Measurement method and associated device
By using machine learning models and a hybrid model based on physical input data bias in lithography technology, the problem of sparseness of data in lithography is solved, and the overprinting accuracy and manufacturing performance of submicron semiconductor devices are improved.
Patent Information
- Application Number
- CN202380033855.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-14
- Filing Date
- 2023-03-14
- Publication Date
- 2024-11-22
AI Technical Summary
Existing lithography techniques have sparse data limitations when measuring and mapping overprinting data, making it difficult to accurately control over the overprinting accuracy of submicron semiconductor devices in device manufacturing, especially on key layers at the tens of nanometers or even several nanometers.
Using a hybrid model including machine learning model components, to improve the measurement data acquisition on the patterned substrate before performing processing steps and to use physically based input data biases to infer and map dense overprint data to Engraving accuracy.
More accurate overprint control in lithography processes is achieved, improving the yield and performance of device manufacturing, especially in the etching and exposure steps.
Smart Images

Figure CN119013622A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims priority to EP application 22168324.6 filed on April 11, 2022, and the entire contents of which are incorporated herein by reference. Technical Field
[0003] The present invention relates to metrology methods performed by patterning processes such as photolithography to maintain performance in device manufacturing. The present invention also relates to methods of manufacturing devices using photolithography techniques. The present invention also relates to computer program products for implementing such methods. Background Art
[0004] The photolithography process involves applying a desired pattern onto a substrate (usually onto a target portion of the substrate) using a lithography device, and then various chemical and / or physical processing steps work through the pattern to create the functional features of a complex product. Accurate placement of patterns on a substrate is a major challenge for reducing the size of circuit components and other products that may be produced by photolithography. In particular, the challenge of accurately measuring features on an already placed substrate is a critical step in being able to superimpose successive layers of features accurately enough to produce a working device with a high yield. A particularly important parameter is overlay, which is typically controlled to within tens of nanometers in today's submicron semiconductor devices, and as low as a few nanometers in the most critical layers.
[0005] Therefore, modern lithographic equipment requires a large number of measurement or "mapping" operations before actually exposing or otherwise patterning the substrate at the target location. However, due to throughput requirements, only sparse overlay data is usually available. Therefore, a dense overlay data map can be derived using a fitting model (polynomial) and / or other data indicating overlay (such as alignment data). For example, this method is described in US10990018B2, which is incorporated herein by reference.
[0006] It is desirable to improve the inference and / or mapping of parameter values of interest, such as overlay, critical dimension (CD), or edge placement error (EPE) measured after a processing step such as etching, based on metrology data measured before the processing step. Summary of the invention
[0007] According to a first aspect of the present invention, there is provided a method for inferring second measurement data relating to at least one patterned substrate from first measurement data measured on at least one patterned substrate before performing the at least one processing step, on which a pattern has been exposed and on which at least one processing step has been performed; the method comprising: obtaining a model comprising at least one first model component, the at least one first model component comprising a machine learning model component, the machine learning model component having been trained to map the first measurement data to the second measurement data, the first model component also comprising a physics-based input channel for receiving physics-based input data; and inferring second measurement data from the first measurement data using the first model component, wherein the first model component is biased by the physics-based input data on the physics-based input channel.
[0008] According to a second aspect of the present invention, there is provided a computer program product comprising one or more machine readable instruction sequences for implementing the computing steps in the method of the first aspect of the present invention as described above.
[0009] The invention also provides a processing device and a measuring apparatus comprising the computer program of the second aspect.
[0010] These and other aspects and advantages of the apparatus and methods disclosed herein may be understood by considering the following description of exemplary embodiments and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] 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 and in which:
[0012] Figure 1 A lithographic apparatus suitable for use in one embodiment of the present invention is shown;
[0013] Figure 2 A lithography cell or lithography cluster in which the inspection device according to the invention can be used is shown;
[0014] Figure 3 According to known practice, schematically shown Figure 1 Measurement and exposure processes in devices;
[0015] Figure 4 is a flow chart describing a method of obtaining second measurement data from first measurement data and physics-based input data according to one embodiment;
[0016] Figure 5 (a), 5(b), and 5(c) show three partition configurations for training a model according to one embodiment;
[0017] Figure 6 (a) graphically and (b) schematically illustrate a second model component according to one embodiment;
[0018] Figure 7 is a flowchart describing a method of training a hybrid model component according to one embodiment;
[0019] Figure 8 is a flow chart describing a method of optimizing process step parameter settings (eg, etch settings) according to one embodiment; and
[0020] Fig. 9 is a flow chart describing a method for optimizing exposure step and / or processing step parameter settings according to one embodiment. DETAILED DESCRIPTION
[0021] Before describing the embodiments of the present invention in detail, an example environment in which the embodiments of the present invention can be implemented is introduced.
[0022] Figure 1 A lithographic apparatus LA is schematically shown. The apparatus comprises an illumination system (illuminator) IL configured to condition a radiation beam B (e.g., UV radiation or DUV radiation); a patterning device support or support structure (e.g., mask stage) MT configured to support a patterning device (e.g., mask) MA and connected to a first positioner PM configured to accurately position the patterning device according to certain parameters; two substrate tables (e.g., wafer stages) WTa and WTb, each configured to accommodate a substrate (e.g., a resist-coated wafer) W and each connected to a second positioner PW configured to accurately position the substrate according to certain parameters; and a projection system (e.g., a refractive projection lens system) PS configured to project a pattern imparted by the patterning device MA to the radiation beam B onto a target portion C (e.g., comprising one or more dies) of the substrate W. A reference frame RF connects the various components and serves as a reference for setting and measuring the position of the patterning device and the substrate and features thereon.
[0023] The illumination system may include various types of optical components, such as refractive, reflective, magnetic, electromagnetic, electrostatic or other types of optical components, or any combination thereof, for directing, shaping or controlling the radiation. For example, in devices using extreme ultraviolet (EUV) radiation, reflective optical components are often used.
[0024] The patterning device support holds the patterning device in a manner that depends on the orientation of the patterning device, the design of the lithographic apparatus, and other conditions, such as whether the patterning device is held in a vacuum environment. The patterning device support may use mechanical, vacuum, electrostatic or other clamping techniques to hold the patterning device. The patterning device support MT may be, for example, a frame or a table that may be fixed or moved as required. The patterning device support may ensure that the patterning device is in a desired position, for example relative to a projection system.
[0025] The term "patterning device" as used herein should be broadly interpreted as referring to any device that can be used to impart a pattern to a cross-section of a radiation beam, such as to create a pattern in a target portion of a substrate. It should be noted that the pattern imparted to the radiation beam may not exactly correspond to the desired pattern in the target portion of the substrate, for example if the pattern includes phase-shifting features or so called assist features. Typically, the pattern imparted to the radiation beam will correspond to a specific functional layer in a device created in the target portion, such as an integrated circuit.
[0026] As shown here, the device is transmissive (e.g., employing a transmissive patterning device). Alternatively, the device may be reflective (e.g., employing a programmable mirror array of the type described above, or employing a reflective mask). Examples of patterning devices include masks, programmable mirror arrays, and programmable LCD panels. Any use of the term "photoreticle" or "mask" herein may be considered synonymous with the more general term "patterning device." The term "patterning device" may also be interpreted as referring to a device that stores pattern information in digital form for controlling such a programmable patterning device.
[0027] The term "projection system" as used herein should be interpreted broadly to cover any type of projection system, including refractive, reflective, catadioptric, magnetic, electromagnetic and electrostatic optical systems, or any combination thereof, appropriate to the exposure radiation used, or to other factors such as the use of immersion liquid or the use of a vacuum. Any use of the term "projection lens" as used herein may be considered synonymous with the more general term "projection system".
[0028] The lithographic apparatus may also 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 and the substrate. Immersion liquid may also be applied to other spaces in the lithographic apparatus, such as between the mask and the projection system. Immersion techniques are well known in the art and are used to increase the numerical aperture of a projection system.
[0029] In operation, the illuminator IL receives a radiation beam from a radiation source SO. For example, when the source is an excimer laser, the source and the lithographic apparatus may be separate entities. In this case, the source is not considered to be part of the lithographic apparatus, and the radiation beam is transferred from the source SO to the illuminator IL with the help of a beam delivery system BD, which includes, for example, suitable guide mirrors and / or a beam expander. In other cases, the source may be an integral part of the lithographic apparatus, for example, when the source is a mercury lamp. The source SO and the illuminator IL, and the beam delivery system BD if necessary, may be referred to as a radiation system.
[0030] The illuminator IL may, for example, comprise an adjuster AD for adjusting the angular intensity distribution of the radiation beam, an integrator IN and a condenser CO. The illuminator may be used to condition the radiation beam so that its cross-section has a desired uniformity and intensity distribution.
[0031] The radiation beam B is incident on a patterning device MA which is held on a patterning device support MT and is patterned by the patterning device. After passing through the patterning device (e.g. a 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 sensor IF (e.g. an interferometer, a linear encoder, a 2-D encoder or a capacitive sensor), the substrate table WTa or WTb can be precisely moved, for example, in order to position a different target portion C in the path of the radiation beam B. Similarly, a first positioner PM and a further position sensor ( Figure 1 ) can be used to accurately position the patterning device (e.g., mask) MA relative to the path of the radiation beam B, for example, after mechanical retrieval from a mask library or during scanning.
[0032] The pattern forming device (e.g., mask) MA and substrate W can be aligned using mask alignment marks M1, M2 and substrate alignment marks P1, P2. Although the substrate alignment marks shown occupy dedicated target portions, they can be located in the space between target portions (these are called scribe alignment marks). Similarly, where more than one die is provided on the pattern forming device (e.g., mask) MA, the mask alignment marks can be located between the dies. Small alignment marks can also be included within the die, among the device features, in which case it is desirable that the mark be as small as possible and do not require different imaging or process conditions than adjacent features. The alignment system for detecting the alignment marks will be further described below.
[0033] The depicted apparatus can be used in a variety of modes. In scanning mode, the pattern forming device support (e.g., mask table) MT and the substrate table WT are scanned synchronously while the pattern imparted to the radiation beam is projected onto the target portion C (i.e., a single dynamic exposure). The speed and direction of the substrate table WT relative to the pattern forming device support (e.g., mask table) MT can be determined by the (reduced) magnification and image reversal characteristics of the projection system PS. In scanning mode, the maximum size of the exposure field limits the width of the target portion (in the non-scanning direction) in a single dynamic exposure, while the length of the scanning movement determines the height of the target portion (in the scanning direction). As is well known in the art, other types of lithographic equipment and operating modes are also possible. For example, stepping modes are known. In so-called "maskless" lithography, the programmable pattern forming device is held stationary but with a changing pattern, and the substrate table WT is moved or scanned.
[0034] Combinations and / or variations on the above described modes of use or entirely different modes of use may also be employed.
[0035] The lithographic apparatus LA is of a so-called dual-stage type, having two substrate tables WTa, WTb and two stations (exposure station EXP and measurement station MEA), between which the substrate table can be exchanged. When a substrate on one substrate table is exposed at the exposure station, another substrate can be loaded onto another substrate table at the measurement station and various preparation steps can be performed. This can greatly improve the throughput of the apparatus. On a single stage device, it is necessary to perform the preparation step and the exposure step in sequence on a single stage for each substrate. The preparation step may include using a horizontal sensor LS to draw the surface height profile of the substrate and using an alignment sensor AS to measure the position of the alignment mark on the substrate. If the position sensor IF cannot measure the position of the substrate table at the measurement station and the exposure station, a second position sensor may be provided so that the position of the substrate table relative to the reference frame RF can be tracked at the two stations. Other arrangements are known and can be used, and may replace the dual-stage arrangement shown. For example, other lithographic apparatus are known in which a substrate table and a measurement table are provided. When performing preparatory measurements, they are docked together and then undocking when the substrate table is exposed.
[0036] like Figure 2As shown, the lithography equipment LA forms part of a lithography cell LC, sometimes also referred to as a lithography cell or lithography cluster, which also includes devices for performing pre-exposure and post-exposure processes on a substrate. Traditionally, these devices include a spin coater SC for depositing a resist layer, a developer DE for developing the exposed resist, a cooling plate CH, and a baking plate BK. A substrate handler or robot RO picks up substrates from input / output ports I / O1, I / O2, moves them between different process equipment, and then transfers them to a feed station LB of the lithography equipment. These devices are generally collectively referred to as tracks and are controlled by a track control unit TCU, which itself is controlled by a monitoring system SCS, which also controls the lithography equipment through a lithography control unit LACU. Therefore, different devices can be operated to maximize throughput and processing efficiency.
[0037] In order to ensure that the substrates exposed by the lithography equipment can be correctly and consistently exposed, it is necessary to check the exposed substrates to measure characteristics such as overlay errors between subsequent layers, line thickness, critical dimensions (CD), etc. Therefore, the manufacturing facility where the lithography unit LC is located also includes a measurement system MET, which receives part or all of the substrates W that have been processed in the lithography unit. The measurement results are directly or indirectly provided to the monitoring system SCS. If an error is detected, the exposure of the subsequent substrate can be adjusted.
[0038] In a metrology system MET, an inspection device is used to determine the properties of a substrate, in particular how the properties of different substrates or different layers of the same substrate vary from layer to layer. The inspection device may be integrated into the lithographic apparatus LA or the lithography cell LC, or it may be a stand-alone device. In order to achieve the fastest measurement, it may be desirable for the inspection device to measure the properties in the exposed resist layer immediately after exposure. However, not all inspection devices have sufficient sensitivity to make useful measurements of the latent image. Therefore, the measurement can be made after the post-exposure bake step (PEB), which is usually the first step performed on the exposed substrate and increases the contrast between the exposed and unexposed parts of the resist. At this stage, the image in the resist can be called a semi-latent image. Measurements can also be made on the developed resist image - at which point the exposed or unexposed parts of the resist have been removed. In addition, substrates that have been exposed can be stripped and reworked to improve yield, or discarded to avoid further processing of substrates that are known to be defective. In the case where only some target parts of the substrate are defective, further exposure can be performed only on those good target parts.
[0039] The metrology step of the metrology system MET can also be performed after the resist pattern has been etched into the product layer. The latter possibility limits the possibilities for reworking defective substrates but can provide additional information about the performance of the overall manufacturing process.
[0040] Figure 3Shown in Figure 1 The steps of exposing a target portion (e.g., a die) on a substrate W in a dual-stage apparatus of the type shown will be described first. The process according to conventional practice will be described first. The present disclosure is in no way limited to dual-stage apparatus of the type shown. Those skilled in the art will recognize that similar operations are performed in other types of lithographic apparatus, such as lithographic apparatus having a single substrate table and a docking metrology table.
[0041] The steps performed at the measurement station MEA are shown in the dashed box on the left, while the steps performed at the exposure station EXP are shown on the right. Sometimes, one of the substrate tables WTa, WTb is located at the exposure station, while the other is located at the measurement station, as described above. For ease of explanation, it is assumed that the substrate W has been loaded into the exposure station. In step 200, a new substrate W' is loaded into the apparatus by a mechanism not shown. In order to increase the throughput of the lithography apparatus, the two substrates are processed in parallel.
[0042] Reference is first made to a newly loaded substrate W', which may be a previously unprocessed substrate prepared with a new resist for first exposure in the apparatus. However, in general, the lithographic process described will be only one step in a series of exposure and processing steps, so the substrate W' has passed through the apparatus and / or other lithographic apparatus several times and may also require subsequent processing. In particular, with regard to the problem of improving overlay performance, the task is to ensure that the new pattern is applied in the correct position on a substrate that has undergone one or more patterning and processing cycles. Each patterning step may introduce positional deviations in the applied pattern, and subsequent processing steps will gradually introduce deformations in the substrate and / or the pattern applied to the substrate, which must be measured and corrected to achieve satisfactory overlay performance.
[0043] The previous and / or subsequent patterning steps can be performed in other lithographic equipment, as just described, and can even be performed in different types of lithographic equipment. For example, in the device manufacturing process, some layers that have very high requirements for parameters such as resolution and overlay can be performed in more advanced lithographic tools than other layers with lower requirements. Therefore, some layers can be exposed in immersion lithography tools, while other layers are exposed in "dry" tools. Some layers can be exposed in tools that work at DUV wavelengths, while other layers are exposed using EUV wavelength radiation. Some layers can be patterned by alternative or supplementary steps for exposure in the lithographic equipment shown. Such alternative and supplementary technologies include, for example, imprint lithography, self-aligned multiple patterning, and directed self-assembly. Similarly, other processing steps performed on each layer (e.g., CMP and etching) can be performed on different equipment for each layer.
[0044] At 202, alignment measurements are performed using substrate marks P1 and the like and an image sensor (not shown) to measure and record the alignment of the substrate relative to the substrate table WTa / WTb. In addition, an alignment sensor AS will be used to measure a number of alignment marks across the substrate W'. In one embodiment, these measurements are used to build a substrate model (sometimes referred to as a "wafer grid") that maps the distribution of the marks across the substrate very accurately, including any deformations relative to the nominal rectangular grid.
[0045] In step 204, a map of wafer height (Z) versus XY position is also measured using a level sensor LS. Primarily, the height map is only used to achieve accurate focus on the exposure pattern. In addition, it can also be used for other purposes.
[0046] When a substrate W' is loaded, profile data 206 is received, defining the exposure to be performed, as well as the properties of the wafer and the previously produced patterns and the patterns to be produced thereon. When there is a selection of alignment marks on the substrate, and there is a selection of settings for the alignment sensors, these selections are defined in the alignment profile in the profile data 206. Thus, the alignment profile defines how the positions of the alignment marks are measured, and which marks are measured.
[0047] At 210, wafers W' and W are exchanged so that the measured substrate W' becomes the substrate W entering the exposure station EXP. Figure 1 In the example apparatus of , this exchange is performed by exchanging supports WTa and WTb within the apparatus so that the substrates W, W' remain accurately clamped and positioned on these supports to maintain the relative alignment between the substrate table and the substrate itself. Therefore, once the worktables have been exchanged, determining the relative position between the projection system PS and the substrate table WTb (formerly WTa) is all that is required to control the exposure steps using the measurement information 202, 204 of the substrate W (formerly W'). At step 212, reticle alignment is performed using the mask alignment marks M1, M2. In steps 214, 216, 218, scanning movements and radiation pulses are applied at consecutive target positions across the substrate W so as to complete the exposure of multiple patterns.
[0048] By using the alignment data and height maps obtained at the metrology station while performing the exposure step, these patterns can be accurately aligned relative to desired locations, particularly relative to features previously placed on the same substrate. At step 220, the exposed substrate (now labeled W") is unloaded from the apparatus for etching or other processes based on the exposure patterns.
[0049] Currently, overlay information is extracted using either direct metrology or indirect metrology. Direct metrology (such as open-cap scanning electron microscope (SEM) metrology and / or high-voltage SEM) is too slow for high-frequency in-line overlay metrology and, in the case of open-cap metrology, destroys the device being measured. Indirect metrology (such as scatterometry based on optical targets) is typically performed on a set of discrete locations, and full-field and / or full-wafer overlay maps are constructed by interpolating the measured overlay values at these discrete locations.
[0050] In either case, the metrology can be performed pre-etch, before certain processing steps such as etching and / or polishing (chemical mechanical polishing, CMP), where the metrology can be described as post-development inspection (ADI), or post-etch (after these processing steps), where the metrology can be described as post-etch inspection (AEI).
[0051] It is also known that certain processing effects (e.g., etching and CMP steps) can affect parameters of interest, such as, for example, overlay, critical dimension (CD), edge placement error. Therefore, the measurements of the same structure on the same wafer when performing ADI may be different than when performing AEI. For example, overlay may be affected by intra-die system (IDS) variations, which refers to systematic variations repeated on each die, which arise from such manufacturing steps being repeated at the die level. Therefore, various etching parameters and / or environments (e.g., the actual etching tool and / or chamber used) may affect AEI measurements relative to ADI measurements. In addition, the actual effects of such manufacturing processes may also depend on the design layout pattern on the mask, such as local pattern density, target design and / or target location (inside the die and on the wafer). For example, etching is affected by pattern density; the chemistry of the etching plasma above the chip, and the etching rate, selectivity and anisotropy depend on the proportion of resist and the proportion of etching waste generated during the etching process.
[0052] ADI metrology is often preferred for a number of reasons. On one hand, ADI metrology provides a more direct link between the metrology images obtained during inspection and the various exposure process conditions (e.g., focus, dose, etc.) that define patterning performance. Inspecting the wafer after development allows for optimization of exposure process conditions. Inspecting the wafer after development allows for rework of the wafer based on inspection metrology results. This is not possible after etching; the wafer can no longer be returned to a previous process state for reprocessing to improve lithography performance.
[0053] However, since etching and / or other processing effects may affect overall lithography or manufacturing performance (e.g., placement performance such as EPE, etc.), AEI measurements will better indicate overall product performance or yield. Therefore, it is desirable to be able to predict AEI measurement data based on ADI measurement data.
[0054] Some models use parameters known from the etch physics and materials to generate a predicted image of the wafer after etching. However, these models may not generate accurate predicted images because the actual physical parameters of etching may not be known in practice. Some models adjust or calibrate parameters to minimize the difference between the physics-based model and the predicted image. But these models may not generate accurate predicted images because they do not have enough capacity to represent the complex relationships present in the data.
[0055] A method for inferring SEM AEI measurement data from SEM ADI measurement data is described in WO2021 / 052918, which is incorporated herein by reference. The method describes the use of a convolutional neural network-based machine learning model, in particular an encoder-decoder-based model, to map SEM ADI measurement data to SEM AEI measurement data (in particular random measurement data such as defects, line edge roughness LER, line width roughness LWR, local critical dimension uniformity LCDU).
[0056] A problem with using purely data-driven models (e.g., using a neural network or other machine learning model without any physics-based input) to map ADI data to AEI data can be that the mapping is generic and may not accurately predict the effects of process variations on AEI measurements not encountered in training (i.e., the model does not generalize beyond the training data). Furthermore, while such purely data-driven models have been shown to perform satisfactorily for fixed patterns, they may perform poorly when there are significant pattern variations (such as changing pattern density, etc.).
[0057] In view of these drawbacks, a hybrid modeling method and a hybrid model are proposed for inferring second metrology data related to at least one patterned substrate from first metrology data measured on at least one patterned substrate before performing the at least one processing step, on which a pattern has been exposed and at least one processing step has been performed. The method includes obtaining a model including at least one first model component, wherein the (at least one) first model component includes a machine learning or hybrid data-driven model component that has been trained to map the first metrology data to the second metrology data, the first model component also including a physics-based input channel for receiving physics-based input data; and inferring second metrology data from the first metrology data using the first model component, the first model component being biased by the physics-based input data on the physics-based input channel.
[0058] Thus, at least one first model component or hybrid data-driven model component can be configured to output different second measurement data (e.g., different output images) given the same first measurement data input (e.g., the same input image) due to different physics-based input data on the physics-based input channel (e.g., different ion tilt parameter values or pattern density values).
[0059] The physics-based input may receive one or more processing step parameters related to at least one processing step performed on the at least one patterned substrate. The one or more processing step parameters may include etching parameters. In one embodiment, the etching parameters may include ion tilt parameters, for example, which may depend on the substrate (e.g., radial) position. Alternatively or additionally, the physics-based input may receive one or more patterning parameters related to the pattern, for example, patterning parameters related to field and / or tube core pattern. An example of a patterning parameter is pattern density.
[0060] In this way, the idea of hybrid models is to enforce physics-inspired network modifications in order to bias the network output, e.g., tailoring network layers based on physics-inspired inputs.
[0061] The model may include at least one second model component including a physics-based model operable to model the effect of the at least one processing step on the second measurement data to generate the physics-based input data for a physics-based input channel.
[0062] The first measurement data may include a measurement image. Such a measurement image may include one or more of the following: an overlay result / image (IBO) based on an optical image, a scatterometry image measured in a pupil plane and / or an image plane (e.g., a diffraction-based overlay (DBO) or in-device measurement (IDM) image measured in a pupil plane / Fourier plane and / or a micro-diffraction-based overlay (μDBO) image measured in an image plane), and / or a scanning electron microscope (SEM) / electron beam image. The second measurement data may also include one or more of these image types, or may include the values of one or more parameters of interest, or a description of the spatial variation of one or more parameters of interest (e.g., a fitted model or fingerprint of the parameters of interest). The parameters of interest (or fingerprints) output by the model may include overlay (or overlay fingerprints) in one or two dimensions of the substrate plane, such as CD (or CD fingerprint) or edge placement error (or EPE fingerprint). Alternatively, the parameters of interest may be one or more random parameters (e.g., assuming the first metrology data includes sufficient information, such as, for example, SEM images), such as LCDU, LER, LWR, average CD. As a further alternative, the second metrology data may include electronic test data (e.g., densely connected electronic test data).
[0063] The first metrology data may include measurement data, e.g., measurement data from an exposed substrate and / or simulated / modeled first metrology data. In the latter case, a first metrology data simulation model (e.g., a neural network or other data-driven model) may be trained and such simulated / modeled first metrology data may be generated based on scanner setting inputs (e.g., describing scanner settings or exposure parameters such as focus and / or dose).
[0064] The second model may generate the one or more processing step parameters (related to the processing step) and / or patterning parameters (related to the pattern). The patterning parameters may include one or more of the following: target design parameters, such as pitch, cd, design, sub-segmentation parameters, target placement in the die, that is, the coordinates (x, y) of the target and any pattern density data (e.g., pattern density variations around the target or other). The pattern density data may include a pattern density spatial distribution or a pattern density map. The pattern density map is a low-resolution GDS (Graphic Design System) file or GDSII file. Because GDS files are large, they are typically reduced to low-resolution pattern density maps so that they can be processed quickly enough for die-level applications. The typical resolution of the pattern density map is 1 μm pixel size.
[0065] Note that the term "target" in this disclosure may include a metrology target formed for metrology purposes and / or any other structure (including a functional product structure) used for measurement (e.g., measured). Therefore, the patterning parameters may be related to structures other than a specific metrology target, especially when the first metrology data includes images obtained using a technique that does not require a dedicated target for measurement (e.g., SEM, some in-device metrology (IDM) methods).
[0066] The output of the second model component may include data of a data type similar to the output of the model as a whole (e.g., when the parameter of interest output by the model as a whole includes overlay, CD, or EPE (or its fingerprint), the second model component may cause the processing parameters to be related to overlay, CD, or EPE (or its fingerprint), respectively). Alternatively, the output of the second model component may include data of a different data type than the model as a whole (e.g., the output of the second model component may include overlay or CD, while the output of the model as a whole may be EPE or an estimated image or other parameters dependent on the output of the second model component).
[0067] The output of the second model component may include values of one or more second model output parameters (e.g., overprint, CD, or edge placement error), which may be used directly by the first model component. Optionally, the model or fingerprint may be fitted to these second model output parameter values before input to the first model component. Alternatively, the second model component may directly output a second model output fingerprint or a second model output spatial variation description.
[0068] The processing step may include an etching step, and the processing step parameters (based on physical input data) may include one or more etching parameters, such as, for example, one or more of the following: etching pressure, etching flow rate, RF voltage, over-etching, edge ring bias / offset, chamber adjustment, etching tool and / or chamber used. Alternatively or additionally, the process parameters may include ion tilt parameters, which may be an effective proxy for some or all of these etching parameters. As previously mentioned, the physical input data may also include patterning data or pattern density data. AEI overlay and / or CD are highly sensitive to etching conditions, target design and its surrounding pattern density variations.
[0069] The first model component may include a neural network-based model, and the physics-based input channels (e.g., the channels to which the outputs of the second model component / each second model component are connected) may be input to an intermediate layer of the neural network, such as a first fully connected layer.
[0070] At a basic level, the proposed model can be used to examine new wafers that have undergone pre-processing (e.g., pre-etch or ADI) and predict the parameters of interest after processing (e.g., post-etch AEI).
[0071] In one embodiment, the proposed model can be used to optimize a method for one or more processing parameters so as to improve the parameter of interest after processing. This can enable parameters that were out of spec when ADI was measured to be brought into spec through processing, and / or improve performance even when they were initially within spec. The extent to which this is possible depends on the parameter of interest; for example, it may be more feasible to bring parameters such as LCDU or EPE back into spec through etch control than through overlay. However, this may change in the future, for example, as etcher control improves. In any case, at least for fields closer to the edge of the wafer, some etch-based AEI overlay control can be achieved by locally changing the direction of the electric field so as to adjust the tilt of the etch path (ion tilt), which effectively introduces a method to adjust overlay with the previous layer. Alternatively or additionally, etcher control of post-etch CD can be used to compensate for at least part of the ADI overlay error, for example by slightly increasing the CD of the etched contact layer to ensure that there is enough overlay area with the previous layer to absorb or compensate for the ADI overlay error.
[0072] The model can also be used to make rework decisions. For example, if the model predicts a parameter of interest that is out of specification, a decision can be made to perform wafer rework. In one embodiment, a decision can be made only after using the model to determine whether the out-of-specification parameter can be adequately corrected by process or etch control (e.g., modeling for different process parameters and / or environments); for example, wafer rework is only proposed if it is deemed impossible.
[0073] In one embodiment, the model can be used to determine scanner-based exposure parameters or parameter corrections of interest in order to optimize (or at least bring to specification) a post-processing parameter of interest (e.g., AEI) for a reworked wafer. This means that the model allows control of the post-processing / AEI parameter of interest prior to a new (reworked) exposure. To achieve this, the model can be trained on training metrology data, where the relevant scanner control parameters have been varied. Such an embodiment may also include optimizing or bringing the parameter of interest to specification by co-optimization of scanner and process tool calibrations.
[0074] As a specific example, the training of the model can be based on training metrology data including measurements from focus exposure or FEM wafers on which patterns are exposed using dose and focus variations close to optimal energy and optimal focus. Such training metrology data can also optionally include modeled training metrology data, such as data obtained from a model operable to input and output metrology images or first metrology data based on various exposure parameters (e.g., focus / dose). When deciding to rework a wafer, dose retargeting or AEI correction (e.g., to improve a stochastic metric such as average CD or LCDU fingerprint) can be achieved through dose and / or focus correction rather than adjusting the etch recipe.
[0075] The physics-based inputs can enable evaluation of the impact of process step parameters and / or patterning parameters on AEI performance. The physics-based inputs (e.g., including second model outputs or otherwise) can be connected to one or more layers of the first model, so that changing process tool and / or scanner parameters will modify the model (output) to account for process details (e.g., etcher / scanner configuration) in the AEI predictions.
[0076] Figure 4 4 is a flow chart describing a method for inferring post-processing (e.g., AEI) parameter data 410 (or post-processing image) of interest from pre-processing (e.g., ADI) data or pre-processing image 400 using a trained hybrid model 405. The pre-processing image may include, for example, a DBO or μDBO image (one example of the latter is shown), or, for example, a SEM / electron beam image. The hybrid model includes a first model component 415 (e.g., a neural network-based model component) having a physics-based input channel 418. In an embodiment, at least one second model component (e.g., a physics-based model component) 445 is provided to determine physics-based inputs for the physics-based input channel 418. More specifically, in this embodiment, the at least one second model component 445 may include one or both of a wafer-level second model component (e.g., a physical etch chamber model component) 450 and a chip-level second model component (e.g., a pattern model component) 455. Processing data 425 (e.g., etch parameters and / or etch context data) and corresponding layout or patterning data 430 (e.g., wafer-level position data 435 and die-level position data 440) may be fed into each of these second model components 450, 455. The outputs of the second model components 450, 455 are fed into the physically-based input channel 418 and thus into one or more specific layers 420 (e.g., the first fully-connected layer) of the first model component 415.
[0077] In one embodiment, cross-wafer variation can be handled by partitioning the wafer into regions and training a separate model for each region of the wafer. Figure 5 Three purely exemplary embodiments are shown, in which the wafer W is divided into (a) 17 zones Z1, (b) 9 zones Z2 and 3 zones Z3. For example, radial etching fingerprints have been noted; this can reduce the number of zones required in the circular direction. Therefore, the number of zones and their arrangement may differ from what is shown. In each case, a corresponding trained model can be used to infer the parameters of interest from each specific zone. Etching chamber variations can be handled in the same way, i.e., by training a separate model for each chamber.
[0078] Alternatively, a single model can be conditioned on wafer region and / or chamber number. In this case, the model is trained using training data from all etch chambers and / or all wafer regions, while identifying the chambers and / or regions of the corresponding subset of the data as additional inputs. When using the model for inference, the contextual information (chamber / region) is used as input (e.g., included in the physics-based input data), and the model infers the parameters of interest from it. A hybrid of these two approaches is also possible (e.g., separate models trained for corresponding larger regions, the larger regions are subdivided into smaller sub-regions, and each model can be conditioned based on these sub-regions.
[0079] If provided, a (e.g., wafer-level) second module assembly or a physical etch chamber module assembly may be used to numerically solve the governing equations, including, for example, the equations of motion for ions, the law of conservation of ion flux, and Poisson's equation at the plasma reactor scale, to obtain the plasma sheath profile f(Γ,P i ,r), which is the chamber geometry Γ, different physical etching conditions P i = a function of {pressure, power, bias voltage, focus ring height, ...} and wafer position (r). Then the wafer-level ion tilt (ion tilt parameter) can be obtained, i.e., In this way, the ion tilt parameter is determined The parameter space used to train the first model component can be reduced by replacing a large number of etching related parameters into a single one-dimensional function of substrate position or substrate radius (ie, depending only on the radial distance r from the wafer center).
[0080] Figure 6 The physical etch chamber module assembly is shown, Figure 6 (a) is in the operating range OP EPr and OP EPo For the ion tilt parameter Three example values of Etching pressure EP r With etching power EP o , which illustrates how the ion tilt parameter Used as a proxy for etch power and etch pressure (among other parameters). Figure 6 (b) Conceptual illustration of the ion tilt parameter The meaning and impact of ion tilt parameters is related to the plasma sheath profile PSP on the substrate W (eg, determined by its normal). The three ion tilt parameters are shown Each parameter is related to a specific radial position r on the substrate: r = -R, r = 0, and r = R. The bottom of the figure shows an example effect of the ion tilt parameter on the etch quality (e.g., structure or trench symmetry due to the effects of sidewall angle and bottom tilt). When r = 0 (substrate center), The etched structure is symmetrical, but near the substrate edge, the etched substrate is asymmetrical due to the effects of non-perpendicular (relative to the substrate plane) sidewall angles and non-parallel (relative to the substrate plane) bottom tilt. The magnitude of this effect is roughly equal for all locations at the same radial distance.
[0081] In one embodiment, the ion tilt parameter can be used (alone or in combination with one or more other parameters) as physically based input data for input to a physically based input channel. In one embodiment, at least one second model component models the ion tilt parameter Such that the output of the at least one second model component is fed into a physically based input channel; for example, its output is fed into the latent space of the encoder-decoder network of the first model component.
[0082] Figure 7 is a flow chart describing a method for training a first model component. The first model component may be primarily data driven, rather than biased by physics-based inputs (e.g., a hybrid model component). Step 700 may include exposing and measuring one or more training wafers, for example, to expose fields with various exposure parameters or focus and / or dose values (e.g., exposing a focus-dose matrix FEM). This step 700 generates training first measurement data 705 (e.g., training first measurement data or pre-processed / ADI training images). Optionally, the training first measurement data 705 may be densified using an additional neural network or first measurement data simulation model that has been trained to create 702 artificial ADI data for different dose / focus conditions (e.g., not included in the measurement data). The method may also include, for example, training such a first measurement data simulation model on training ADI images (e.g., first measurement data obtained in step 700) and corresponding known focus / dose settings.
[0083] Second metrology data 715 or AEI data may be obtained. For example, in certain embodiments, different process step parameter values (e.g., different ion tilt parameter values) may be used. ) etch 710 training wafers from which first measurement data are obtained. These etched wafers can then be measured to obtain second measurement data 715 or AEI data. For example, using each ion parameter value A corresponding wafer (or wafer group) including a full set of modeled focus / dose parameters is etched and measured to obtain second measurement data 715 including all combinations of set focus values, set dose values, and set ion parameter values (the set focus values, set dose values, and set ion parameter values may describe a physical setup within the scanner / etcher or an input to a simulation model, depending on whether the data is a measurement or a simulation). The training measurement data may also include known post-processing or AEI training values; for example, known values of AEI parameters of interest corresponding to ADI training images (alternatively, if the model is trained to output images, then output (AEI) training images). This enables supervised learning of the first model component.
[0084] Thus, the training data may include pre-processed training images and corresponding post-processed training data or images, wherein the correspondence in this case may include measurements taken from the same wafer location.
[0085] In step 730, the first model component is trained using the first metrology data / ADI data 705 with known exposure parameter values (e.g., known focus / dose values) and the corresponding second metrology data / AEI data 715 so that it can map the first metrology data to the second metrology data based on the physics-based input data on the physics-based input channel (e.g., known ion tilt parameter values). The known values of are provided to the network, for example, in the last layer of the convolutional part of the model network and before the dense network of the model.
[0086] Optionally, in step 720, for different ion tilt parameters Then, in step 725, the second model component or physical etching model can be trained for New (simulated) AEI data is generated by adding values of (eg, values that are not actually etched) to further densify the AEI data 715 .
[0087] As mentioned above, the ion tilt parameter is just one example of a physics-based input. Thus, as an alternative or in addition, AEI data may be generated for variations in another parameter, such as another processing step parameter and / or patterning parameter (e.g., a pattern density parameter). In each case, known values of the processing step parameter and / or patterning parameter may be input to the physics-based input channel during training.
[0088] Thus, the training data may include intra-die fingerprints, such as those obtained by modeling metrology data from different die locations using a physical die-level model (e.g., a die-level second model component) that may further understand the surrounding pattern (layout or patterning parameter data). Such intra-die and / or intra-target fingerprints predicted by the die-level etch model may be used to train a hybrid model for chip-scale AEI parameter variations of interest.
[0089] A similar approach can also be used to train the first model component to obtain a wafer-level process (e.g., etch) fingerprint using a separate wafer-level second model component (e.g., the model may include more than one second model component). In addition, data obtained from different process tools or etch chambers can also be used to train the first model component to obtain a tool / chamber fingerprint.
[0090] Figure 8 8 is a flow chart illustrating a first example application of the hybrid model described herein. In step 800, an ADI inspection is performed on an exposed wafer to obtain ADI image data (first metrology data). In step 805, the ADI image data is combined with a physics-based ion tilt parameter on an input channel. and / or other physics-based input data are input to the first model component; based on this, the first model component predicts the AEI image or AEI image data (second measurement data) 810. Ion tilt parameter The physics-based input data may be known or may be provided by one or more second model components, as already described. In step 815, the predicted AEI image data is evaluated, for example, by comparing it to an expected equivalent value or range across the wafer. In step 820, the predicted AEI image data may be evaluated for different physical parameter values (e.g., for different ion tilt parameters). Steps 805 to 815 are repeated until an optimal or at least acceptable value of the physical parameter is found.
[0091] Fig. 9is a flow chart illustrating a first example application of the hybrid model described herein. Steps 900 to 920 correspond to steps 800 to 820 and will not be described again, except that the first model component now receives (e.g., at step 905) simulated ADI data 925 generated for different exposure parameters (e.g., focus / dose conditions), e.g., the model is simulated using the first measurement data. Additional optimization is performed at step 930, which repeats the other steps 905-925 using different exposure parameters until an optimal (or at least acceptable) AEI result is predicted. Based on the prediction, a scanner rework exposure setting can be determined, and / or an alarm can be issued if acceptable performance cannot be found.
[0092] The proposed method can be used in conjunction with reinforcement learning methods to improve the generalization ability of reinforcement learning methods. For example, when reinforcement learning methods are used for high-volume manufacturing (HVM), it may not learn much when the etching process is stable due to the lack of variability in the data. On the other hand, the present method will learn all the necessary information during minimal training (because it is physically inspired) and be used to improve the performance of HVM through reinforcement learning.
[0093] This hybrid model approach enables optimal configuration of the relevant ADI post-process (etch) to improve the expected AEI performance (based on the ADI input). Alternatively, if the ADI performance of the wafer indicates that the AEI performance will never meet the requirements, it can be sent for rework; optionally, the model can also optimize the scanner settings / calibrations (or jointly optimize the scanner and etcher settings / calibrations) for rework to improve the AEI performance.
[0094] In association with the hardware of the lithographic apparatus and the lithographic cell LC, embodiments may include a computer program comprising one or more machine-readable instruction sequences for causing a processor of the lithographic manufacturing system to implement the model mapping and control method as described above. The computer program may be executed, for example, in a separate computer system for the image calculation / control process. Alternatively, the calculation steps may be fully or partially implemented in the processor, the measurement tool and / or Figure 1 and Figure 2 The computer program is executed in the control unit LACU and / or the monitoring system SCS. A data storage medium (eg a semiconductor memory, a magnetic disk or an optical disk) may also be provided, in which such a computer program is stored in a non-transitory form.
[0095] Although specific reference may have been made above to embodiments of the present invention being used in the context of optical lithography, it should be understood that the present invention may be used in other patterning applications, such as imprint lithography. In imprint lithography, the topography in the patterning device defines the pattern produced on the substrate. The topography of the patterning device may be pressed into a layer of resist supplied to the substrate, and the resist is then cured by applying electromagnetic radiation, heat, pressure, or a combination thereof. After the resist is cured, the patterning device is removed from the resist, leaving a pattern therein.
[0096] Further embodiments are disclosed in the following list of numbered clauses:
[0097] 1. A method for inferring second metrology data relating to at least one patterned substrate from first metrology data measured on at least one patterned substrate before performing at least one processing step, wherein the at least one patterned substrate has been exposed with a pattern and at least one processing step has been performed; the method comprising:
[0098] obtaining a model comprising at least one first model component and at least one second model component, wherein the at least one first model component comprises a machine learning model component that has been trained to map the first measurement data to the second measurement data, and the at least one second model component comprises a physics-based model operable to model an effect of the at least one processing step on the second measurement data; and
[0099] Inferring second measurement data based on the first measurement data using the first model component;
[0100] The output of the at least one second model component is used by the first model component to infer the second measurement data.
[0101] 2. A method according to clause 1, wherein the first measurement data comprises a measurement image or a representation thereof.
[0102] 3. The method of clause 2, wherein the metrology image may include one or more of the following: an image-based overlay image, a scatterometry image captured in a Fourier plane, a scatterometry image captured in an image plane, or a scanning electron microscope image. [Comments]
[0103] 4. A method according to any of the preceding clauses, wherein the second measurement data comprises values of one or more parameters of interest or a description of the spatial variation of one or more parameters of interest.
[0104] 5. The method of clause 4, wherein the parameters of interest include one or more of: overlay, CD, edge placement error, line edge roughness, line width roughness, local critical dimension uniformity, average critical dimension.
[0105] 6. The method of any one of clauses 1 to 3, wherein the second metrology data comprises electronic test data.
[0106] 7. A method according to any of clauses 1 to 3, wherein the second metrology data comprises a metrology image or a representation thereof.
[0107] 8. A method according to any of the preceding clauses, wherein at least one second model relates to the influence of one or more processing step parameters related to the processing step and / or patterning parameters related to the pattern on the second measurement data.
[0108] 9. The method of clause 8, wherein the patterning parameters include at least target design parameters.
[0109] 10. A method according to clause 9, wherein the target design parameters include one or more of the following: target pitch, critical dimensions, target design, target sub-segment parameters, target placement within the die, target position, any parameters of surrounding and / or adjacent structures, wherein the surrounding and / or adjacent structures surround and / or are adjacent to the target.
[0110] 11. A method according to clause 8, 9 or 10, wherein the patterning parameters include one or more pattern density parameters associated with any of the patterns.
[0111] 12. A method according to any of the preceding clauses, wherein the output of the second model component comprises values of one or more second model output parameters.
[0112] 13. A method according to clause 12, wherein, in the inferring step, the first model component directly uses the value as one or more second model output parameters.
[0113] 14. A method according to clause 13, comprising the step of fitting a model to said values of one or more second model output parameters; wherein said first model component uses said fitted model when inferring said second measurement data.
[0114] 15. A method according to any of the preceding clauses, wherein the processing step comprises at least one etching step; and the process parameters comprise one or more etching parameters.
[0115] 16. The method of clause 15, wherein the etching parameters include one or more of: etching pressure, etching flow rate, RF voltage, over-etching, edge ring bias / offset, chamber conditioning, etching tool and / or chamber used.
[0116] 17. A method according to any of the preceding clauses, wherein the at least one first model component comprises a neural network based model.
[0117] 18. A method according to clause 17, wherein the output of the second model component is input to an intermediate layer of a neural network.
[0118] 19. A method according to clause 18, wherein the output of the second model component is input to a first fully connected layer of a neural network.
[0119] 20. A method according to any of the preceding clauses, comprising determining settings and / or corrections of one or more processing parameters associated with the processing step to optimize and / or bring the second metrology data into specification.
[0120] 21. The method according to clause 20, comprising the step of performing said processing steps based on the determined settings and / or corrections.
[0121] 22. A method according to any of the preceding clauses, comprising determining whether the patterned substrate should be reworked based on the second metrology data.
[0122] 23. The method of clause 22, comprising determining that the wafer should be reworked only after using the model to determine that second metrology data inferred by the model cannot be adequately corrected by the processing step.
[0123] 24. A method according to clause 22 or 23, comprising using the model to determine settings and / or corrections for one or more exposure parameters for exposing a pattern on a reworked substrate to optimise and / or bring the second metrology data within specification.
[0124] 25. A method according to clause 22 or 23, comprising using the model to jointly optimize the settings and / or corrections of one or more exposure parameters for exposing a pattern on a reprocessed substrate, and determining the settings and / or corrections of one or more processing parameters associated with the processing step to optimize and / or bring the second measurement data into compliance with specifications.
[0125] 26. A method according to clause 24 or 25, comprising performing said reworking based on said setting and / or correction of one or more exposure parameters.
[0126] 27. A method according to any of the preceding clauses, comprising performing an initial training step to train the model, the training step comprising:
[0127] The at least one first model component is trained on training measurement data and corresponding second model component output training data or equivalent measurement data, wherein the training measurement data includes known data similar to the first measurement data and the second measurement data.
[0128] 28. A method according to clause 27, wherein the second model component outputs training data or equivalent measurement data including one or more of the following: a spatial description or value of one or more parameters of interest, wherein the spatial description or value is described within a die, within a target, between wafers, or within each processing tool / chamber.
[0129] 29. A method according to clause 27 or 28, comprising training the model based on training data relating to changes in one or more exposure tool parameters and / or processing tool parameters.
[0130] 30. A method according to any of clauses 27 to 29, comprising training separate instances of the model to obtain respective trained models for each of two or more substrate regions and / or two or more processing tools or chambers; and
[0131] The inferring step includes using a corresponding trained model for its respective substrate region and / or processing tool or chamber.
[0132] 31. A method according to any of clauses 27 to 29, wherein training comprises adjusting a single model for the entire substrate and / or all processing tools and chambers by including within the training data the wafer regions and / or chamber numbers to which the training data relates.
[0133] 32. The method of any of the preceding clauses, wherein the at least one second model component comprises one or more of a die-level second model component and a wafer-level second model component.
[0134] 33. A computer program comprising program instructions which, when run on a suitable device, are capable of performing any of the methods described above.
[0135] 34. A non-transitory computer program carrier comprising the computer program described in clause 33.
[0136] 35. A processing device comprising:
[0137] a computer program carrier comprising a computer program as referred to in clause 33; and
[0138] A processor operable to execute said computer program.
[0139] 36. A measuring device comprising the processing device of clause 35.
[0140] 37. A method for inferring second measurement data related to at least one patterned substrate from first measurement data measured on at least one patterned substrate before performing the at least one processing step, wherein a pattern has been exposed and at least one processing step has been performed on the at least one patterned substrate; the method comprising: obtaining a model including at least one first model component, the at least one first model component including a machine learning model component, the machine learning model component has been trained to map the first measurement data to the second measurement data, the first model component also including a physics-based input channel for receiving physics-based input data; and using the first model component to infer second measurement data from the first measurement data, wherein the first model component is biased by the physics-based input data on the physics-based input channel.
[0141] 38. A method according to clause 37, wherein the first metrology data comprises a metrology image or a representation thereof.
[0142] 39. A method according to clause 38, wherein the metrology image may include one or more of the following: an image-based overlay image, a scatterometry image captured in a Fourier plane, a scatterometry image captured in an image plane, or a scanning electron microscope image.
[0143] 40. A method according to any of clauses 37 to 39, wherein the second measurement data comprises values of one or more parameters of interest or a description of a spatial variation of one or more parameters of interest.
[0144] 41. A method according to clause 40, wherein the parameters of interest include overlay, CD or edge placement errors in each case in one or two dimensions of the substrate plane.
[0145] 42. The method of clause 40, wherein the parameter of interest comprises one or more of: line edge roughness, line width roughness, local critical dimension uniformity, average critical dimension.
[0146] 43. A method according to any of clauses 37 to 39, wherein the second metrology data comprises electronic test data.
[0147] 44. A method according to any of clauses 37 to 39, wherein the second metrology data comprises a metrology image or a representation thereof.
[0148] 45. A method according to any one of clauses 37 to 44, wherein the physically based input data comprises one or more processing step parameters relating to the processing step and / or one or more patterning parameters relating to the pattern.
[0149] 46. The method of clause 45, wherein the patterning parameters include at least target design parameters.
[0150] 47. A method according to clause 46, wherein the target design parameters include one or more of the following: target pitch, critical dimensions, target design, target sub-segment parameters, target placement within the tube core, target position, any parameters of surrounding and / or adjacent structures, and the surrounding and / or adjacent structures surround and / or are adjacent to the target.
[0151] 48. A method according to clause 45, 46 or 47, wherein the patterning parameters include one or more pattern density parameters associated with any of the patterns.
[0152] 49. A method according to any of clauses 45 to 48, wherein the processing step comprises at least one etching step; and the one or more processing step parameters comprise one or more etching parameters.
[0153] 50. The method of clause 49, wherein the etching parameters include one or more of: etching pressure, etching flow rate, RF voltage, over-etching, edge ring bias / offset, chamber conditioning, etching tool and / or chamber used.
[0154] 51. A method according to clause 49 or 50, wherein the one or more etching parameters include an ion tilt parameter.
[0155] 52. A method according to clause 51, wherein the ion tilt parameter depends only on the radial position on the substrate.
[0156] 53. A method according to any of clauses 37 to 52, wherein, in the inferring step, a first model component directly uses the physically based input data.
[0157] 54. A method according to any one of clauses 37 to 52, comprising the step of fitting a model to the physics-based input data; wherein the first model component uses the fitted model when inferring the second measurement data.
[0158] 55. A method according to any one of clauses 37 to 54, wherein the model comprises at least one second model component, each of the at least one second model component comprises a corresponding physics-based model, and the physics-based model can be used to generate the physics-based input data for the physics-based input channel.
[0159] 56. The method of clause 55, wherein the at least one second model component comprises one or more of: a die-level second model component and / or a wafer-level second model component.
[0160] 57. The method of clause 55 or 56, wherein the at least one second model component comprises a physical etch chamber model component.
[0161] 58. A method according to clause 55, 56 or 57, wherein at least one second model component comprises a mode model component.
[0162] 59. A method according to any of clauses 37 to 58, wherein the at least one first model component comprises a neural network based model.
[0163] 60. A method according to clause 59, wherein the physically based input data is input to an intermediate layer of the neural network.
[0164] 61. A method according to clause 59 or 60, wherein the physically based input data is input to a first fully connected layer of the neural network.
[0165] 62. A method according to any of clauses 37 to 61, comprising determining settings and / or corrections of one or more processing step parameters associated with the processing step to optimize and / or bring the second measurement data into compliance with specifications.
[0166] 63. A method according to clause 62, comprising the step of performing said processing steps based on the determined settings and / or corrections.
[0167] 64. A method according to any of clauses 37 to 63, comprising determining whether the patterned substrate should be reworked based on the second metrology data.
[0168] 65. The method of clause 64, comprising determining that the wafer should be reworked only after using the model to determine that second metrology data inferred by the model cannot be adequately corrected by the processing step.
[0169] 66. A method according to clause 64 or 65, comprising using the model to determine settings and / or corrections for one or more exposure parameters for exposing a pattern on a reworked substrate to optimize and / or bring the second metrology data into specification.
[0170] 67. A method according to clause 64 or 65, comprising using the model to jointly optimize the settings and / or corrections of one or more exposure parameters for exposing a pattern on a reprocessed substrate, and determining the settings and / or corrections of one or more processing step parameters associated with the processing step to optimize and / or bring the second measurement data into compliance with specifications.
[0171] 68. A method according to clause 66 or 67, comprising performing said reworking based on said setting and / or correction of one or more exposure parameters.
[0172] 69. The method according to any one of clauses 37 to 68 comprises performing an initial training step to train the model, the training step comprising: training at least one first model component on training measurement data and corresponding training physics-based input data, wherein the training measurement data comprises known data similar to the first measurement data and the second measurement data.
[0173] 70. A method according to clause 69, wherein the training physics-based input data includes one or more of the following: a spatial description or value of one or more parameters of interest, wherein the spatial description or value is described within a field, within a die, within a target, across a wafer, or within each processing tool / chamber.
[0174] 71. A method according to clause 69 or 70, wherein the training physics-based input data comprises data relating to one or more patterning parameters and / or one or more etching parameters related to the pattern.
[0175] 72. The method of clause 71, wherein the etching parameters include one or more of: etching pressure, etching flow rate, RF voltage, over-etching, edge ring bias / offset, chamber adjustments, etching tool and / or chamber used.
[0176] 73. A method according to clause 71 or 72, wherein the one or more etching parameters include an ion tilt parameter.
[0177] 74. A method according to any of clauses 69 to 73, comprising training the model based on training data relating to changes in one or more exposure tool parameters and / or processing tool parameters.
[0178] 75. A method according to any one of clauses 69 to 74, comprising training a separate instance of the model to obtain a corresponding training model for each of two or more substrate regions and / or two or more processing tools or chambers; and the inference step comprises using the corresponding training model for its corresponding substrate region and / or processing tool or chamber.
[0179] 76. A method according to any of clauses 69 to 74, wherein training comprises adjusting a single model for the entire substrate and / or all processing tools and chambers by including within the training data the wafer regions and / or chamber numbers to which the training data relates.
[0180] 77. A computer program comprising program instructions which, when run on suitable equipment, are operable to perform the method of any one of clauses 37 to 76.
[0181] 78. A non-transitory computer program carrier comprising the computer program described in Article 77.
[0182] 79. A processing device comprising:
[0183] a computer program carrier containing a computer program as defined in clause 77; and
[0184] A processor operable to execute said computer program.
[0185] 80. A measurement device comprising a processing device according to clause 79.
[0186] 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, such 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.
[0187] 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 inferring second metrology data relating to at least one patterned substrate from first metrology data measured on the at least one patterned substrate before performing the at least one processing step, wherein the at least one patterned substrate has been exposed with a pattern and at least one processing step has been performed, the method comprising: Obtaining a model comprising at least one first model component, the at least one first model component comprising a machine learning model component that has been trained to map the first measurement data to the second measurement data, the first model component also comprising a physics-based input channel for receiving physics-based input data; as well as Second measurement data is inferred from the first measurement data using the first model component, wherein the first model component is biased by the physics-based input data on the physics-based input channel. 2 . The method according to claim 1 , wherein the first measurement data comprises one or more measurement images or representations thereof.
3. The method of claim 2, wherein the one or more metrology images may include one or more of: an image-based overlay image, a scatterometry image captured at a Fourier plane, a scatterometry image captured at an image plane, or a scanning electron microscope image. 4 . The method of claim 1 , wherein the second measurement data comprises values of one or more parameters of interest or a description of spatial variation of one or more parameters of interest.
5. The method of claim 4, wherein the parameters of interest include overlay, CD or edge placement error EPE. 6 . The method of claim 4 , wherein the parameter of interest comprises one or more of the following: line edge roughness, line width roughness, local critical dimension uniformity, average critical dimension. The method of claim 1 , wherein the second metrology data comprises electronic test data. The method according to claim 1 , wherein the second metrology data comprises a metrology image or a representation thereof.
9. The method of claim 1, wherein the physically based input data comprises one or more processing step parameters related to the processing step and / or one or more patterning parameters related to the pattern.
10. The method of claim 9, wherein the patterning parameters include at least target design parameters.
11. The method of claim 10, wherein the target design parameters include one or more of: a target pitch, a target critical dimension, a target design, a target sub-segmentation parameter, a target placement within a die, a target position, any parameters of surrounding and / or adjacent structures that surround and / or are adjacent to the target.
12. The method of claim 9, wherein the patterning parameters include one or more pattern density parameters associated with any of the patterns.
13. The method of claim 9, wherein the processing step comprises at least one etching step; and the one or more processing step parameters comprise one or more etching parameters.
14. The method of claim 1, wherein the model comprises at least one second model component, each of the at least one second model component comprising a corresponding physics-based model, the physics-based model being operable to generate the physics-based input data for the physics-based input channel.
15. A computer program comprising program instructions operable, when run on suitable means, to perform the method according to any one of claims 1 to 14.
