Electric vehicle

Abstract

A method of training an inference model for determining one or more parameters of a product of a manufacturing process from measurements of the product. The method includes acquiring a set of measurement data for one or more products of a manufacturing process, each measurement result including an array of values obtained by measuring a corresponding one of the products. The method further includes selecting an appropriate subset of the data set for training the inference model, the subset being selected by applying an optimization program to an objective function that provides a difference metric between each measurement in the data set and a corresponding rendered value of the measurement obtained using the rendered function, the rendering function has a domain that includes measurements in a subset and excludes measurements not in the subset. The method also includes training an inference model using an appropriate subset of the data sets.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to EP application 22189411.6 filed on August 9, 2022 and EP application 22203256.7 filed on October 24, 2022, and the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present invention relates to a method and system for training an inference model that determines one or more parameters of a product of a manufacturing process. Background Art

[0004] A lithographic apparatus is a machine used to apply a desired pattern onto a substrate. A lithographic apparatus can be used to manufacture integrated circuits (ICs). A lithographic apparatus can, for example, project a pattern (also often referred to as a "design layout" or "design") at a patterning device (e.g., a mask) onto a layer of radiation-sensitive material (resist) on a substrate (e.g., a wafer).

[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 a substrate. Typical wavelengths currently used are 365 nm (i-line), 248 nm, 193 nm, and 13.5 nm. A lithographic apparatus using extreme ultraviolet (EUV) radiation with a wavelength in the range of 4-20 nm (e.g., 6.7 nm or 13.5 nm) can be used to form smaller features on a substrate than a lithographic apparatus using, for example, radiation with a wavelength of 193 nm.

[0006] Low-k1 lithography can be used to process features whose size is smaller than the traditional resolution limit of the lithographic equipment. In this process, the resolution formula can be expressed as CD = k1 × λ / NA, where λ is the wavelength of the radiation used, NA is the numerical aperture of the projection optics in the lithographic equipment, CD is the "critical dimension" (usually the smallest feature size printed, but in this case it is half pitch), and k1 is an empirical resolution factor. In general, the smaller the k1, the more difficult it is to reproduce on the substrate a pattern similar in shape and size to that planned by the circuit designer to achieve a specific electrical function and performance.

[0007] To overcome these difficulties, complex fine-tuning steps can be applied to the lithographic projection equipment and / or the design layout. These include, for example, but not limited to, optimization of NA, customized illumination schemes, use of phase-shifting patterning devices, various optimizations of the design layout, such as optical proximity correction (OPC, sometimes also called "optical and process correction") in the design layout, or other methods generally defined as "resolution enhancement technology" (RET). Alternatively, a tight control loop for controlling the stability of the lithographic equipment can be used to improve pattern reproduction at low k1.

[0008] In photolithography processes, it is often necessary to make measurements of the structures being created, for example for process control and verification. Various tools are known for making such measurements, including scanning electron microscopes, which are commonly used to measure critical dimensions (CDs), and specialized tools for measuring overlay (i.e., the alignment accuracy of two layers in a device). More recently, various forms of scatterometers have been developed for use in the field of photolithography.

[0009] Examples of known scatterometers generally rely on providing a dedicated measurement target. For example, one approach may require a target in the form of a simple grating that is large enough so that the measurement beam produces a smaller spot than the grating (i.e. the grating is underfilled). In a so-called reconstruction approach, the properties of the grating can be calculated by simulating the interaction of the scattered radiation with a mathematical model of the target structure. The parameters of the model are adjusted until the simulated interaction produces a diffraction pattern similar to that observed from a real target.

[0010] In addition to measuring feature shapes by reconstruction, diffraction-based overlay can also be measured using such devices, as described in published patent application US2006066855A1. Diffraction-based overlay metrology using diffraction-order dark-field imaging allows overlay measurement of smaller targets. These targets can be smaller than the illumination spot and can be surrounded by product structures on the wafer. Examples of dark-field imaging metrology can be found in many published patent applications, such as US2011102753A1 and US20120044470A. Using composite grating targets, multiple gratings can be measured in one image. Known scatterometers tend to use light in the visible or near-infrared (IR) wave range, which requires the spacing of the grating to be much coarser than the actual product structure (whose characteristics are actually of interest). Such product features can be defined using much shorter wavelength deep ultraviolet (DUV), extreme ultraviolet (EUV) or X-ray radiation. Unfortunately, such wavelengths are generally not suitable for metrology.

[0011] On the other hand, the dimensions of modern product structures are so small that they cannot be imaged by optical metrology techniques. Small features include, for example, features formed by multiple patterning processes and / or pitch multiplication. Therefore, targets for high-volume metrology often use features that are much larger than the product, where the product overlay error or critical dimensions are the attributes of interest. The measurement results are only indirectly related to the dimensions of the actual product structure and may not be accurate because the metrology target does not suffer the same distortions under the optical projection of the lithography equipment and / or is not subjected to different treatments in other steps of the manufacturing process. Although scanning electron microscopes (SEMs) are able to directly resolve these modern product structures, SEMs are much more time-consuming than optical measurements. In addition, electrons cannot penetrate thick process layers, which makes them less suitable for metrology applications. Other techniques, such as measuring electrical properties using contact pads are also known, but only provide indirect evidence of the real product structure. By reducing the wavelength of the radiation used in the metrology process, smaller structures can be resolved, the sensitivity to structural structural changes can be increased and / or further penetrated into the product structure. One such method of generating suitable high frequency radiation (e.g. hard X-rays, soft X-rays and / or EUV radiation) may be to use pump radiation (e.g. infrared IR radiation) to excite the generation medium, thereby generating emission radiation, optionally including higher harmonics of the high frequency radiation. Summary of the invention

[0012] According to one aspect of the invention, there is provided a method for training an inference model to determine one or more parameters of a product of a manufacturing process based on measurement results of the product. The method comprises: obtaining a measurement data set of one or more products of a manufacturing process. Each measurement result comprises an array of values ​​obtained by measuring a corresponding one of the products. The method further comprises selecting an appropriate subset of the data set for training the inference model. The subset is selected by applying an optimization procedure to an objective function, the objective function providing a difference measure between each measurement result in the data set and a corresponding reproduced value of the measurement result obtained using a reproduction function, the reproduction function having a domain of the measurement results included in the subset and excluding the measurement results not included in the subset. The inference model is trained using (only) an appropriate subset of the measurement data set; that is, the portion of the measurement data set not included in the appropriate subset is not used to train the inference model.

[0013] The benefit of this is that it is easier to train the perturbation model because the number of measurements used as input to the training model may be smaller than when using the entire dataset. Experiments have found that the accuracy of the perturbation results is almost as high as using the entire measurement dataset during training, but the computational cost is greatly reduced.

[0014] The reproduction function can generate the reproduction value of the measurement as the weighted sum of the combination of the arrays of the measurement results in the subset. The array combination can be a linear or nonlinear combination, for example, the weighted sum of the array of the measurement results in the subset or the weighted sum of the product of the corresponding values ​​of the arrays of two or more measurement results in the subset. The reproduction value of the measurement result in the data set can be an approximation of the array value of the measurement result in the data set obtained by projecting the array of the measurement results onto the space spanned by the measurement in the subset. The difference measure between each measurement result in the data set and the corresponding reproduction value of the measurement result can, for example, include the residual between the value of the array of the measurement result and the corresponding reproduction value of the array of the measurement result, such as the sum of the squares of the residuals. For example, the metric used can be the Frobenius norm. The array can be one-dimensional (e.g., a vector), two-dimensional (e.g., a 2D matrix), or higher-dimensional (e.g., a 3D, 4D, etc. matrix). The dimensionality of the array is usually equal to or less than the dimensionality of the measurement result in the data set.

[0015] In another aspect of the invention, once an appropriate subset of measurements is selected, when further operations of the manufacturing process are performed to manufacture additional products, the measurement results of the additional products may omit at least one measurement result whose measurement result is included in the measurement data set but omitted from the appropriate subset of measurements. In this way, the workload required to perform the measurements is reduced while still capturing enough information to characterize the manufacturing process. For example, the measurement results of the additional products of the manufacturing process may include only the measurement results corresponding to the measurement results of the selected subset. The measurement results of the additional products of the manufacturing process can be used in an inference model trained on the selected measurement results of the products of the earlier manufacturing process, or can be used to train a new inference model. Reducing the number of measurements performed can speed up the measurement process and improve the throughput of the manufacturing process. Again, experimental results show that this is possible without significantly reducing the quality of the inspection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings are incorporated in and constitute a part of this specification to illustrate one or more embodiments and together with the specification explain these embodiments. Embodiments of the present invention will now be described by way of example only with reference to the accompanying drawings, in which corresponding reference characters indicate corresponding parts, wherein:

[0017] Figure 1 A lithographic apparatus according to an embodiment is shown.

[0018] Figure 2 A schematic diagram of a lithography cell according to one embodiment is shown.

[0019] Figure 3 A schematic diagram depicting overall lithography according to one embodiment represents the cooperation between three technologies to optimize semiconductor manufacturing.

[0020] Figure 4 An example metrology device, such as a scatterometer, is shown according to one embodiment.

[0021] Figure 5 is a flowchart of a method for training an inference model according to one embodiment.

[0022] Figure 6 is a schematic diagram of a top view of the wafer showing where measurements are taken on the wafer and the division of the wafer into quadrants.

[0023] Figure 7 A method of sampling measurements from a data set according to one embodiment is described.

[0024] Figure 8 A method of sampling measurements from a data set according to one embodiment is described.

[0025] Fig. 9 A method of sampling measurements from a data set according to one embodiment is described.

[0026] Fig. 10A is a schematic diagram of a top view of a wafer showing inferred tilt values ​​obtained by a tilt inference model trained using pupil image measurements at each of a plurality of locations on the wafer.

[0027] Fig. 10B Yes Display Fig. 10A A graph of the relationship between the tilt value and the radial distance from the center of the wafer.

[0028] Fig. 10C yes Fig. 10A Schematic diagram of a top view of a wafer showing measured tilt values ​​for a portion of locations selected for training an additional tilt inference model.

[0029] Fig. 10D Yes Display Fig. 10C A graph of the relationship between the tilt value and the radial distance from the center of the wafer.

[0030] Fig.10E yes Fig. 10A and 10B Schematic diagram of a top view of a wafer showing the inferred tilt values ​​for all locations on the wafer determined from pupil image measurements using a trained further tilt inference model.

[0031] Fig.10F Yes Display Fig.10E A plot of the tilt value as a function of radial distance from the center of the wafer.

[0032] Fig.11 is a block diagram of an example computer system according to one embodiment. DETAILED DESCRIPTION

[0033] As a brief introduction, Figure 1 A lithographic apparatus LA is schematically shown. 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) T configured to support a patterning device (e.g., a mask) MA and connected to a first positioner PM configured to precisely position the patterning device MA according to certain parameters, a substrate support (e.g., a wafer stage) WT configured to support a substrate (e.g., a resist-coated wafer) W and connected to a second positioner PW configured to precisely 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 (e.g., comprising one or more dies) of the substrate W.

[0034] 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 refractive, reflective, magnetic, electromagnetic, electrostatic and / or other types of optical elements, or any combination thereof, for directing, shaping and / or controlling the radiation. The illuminator IL may be used to condition the radiation beam B so that it has a desired spatial and angular intensity distribution in its cross-section at the plane of the patterning device MA.

[0035] The term "projection system" PS as used herein should be interpreted broadly to cover various types of projection systems, including refractive, reflective, catadioptric, anamorphic, magnetic, electromagnetic and / or electrostatic optical systems, or any combination thereof, depending on the exposure radiation used and / or other factors, such as the use of immersion liquid or the use of a vacuum. Any use of the term "projection lens" herein may be considered synonymous with the more general term "projection system" PS.

[0036] The lithographic apparatus LA may be of a type in which at least a portion of the substrate may be covered by a 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. For more information on immersion technology, see US6952253, which is incorporated herein by reference.

[0037] The lithographic apparatus LA may also be of a type having two or more substrate supports WT (also referred to as "dual stage"). In such a "multi-stage" machine, the substrate supports WT may be used in parallel, and / or steps for preparing a substrate W for subsequent exposure may be performed on a substrate W located on one substrate support WT while another substrate W on another substrate support WT is used to expose a pattern on another substrate W.

[0038] In addition to the substrate support WT, the lithographic apparatus LA may comprise a measurement table. The measurement table is arranged to hold sensors and / or cleaning means. The sensors may be arranged to measure a property of the projection system PS or a property of the radiation beam B. The measurement table may hold a plurality of sensors. The cleaning means may be arranged to clean a part of the lithographic apparatus, such as a part of the projection system PS or a system for providing immersion liquid. The measurement table may be moved under the projection system PS when the substrate support WT is away from the projection system PS.

[0039] In operation, a radiation beam B is incident on a patterning device, e.g. a mask MA, which is held on a mask support MT and is patterned by a pattern (design layout) on the patterning device MA. After passing through the mask MA, the radiation beam B passes through a projection system PS which focuses the beam onto a target portion C of the substrate W. With the aid of a second positioner PW and a position measurement system IF the substrate support WT can be precisely moved, e.g. in order to position different target portions C in the path of the radiation beam B in a focused and aligned position. A first positioner PM and possibly another position sensor ( Figure 1 The patterning device MA may be accurately positioned relative to the path of the radiation beam B using mask alignment marks M1, M2 and substrate alignment marks P1, P2. Although the substrate alignment marks P1, P2 are shown occupying dedicated target portions, they may also be located in spaces between target portions. When the substrate alignment marks P1, P2 are located between target portions C, they are referred to as scribe lane alignment marks.

[0040] Figure 2 Schematic diagram of the lithography unit LC is shown. Figure 2As shown, the lithography apparatus LA may form part of a lithography cell LC, sometimes also referred to as a lithography cell or (lithography) cluster, which typically also includes equipment for performing pre-exposure and post-exposure processes on a substrate W. Conventionally, these include a spin coater SC configured to deposit a resist layer, a developer DE for developing the exposed resist, a cooling plate CH and a bake plate BK, for example for regulating the temperature of the substrate W, for example for regulating a solvent in the resist layer. A substrate handler or robot RO picks up substrates W from input / output ports I / O1, I / O2, moves them between different process equipment, and transfers the substrates W to a loading bay LB of the lithography apparatus LA. The equipment in the lithography cell is also often collectively referred to as a track, typically by a track control unit TCU which itself may be controlled by a monitoring system SCS, which may also control the lithography apparatus LA, for example via a lithography control unit LACU.

[0041] In order to expose the substrate W ( Figure 1 ) can be correctly and consistently exposed, the substrate needs to be inspected to measure the properties of the patterned structure, such as overlay error between subsequent layers, line thickness, critical dimension (CD), etc. For this purpose, an inspection tool (not shown) may be included in the lithography unit LC. If detected, the exposure of subsequent substrates or other processing steps to be performed on the substrate W can be adjusted, especially if the inspection is performed before other substrates W of the same batch are still to be exposed or processed.

[0042] The inspection device, which may also be referred to as a metrology device, is used to determine the substrate W( Figure 1 ), in particular how properties vary from one substrate to another W, or how related properties vary from one layer to another of different layers of the same substrate W. The inspection apparatus may alternatively be 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 stand-alone apparatus. The inspection apparatus may measure properties of a latent image (the image in the resist layer after exposure) or a semi-latent image (the image in the resist layer) on the layer after a post-exposure bake step (PEB), or on a developed resist image (where exposed or unexposed parts of the resist have been removed), or even on an etched image (after a pattern transfer step such as etching).

[0043] Figure 3 A schematic diagram depicting overall lithography represents the collaboration between the three technologies to optimize semiconductor manufacturing. Typically, the patterning process in the lithography apparatus LA is one of the most critical steps in the processing, requiring high-precision dimensioning and placement of structures on the substrate W ( Figure 1 ). To ensure this high level of accuracy, the three systems (in this case) can be combined in a so-called “holistic” control environment such as Figure 3Schematically shown in . One of these systems is a lithography apparatus LA, which is (virtually) connected to a metrology device (e.g., a metrology tool) MT (a second system) and a computer system CL (a third system). The "holistic" environment can be configured to optimize the cooperation between these three systems to enhance the overall process window and provide a tight control loop to ensure that the patterning performed by the lithography apparatus LA remains within the process range. The process window defines the range of process parameters (e.g., dose, focus, overlay). A specific manufacturing process produces a determined result (e.g., a functional semiconductor device) - typically, the process parameters in the lithography process or patterning process can vary.

[0044] The computer system CL can use (part of) the design layout to be patterned to predict which resolution enhancement techniques are used and perform computational lithography simulations and calculations to determine which mask layout and lithography equipment settings will achieve the maximum overall process window for the patterning process ( Figure 3 Typically, the resolution enhancement technique is arranged to match the patterning possibilities of the lithographic apparatus LA. The computer system CL may also be used to detect whether defects due to suboptimal processing, etc., are within the process window in which the lithographic apparatus LA is currently operating (e.g. using information from a geometry tracker (e.g. metrology tool MT) to predict whether defects due to suboptimal processing, etc., are present. Figure 3 Indicated by the arrow pointing to “0” in the second scale SC2).

[0045] The metrology device (tool) MT can provide input to the computer system CL for accurate simulation and prediction, and can provide feedback to the lithographic apparatus LA to identify possible drift, for example in the calibration state of the lithographic apparatus LA (in Figure 3 (indicated by multiple arrows in the third scale SC3).

[0046] In a lithography process, it is frequently necessary to measure the created structures, for example for process control and verification. Tools for performing such measurements include a metrology tool (device) MT. Different types of metrology tools MT are known for performing such measurements, including scanning electron microscopes or various forms of scatterometer metrology tools MT. A scatterometer is a versatile instrument that allows measuring parameters of a lithography process by placing a sensor in the pupil of a scatterometer objective or in a plane conjugate to the pupil, such measurements are usually referred to as pupil-based measurements, or by placing the sensor in an image plane or in a plane conjugate to the image plane, in which case the measurements are usually referred to as image- or field-based measurements. Such scatterometers and related 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. Such scatterometers can measure characteristics of a substrate, such as, for example, using gratings of soft x-rays and light in the visible to near infrared wavelength range.

[0047] In some embodiments, the scatterometer MT is an angle-resolved scatterometer. In these embodiments, scatterometer reconstruction methods may be applied to the measured signals to reconstruct or calculate properties of the grating and / or other features in the substrate. Such a reconstruction may be obtained, for example, by simulating the interaction of the scattered radiation with a mathematical model of the target structure and comparing the simulation results with the measurement 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.

[0048] In some embodiments, the scatterometer MT is a spectroscopic scatterometer MT. In these embodiments, the spectroscopic scatterometer MT can be configured so that radiation emitted by the scatterometer is directed from a radiation source to a target feature of the substrate, and radiation reflected or scattered by the target is directed to a spectrometer detector, which measures the spectrum of the specularly reflected radiation (i.e., a measurement of the intensity as a function of wavelength). From these data, the structure or profile of the target can be reconstructed to produce the detected spectrum, for example by rigorous coupled wave analysis and nonlinear regression or by comparison with a library of simulated spectra.

[0049] In some embodiments, the scatterometer MT is an ellipsometer. An ellipsometer allows the parameters of a lithography process to be determined by measuring the scattered radiation of each polarization state. Such a measurement device (MT) emits polarized light (e.g., linear, circular, or elliptical) by using, for example, an appropriate polarization filter in the illumination portion of the measurement device. A source suitable for the measurement device can provide polarized radiation. Various embodiments of existing ellipsometers 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.

[00214] In some embodiments, the scatterometer MT is adapted to measure the overlay of two misaligned gratings or periodic structures (and / or other target features of the substrate) by measuring a reflection spectrum and / or an asymmetry in the detection configuration, the asymmetry being related to the degree of overlay. The two (usually overlapping) grating structures may be applied to two different layers (not necessarily continuous layers) and may be formed at substantially the same position on the wafer. The scatterometer may have a symmetrical detection configuration, such as described in patent application EP1,628,164A, so that any asymmetry can be clearly distinguished. This provides a way of measuring grating misalignment. Further examples of measuring overlay may be found in PCT patent application publication number WO2011 / 012624 or in US patent application US20160161863, the entire contents of which are incorporated herein by reference.

[0050] Other parameters of interest may be focus and dose. Focus and dose can be determined simultaneously by scatterometry (or by scanning electron microscopy) measurements, as described in U.S. Patent Application US2011-0249244, which is incorporated herein by reference in its entirety. A single structure (e.g., a feature in a substrate) can be used that has a unique combination of critical dimensions and sidewall angle measurements for each point in a focus energy matrix (FEM—also called a focus exposure matrix). If a unique combination of these critical dimensions and sidewall angles is available, then focus and dose values ​​can be uniquely determined by these measurements.

[0051] The metrology target may be a collection of composite gratings and / or other features in a substrate, formed by a lithography process, typically in a resist, but may also be formed after an etching process. In some embodiments, one or more target groups may be gathered at different locations around the wafer. Typically, the pitch and line width of the structures in the gratings depend on the ability of the measurement optics (particularly the NA of the optics) to capture diffraction orders from the metrology target. The diffraction signal may be used to determine the offset between two layers (also known as "overlay"), or may be used to reconstruct at least a portion of the original grating produced by the lithography process. This reconstruction may be used to provide guidance on the quality of the lithography process and may be used to control at least a portion of the lithography process. The target may have smaller sub-segments that are configured to mimic the size of a functional portion of the design layout in the target. Due to this sub-segmentation, the behavior of the target will be more similar to that of the functional portion of the design layout, making the overall process parameter measurement similar to that of the functional portion of the design layout. The target may be measured in an underfill mode or an overfill mode. In the underfill mode, the measurement beam produces a smaller spot than the overall target. In the overfill mode, the measurement beam produces a spot that is larger than the overall target. In this overfill mode, different targets may also be measured simultaneously, thereby determining different processing parameters simultaneously.

[0052] The overall quality of the measurement of the lithography parameters using a specific target depends at least in part on the measurement scheme used to measure the lithography parameters. The term "substrate measurement recipe" may include one or more parameters of the measurement itself, one or more parameters of the one or more patterns measured, or both. For example, if the measurement used in the substrate measurement recipe is an optical measurement based on diffraction, the one or more parameters measured may include the wavelength of the radiation, the polarization of the radiation, the angle of incidence of the radiation relative to the substrate, the direction of the radiation relative to the pattern on the substrate, etc. One of the criteria for selecting a measurement recipe may be, for example, the sensitivity of one of the measurement parameters to processing changes. More examples are described in U.S. Patent Application US2016-0161863 and Published U.S. Patent Application US2016 / 0370717A1. The entire contents are incorporated herein by reference.

[0053] Figure 4 An exemplary metrology device (tool or platform) MT is shown, such as a scatterometer. MT comprises a broadband (white light) radiation projector 40 which projects radiation onto a substrate 42. The reflected or scattered radiation is passed to a spectrometer detector 44 which measures a spectrum 46 of the specularly reflected radiation (i.e. a measure of the intensity as a function of wavelength). From this data, the structure or profile that caused the detected spectrum to be determined can be reconstructed 48 by a processing unit PU, for example by rigorous coupled wave analysis and nonlinear regression or by comparing it to a spectrum obtained by a process. Figure 3The simulated spectral library shown at the bottom is used for comparison. Typically, for the reconstruction, the general form of the structure is known and some parameters are assumed from knowledge of the structure fabrication process, leaving only a few parameters of the structure to be determined from the scatterometry data. Such a scatterometer can be configured, for example, as a normal incidence scatterometer or an oblique incidence scatterometer.

[0054] Often, one would like to be able to computationally determine how a patterning process produces a desired pattern on a substrate. For example, computational determination may include simulation and / or modeling. Models and / or simulations may be provided for one or more portions of a manufacturing process. For example, one would like to be able to simulate a lithography process to transfer a pattern of a pattern forming device to a resist layer of a substrate, as well as the pattern produced in the resist layer after the resist is developed, simulate metrology operations (e.g., determining overlay), and / or perform other simulations. The purpose of the simulation may be to accurately predict metrology metrics (e.g., overlay, critical dimensions, reconstruction of a three-dimensional profile of a substrate feature, dose or focal length of a lithography device at a certain moment when a substrate feature is printed by a lithography device, etc.), manufacturing process parameters (e.g., edge placement, spatial image intensity slope, sub-resolution assist features (SRAF), etc.), and / or other information, which may then be used to determine whether an intended or target design has been achieved. The intended design is typically defined as a pre-optical proximity correction design layout, which may be provided in a standardized digital file format, such as GDSII, OASIS, or other file formats.

[0055] Simulation and / or modeling may be used to determine one or more metrology metrics (e.g., perform overlay and / or other metrology measurements), configure one or more features of a pattern forming device pattern (e.g., perform optical proximity correction), configure one or more features of an illumination (e.g., change one or more characteristics of a spatial / angular intensity distribution, optimize illumination, such as changing a shape), configure one or more features of a projection optics (e.g., numerical aperture, etc.), and / or for other purposes. Such determination and / or configuration may generally be referred to as mask optimization, source optimization, and / or projection optimization. Such optimizations may be performed individually or combined in various combinations. One such example is source mask optimization (SMO), which involves configuring one or more features of a pattern forming device pattern and one or more features of an illumination. For example, the optimization may use the parameterized models described herein to predict values ​​of various parameters (including images, etc.).

[0056] In some embodiments, the optimization process of the system can be expressed as a cost function. The optimization process can include finding a set of system parameters (design variables, process variables, inspection operation variables, etc.) that minimize the cost function. The cost function can have any suitable form, depending on the optimization target. For example, the cost function can be the weighted root mean square (RMS) of the deviation of certain characteristics (evaluation points) of the system relative to the expected values ​​(e.g., ideal values) of these characteristics. The cost function can also be the maximum value (i.e., the worst deviation) among these deviations. The term "evaluation point" should be broadly interpreted as including any characteristic of the system or manufacturing method. The design and / or process variables of the system can be limited to a limited range and / or interdependent due to the actual situation of the implementation of the system and / or method. In the case of lithographic projection and / or inspection devices, the constraints are usually related to the properties and characteristics of the hardware, such as adjustable range and / or pattern forming device manufacturability design rules. The evaluation points can include physical points on the resist image on the substrate, as well as non-physical characteristics, such as dose and focal length.

[0057] The inference model can be used to determine one or more characteristics of a product of a manufacturing process based on measurements of the product. Examples of such inference models will be described below with reference to wafers obtained from a photolithography process, but it should be understood that other inference models can be used for products other than wafers or products of manufacturing processes other than photolithography processes. The one or more parameters determined by the inference model can characterize the height or depth of a feature, the width of a feature of a structure (e.g., a periodic structure) on a wafer, etc.

[0058] The training of inference models is affected by the location of measurement targets on the product. For example, if the measurement targets used to train the inference model are concentrated in areas of the product where the attribute variations are small, then the trained inference model may not be able to well determine the parameters that characterize the variations of these attributes in other parts of the product. In other words, the inference model may be biased to not accurately capture the variations in product attributes that may appear in certain areas of the product that are sparsely represented in the dataset. In this case, increasing the number of measurements / measurement targets does not necessarily improve the accuracy of the inference model and may make it difficult to train the model.

[0059] The wafer is typically equipped with a large number of measurement targets, and a data set containing the measurement results of each target is typically used when training the inference model, although some filtering may be performed on the data set to remove failed and / or outlier measurement results. The large size of the data set means that a large amount of computing resources are required to train the inference model. The inference model can typically be an artificial neural network, such as a deep neural network and / or a convolutional neural network, which takes the measurement results of the product as input and provides one or more parameters of the product as output. The artificial neural network may, for example, include an autoencoder. The autoencoder can be configured to compress the measurement result data set (e.g., pupil image) into an effective low dimension which can then be used for parameter inference (i.e., regression).

[0060] As described above, the measurement target may include a diffraction grating including parallel lines formed on the surface of a wafer, the lines having sidewalls extending in a depth direction perpendicular to the surface of the wafer. Variations in the manufacturing process may cause the sidewalls of the lines to tilt a small amount relative to the depth direction, i.e., the sidewalls deviate from being perpendicular to the surface of the wafer. This deviation may be referred to as "tilt" or "pattern tilt" and typically has a magnitude and direction that varies across the surface of the wafer. The tilt may be measured from an image of the measurement target obtained in the pupil plane of the scatterometer. The tilt measurements obtained in this manner may then be used to train an inference model to determine the tilt of the wafer based on a pupil plane image of the measurement target provided on the wafer. For example, a measurement data set suitable for training an inference model may be obtained by producing wafers under different etching conditions, such that the tilt of the wafer varies significantly throughout the data set. For some data sets, the tilt inference model may be difficult to train accurately because the largest magnitude of tilt occurs near the edge of the wafer, so using additional measurements of measurement targets away from the edge of the wafer has little effect on improving the accuracy of the inference model in areas where the tilt is more pronounced, and may instead reduce the model's ability to accurately determine the tilt of the wafer edge.

[0061] Another relevant parameter of a wafer (or more generally, a product obtained by a lithographic process) relates to the displacement (or "overlay") of one structure on the wafer in a direction lateral to the wafer surface relative to another structure spaced from the structure in a depth direction perpendicular to the surface. Measurements that indicate such displacements of pairs of structures may be referred to as overlay measurements. These measurements may be used to train an inference model to determine one or more parameters that characterize wafer overlay from other such measurements.

[0062] As described herein, instead of using measurements from all measurement targets to train an inference model, a subset of a dataset may be used to train an inference model, i.e., a "proper" subset of the dataset that contains some but not all (and typically much fewer than the dataset) of the measurements in the dataset. The subset is preferably selected to effectively capture the distribution of measurements within the dataset. For example, the measurements in the subset may be obtained based on the importance of the measurements in the subset in representing the overall information contained in the dataset, while measurements that are relatively uninformative about the distribution may be omitted from the subset.

[0063] The measurement data set can be obtained through the matrix D N (or more generally, a tensor), where N is the number of measurements in the data set, the matrix has N columns, and the values ​​of each column together define a measurement, i.e., each measurement is represented by a column vector of values. The problem of selecting a subset SM of measurements that best represents the N measurement data set (where M is the number of measurements in the subset; M is an integer less than the integer N) can be defined by the following optimization problem:

[0064]

[0065] where "argmin" is the independent variable (i.e., subset) corresponding to the minimum value of the matrix norm in the equation, is a projection operator that projects each measurement in the data set onto the measurement in the subset, and ‖·‖ represents the norm operation. In a simple example, the projection operator can be a linear operator. In this case, the optimal subset is the subset whose column vector of each measurement result in the data set can be most accurately represented by a linear combination of the column vectors of the measurement results in the subset, that is, where D N ≈D M ×W,D M is a matrix of column vectors for each measurement result in the subset, W is a matrix of column vectors, each column vector defines D M The most suitable for D N The matrix norm used to provide a measure of the difference (“projection error”) between each column vector in the dataset and its projection onto the column vector of the subset can in this case be the Frobenius norm, i.e., from D N Each element in the matrix is ​​related to its D N The sum of the squares of the differences between corresponding elements in the projection. However, other matrix norms can be used in some cases. It is also possible to define a generalized version of the above equation where the measurements in the dataset are projected onto the measurements in the subset using a nonlinear mapping instead of a linear projection operator.

[0066] Due to the huge number of measurements in most data sets, solving equation (1) to find the best subset of measurement results is often challenging. Despite this, there are still a variety of selection algorithms to obtain the best (or near-optimal) subset. One such selection algorithm is called "kernel spectrum tracking", which iteratively selects M column vectors (measurements) that span close to the first M spectral components of the data set. The algorithm is described in "Select to Better Learn: Fast and Accurate Deep Learning Using Data Selection From Nonlinear Manifolds" by M. Joneidi et al., 2020 IEEE / CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 7816-7826, doi: 10.1109 / CVPR42600.2020.00784, the contents of which are incorporated herein by reference.

[0067] Figure 5 is a flow chart showing steps of a method 500 for training an inference model to determine one or more parameters of a product of a manufacturing process (e.g., a lithography process) based on measurement results of the product. As an example, the product may be a wafer produced by the lithography process, and the inference model may be trained to determine one or more parameters of the wafer, such as parameters characterizing tilt, based on an image of a metrology target on the wafer obtained in a pupil plane of a scatterometer.

[0068] The first step 502 of method 500 includes obtaining a measurement data set of one or more products in a manufacturing process, each measurement result comprising an array of values ​​obtained by measuring the corresponding product.

[0069] The second step 504 (which may be referred to as "subset selection") includes selecting an appropriate subset of the data set for training the inference model, the selection of the subset being achieved by applying an optimization procedure to an objective function that provides a difference measure between each measurement in the data set and a corresponding reproduction value of the measurement obtained using a reproduction function, the reproduction function having a domain that includes the measurements in the subset and excludes the measurements not included in the subset. For example, kernel spectral tracking may be applied to the objective function defined by equation (1) above to select the measurements in the subset.

[0070] A third step 506 of method 500 may include training an inference model using the measurements in the subset, preferably in conjunction with known values ​​of one or more parameters for each of the one or more products, ie, by supervised learning.

[0071] Figure 6A top view of an exemplary wafer 600 is shown with the locations of metrology targets on the wafer surface marked from which measurements can be obtained. For example, each measurement can be an image of the metrology target obtained in the pupil plane of the scatterometer. For example, the measurement can indicate the tilt at the location of the metrology target. The surface of the wafer 500 is subdivided into four equal quadrants AD, such that each metrology target belongs to one of the quadrants. Note that in alternative embodiments, the surface of the wafer can be subdivided into regions of circular and / or annular sectors defined relative to the normal axis of the wafer surface or layer.

[0072] Figure 7 FIG. 6 is a schematic diagram of how to select measurement results of a wafer 600 for training an inference model according to an embodiment of the present invention.

[0073] The measurement results of the wafer together form a data set DS. The measurement results are then grouped into four non-overlapping groups, GA, GB, GC, and GD. Then, for each group, a subset of the measurement results in the group is selected. Each subset SS-A, SS-B, SS-C, and SS-D is selected by solving equation (1) as described above (the data set is replaced by the measurement results belonging to the group). Then, the subsets SS-A, SS-B, SS-C, and SS-D of each group are combined together, for example by union of the subsets, to obtain a subset SS of the entire data set for training the inference model.

[0074] Grouping the measurements and then applying subset selection to the measurements in each group can ensure, for example, that the measurements selected for training the inference model are representative of each group. This approach can ensure that feature measurements from each group are not omitted from the subset used to train the inference model. For example, the lithography process used to produce wafer 500 may be known or expected to produce features with skewed variations in distribution across the wafer. Figure 6 In the case of wafer 600, these variations may result in different tilt distributions in each of the four quadrants. By grouping the measurements by quadrant, measurements that are representative of the tilt distribution within each quadrant can be selected, thereby ensuring that the inference model is appropriately trained for the measurements in each quadrant.

[0075] More generally, measurements can be grouped according to metadata associated with the measurements. For example, metrology targets can be arranged in clusters on a wafer surface (e.g., clusters can be associated with Figure 6 The measurements may be grouped according to metadata indicating which cluster the measurements were obtained from. For example, such an approach may allow the best metrology target in each cluster to be selected for training an inference model.

[0076] Figure 8 is a schematic diagram of how to select measurement results of a wafer for training an inference model according to an embodiment of the present invention, wherein a subset selection method can be used to identify which clusters are important for representing a data set (e.g., which clusters are responsible for most of the variation in the data set). The measurement results in the data set DS can be grouped according to which cluster the measurement results come from to obtain multiple subsets, one for each cluster, C1-CN. The measurement results within each cluster subset C1-CN can then be connected so that the data set can be represented by a matrix having corresponding columns formed by the connected measurement results of each cluster subset C1-CN. That is, each column is a connected measurement result of one of the cluster subsets. A subset selection method can then be applied to select a subset of columns SS that best represents the data set DS. For example, this method can allow the selection of the best cluster (or combination thereof) of the measurement target to train the inference model. Additional subset selection can be applied to the measurement results in each selected cluster by applying subset selection to each selected cluster group separately and / or applying subset selection across the selected cluster group.

[0077] Fig. 9 is a schematic diagram of how to select measurement results of a wafer for training an inference model according to one embodiment of the present invention, wherein the measurement process for obtaining the measurement results in the data set DS may have multiple acquisition channels AC1, AC2. The acquisition channels may include, for example, different wavelengths and / or different polarizations of electromagnetic radiation, which the scatterometer may use to obtain the measurement results. Of course, the measurement results of different acquisition channels may be grouped, for example, based on whether the measurement results are from the same area or measurement target. In some examples, the measurement results of different acquisition positions may be connected to produce a set of connected measurement results CC, to which subset selection may be applied. A subset selection method may then be applied to select a subset of columns SS that best represents the data set DS.

[0078] In another example, measurements taken using different acquisition channels can be grouped by channel, and a subset selection method applied to identify which measurements in each channel are most representative. This means that for some locations on the product, all channels are included in the appropriate subset of measurements; for other locations, only an appropriate subset of channels are included in the appropriate subset of measurements; and for other locations, no channels are included in the appropriate subset of measurements. Alternatively, or in addition, the measurements in each channel can be concatenated together, and subset selection can be collectively applied to the concatenated channel measurements in order to identify which channel best represents the variation of the entire data set. For example, such an approach can allow the best channel (or combination thereof) of a measurement process to be selected to train an inference model.

[0079] Fig. 10AA representation of the tilt values ​​measured from each of a plurality of regions of a wafer is shown. The tilt values ​​are determined by training a tilt inference model on measurements from a plurality of wafers, with some of the wafers used for training and others used for validation. The measurements are pupil plane images obtained using a scatterometer with two acquisition channels.

[0080] Fig. 10B A graph is shown in which the horizontal axis represents the radial distance from the center of the wafer (in arbitrary units) and the vertical axis represents the inferred tilt value (also in arbitrary units) obtained using the trained tilt inference model. Negative tilt values ​​mainly come from Fig. 10A The positive tilt values ​​are obtained from the lower left quadrant of the wafer shown in , while the positive tilt values ​​are mainly obtained from the upper right quadrant.

[0081] Fig. 10C and Fig. 10D and Fig. 10A and Fig. 10B Similar, except that only the tilt values ​​for a subset of wafer measurements are shown. This subset is obtained by selecting the 60 most important (i.e., most representative) measurements for each channel for each wafer used to train the inference model. The selected measurements are then combined (by taking their union) to form a subset. The number of measurements in this subset is approximately 5% of the original data set.

[0082] Fig.10E and Fig.10F and Fig. 10A and Fig. 10B Similar, except that the inclination values ​​are obtained using an inclination inference model trained using only a subset of the measurements. It should be noted that because each inference model in these examples outputs inclination values ​​in (different) arbitrary units, the magnitude of the inclination values ​​determined by the two inclination inference models cannot be directly compared. However, it can be seen that Fig. 10B and Fig.10F The respective shapes of the envelopes of the medium tilt values ​​are very close to each other, and very good agreement is observed between the tilt values ​​determined by the inference model trained using the full dataset and the corresponding tilt values ​​determined by the inference model trained using only the subset.

[0083] Once the optimal (or near-optimal) subset of measurement results has been determined, the information obtained from selecting the subset of measurement results can be used to ensure that only these specific measurements (or measurements very similar to them) are made when training subsequent inference models. Specifically, a subset of optimal measurement results can be provided for a particular parameter or parameter type that characterizes a wafer when a new inference model for that parameter (or parameter type) needs to be trained. Information characterizing such an optimal subset of measurement results can be provided, for example, in the form of a list of measurement targets from which measurements should be made, the list identifying only a subset of the measurement targets provided on the wafer. Optionally, the list can be different depending on the parameters determined by the inference model, for example, the list may be different for training a tilt inference model compared to training an overlay inference model. In some implementations, the subset selection method can be used to reduce the number of measurement targets provided on a wafer, for example by including only the measurement targets selected for the subset on subsequent wafers.

[0084] Fig.11 1 is a block diagram illustrating a computer system 100 that may perform and / or assist in implementing the methods, processes, systems, or devices disclosed herein. The computer system 100 includes a bus 102 or other communication mechanism for communicating information, and a processor 104 (or multiple processors 104 and 105) coupled to the bus 102 for processing information. The computer system 100 also includes a main memory 106, such as a random access memory (RAM) or other dynamic storage device, coupled to the bus 102 for storing information and instructions executed by the processor 104. The main memory 106 may also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processor 104. The computer system 100 also includes a read-only memory (ROM) 108 or other static storage device coupled to the bus 102 for storing static information and instructions for the processor 104. A storage device 110, such as a magnetic disk or optical disk, is provided and coupled to the bus 102 for storing information and instructions.

[0085] The computer system 100 may be coupled via the bus 102 to a display 112, such as a cathode ray tube (CRT) or a flat panel or touch panel display, for displaying information to a computer user. An input device 114, including alphanumeric and other keys, is coupled to the bus 102 for communicating information and command selections to the processor 104. Another type of user input device is a cursor control 116, such as a mouse, trackball, or cursor direction keys, for communicating directional information and commands to the processor 104 to receive selections and control cursor movement on the display 112. The input device typically has two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), allowing the device to specify a position in a plane. A touch panel (screen) display may also be used as an input device.

[0086] According to one embodiment, portions of one or more methods described herein may be performed by computer system 100 in response to processor 104 executing one or more sequences of one or more instructions contained in main memory 106. These instructions may be read into main memory 106 from another computer-readable medium (e.g., storage device 110). Execution of the sequences of instructions contained in main memory 106 causes processor 104 to perform the processing steps described herein. One or more processors in a multi-processing arrangement may also be used to execute the sequences of instructions contained in main memory 106. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions. Therefore, the description herein is not limited to any specific combination of hardware circuitry and software.

[0087] As used herein, the term "computer-readable medium" or "machine-readable" refers to any medium that participates in providing instructions to processor 104 for execution. Such media can take a variety of forms, including but not limited to non-volatile media, volatile media, and transmission media. Non-volatile media include, for example, optical or magnetic disks, such as storage device 110. Volatile media include dynamic memory, such as main memory 106. Transmission media include coaxial cables, copper wires, and optical fibers, including wires that make up bus 102. Transmission media can also take the form of sound waves or light waves, such as in radio frequency (RF) and infrared (IR) data communications. Common forms of computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs, any other optical media, punch cards, paper tapes, any other physical media with hole patterns, RAM, PROMs and EPROMs, FLASH-EPROMs, any other memory chips or cartridges, carrier waves described below, or any other computer-readable media.

[0088] Various forms of computer readable media may be used to carry one or more sequences of one or more instructions to processor 104 for execution. For example, the instructions may initially be carried on a disk of a remote computer. The remote computer may load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. The local modem computer system 100 may receive the data on the telephone line and use an infrared transmitter to convert the data to an infrared signal. An infrared detector coupled to bus 102 may receive the data carried in the infrared signal and place the data on bus 102. Bus 102 carries the data to main memory 106, from which processor 104 retrieves and executes the instructions. The instructions received by main memory 106 may optionally be stored on storage device 110 before or after execution by processor 104.

[0089] The computer system 100 may also include a communication interface 118 coupled to the bus 102. The communication interface 118 provides a two-way data communication coupled to a network link 120 connected to a local network 122. For example, the communication interface 118 may be an integrated services digital network (ISDN) card or a modem to provide a data communication connection to a corresponding type of telephone line. As another example, the communication interface 118 may be a local area network (LAN) card to provide a data communication connection to a compatible LAN. A wireless link may also be implemented. In any such implementation, the communication interface 118 sends and receives electrical, electromagnetic, or optical signals carrying digital data streams representing various types of information.

[0090] The network link 120 typically provides data communication through one or more networks to other data devices. For example, the network link 120 may provide a connection to a host computer 124 or to data equipment operated by an Internet Service Provider (ISP) 126 through a local network 122. The ISP 126, in turn, provides data communication services through a global packet data communication network, now commonly referred to as the "Internet" 128. Both the local network 122 and the Internet 128 use electrical, electromagnetic or optical signals that carry digital data streams. The signals on the various networks and network links 120 and through the communication interface 118 that carry the digital data from the computer system 100 are exemplary forms of carrier waves that transport the information.

[0091] The computer system 100 can send messages and receive data (including program code) through the network, network link 120, and communication interface 118. In the Internet example, the server 130 can transmit the requested code of the application program through the Internet 128, ISP 126, local network 122, and communication interface 118. For example, one such downloaded application program can provide all or part of the methods described herein. The received code can be executed when received by the processor 104, and / or stored in the storage device 110 or other non-volatile memory for later execution. In this way, the computer system 100 can obtain the application program code in the form of a carrier wave.

[0092] Additional embodiments are disclosed in the following list of numbered clauses:

[0093] 1. A method for training an inference model for determining one or more parameters of a product of a manufacturing process based on measurement results of the product, the method comprising:

[0094] obtaining a measurement data set of one or more products in the manufacturing process, each measurement result comprising an array of values ​​obtained by measuring a corresponding one of the products;

[0095] selecting an appropriate subset of the data set for training the inference model, the subset being selected by applying an optimization procedure to an objective function that provides a measure of difference between each measurement in the data set and a corresponding reproduced value of the measurement obtained using a reproduction function, the reproduction function having a domain that includes the measurements in the subset and excludes the measurements not in the subset; and

[0096] The inference model is trained using an appropriate subset of the dataset.

[0097] 2. The method of clause 1, wherein selecting the appropriate subset of the dataset for training the inference model comprises:

[0098] assigning each of the measurement results of the data set to one or more of a plurality of groups based on metadata associated with one or more of: the measurement results; the product corresponding to the measurement results; and the manufacturing process for producing the product corresponding to the measurement results; and

[0099] The optimization procedure is applied to select measurements from each group respectively such that the subset includes measurements from each of the groups.

[0100] 3. The method of clause 2, wherein applying the optimization procedure to select measurements from each of the groups separately comprises:

[0101] for each of the one or more groups, selecting a subset of the measurements assigned to the group by applying the optimization procedure to optimize the objective function of the measurements assigned to the group over a subdomain including the reproduction function of the measurements assigned to the group and excluding the measurements not assigned to the group; and

[0102] The measurements from two or more respective subsets of the group are combined to select a subset of the data set for training the inference model.

[0103] 4. A method according to clause 3, wherein the training data is obtained from the union or the intersection of two or more corresponding subsets of the group.

[0104] 5. The method of clause 1, wherein selecting an appropriate subset of the dataset for training the inference model comprises:

[0105] assigning each measurement result of the data set to one or more of a plurality of groups based on metadata associated with one or more of: the measurement result; the product corresponding to the measurement result; and the manufacturing process for producing the product corresponding to the measurement result; and

[0106] The optimization procedure is applied to select one or more measurement groups such that the subset includes measurements from each of the one or more selected groups and excludes measurements from each of the one or more unselected groups.

[0107] 6. The method of clause 5, wherein applying the optimization procedure to select one or more measurement groups comprises:

[0108] For each group, combining the measurement results assigned to the group to obtain the connection measurement result array corresponding to the group; and

[0109] selecting a subset of the connected arrays by applying the optimization procedure to optimize the objective function, the objective function providing a measure of the difference between the value of each connected array and the corresponding value of the reproduction of the connected array obtained using the reproduction function, the domain of the reproduction function including the connected arrays of the measurement results in the subset and excluding the connected arrays of the measurement results not in the subset; and

[0110] The dataset subset for training the inference model is selected in conjunction with the measurements of the connection arrays in the subset.

[0111] 7. A method according to any of clauses 2 to 6, wherein the metadata indicates from which of a plurality of areas of the product the measurement result was obtained.

[0112] 8. A method according to clause 7, wherein the region is the angular sector, such as the quadrant, defined relative to the axis perpendicular to the surface or layer of the product.

[0113] 9. The method according to clause 7, wherein the area is the circular and / or annular sector defined relative to the axis perpendicular to the surface or layer of the product.

[0114] 10. A method according to any of clauses 2 to 9, wherein the metadata indicates from which of a plurality of the metrology targets formed on or within the product the measurement result was obtained.

[0115] 11. A method according to any of clauses 2 to 10, wherein the metadata indicates from which of the products the measurements were obtained.

[0116] 12. The method according to any of clauses 2 to 11, wherein the metadata indicates which of the plurality of acquisition channels of the measurement process was used to obtain the measurement result.

[0117] 13. A method according to clause 12, wherein each acquisition channel corresponds to one or more wavelengths of the electromagnetic radiation used in the measurement process and / or the polarization and / or a corresponding rotation of the product around the axis related to the measurement process, such as the imaging direction.

[0118] 14. A method according to any of the preceding clauses, wherein the manufacturing process is the photolithography process.

[0119] 15. The method of clause 14, wherein the metadata indicates the etching conditions used in the photolithography process.

[0120] 16. A method according to any preceding clause, wherein the manufacturing process is the wafer manufacturing process and each of the one or more products is the manufactured wafer.

[0121] 17. A method according to any preceding clause, wherein each product comprises a plurality of said layers spaced apart in said depth direction, and the one or more parameters of said product determined by said inference model comprise one or more numerical parameters respectively indicating one of the following:

[0122] said translational offset transverse to said depth direction of two said structures in different corresponding said layers in said product;

[0123] said angle between one or more walls of said structure and said depth direction;

[0124] The transverse angular offset between the length direction and the depth direction of each of two of the elongated structures in different corresponding layers in the product; or

[0125] The distance between two of the structures in the product in the depth direction.

[0126] 18. A method according to any of the preceding clauses, wherein each measurement comprises one or more pupil images corresponding to the product captured in the pupil plane of the scatterometer.

[0127] 19. A method according to any preceding clause, wherein the number of measurements in the subset is less than 50%, preferably less than 20%, more preferably less than 10% of the number of measurements in the data set.

[0128] 20. A method according to any of the preceding clauses, wherein the training of the inference model uses a combination of the subset of the data set and known values ​​of one or more parameters for each of one or more products.

[0129] 21. The method of any of the preceding clauses, further comprising using the trained inference model to determine one or more parameters of the product of the manufacturing process based on the measurement results of the product.

[0130] 22. The method of clause 21, further comprising adjusting the manufacturing process in response to one or more parameters determined by the trained inference model.

[0131] 23. A method according to any of the preceding clauses, wherein the inference model comprises the artificial neural network, preferably the autoencoder.

[0132] 24. A method according to any of the preceding clauses, wherein the data set is the first data set, the method further comprising training the second inference model to determine one or more parameters of the manufacturing process product based on the measurement results of the manufacturing process product, wherein training the second inference model comprises:

[0133] acquiring a second data set of the measurement results of one or more products in the manufacturing process, each measurement result in the second data set corresponding to a respective one of the measurement results in the subset selected from the first data set; and

[0134] The second inference model is trained using the measurements in the second data set.

[0135] 25. The method of clause 24, wherein the second inference model is trained using the measurements in the second data set in combination with known values ​​of one or more parameters for each of one or more products of the second manufacturing process.

[0136] 26. A method according to clause 24, wherein the second inference model is trained using the measurement results in the second data set in combination with values ​​of one or more parameters of each of one or more products determined by the first inference model based on the measurement results in the second data set.

[0137] 27. A method according to any one of clauses 24 to 26, wherein each of the measurement results of the second data set and the measurement results of the first data set corresponding to the measurement results of the second data set are obtained from one or more of: the corresponding area of ​​the product; the corresponding measurement target on or in the product; and / or the corresponding acquisition channel of the measurement process.

[0138] 28. A method for determining one or more parameters of a product of a manufacturing process based on measurement results of the product, the method comprising:

[0139] obtaining a data set of measurements of one or more products of the manufacturing process, each measurement result comprising an array of values ​​obtained by measuring the corresponding product, wherein each measurement result in the data set corresponds to a respective one of a plurality of measurement results selected from another data set of measurements of the one or more products of the manufacturing process, each measurement result comprising an array of values ​​obtained by measuring the corresponding product of the manufacturing process, the subset being selected by applying an optimization procedure to a target function region that provides a measure of difference between each measurement result in the another data set and a corresponding reproduced value of the measurement result obtained using a reproduction function, the reproduction function having a range that includes the measurements in the subset and excludes the measurements not in the subset; and

[0140] One or more parameters of each of the one or more products are determined using an inference model trained using measurements of the data set and / or measurements of an appropriate subset of another data set.

[0141] 29. The method of clause 28, wherein the inference model is trained using measurements in the data set in combination with known values ​​of one or more parameters for each of one or more products of the manufacturing process.

[0142] 30. A method according to clause 28, wherein the measurements in the data set are used in combination with an inference model trained using the measurements in the subset, the inference model being derived from values ​​of one or more parameters for each of the one or more products determined based on the measurements in the data set.

[0143] 31. A method according to any one of clauses 28 to 30, wherein each measurement result and another measurement result corresponding to the measurement result are obtained from one or more of: a corresponding area of ​​the product; a corresponding measurement target on or in the product; and / or a corresponding acquisition channel of the measurement process.

[0144] 32. A method according to clause 27 or 31, wherein the measuring process comprises an imaging process having an imaging direction, and the acquisitions of the measuring process each indicate:

[0145] the orientation of one or more products relative to the imaging direction;

[0146] The spacing between repeating elements of a periodic structure (e.g., a diffraction grating) formed on or in one or more products;

[0147] the translational position of one or more products transverse to the imaging direction; or

[0148] The wavelength and / or polarization of the electromagnetic radiation used in the imaging process.

[0149] 33. A computing system comprising a processor and a memory, the memory storing program instructions operable when executed by the processor to cause the processor to perform the method of any of clauses 1 to 24.

[0150] 34. A metrology system for determining one or more parameters of a product of a manufacturing process based on a measurement signal characterizing the product, the metrology system comprising a computing system according to clause 33 and a metrology device configured to measure the product to obtain the measurement signal.

[0151] 35. A computer program product storing program instructions which, when executed by the processor, cause the processor to perform the method according to any one of clauses 1 to 24.

[0152] As used herein, the terms "radiation" and "beam" encompass all types of electromagnetic radiation, including ultraviolet (UV) radiation (e.g., having a wavelength of around 365 nm, 355 nm, 248 nm, 193 nm, 157 nm, or 126 nm) and extreme ultraviolet (EUV) radiation (e.g., having a wavelength in the range of 5-20 nm), as well as particle beams, such as ion beams or electron beams.

[0153] The term "lens," where the context permits, may refer to any one or combination of various types of optical elements, including refractive, reflective, magnetic, electromagnetic, and electrostatic optical elements.

[0154] The term "target" should not be construed to refer only to specialized targets formed for specific metrology purposes. The term "target" should be understood to encompass other structures, including product structures, having properties suitable for metrology applications.

[0155] The description of the above specific embodiments will fully reveal the general nature of the present invention, and others can easily modify and / or adjust these specific embodiments for various applications without departing from the general concept of the present invention by applying knowledge within the technical scope of the art, without excessive experimentation. Therefore, based on the teachings and guidance provided herein, these adjustments and modifications are intended to belong to the meaning and scope of the equivalents of the disclosed embodiments. It should be understood that the wording or terminology herein is for illustration, not for limitation, and therefore, those skilled in the art should interpret the terms or wording of this specification in accordance with the teachings and guidance.

[0156] The breadth and scope of the present invention should not be limited by any of the above-described exemplary embodiments, but should be defined only in accordance with the following claims and their equivalents.

Claims

1. A method for training an inference model for determining one or more parameters of a product of a manufacturing process based on measurement results of the product, the method comprising: acquiring a measurement data set of one or more products of the manufacturing process, each of the measurement results comprising an array of values ​​obtained by measuring a corresponding one of the products; selecting an appropriate subset of the data set for use in training the inference model, the subset being selected by applying an optimization procedure to an objective function that provides a measure of difference between each measurement in the data set and a corresponding reproduced value of the measurement obtained using a reproduction function, the reproduction function having a domain that includes the measurements in the subset and excludes the measurements not in the subset; as well as The inference model is trained using the appropriate subset of the dataset.

2. The method of claim 1 , wherein selecting the appropriate subset of the dataset for training the inference model comprises: assigning each of the measurements of the data set to one or more of a plurality of groups based on metadata associated with one or more of: the measurements; said product corresponding to said measurement result; and said manufacturing process for producing said product corresponding to said measurement result; as well as The optimization procedure is applied to select measurements from each of the groups, respectively, such that the subset includes measurements from each of the groups.

3. The method of claim 2, wherein applying the optimization procedure to select measurements from each of the groups separately comprises: for each of the one or more groups, selecting a subset of the measurement results assigned to the group by applying the optimization procedure to optimize the objective function of the measurement results assigned to the group over a subdomain of the reproduction function including the measurement results assigned to the group and excluding the measurement results not assigned to the group; as well as Measurements from two or more corresponding subsets of the set are combined to select the subset of the data set for training the inference model.

4. The method of claim 3, wherein the training data is obtained from a union or intersection of the two or more corresponding subsets of the group.

5. The method of claim 1 , wherein selecting the appropriate subset of the data set for training the inference model comprises: assigning each of the measurements of the data set to one or more of a plurality of groups based on metadata associated with one or more of: the measurements; said product corresponding to said measurement result; and said manufacturing process for producing said product corresponding to said measurement result; as well as The optimization procedure is applied to select one or more of the groups of measurements such that the subset includes measurements from each of the one or more selected groups and excludes measurements from each of the one or more unselected groups.

6. The method of claim 5, wherein applying the optimization procedure to select one or more of the groups of measurements comprises: for each of the groups, combining the measurement results assigned to the group to obtain a corresponding concatenated array of measurement results for the group; as well as selecting a subset of the connected arrays by applying the optimization procedure to optimize the objective function, the objective function providing a measure of the difference between a value of each connected array and a corresponding reproduced value of the connected array obtained using the reproduction function, the domain of the reproduction function including the connected arrays of the measurements in the subset and excluding the connected arrays of measurements not in the subset; as well as The measurements from the connection arrays in the subset are combined to select the subset of the data set for training the inference model.

7. The method of any one of claims 2 to 6, wherein the metadata indicates from which region of a plurality of regions of the product the measurements were obtained.

8. The method according to any one of claims 2 to 7, wherein the metadata indicates from which metrology target of a plurality of metrology targets formed on or within the product the measurement result was obtained.

9. The method according to any one of claims 2 to 8, wherein the metadata indicates from which product the measurement result was obtained.

10. The method according to any one of claims 2 to 9, wherein the metadata indicates which acquisition channel of a plurality of acquisition channels of a measurement process was used to obtain the measurement result.

11. A method for determining one or more parameters of a product of a manufacturing process based on measurement results of the product, the method comprising: obtaining a measurement data set of one or more products of the manufacturing process, each of the measurement results comprising an array of values ​​obtained by measuring a corresponding one of the products, wherein each of the measurement results in the data set corresponds to a respective one of a plurality of measurement results selected from another measurement data set of one or more products of the manufacturing process, each of the measurement results comprising an array of values ​​obtained by measuring a corresponding one of the products of the manufacturing process, selecting an appropriate subset of the data set for use in training the inference model, the subset being selected by applying an optimization procedure to an objective function that provides a measure of difference between each measurement in another data set and a corresponding reproduced value of the measurement obtained using a reproduction function, the reproduction function having a domain that includes the measurements in the subset and excludes the measurements not in the subset; as well as The one or more parameters of each of the one or more products are determined using the inference model trained using the measurements of the data set and / or the measurements of the appropriate subset of the further data set.

12. The method of claim 11, wherein the inference model is trained using the measurements in the data set combined with known values ​​of the one or more parameters for each of the one or more products of the manufacturing process.

13. A computing system comprising a processor and a memory, the memory storing program instructions which, when executed by the processor, enable the processor to perform the method according to any one of claims 1 to 12.

14. A metrology system for determining one or more parameters of a product of a manufacturing process based on a metrology signal characterizing the product, the metrology system comprising a computing system according to claim 13 and a metrology device configured to perform measurements on the product to obtain the metrology signal.

15. A computer program product storing program instructions, which, when executed by the processor, causes the processor to perform the method according to any one of claims 1 to 12.

Images