.

[0032] like Figure 2As shown in , the lithography apparatus LA may form part of a lithography cell LC, which is sometimes also referred to as a lithocell or lithography cluster, which often also includes equipment for performing pre-exposure and post-exposure processes on a substrate W. Conventionally, these include a spin coater SC for depositing a resist layer, a developer DE for developing the exposed resist, a chill plate CH and a bake plate BK, for example, for regulating the temperature of the substrate W (e.g., for regulating the solvent in the resist layer). A substrate handling device or robot RO picks up substrates W from input / output ports I / O1, I / O2, moves the substrates between different process equipment, and transfers the substrates W to a feed station LB of the lithography apparatus LA. The devices in the lithography cell, which are often also collectively referred to as a track or coating and developing system, are usually under the control of a track or coating and developing system control unit TCU, which itself may be controlled by a management control system SCS, which may also control the lithography apparatus LA, for example via a lithography control unit LACU.

[0033] In order to correctly and consistently expose a substrate W exposed by the lithographic apparatus LA, it is desirable to inspect the substrate to measure properties of the patterned structure, such as overlay errors between subsequent layers, line thickness, critical dimensions (CD), etc. For this purpose, an inspection tool (not shown) may be included in the lithography cell LC. If errors are detected, then, for example, adjustments may be made to the exposure of subsequent substrates or to other processing steps to be performed on the substrate W, particularly if the inspection is performed before other substrates W of the same batch or lot are yet to be exposed or processed.

[0034] An inspection apparatus, which may also be referred to as a metrology apparatus, is used to determine properties of a substrate W, and in particular, to determine how properties of different substrates W vary or how properties associated with different layers of the same substrate W vary from layer to layer. The inspection apparatus is alternatively configured to identify defects on the substrate W, and may for example be part of the lithography cell LC, or may be integrated into the lithography apparatus LA, or may even be a separate device. The inspection apparatus may measure properties on a latent image (the image in the resist layer after exposure), or on a semi-latent image (the image in the resist layer after a post-exposure bake step PEB), or on a developed resist image (wherein exposed or unexposed parts of the resist have been removed), or even on an etched image (after a pattern transfer step such as etching).

[0035] Figure 3An alternative type of lithography system using EUV radiation for exposure is shown. The lithography system comprises a radiation source SO and a lithography apparatus LA. The radiation source SO is configured to generate an EUV radiation beam B and for supplying the EUV radiation beam B to the lithography apparatus LA. The lithography apparatus LA comprises an illumination system IL, a support structure MT configured to support a patterning device MA (e.g. a mask), a projection system PS, and a substrate table WT configured to support a substrate W.

[0036] The illumination system IL is configured to condition the EUV radiation beam B before it is incident on the patterning device MA. Furthermore, the illumination system IL may comprise a faceted field mirror arrangement 10 and a faceted pupil mirror arrangement 11. The faceted field mirror arrangement 10 and the faceted pupil mirror arrangement 11 together provide a desired cross-sectional shape and a desired intensity distribution for the EUV radiation beam B. The illumination system IL may comprise other mirrors or arrangements in addition to or instead of the faceted field mirror arrangement 10 and the faceted pupil mirror arrangement 11.

[0037] After being so adjusted, the EUV radiation beam B interacts with the pattern forming device MA. As a result of this interaction, a patterned EUV radiation beam B' is generated. The projection system PS is configured to project the patterned EUV radiation beam B' onto the substrate W. To this end, the projection system PS may include a plurality of mirrors 13, 14, which are configured to project the patterned EUV radiation beam B' onto the substrate W held by the substrate table WT. The projection system PS may apply a reduction factor to the patterned EUV radiation beam B', thereby forming an image MA whose features are smaller than corresponding features on the pattern forming device MA. For example, a reduction factor of 4 or 8 may be applied. Although the projection system PS is Figure 1 1. Although illustrated in FIG. 1 as having only two mirrors 13, 14, the projection system PS may comprise a different number of mirrors (eg six or eight mirrors).

[0038] The substrate W may include a previously formed pattern. In this case, the lithographic apparatus LA aligns an image formed by the patterned EUV radiation beam B′ with a pattern previously formed on the substrate W.

[0039] A relative vacuum, ie a small amount of gas (eg hydrogen) at a pressure well below atmospheric pressure, may be provided in the radiation source SO in the illumination system IL and / or the projection system PS.

[0040] The radiation source SO may therefore be a laser produced plasma (LPP) source, a discharge produced plasma (DPP), a free electron laser (FEL), or any other radiation source capable of generating EUV radiation.

[0041] Scatterometers are general instruments that allow measuring parameters of a lithography process by having a sensor in the pupil or a plane conjugate to the pupil of the scatterometer's objective, the measurement being generally referred to as pupil-based measurement, or by having a sensor in the image plane or a plane conjugate to the image plane, in which case the measurement is generally referred to as image-based or field-based measurement. Such scatterometers and associated measurement techniques are further described in patent applications US20100328655, US2011102753A1, US20120044470A, US20110249244, US20110026032 or EP1,628,164A, which are incorporated herein by reference in their entirety. The above scatterometers can measure multiple targets from multiple gratings in one image using light from soft x-rays and visible to near IR bands.

[0042] exist Figure 4 A measurement device such as a scatterometer MT is depicted in . It comprises a radiation source 2 (e.g. a broadband (white light) radiation source) which projects radiation 5 onto a substrate W via a projection optical system 6. Reflected or scattered radiation 8 is collected by an objective system 8 and passed to a detector 4. The scattered radiation 8 detected by the detector 4 can then be processed by a processing unit PU. A pupil plane PP and an image plane IP of the objective system 8 are also shown. The terms "pupil plane" and "field plane" in this specification may refer to these planes or any plane conjugate thereto, respectively. Such a scatterometer may be configured as a normal incidence scatterometer or (as shown) an oblique incidence scatterometer. In some embodiments, the projection optical system 6 and the objective system 8 are combined; that is, the same objective system is used to illuminate the substrate and collect scattered radiation from the substrate.

[0043] In a first embodiment, the scatterometer MT is an angle-resolved scatterometer. In such a scatterometer, reconstruction methods can be applied to the measured signals to reconstruct or calculate the properties of the grating. Such a reconstruction may, for example, result from simulating the interaction of the scattered radiation with a mathematical model of the target structure and comparing the simulation results with the measured results. The parameters of the mathematical model are adjusted until the simulated interaction produces a diffraction pattern similar to the diffraction pattern observed from a real target.

[0044] In a second embodiment, the scatterometer MT is a spectroscopic scatterometer MT. In such a spectroscopic scatterometer MT, radiation emitted by a radiation source is directed onto a target and reflected or scattered radiation from the target is directed to a spectrometer detector which measures the spectrum of the specularly reflected radiation (i.e., intensity measurements as a function of wavelength). From such data, the structure or profile of the target producing the detected spectrum can be reconstructed, for example, by rigorous coupled wave analysis and non-linear regression or by comparison with a library of simulated spectra.

[0045] In a third embodiment, the scatterometer MT is an ellipsometric scatterometer. The ellipsometric scatterometer allows the parameters of the lithographic process to be determined by measuring the scattered radiation for each polarization state. Such a measurement device emits polarized light (such as linear, circular or elliptical) by using, for example, an appropriate polarization filter in the illumination cross section of the measurement device. A source suitable for the measurement device can also provide polarized radiation. Various embodiments of existing ellipsometric scatterometers are described in U.S. patent applications 11 / 451,599, 11 / 708,678, 12 / 256,780, 12 / 486,449, 12 / 920,968, 12 / 922,587, 13 / 000,229, 13 / 033,135, 13 / 533,110 and 13 / 891,410, which are incorporated herein by reference in their entirety.

[0046] In one embodiment of the scatterometer MT, the scatterometer MT is adapted to measure the overlap of two misaligned gratings or periodic structures by measuring asymmetry in the reflection spectrum and / or in the detection configuration, the asymmetry relating to the degree of overlap. Two (typically superimposed) grating structures may be applied in two different layers (not necessarily continuous layers) and the two grating structures may be formed at approximately the same position on the wafer. The scatterometer may have a symmetrical detection configuration as described, for example, in the co-owned patent application EP1,628,164A, so that any asymmetry is clearly identifiable. This provides a simple and direct way to measure misalignment in gratings. Further examples of measuring the overlay error between two layers comprising a periodic structure as a target by means of an asymmetry of the periodic structure may be found in PCT patent application number WO 2011 / 012624 or in US patent application number US 20160161863, which are incorporated herein by reference in their entirety.

[0047] Other parameters of interest may be focus and dose. Focus and dose may be determined simultaneously by a scatterometer (or alternatively by a scanning electron microscope), as described in U.S. Patent Application US2011-0249244, the entire contents of which are incorporated herein by reference. A single structure may be used that has a unique combination of critical dimension and sidewall angle measurements for each point in a focal energy matrix (FEM - also referred to as a focal exposure matrix). If these unique combinations of critical dimensions and sidewall angles can be obtained, then focus and dose values ​​may be uniquely determined from these measurements.

[0048] In the lithography and metrology machines just described, silicon wafers are moved at extremely high speeds to maximize productivity. In particular, in machines such as Figure 3 In the case of the EUV lithography machine shown in FIG. , the wafer is exposed to an EUV light pattern (e.g., patterned using a reticle or mask) below a projection optical box (POB). Within the POB, mirrors (e.g., Figure 3 A first component such as the reflectors 13, 14 in the POB is used to transfer and manipulate / guide EUV light to produce a sharp, i.e., clear image on the wafer. These reflectors can be actuated by electromagnetic actuators and can be supported by gravity compensators for stabilization purposes. When a second component such as a wafer platform approaches the POB during exposure, the electromagnetic field produces a disturbance that acts on the components within the POB. The rapid movement and high acceleration of the wafer platform cause this changing electromagnetic field. The electromagnetic crosstalk (EMX) disturbance force generated by these magnetic fields may cause overlay errors. The magnitude of the disturbance force depends on many different factors. The main contributing factors are the current through the long stroke coil of the wafer platform and the position of the exposure platform. The impact of the disturbance is also affected by several smaller contributing factors, such as the presence of magnetic (e.g., shielding) or conductive parts (e.g., eddy current effects) in the path of the magnetic field and the design configuration of the POB.

[0049] In other examples, the first component and the second component may be located in the same device.For example, a wafer stage long stroke movement may also introduce EMX into a wafer stage short stroke, and vice versa.

[0050] Feedback controllers of the current prior art are able to reduce the effects caused by the disturbance force caused by the magnetic crosstalk from the second component. However, a disadvantage of feedback controllers is that they are reactive in nature: that is, an error must exist before a correction can be performed. The effects of the disturbance force can be further reduced by using a feedforward controller. However, in order to produce a functioning feedforward controller, an estimate of the disturbance force is required as an input to the feedforward controller. Producing a physics-based prediction model is a difficult task because many different factors are at work and it takes a lot of time to generate simulation data.

[0051] More specifically, the disturbance force on each actuator may depend on:

[0052] The current through each wafer platform long-stroke coil;

[0053] The position of the exposure stage (i.e. the wafer stage used for exposure. Some lithography machines include multiple separate wafer stages for measurement / alignment and for exposure);

[0054] The position of the wafer stage counterweight mass due to the strong permanent magnets that generate the magnetic field at the actuator site, and in particular because it affects the commutation (current distribution) in the long-stroke actuator of the exposure stage;

[0055] The position of the measuring platform, as it has an impact on the balancing quality and may also have direct crosstalk to the mirror actuators (although the latter effect is usually small);

[0056] The presence of ferromagnetic components in the path between the magnetic source and the actuator, either as a shield or as a field focusing mechanism. In other embodiments, ferromagnetic components may also be present on the mirror itself;

[0057] The presence of conductive parts, which produce eddy current effects. Modeling eddy current effects is extremely difficult in this area;

[0058] • The configuration of the actuators and magnetic gravity compensators (which provide a constant lift force to the mirror so that the actuators don't have to) themselves, making them more or less sensitive to magnetic fields.

[0059] The relationship between these sources, elements in the magnetic path, and the resulting disturbance forces is very difficult to model, as the physical model is highly affected by small details, and the computation time for a static configuration can take days on a high-performance cluster. Thus, the computation time required by current models to achieve the desired accuracy is too long to be suitable for feed-forward implementation.

[0060] For this reason, it is proposed to use artificial intelligence (AI), and in particular a regression tree-based AI model, to infer corrections (e.g., feed-forward corrections) for controlling at least one first component of a lithographic or metrology apparatus that compensates for the effects of at least one second component (e.g., movement or actuation) of the lithographic or metrology apparatus. The regression tree AI model can learn, based on simulations or measurements, the relationship between a (e.g., electromagnetic) perturbation force on the first component caused by the second component and an input parameter related to the second component, without explicitly describing the underlying physics.

[0061] A regression tree is a decision tree based model used in a regression context. It uses a tree structure to convert the observed relationships in the input data into output values. The tree structure can be described as using branches and leaves. At the root of the tree, the regression tree begins to split into multiple branches based on the conditions of the input data at each branch split. This process continues until a leaf is reached. The leaves are the endpoints of the tree.

[0062] Figure 5 FIG. 1 shows an exemplary (simplified) regression tree model RTM, such as may be used to determine a perturbation parameter DP (e.g., a compensation force of one or more mirrors, which may be a compensation force described in terms of one or more degrees of freedom, where the degrees of freedom may include three spatial dimensions (e.g., x, y, z) and a plurality of spatial dimensions (e.g., x, y, z) based on input data DAT IN (For example, one or more second components or transmission platform variables, such as, depending on the type of platform, platform motor current, platform position, platform counterweight position, and possibly platform speed). IN The input data at the root node RN is mapped to a corresponding disturbance parameter value at one of the leaf nodes LN1-LN14, based on the value of the disturbance parameter. In the case where the disturbance parameter value is a (e.g., feed-forward) correction (such as a compensation force for a mirror), this force can then be applied to the mirror to compensate for the predicted / expected disturbance force based on the known platform variables.

[0063] The regression tree may be trained in a training phase based on the training data. Such training may be supervised based on known input data values ​​and corresponding perturbation parameter values. In order to establish a well-performing prediction model using the decision tree method, the training data should include a sufficiently representative measurement data set. Ideally, this data set covers all relevant system dynamics present on the machine.

[0064] The regression tree can be trained using an algorithm known as recursive binary splitting. At each branch, an evaluation is performed to determine whether splitting a branch into two branches will result in a reduction in the variance or error metric (e.g., mean square error) within the data. The resulting variance or mean square error of the data at the new branch should be lower than the variance or mean square error at the original branch. This process is repeated for each branch until all branches have reached a stopping parameter (such as a minimum leaf size). In each case, deciding where to split a branch can include an iterative process that calculates the variance or mean square error of all possible splits within the data of each branch and selects the split with the minimum value of the variance or mean square error.

[0065] The training may include an initial step of defining a minimum leaf size and / or maximum depth for the regression tree. The minimum leaf size determines the minimum number of individual values ​​in the training data that can be grouped together on a single leaf. As such, the minimum leaf size is a stopping parameter. Increasing this number results in a more general regression tree, while a lower number results in a specific prediction model that fits the training data more closely. The risk of a low minimum leaf size is that the tree will fit the features of the training data in a direct way, which may lead to overfitting. When such a tree will be used on unseen data, the predictions may become inaccurate compared to a more generalized model.

[0066] The ideal minimum leaf size will depend on the data. When the training data is representative enough, a smaller leaf size will lead to a more accurate model, and overfitting is not an immediate risk. If a more general model is being built, or the training data is sparsely distributed, a larger leaf size will provide better performance. It is difficult to precompute the minimum leaf size, so parameter sweeping or trial and error can be used to decide the best value.

[0067] Another parameter that can be tuned is the maximum depth of the tree. This determines the number of branching layers between the starting root and the leaves. This is another way to control the generality of the tree; a small depth allows less detail, while a large depth allows a more specific branching structure.

[0068] When the perturbation parameters are compensating mirror forces (to compensate for electromagnetic forces on the mirror), these electromagnetic forces cannot be measured directly. As such, methods for determining or extracting such mirror forces for training (ie, to label input training data) will also be described.

[0069] Figure 6 is a reflector feedback flow chart that illustrates in a conceptual manner, ie, in a conceptual manner, how the inverse process sensitivity plant model may be used and the control output Ce (assuming reference input r) to extract the derived mirror disturbance force (e.g. 6-DOF mirror perturbation forces).

[0070] As a force perturbation F dist The control error e as a function of the process sensitivity is defined by:

[0071]

[0072] Where P is the plant model and C is the control algorithm. is the process sensitivity. Therefore, the disturbance force can be estimated as:

[0073]

[0074] Since nominally (ignoring any dynamics), P = 1 / ms 2 Therefore, it is believed where m is the mirror mass (or inertia in the case of rotation) and s is the Laplace operator. Therefore, the inverse process sensitivity acting on the error e can be given by the inverse plant model P acting on e -1 Instead, it is combined with the controller output Ce to extract the estimated disturbance force It should be understood that any feed-forward control force should not be used to extract the force disturbance, only the feedback controller output Ce should be used. It should be noted that this concept can be generalized to any disturbance force that cannot be measured directly.

[0075] During the training of the platform-induced magnetic crosstalk of the POB mirror, as already described above, a distinction can be made between EMX-related disturbances and mechanically induced disturbances (so that the model is trained to correct only for the former). These phenomena can be distinguished, for example, by moving the measurement platform and noting the disturbance on the mirror before making a similar movement of the exposure platform, and again noting the disturbance force on the mirror. Since the mechanical disturbance is largely the same in both cases, but only the movement of the exposure platform causes the EMX-related disturbance (or at least the exposure platform movement is responsible for most of the EMX-related disturbance), the mechanical disturbance can be removed, for example, by taking the difference of the two disturbances. However, since the measurements show that the mechanically induced disturbance is much smaller than the EMX-related disturbance, the risk of learning the mechanical disturbance by the model is minimal. In this way, it is possible that this distinction of disturbances is not necessary. In any case, even if the training algorithm learns mechanical disturbances in addition to EMX disturbances, it does not automatically mean a loss of performance. In general, the correlation between input and output can be compensated by this mechanism.

[0076] By combining the computational power of multiple trained trees, regression trees can be made more robust, i.e. more robust and accurate. These are referred to as sets of regression trees. In an embodiment, it is proposed to train multiple regression trees or sets of regression trees, and once trained, they are used together for inference of the perturbation parameters. To prevent overfitting, assuming a bagged tree approach (see below), multiple different trees can be trained using (e.g. randomly selected) subsets of the learning data. The predicted perturbation force can then be the result (e.g., mean or other average) of the combined outputs of these trees.

[0077] A single regression tree may be greatly affected by changes to the training data. For example, if a portion of the training data is removed, the resulting tree may differ significantly from the original tree trained using the complete training data. Ensemble methods exploit this difference by training multiple trees by training them on subsets of the training data. Ensemble methods can be divided into two main techniques, either of which can be used in the methods disclosed herein: bagged trees and boosted trees.

[0078] Figure 7 is a diagram showing how such an integration method can be used to extract the input data DAT IN Flowchart for determining the perturbation parameter DP. Bag trees train multiple trees using a subset of the training data. The resulting set of trees RT1 to RT5 (only five trees are shown here for example only) first predicts the desired output value based on the input data of each individual tree. The final output DP of the bagged trees can be the average AV of all predicted values. This method effectively utilizes the differences between each tree and provides a more continuous output compared to a single regression tree. In addition, it reduces overfitting because each tree is trained using a different part of the training data. The advantage of this methodology is that bagged trees produce a more general prediction model that can also perform well for previously unseen data.

[0079] Boosting trees is a method that uses sequentially positioned trees to reduce the prediction error of each previous tree. Each successive tree aims to minimize the prediction error of the previous tree. This method is effective in reducing variance but has a higher risk of overfitting.

[0080] In such an ensemble model-based approach, the number of regression trees trained and / or used to infer perturbation parameters can be a user-adjustable variable, and the method can include tuning the number of regression trees. As with the minimum leaf size, the optimal value for the number of parameters is data dependent and can be found, for example, using parameter sweeps or trial and error.

[0081] Figure 8 is a flow chart that conceptually indicates how perturbation parameters inferred using a trained regression tree based on the methods described herein can be used to control a mirror (e.g., a mirror used to direct a patterned beam to a wafer during exposure). Here, the perturbation parameter is the compensating mirror force F comp , which is modeled to compensate for the perturbation force F on the mirror dist It can be seen that before the perturbation force is applied, F dist Determine the compensating mirror force F comp , and if so is applied to the control as a feedforward correction.

[0082] Due to the nature of decision trees, significant outliers can often be observed in the predicted output. In order to use the prediction model in practice, large errors should be avoided as much as possible. Since the prediction will be used in a feedforward controller, inaccurate predictions may lead to unexpected behavior of the machine. Therefore, in an embodiment, outlier removal can be performed. For example, a low-pass filter can be provided to filter the model output. For example, the regression tree modeling can include static modeling that does not take time into account. However, in a dynamic situation, there will not be a large change from one sample to the next, resulting in spikes in the model output. These spikes can be filtered using a (e.g., 500Hz) low-pass filter. Alternative methods can include training the model not to output outliers or sudden spikes.

[0083] Although the concepts disclosed herein are primarily described in the context of feed-forward mirror control within a projection optics box to compensate for EMX-related disturbance forces, the methods are not limited to such contexts. The regression tree-based model can be used in other contexts in a lithography or metrology control environment, and in particular for any control aspect where there is a strong nonlinear relationship between the measured (input) parameters and the modeled parameters. For example, any interaction in which a force or actuation of a second component induces an error (e.g., induces an electromagnetic field) that in turn affects the control of at least one first component can use the concepts disclosed herein.

[0084] Such a second component may include any stage (e.g., a wafer stage or a reticle stage) in a lithographic apparatus or a metrology apparatus, the movement of which may cause an electromagnetic field to be generated or changed, which may then cause a perturbation force or a change in the perturbation force on the first component. Such one or more first components may include one or more guide mirrors or one or more other components for guiding a beam (e.g., a patterned beam or a measurement beam), or other components such as a transport platform. Thus, in general, a regression tree-based model (or a collection thereof) may be used to determine a feed-forward correction for a first component based on relevant control and / or measurement data for a second component.

[0085] In another specific example, the wafer stage typically includes a long-stroke actuator for coarse control and a short-stroke actuator for fine control, respectively. This is a common case for lithography and metrology equipment, including Figure 1 , Figure 3 and Figure 4 , or other types of metrology devices such as scanning electron microscopes / electron beam devices. Actuation of the long-stroke actuator may produce unwanted fields that affect / perturb actuation of the short stage, and vice versa. A regression tree-based model (or set thereof) may be used to determine a feedforward correction for one of the actuators based on related control and / or measurement data for the other actuators.

[0086] The method disclosed herein has been demonstrated to reduce the disturbance forces on the reflectors within the POB to a very significant degree. For example, using bagged trees as described, the method is able to reduce the disturbance forces by 96% or more. The predicted disturbance forces can be used as input to a feedforward controller or control loop to mitigate the effects of electromagnetic disturbances on the POB.

[0087] Another advantage of regression trees is that they are interpretable. For each predicted outcome, a specific decision process is followed. Therefore, when an inaccurate prediction occurs, its cause can be traced back through the regression tree. As the number of dimensions of the training data increases, the difficulty of tracing back decisions becomes more complicated, but the principle remains the same.

[0088] While the foregoing embodiments use a regression tree-based AI model to infer a correction (e.g., a feedforward correction) for controlling at least one first component of a lithographic apparatus or a metrology apparatus that compensates for the effects of at least one second component (e.g., movement or actuation) of a lithographic apparatus or a metrology apparatus, it has been found that a neural network (NN)-based AI model can also achieve similar effects. Different types of NNs may be suitable for achieving similar correction performance.

[0089] In an embodiment, the NN may be an artificial neural network (ANN) known as a feedforward neural network. The ANN is based on an interconnected group of artificial neurons AN or nodes. The AN is organized into multiple layers. The AN of one layer is only connected to the AN of the immediately preceding and immediately succeeding layers. Fig. 9 The example ANN shown in is a feed-forward neural network. Fig. 9 As shown in FIG. 1 , receiving external data DAT IN The layer is the input layer L IN The layer that outputs the final result (eg, the perturbation parameter DP in the context of the present disclosure) is the output layer L OUT In the input layer L IN With the output layer L OUT There are two hidden layers L1 HID 、L2 HID In other example ANNs, no hidden layers may be included, or any other number of hidden layers may be included (e.g., one, three, four, or more hidden layers). Between any two adjacent layers, the ANN may be connected in many different ways, which may be used to determine the type of ANN. Fig. 9 The example ANN shown in is a fully connected ANN, that is, each AN in one layer is connected to each neuron in the adjacent layer. In other examples such as pooling ANNs, a group of ANs in one layer are connected to a single AN in an adjacent layer, and this can reduce the dimensionality of the adjacent layers.

[0090] Each AN has an input and produces a single output, which can be sent to multiple other ANs (see above). Each AN contains a weight vector of the same dimension as the input vector (e.g., Fig. 9 W shown in 1 To W 6 ), where each weight determines the strength of the corresponding input. Each AN is configured to calculate a weighted sum of all inputs using the weights associated with the corresponding input connections, and to add a bias to the weighted sum. The AN then passes such a biased weighted sum through an activation function (e.g., a sigmoid function) to obtain the output signal. The weights are optimized using an optimization algorithm. The optimization process depends on the optimization algorithm selected. A suitable optimization algorithm may be stochastic gradient descent (SGD). The goal of the optimization algorithm is to minimize a loss function (e.g., mean square error (MSE)). The optimization algorithm may determine and update the weights of the model in such a way that the loss function is as close to a minimum as possible.

[0091] The accuracy of the network may depend on the network structure (e.g., the number of ANs, the number of layers, and the connection pattern between layers) and the optimization algorithm. Similar to the regression tree-based model, the ANN-based model can also be trained in a training phase based on training data. Such training can be supervised based on known input data values ​​and corresponding perturbation parameter DP values.

[0092] refer to Figure 6 and Fig. 9 , by taking input data DAT related to, for example, wafer stage long stroke (LoS) current, wafer stage position, balance quality, wafer stage velocity, and / or n-sample delay signal IN Input to the input layer L IN (e.g., four ANs in the input layer), the (trained) ANN is able to infer the compensating mirror force F comp (or any other disturbance parameter DP). When this predicted disturbance force is used as an input to a feedforward controller or control loop, the electromagnetic disturbance (e.g., disturbance force F dist )The impact on reflectors within the POB can be reduced by 22% or more.

[0093] Further embodiments of the invention are disclosed in the following list of numbered aspects:

[0094] 1. A method of determining a correction for controlling at least one first component of a lithographic apparatus or a metrology apparatus, the method comprising:

[0095] obtaining a trained artificial intelligence model, the trained artificial intelligence model having been trained to map input data relating to at least one second component of the lithographic apparatus or metrology apparatus to disturbance parameters relating to a disturbance of the at least one first component, the disturbance having been caused by the at least one second component;

[0096] obtaining input data related to the at least one second component; and

[0097] A compensating correction to the disturbance parameter is determined based on the input data using the trained artificial intelligence model.

[0098] 2. A method according to aspect 1, wherein the compensating correction is determined as a feed-forward correction, the determining step being performed before the disturbance occurs.

[0099] 3. A method according to aspect 1 or 2, wherein the disturbance parameter is caused by movement or actuation of the at least one second component.

[0100] 4. A method according to any preceding aspect, wherein the disturbance parameter comprises an electromagnetic force on the at least one first component.

[0101] 5. A method according to any preceding aspect, wherein the input data comprises data of one or more parameters relating to movement and / or actuation of the at least one second component.

[0102] 6. The method of aspect 5, wherein the one or more parameters include one or more of: velocity, acceleration, actuation current and position of one or more subcomponents of the at least one second component.

[0103] 7. The method according to any preceding aspect, wherein the at least one second component comprises one or more of: a substrate stage and a reticle stage.

[0104] 8. A method according to any preceding clause, wherein the at least one first component comprises at least one mirror for steering and / or directing the patterned beam to the substrate during a lithographic exposure process.

[0105] 9. The method according to any one of clauses 1 to 6, wherein the at least one first component comprises a first actuator for a substrate stage or a reticle stage, and the at least one second component comprises a second actuator for the substrate stage or the reticle stage.

[0106] 10. A method according to any of the preceding aspects, wherein the trained artificial intelligence model comprises a trained regression tree model.

[0107] 11. A method according to aspect 10, wherein the trained regression tree model comprises a set of trained regression trees, each trained regression tree in the set has been trained on a different subset of training data; and the method comprises determining the compensating correction based on the combined output of the set of trained regression trees.

[0108] 12. A method according to clause 11, wherein the compensation correction is determined as the average of the corresponding outputs of the set of trained regression trees.

[0109] 13. A method according to any preceding aspect, comprising performing an outlier removal step on the compensation correction.

[0110] 14. A method according to clause 13, wherein the outlier removal step comprises low pass filtering the compensation correction.

[0111] 15. A method according to any preceding aspect, comprising the step of training a regression tree model on training data to obtain the trained regression tree model, the training comprising applying a recursive binary splitting algorithm to the training data.

[0112] 16. The method of clause 15, wherein the recursive binary splitting algorithm uses a reduction in error metric or variance to determine whether and where to split branches of the training data.

[0113] 17. A method according to aspect 15 or 16, comprising an initial step of determining a minimum leaf size for the model.

[0114] 18. A method according to any one of clauses 15 to 17, comprising training a collection of regression trees, each regression tree being trained on a different subset of the training data.

[0115] 19. A method according to any one of aspects 1 to 9, wherein the trained artificial intelligence model comprises a trained neural network model.

[0116] 20. The method of aspect 19, wherein the trained neural network model is a feed-forward neural network model.

[0117] 21. The method according to aspect 19 or 20, comprising the step of training a neural network model on training data to obtain the trained neural network model, wherein the training comprises applying a stochastic gradient descent algorithm to the training data.

[0118] 22. A method according to any one of aspects 15 to 21, comprising processing the training data by removing mechanical disturbance parameter contributions to the disturbance parameters.

[0119] 23. A method according to any one of aspects 15 to 22, wherein the training data relates to a disturbance parameter that cannot be directly measured, and the method comprises extracting the disturbance parameter from an inverse process sensitivity of a feedback control process of the at least one first component, a control action for the feedback control process, and an error signal.

[0120] 24. A method according to any preceding aspect, comprising applying the compensation correction to a control action of the at least one first component.

[0121] 25. A computer program comprising program instructions operable to perform a method according to any preceding aspect when run on a suitable device.

[0122] 26. A non-transitory computer program carrier comprising a computer program according to clause 25.

[0123] 27. A processing device comprising:

[0124] a computer program carrier comprising a computer program according to clause 26; and

[0125] A processor is operable to execute the computer program.

[0126] 28. A lithography tool for patterning a substrate or measuring a patterned structure on a substrate, the lithography tool comprising:

[0127] at least one first component;

[0128] at least one second component;

[0129] Processor; and

[0130] A non-transitory computer program carrier comprising a trained artificial intelligence model, the trained artificial intelligence model having been trained to map input data relating to the at least one second component to disturbance parameters relating to a disturbance of the at least one first component, the disturbance having been caused by the at least one second component.

[0131] 29. A lithographic tool according to clause 28, wherein the computer program carrier further comprises program instructions operable to perform a method of determining a correction for controlling the at least one first component; the method comprising:

[0132] obtaining input data related to the at least one second component; and

[0133] A compensating correction to the disturbance parameter is determined based on the input data using a trained artificial intelligence model.

[0134] 30. The lithographic tool of clause 29, wherein the processor is operable to determine the compensation correction as a feed-forward correction, the determining step being performed before the perturbation occurs.

[0135] 31. A lithographic tool according to clause 30 or 29, wherein the processor is operable to apply the compensating correction to the control action of the at least one first component.

[0136] 32. A lithography tool according to clause 29, 30 or 31, wherein the processor is operable to perform an outlier removal step on the compensation correction.

[0137] 33. The lithography tool of clause 32, wherein the outlier removal step comprises low pass filtering the compensation correction.

[0138] 34. A lithographic tool according to any of clauses 28 to 33, wherein the disturbance parameter is caused by movement or actuation of the at least one second component.

[0139] 35. A lithographic tool according to any one of clauses 28 to 34, wherein the perturbation parameter comprises an electromagnetic force on the at least one first component.

[0140] 36. A lithographic tool according to any of clauses 28 to 35, wherein the input data comprises data of one or more parameters relating to movement and / or actuation of the at least one second component.

[0141] 37. The lithography tool of clause 36, wherein the one or more parameters include one or more of: velocity, acceleration, actuation current, and position of subcomponents of one or more of the at least one second component.

[0142] 38. The lithographic tool of any of clauses 28 to 37, wherein the at least one second component comprises one or more transport platforms of the lithographic tool.

[0143] 39. The lithography tool of clause 38, wherein the at least one first component comprises a first actuator for the one or more transport stages, and the at least one second component comprises a second actuator for the one or more transport stages.

[0144] 40. The lithography tool of clause 38 or 39, wherein the one or more transport stages comprises a reticle stage of the lithography tool.

[0145] 41. The lithography tool of clause 38 or 39, wherein the one or more transfer stages include one or more substrate stages of the lithography tool.

[0146] 42. The lithography tool of any one of clauses 28 to 41, wherein the lithography tool is an exposure tool for patterning a substrate.

[0147] 43. A lithography tool according to aspect 42, wherein the lithography tool also includes a projection optical device for manipulating and / or directing the patterned beam to the substrate; and the at least one first component includes at least one mirror contained within the projection optical device.

[0148] 44. The lithography tool of any of clauses 28 to 41, wherein the lithography tool comprises a metrology tool for measuring the patterned structure.

[0149] 45. The lithography tool of clause 44, wherein the metrology tool is a scatterometer.

[0150] 46. ​​The lithography tool of clause 44, wherein the metrology tool is a scanning electron microscope or an electron beam tool.

[0151] 47. The lithography tool of any one of clauses 28 to 46, wherein the trained artificial intelligence model comprises a trained regression tree model.

[0152] 48. The lithography tool of any of clause 47, wherein the trained regression tree model comprises a collection of trained regression trees, each of the trained regression trees having been trained on a different subset of training data.

[0153] 49. A lithography tool according to any one of clauses 28 to 46, wherein the trained artificial intelligence model comprises a trained neural network model.

[0154] 50. The lithography tool of clause 49, wherein the trained neural network model is a feed-forward neural network model.

[0155] 51. The lithography tool of clause 49 or 50, comprising the step of training a neural network model on training data to obtain the trained neural network model, the training comprising applying a stochastic gradient descent algorithm to the training data.

[0156] 52. A component controller configured to control a first component, the controller being configured to implement the method according to any one of aspects 1 to 24.

[0157] Although specific reference may be made herein to the use of lithographic apparatus in IC manufacturing, it should be understood that the lithographic apparatus described herein may have other applications. Possible other applications include the manufacture of integrated optical systems, guidance and detection patterns for magnetic domain memories, flat panel displays, liquid crystal displays (LCDs), thin film magnetic heads, etc.

[0158] Although specific reference is made herein to embodiments of the invention in the context of lithographic equipment, embodiments of the invention may be used in other equipment. Embodiments of the invention may form part of a mask inspection apparatus, a metrology apparatus, or any apparatus for measuring or processing an object such as a wafer (or other substrate) or a mask (or other pattern forming device). These apparatus may generally be referred to as lithographic tools. Such lithographic tools may use vacuum conditions or ambient (non-vacuum) conditions.

[0159] Although specific reference has been made above to the use of embodiments of the invention in the context of optical lithography, it will be appreciated that the invention is not limited to optical lithography and may be used in other applications, such as imprint lithography, where the context permits.

[0160] Where context permits, embodiments of the present invention may be implemented in hardware, firmware, software, or any combination thereof. Embodiments of the present invention may also be implemented as instructions stored on a machine-readable medium that can be read and executed by one or more processors. A machine-readable medium may include any mechanism for storing or transmitting information in a form that can be read by a machine (e.g., a computing device). For example, a machine-readable magnetic storage medium may include a read-only memory (ROM); a random access memory (RAM); a magnetic storage medium; an optical storage medium; a flash memory device; an electrical, optical, acoustic, or other form of propagation signal (e.g., a carrier wave, an infrared signal, a digital signal, etc.), etc. In addition, herein, firmware, software, routines, instructions may be described as performing certain actions. However, it should be understood that such descriptions are for convenience only, and that these actions are actually generated by a computing device, a processor, a controller, or other device that executes firmware, software, routines, instructions, etc., and in doing so may cause an actuator or other device to interact with the physical world.

[0161] Although specific embodiments of the present invention have been described above, it will be appreciated that the present invention may be practiced in a manner different from that described. The above description is intended to be exemplary and not restrictive. Therefore, it will be appreciated by those skilled in the art that modifications may be made to the present invention as described without departing from the scope of the claims set forth below.

Claims

1. A method of determining a correction for controlling at least one first component of a lithographic apparatus or a metrology apparatus, the method comprising: obtaining a trained artificial intelligence model, the trained artificial intelligence model having been trained to map input data relating to at least one second component of the lithographic apparatus or metrology apparatus to disturbance parameters relating to a disturbance of the at least one first component, the disturbance having been caused by the at least one second component; obtaining input data related to the at least one second component; as well as A compensating correction to the disturbance parameter is determined based on the input data using the trained artificial intelligence model.

2. The method according to claim 1, wherein: The compensating correction is determined as a feed-forward correction, the determining step being performed before the disturbance occurs.

3. A method according to any preceding claim, wherein: The disturbance parameter comprises an electromagnetic force on the at least one first component.

4. A method according to any preceding claim, wherein: The input data comprises data of one or more parameters related to the movement and / or actuation of the at least one second component.

5. A method according to any preceding claim, wherein: The at least one second component includes one or more of: a substrate stage and a reticle stage; and The at least one first component comprises at least one mirror for steering and / or directing the patterned beam to the substrate during a lithographic exposure process.

6. A method according to any preceding claim, wherein: The trained artificial intelligence model includes: Regression tree models; and / or a set of trained regression trees, each trained regression tree in the set having been trained on a different subset of the training data; and the method includes determining the compensation correction based on a combined output of the set of trained regression trees.

7. A method according to any preceding claim, comprising the step of training an artificial intelligence model on training data to obtain the trained artificial intelligence model, the training comprising applying a recursive binary splitting algorithm to the training data.

8. A computer program comprising program instructions operable to perform a method according to any preceding claim when run on a suitable device.

9. A lithography tool for patterning a substrate or measuring a patterned structure on a substrate, the lithography tool comprising: at least one first component; at least one second component; processor; as well as A non-transitory computer program carrier comprising a trained artificial intelligence model, the trained artificial intelligence model having been trained to map input data relating to the at least one second component to disturbance parameters relating to a disturbance of the at least one first component, the disturbance having been caused by the at least one second component.

10. The lithography tool of claim 9, wherein: The computer program carrier further comprises program instructions operable to perform a method of determining a correction for controlling the at least one first component; the method comprising: obtaining input data related to the at least one second component; and A compensating correction to the disturbance parameter is determined based on the input data using a trained artificial intelligence model.

11. The lithography tool of claim 10, wherein: The processor is operable to determine the compensation correction as a feed-forward correction, the determining step being performed before the disturbance occurs.

12. The lithography tool according to claim 10 or 11, wherein: The processor is operable to apply the compensation correction to the control action of the at least one first component.

13. A lithography tool according to any one of claims 9 to 12, wherein: The at least one second component comprises one or more transport platforms of the lithography tool.

14. A lithography tool according to any one of claims 9 to 12, wherein: The photolithography tool is an exposure tool for patterning a substrate, and The lithography tool further comprises projection optics for steering and / or directing the patterned beam to the substrate; and the at least one first component comprises at least one mirror contained within the projection optics.

15. A lithographic tool according to any one of claims 9 to 12, wherein: The lithography tool includes a metrology tool for measuring the patterned structures.

Citation Information

Patent Citations

  • A method of determining a correction for control of a lithography and / or metrology process, and associated devices

    EP4261618A1

  • Diffraction Based Overlay Metrology Tool and Method

    US20100328655A1

  • Method of Assessing a Model of a Substrate, an Inspection Apparatus and a Lithographic Apparatus

    US20110026032A1

  • Apparatus and Method of Measuring a Property of a Substrate

    US20110102753A1

  • Lithographic Focus and Dose Measurement Using A 2-D Target

    US20110249244A1

Cited By

  • Immersion flow field disturbing force measuring device and measuring method

    CN117804728A