Determine the lithography matching performance

Abstract

A method for determining matching performance between tools for semiconductor manufacturing and associated tools are described. The method includes obtaining a plurality of data sets associated with a plurality of tools, and a representation of the data sets in a reduced space, the reduced space having reduced dimensions. Based on matching the reduced data sets in the reduced space, a matching metric and / or matching correction is determined.

Description

[0001] Cross - Reference to Related Applications

[0002] This application claims the priority of EP application 20157301.1 filed on February 14, 2020 and EP application 20176415.6 filed on May 26, 2020, which are incorporated herein by reference in their entirety. Field of the Invention

[0003] The present invention relates to a method for determining lithography matching performance between lithography apparatuses for semiconductor manufacturing, a semiconductor manufacturing method, a lithography apparatus, a lithography cell, and an associated computer program product. Background Art

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

[0005] To project a pattern onto a substrate, a lithography apparatus can use electromagnetic radiation. The wavelength of this radiation determines the minimum feature size that can be formed on the substrate. Typical wavelengths currently in use are 365 nm (i-line), 248 nm deep ultraviolet (DUV), 193 nm deep ultraviolet (DUV), and 13.5 nm. Compared with DUV lithography apparatuses using radiation with a wavelength of, for example, 193 nm, a lithography 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.

[0006] Low-k1 lithography can be used to process features smaller than the classical resolution limit of a lithography apparatus. In such a process, the resolution formula can be expressed as CD = k1 × λ / NA, where λ is the wavelength of the radiation employed, NA is the numerical aperture of the projection optics in the lithography apparatus, CD is the "critical dimension" (usually the smallest feature size printed, but in this case the half-pitch), and k1 is an empirical resolution factor. Generally, the smaller k1 is, the more difficult it is to replicate on a substrate a pattern similar in shape and size to what the circuit designer has planned for a particular electrical function and performance. To overcome these difficulties, complex fine-tuning steps can be applied to the lithography projection apparatus 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 referred to as "optical process correction") in the design layout), or other methods generally defined as "resolution enhancement techniques" (RET). Alternatively, a strict control loop for controlling the stability of the lithography apparatus can be used to improve the reproduction of patterns at low k1..

[0007] Cross-platform (e.g., DUV to EUV) matching performance between lithography apparatuses is crucial for overlay performance on a product. Generally, this is achieved using dedicated verification tests. Such tests require a specific machine setup procedure as a prerequisite, which takes several hours. Additional scanner and metrology time is also required for pre-setup, exposure, and overlay measurement. Such tests are only performed when absolutely necessary and thus cannot be used for routine monitoring purposes, which are necessary for high-volume manufacturing. SUMMARY OF THE INVENTION

[0008] There is a desire to provide a method for determining lithography matching performance between lithography apparatuses that solves the above problems.

[0009] Embodiments of the present invention are disclosed in the claims and the detailed description.

[0010] In a first aspect of the present invention, there is provided a method for determining matching performance between tools for semiconductor manufacturing, the method comprising: obtaining a plurality of data sets related to a plurality of tools; obtaining a representation of the data sets in a reduced space, the reduced space having reduced dimensions; and determining a matching metric and / or a matching correction based on matching the reduced data sets in the reduced space.

[0011] In a second aspect of the present invention, there is provided a semiconductor manufacturing process comprising the method for determining lithography matching performance according to the first aspect.

[0012] In a third aspect of the present invention, there is provided a lithography apparatus comprising:

[0013] Irradiation system, configured to provide a projection beam of radiation;

[0014] Support structure, configured to support a patterning device, which is configured to pattern the projection beam according to a desired pattern;

[0015] Substrate table, configured to hold a substrate;

[0016] Projection system, configured to project the patterned beam onto a target portion of the substrate; and

[0017] Processing unit, configured to determine lithography matching performance according to the method of the first aspect.

[0018] In a fourth aspect of the present invention, there is provided a computer program product comprising machine-readable instructions for causing a general-purpose data processing device to perform the steps of the method according to the first aspect. Description of the Drawings

[0019] Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying schematic diagrams, in which:

[0020] Figure 1 A schematic diagram of a lithography apparatus is depicted;

[0021] Figure 2 A schematic diagram of a lithography cell is depicted;

[0022] Figure 3 A schematic representation of overall lithography is depicted, showing the collaboration between three key technologies for optimizing semiconductor manufacturing;

[0023] Figure 4 is a flowchart of a decision-making method;

[0024] Figure 5 is a schematic diagram of a control mechanism in a lithography process using a scanner stability module;

[0025] Figure 6 A schematic diagram of the normal operation of a set of DUV and EUV lithography apparatuses with loop monitoring for stability control is depicted;

[0026] Figure 7 A problem of unavailability of a lithography apparatus that requires cross-platform lithography matching is depicted;

[0027] Figure 8 A test of determining cross-platform lithography matching performance using a conventional method is depicted;

[0028] Figure 9 Includes three graphs related to a common time frame: Figure 9(a) is a graph of the original parameter data, more specifically the reticle alignment (RA) data with respect to time t; Figure 9 (b) is an equivalent non - linear model function mf derived according to the method described herein; and Figure 9 (c) includes Figure 9 (a) and Figure 9 (b) the residual Δ between the graphs, showing the class indicator of the method according to an embodiment of the present invention;

[0029] Figure 10 is a schematic diagram of the encoder / decoder network used in an embodiment of the present invention;

[0030] Figure 11 is a flowchart of an embodiment according to a third main embodiment of the present invention;

[0031] Figure 12 a, b, c and d conceptually illustrate the concepts of clustering and manifold learning;

[0032] Figure 13 a, b and c conceptually illustrate Figure 11 the production monitoring application of the basic method; and

[0033] Figure 14 depicts a block diagram of a computer system for controlling the systems and / or methods disclosed herein. Detailed Description

[0034] In this document, the terms "radiation" and "beam" are used to cover all types of electromagnetic radiation, including ultraviolet radiation (e.g., wavelengths of 365, 248, 193, 157 or 126 nm) and EUV (extreme ultraviolet radiation, e.g., wavelengths in the range of about 5 - 100 nm).

[0035] The terms "reticle", "mask" or "patterning device" as used herein can be broadly interpreted to mean a general patterning device that can be used to impart a patterned cross - section corresponding to a pattern to be produced in a target portion of a substrate. The term "light valve" can also be used in this context. In addition to classical masks (transmission or reflection masks, binary masks, phase - shift masks, hybrid masks, etc.), examples of other such patterning devices include programmable mirror arrays and programmable LCD arrays.

[0036] Figure 1Schematically depicts a lithographic apparatus LA. The lithographic apparatus LA includes: 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 table) MT configured to support a patterning device (e.g., a mask) MA and connected to a first positioner PM configured to accurately position the patterning device MA according to certain parameters; a substrate support (e.g., a wafer table) WT configured to hold a substrate (e.g., a wafer coated with resist) W and connected to a second positioner PW configured to accurately position the substrate support according to certain parameters; and a projection system (e.g., a refractive projection lens system) PS configured to project the pattern imparted to the radiation beam B by the patterning device MA onto a target portion C (e.g., including one or more dies) of the substrate W.

[0037] In operation, the illumination system IL receives a radiation beam from a radiation source SO, e.g., via a beam delivery system BD. The illumination system IL may include various types of optical components for guiding, shaping, and / or controlling the radiation, such as refractive, reflective, magnetic, electromagnetic, electrostatic, and / or other types of optical components, or any combination thereof. The illuminator IL may be used to condition the radiation beam B such that it has a desired spatial and angular intensity profile in the plane of the patterning device MA in its cross-section.

[0038] The term "projection system" PS as used herein should be broadly interpreted to cover various types of projection systems including: refractive, reflective, catadioptric, anamorphic, magnetic, electromagnetic, and / or electrostatic optical systems, or any combination thereof, which are suitable for the exposure radiation used and / or other factors (such as the use of an 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.

[0039] The lithographic apparatus LA may be of the type in which at least a portion of the substrate may be overlapped by a liquid (e.g., water) having a relatively high refractive index to fill the space between the projection system PS and the substrate W, which is also referred to as immersion lithography. More information on immersion techniques is given in US6952253 (which is incorporated herein by reference).

[0040] The lithographic apparatus LA may also be of the type having two (also referred to as "dual stage") or more substrate supports WT. In such a "multi-stage" machine, the substrate supports WT may be used in parallel, and / or steps of subsequent exposure preparation of the substrate W on one of the substrate supports WT may be performed while another substrate W on another substrate support WT is being used to expose the pattern on another substrate W.

[0041] In addition to the substrate support WT, the lithographic apparatus LA may include a measurement stage. The measurement stage is arranged to hold sensors and / or cleaning devices. The sensors may be arranged to measure properties of the projection system PS or of the radiation beam B. The measurement stage may hold a plurality of sensors. The cleaning device may be arranged to clean a part of the lithographic apparatus, such as a part of the projection system PS or a part of the system providing the immersion liquid. The measurement stage may move beneath the projection system PS when the substrate support WT is moved away from the projection system PS.

[0042] In operation, the radiation beam B is incident on a patterning device MA (e.g., a mask) held on a mask support T and is patterned by the pattern (design layout) on the patterning device MA. After passing through the mask MA, the radiation beam B passes through the projection system PS which focuses the beam onto a target portion C of the substrate W. By means of the second positioner PW and the position measurement system IF, the substrate support WT can be accurately moved, e.g., so as to position different target portions C in the path of the radiation beam B at the focus and alignment positions. Similarly, the first positioner PM and possibly another position sensor (which is not explicitly shown in Figure 1 are available to accurately position the patterning device MA relative to the path of the radiation beam B. Mask alignment marks M1, M2 and substrate alignment marks P1, P2 may be used to align the patterning device MA and the substrate W. Although the illustrated substrate alignment marks P1, P2 occupy dedicated target portions, they may be located in the space between the target portions. When the substrate alignment marks P1, P2 are located between the target portions C, they are referred to as scribe alignment marks.

[0043] As Figure 2 shown, the lithographic apparatus LA may form part of a lithographic cell LC, which is sometimes also referred to as a lithocell or (lithographic) cluster. The lithographic cell LC typically also includes equipment for performing pre-exposure and post-exposure processes on the substrate W. Generally, this equipment includes a spin coater SC for depositing a resist layer, a developer DE for developing the exposed resist, a chill plate CH and a bake plate BK, e.g., for adjusting the temperature of the substrate W, e.g., for adjusting the solvent in the resist layer. A substrate transfer device (or robot) RO picks up the substrate W from the input / output ports I / O1, I / O2, moves the substrate W between different processing devices, and transports the substrate W to the load port LB of the lithographic apparatus LA. The devices in the lithographic cell (which are generally also collectively referred to as the track) are typically controlled by a track control unit TCU which itself may be controlled by a management control system SCS which may also control the lithographic apparatus LA, e.g., via a lithography control unit LACU.

[0044] To correctly and consistently expose the substrate W exposed by the lithographic apparatus LA, it is desirable to inspect the substrate to measure properties of the patterned structure, such as overlay errors between subsequent layers, line thickness, critical dimension (CD), etc. For this purpose, an inspection tool (not shown) may be included in the lithography cell LC. If an error is detected, especially if the inspection is performed before exposing or processing other substrates W of the same batch or lot, then, for example, the exposure of subsequent substrates and / or other processing steps to be performed on the substrate W may be adjusted.

[0045] The inspection device (which may also be referred to as a metrology device) is used to determine properties of the substrate W, in particular how the properties of different substrates W vary or how the properties associated with different layers of the same substrate W vary between the layers. Alternatively, the inspection device may 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 lithographic apparatus LA, or may even be a stand-alone device. The inspection device may measure properties on a latent image (the image in the resist layer after exposure), or a semi-latent image (the image in the resist layer after the post-exposure bake step PEB), or a developed resist image (where the exposed or unexposed portions of the resist have been removed), or even an etched image (after a pattern transfer step such as etching).

[0046] Generally, the patterning process in the lithographic apparatus LA is one of the most critical steps in a process that requires high precision in the dimensions of the structures formed on the substrate W and the positions at which the structures are located. To ensure such high precision, three systems may be combined into a so-called "integrated" control environment, as Figure 3 shown. One of these systems is the lithographic apparatus LA which is (effectively) connected to a metrology tool MT (the second system) and a computer system CL (the third system). The key to such an "integrated" environment is to optimize the collaboration between these three systems to enhance the overall process window and to provide a tight control loop to ensure that the patterning performed by the lithographic apparatus LA remains within the process window. The process window defines the range of process parameters (e.g., dose, focus, overlay accuracy) within which a particular manufacturing process produces a defined result (e.g., a functional semiconductor device), and typically allows the process parameters in the lithography process or patterning process to vary within this range of process parameters.

[0047] The computer system CL may use (a portion of) the design layout to be patterned to predict which resolution enhancement techniques are to be used and perform computational lithography simulations and calculations to determine which mask layouts and lithographic apparatus settings achieve the maximum overall process window for the patterning process (in Figure 3represented by the double arrow in the first scale SC1). Generally, the resolution enhancement technique is arranged to match the patterning capabilities of the lithography apparatus LA. The computer system CL can also be used to detect at which position within the process window the lithography apparatus LA is currently operating (e.g., by using the input from the metrology tool MT) to predict whether there may be defects due to, for example, sub-optimal processes (in Figure 3 represented by the arrow pointing to "0" in the second scale SC2).

[0048] The metrology tool MT can provide input to the computer system CL to enable accurate simulation and prediction, and can provide feedback to the lithography apparatus LA to identify possible drifts in, for example, the calibration state of the lithography apparatus LA (in Figure 3 represented by the multiple arrows in the third scale SC3).

[0049] Accordingly, the proposed method includes making a decision as part of the manufacturing process, the method comprising: obtaining scanner data related to one or more parameters of the lithography exposure step of the manufacturing process; deriving a class indicator from the scanner data, the class indicator indicating the quality of the manufacturing process; and, based on the class indicator, deciding on an action. The scanner data related to one or more parameters of the lithography exposure step may include: data generated by the scanner itself during the exposure step or during the exposure preparation step, and / or data generated by other stations (e.g., independent measurement / alignment stations) during the exposure preparation step. Thus, it does not necessarily have to be generated by or within the scanner. The term scanner is generally used to describe any lithography exposure apparatus.

[0050] Figure 4 is a flow chart depicting a method of making a decision during a manufacturing process using a fault detection and classification (FDC) method / system. Scanner data 400 is generated during exposure (i.e., exposure scanner data), or after a maintenance action (or by any other means). This scanner data 400 (which is digital in nature) is fed into the FDC system 410. The FDC system 410 converts the data into functional indicators based on the physical knowledge of the scanner, and aggregates these functional indicators according to the system physical knowledge in order to determine the class system indicator for each substrate. The class indicator can be binary, such as meeting the quality threshold (OK) or not meeting the quality threshold (NOK). Optionally, there may be more than two classes (e.g., based on statistical binning techniques).

[0051] Based on the scanner data 400, and more specifically, based on the class indicator assigned to the substrate, an inspection decision 420 is made to determine whether to inspect / detect the substrate. If it is decided not to inspect the substrate, the substrate is sent for processing 430. Some of these substrates may still undergo a metrology step 440 (e.g., input data for a control loop and / or verify the decision made in step 420). If it is decided to inspect in step 420, the substrate is measured 440, and based on the results of the measurement, a rework decision 450 is made to determine whether to rework the substrate. In another embodiment, the rework decision is made directly based on the class quality value determined by the FDC system 410 without an inspection decision. Depending on the result of the rework decision, the substrate is reworked 460, or is considered OK and is transferred for processing 430. If the latter, this would indicate that the class indicator assigned to the substrate is incorrect / inaccurate. Note that the actual decisions shown (inspection and / or rework) are merely exemplary, and other decisions can be based on the class values / suggestions output from the FDC, and / or the FDC output can be used to trigger an alert (e.g., to indicate poor scanner performance). The result of the rework decision 450 for each substrate is fed back to the FDC system 410. The FDC system can use this data to refine and validate its class and decision suggestions (the assigned class indicator). In particular, the FDC system can verify the assigned class indicator against the actual decision and, based on this, make any appropriate changes to the class criteria. For example, the FDC system can change / set any class thresholds based on the verification. Thus, all rework decisions made by the user at step 450 should be fed back in order to verify all inspection decisions of the FDC system 410. In this way, the class classifier within the FDC system 410 is continuously trained during production such that the class classifier receives more data and thus becomes more accurate over time.

[0052] The scanner generates digital scanner or exposure data, which includes digital data parameters or indicators generated by the scanner during exposure. For example, the scanner data may include any data generated by the scanner that may influence decisions to be recommended by the FDC system. For example, the scanner data may include measurement data from measurements routinely performed during exposure (or in preparation for exposure), such as reticle and / or wafer alignment data, leveling data, lens aberration data, any sensor output data, etc. The scanner data may also include less routine measurement data (or estimated data), such as data from less routine maintenance steps, or data inferred therefrom. Specific examples of such data may include source collector contamination data for EUV systems. The FDC system derives digital function indicators based on the scanner data. These function indicators can be trained on production data to reflect the actual usage of the scanner (e.g., temperature, exposure interval, etc.). Function indicators can be trained using, for example, statistical, linear / nonlinear regression, deep learning, or Bayesian learning techniques. For example, reliable and accurate function indicators can be constructed based on scanner parameter data and domain knowledge, where the domain knowledge may include measuring the deviation of scanner parameters from nominal. The nominal can be based on the known physics of the system / process and scanner behavior.

[0053] Then, a model can be defined that links these indicators to product-level indicators. The categories can be binary (e.g., OK / NOK), or more advanced categories based on measurement binning or patterns. The linking model binds the physics-driven function indicators to the product impacts observed for a particular user application and mode of operation. The category indicators aggregate the function indicators according to the physics of the system. There can be two or more levels or hierarchies of category indicators, each level or hierarchy for a specific error contribution term. For example, the first level may include overlap contribution terms (e.g., reticle alignment contribution terms for in-field overlap in the X direction, reticle alignment contribution terms for between-field overlap in the Y direction, leveling contribution terms for in-field CD, etc.). The category indicators at the second level can aggregate the category indicators at the first level (e.g., according to direction, and / or according to between-field to in-field for overlap, and / or according to between-field to in-field for CD). These can be further aggregated at the third level; for example, overlap OK / NOK, and / or CD OK / NOK. The above category indicators are merely examples, and any suitable alternative indicators can be used. Then, these indicators can be used to provide recommendations and / or make processing decisions, such as whether to inspect and / or rework the substrate.

[0054] A class indicator can be derived from a model / simulator based on machine learning techniques. Such a machine learning model can be trained with historical data (previous indicator data) labeled according to its appropriate class (i.e., should be reworked). The labeling can be based on in-line (expert) data (e.g., input from the user) and / or measurement results (e.g., based on), such that the model is taught to provide a prediction of the substrate quality that is valid and reliable based on future digital data inputs from scanner data. For example, system class indicator training can use feed-forward neural networks, random forests, and / or deep learning techniques. Note that the FDC system does not need to know any user-sensitive data used for this training; only high-level classes, tolerances, and / or decisions (e.g., whether to rework the substrate) are required.

[0055] Figure 5 Depicts an overall lithography and metrology method incorporating a stability module 500 (in this example, an application running essentially on a server). Three main process control loops are shown, labeled 1, 2, 3. The first loop provides a cyclic monitoring for the stability control of the lithography equipment using the stability module 500 and a monitor wafer. The monitor wafer (MW) 505 is shown passing through the lithography cell 510 and has been exposed to set baseline parameters for focus and overlay. At a later time, the metrology tool (MT) 515 reads these baseline parameters, which are then interpreted by the stability module (SM) 500 to calculate a correction route, thereby providing scanner feedback 550, which is passed to the main lithography equipment 510 and used during further exposure. The exposure of the monitor wafer can involve printing a mark pattern on top of a reference mark. By measuring the overlay error between the top and bottom marks, the performance deviation of the lithography equipment can be measured even when the wafer has been removed from the equipment and placed in the metrology tool.

[0056] The second (APC) loop is for local scanner control of the product (determining focus, dose, and overlay of the product wafer). The exposed product wafer 520 is passed to the metrology unit 515, which determines information related to parameters such as critical dimension, sidewall angle, and overlay, and is passed to the advanced process control (APC) module 525. This data is also passed to the stability module 500. Process correction 540 is performed before the manufacturing execution system (MES) 535 takes over, thereby providing control of the main lithography equipment 510 and communicating with the scanner stability module 500.

[0057] The third control loop allows the integration of metrology into the second (APC) loop (e.g., for double patterning). The etched wafer 530 is transferred to the metrology unit 515, which measures again the parameters read from the wafer (such as critical dimension, sidewall angle, and overlay). These parameters are transferred to the advanced process control (APC) module 525. Next, the loop is the same as the second loop.

[0058] Figure 6 A schematic diagram depicting the normal operation of a set of lithographic apparatuses in a loop monitoring scenario for stability control is shown. In the example given below, the lithographic apparatus is a scanner. Four deep ultraviolet scanners DUV1 to DUV4 are shown processing four wafer lots WL1 to WL4 in a lithographic exposure step n - 1. These wafer lots are then processed in four extreme ultraviolet scanners EUV1 to EUV4 in the next lithographic exposure step n. The wafer lots have dedicated routes. For example, wafer lot WL1 is exposed in deep ultraviolet scanner DUV1 and then in extreme ultraviolet scanner EUV1.

[0059] Each scanner has a process for loop monitoring for stability control, as described in reference Figure 5 Monitoring data is obtained by measuring one or more monitoring substrates processed periodically on the respective lithographic apparatuses. In Figure 6 , for example, extreme ultraviolet scanner EUV2 processes the monitoring wafer EMW, which is measured by the metrology tool MW, and the metrology tool MW outputs the overlay measurement result OV to the stability module SM. The overlay measurement result OV is recorded as the wafer map E2M, and the wafer map E2M includes a grid of overlay measurement results (which can be represented as overlay residuals). Thus, the first monitoring data E2M is obtained from the loop monitoring for stability control of the first lithographic apparatus EUV2. The first monitoring data E2M is in the first layout. For example, each fiducial has a specific location on the substrate where measurements are made. Also, as depicted in this example, deep ultraviolet scanner DUV2 processes the monitoring wafer DMW, which is measured by the metrology tool MW to output the overlay measurement result OV to the stability module SM. The overlay measurement result OV is recorded as the wafer map D2M, and the wafer map D2M includes a grid of overlay measurement results. Thus, the second monitoring data D2M is obtained from the loop monitoring for stability control of the second lithographic apparatus DUV2. The second monitoring data D2M is in a second layout different from the first layout. This difference results from the different layouts and densities of features on the monitoring wafers EMW and DMW, as well as the difference in the sampling scheme for overlay measurement. This is expected for different platforms such as DUV and EUV.

[0060] Figure 7Depicts the problem that the lithography equipment required for cross-platform lithography matching is unavailable. Figure 6 Shows the selected scanners. One of the EUV scanners, EUV2, is unavailable for production, possibly because it is down for preventive maintenance. Thus, the following questions arise: Where should the wafer lot WL2 from the second DUV scanner DUV2 be processed next? Which of the available EUV scanners EUV1, EUV3, or EUV4 should be used? The answer can be found by determining which of the EUV scanners has the best overlap matching performance with the DUV scanner DUV2.

[0061] Figure 8 Depicts determining cross-platform lithography matching performance using a conventional method. The cross-platform test wafer XW is exposed on the second DUV scanner DUV2, and the metrology tool MT measures the overlap OV2. The test wafer XW is reworked RW1 and exposed in the first EUV tool EUV1. Next, the metrology tool MT measures the overlap OV1. The test wafer XW is reworked RW2 and exposed in the third EUV tool EUV3. Then, the metrology tool MT measures the overlap OV3. Finally, the test wafer XW is reworked RW3 and exposed in the fourth EUV tool EUV4. Then, the metrology tool MT measures the overlap OV4. The cross-platform overlap matching performance between the second DUV scanner DUV2 and the first EUV scanner EUV1 is determined by calculating the difference between the corresponding overlap measurement results OV2 and OV1. This process is repeated for each of the remaining EUV scanners (i.e., OV2 - OV3 and OV2 - OV4). The differences are ranked, and the EUV scanner with the smallest difference is determined to have the best overlap matching performance. Then, the wafer lot WL2 has a route through that scanner. Refer to Figure 8 The dedicated verification test scanner setup procedure described as a prerequisite, which takes several hours. It is only performed when absolutely necessary and thus cannot be used for routine monitoring purposes, which are necessary for a high-volume manufacturing environment.

[0062] Other known matching methods use the output from cyclic monitoring for stability control (drift control, DC), such as that described regarding Figure 5 This method requires very complex models to extract the correct parameters from each calibration dataset and map these parameters to the scanner parameters. Any change in the scanner capabilities requires a fine change to the model. Any error contribution that is not part of the model can potentially introduce an unwanted drift between the systems.

[0063] To address one or more of these problems, improved matching methods have been proposed. Such methods include: obtaining a plurality of data sets associated with a plurality of tools; obtaining a model configured to represent the data sets as reduced data sets in a reduced space, the reduced space including reduced dimensions; and determining a matching metric based on matching the reduced data sets in the reduced space.

[0064] Three main embodiments will be described, a first physics-based method and second and third data-driven methods. The first method is based in part on Figure 4 the FDC system, and in particular on the functional indicators derived therefrom.

[0065] This embodiment is based on the fact that scanner functional indicators are related to scanner data (e.g., alignment data / levelling data / lens data / etc.) using physics / domain knowledge. The relationships of various functional indicators or the functional fingerprints defined thereby are scanner- and product-specified (trained). The functional indicators or fingerprints are represented in a reduced (or latent) feature space such that similar scanners appear as clusters in this feature space.

[0066] Figure 9 Includes three figures that illustrate the derivation of functional (and class) indicators and their effectiveness compared to currently used statistical indicators. Figure 9 (a) is a graph of the original parameter data, more specifically a graph of reticle alignment (RA) versus time t. The original parameter data can be related to any parameter of the scanner and / or the lithography process. Figure 9 (b) is an equivalent (e.g., for reticle alignment) non-linear model function (or fit) mf derived according to the method described herein. As described, such a model can be derived based on knowledge of scanner physics and can be further trained on production data (e.g., in this particular case, performing reticle alignment measurements when performing a particular manufacturing process of interest). For example, training of this model can use statistical, regression, Bayesian learning, or deep learning techniques. Figure 9 (c) includes Figure 9 (a) and Figure 9 (b) the residual Δ between the graphs, which can be used as a functional indicator of the method disclosed herein. One or more thresholds ΔT (e.g., initially based on user knowledge / industry opinion, and / or trained as described) can be set and / or learned to provide a class indicator. In particular, during the training phase of training a class classifier, by the class classifier block 430( Figure 4) Learning thresholds ΔT. These thresholds may actually be unknown or hidden (e.g., when implemented by a neural network). For example, a class indicator may be associated with one or more of overlap, focus, critical dimension, critical dimension uniformity (e.g., based on OK / NOK on which side of the threshold the value is, but non-binary class indicators are also possible and conceivable).

[0067] It is beneficial to compare this with the statistical control techniques that are currently commonly applied to raw data. Set the statistical threshold RAT to Figure 9 (a)'s raw data will result in an outlier being identified at time t1, but not at time t3. Additionally, the point at time t2 will be incorrectly identified as an outlier, but in fact, according to the class indicators disclosed herein, the point at time t2 is not an outlier (i.e., it is OK) (as Figure 9 (c) shows).

[0068] Functional indicators can be defined along the lifetime of the scanner and / or other tools on the wafer (e.g., self-loading, measurement (alignment / leveling, etc.), exposure, etc.). Thus, the raw data related to multiple scanners and process parameters can be processed in the same way as Figure 9 shown to obtain a functional indicator for each data, where the functional indicator includes the residuals (e.g., over time) relative to the expected behavior, nominal behavior, or average behavior. These functional indicators can be combined and / or aggregated by tool (and / or by process) to obtain a scanner functional fingerprint, which includes a model of the functions that define the performance of the scanner on the product.

[0069] In particular, semi-supervised machine learning techniques can be applied to the functional indicators to identify the scanner functional fingerprint. Such a fingerprint will be different for each scanner and optionally according to the product and layer. By examining the different indicators during the lifetime of the wafer, in-line rules can also determine the most critical matching functional indicators to be used (i.e., determine which functional indicators are more relevant to the match), and / or the changes most likely caused by the process, and thus the scanner match should not be used.

[0070] Then the functional indicators or functional fingerprints can be ranked; for example, according to their similarity to the tool of interest (such as the tool being matched or replaced (e.g., for consecutive layers)). Thus, if there is a need to match (or replace) one machine with another, other machines can be ranked in the reduced (or latent) space representing the functional indicators or fingerprints according to the proximity of the other machines to the machine being matched and replaced (e.g., based on the measurement of the similarity of the functional indicators or functional fingerprints).

[0071] Machine learning methods (such as clustering algorithms, e.g., unsupervised or semi-supervised) can be applied to scanner fingerprint data or functional indicators within a reduced or latent feature space (the terms reduced and latent feature space can be used interchangeably in the document). Such clustering algorithms can learn "normal" regions (e.g., to describe nominal or average scanner behavior) that have a high density of data points. In this reduced space, the distance or other matching metric between tools / scanners represents the degree of match of the machines.

[0072] In one embodiment, a trained model and decision-making framework such as that described with respect to Figure 4 can be used for scanner matching and / or scanner selection, e.g., to verify the results of the method just described, or as an alternative thereto. The method can include using a (e.g., scanner-specific) classifier (e.g., a neural network) that is trained to predict per-wafer performance based absolutely on scanner functional indicators. For example, a batch exposed to a first scanner can be run through an FDC engine associated with a second scanner being evaluated to determine if the second scanner is a good match for the first scanner. The FDC engine can return a failure probability prediction for each inspection type for that scanner combination. The difference between the prediction results (likelihood in percentage form) combined with the functional indicator values can provide further insight into the expected on-product performance match. By repeating this process over multiple batches, statistical information can be collected.

[0073] For example, if a scanner is unavailable, the most recent data from that scanner can be converted into a series of functional indicators (e.g., using the method already described with respect to Figure 4 ). The functional indicators can be input into a neural network associated with a different scanner (e.g., trained for a different scanner), and the resulting class indicators can be compared to values associated with the unavailable scanner. If the class indicators match or show a high correlation, it can be concluded that the scanners match well.

[0074] By combining the results of in-line rules, semi-supervised learning, and statistical comparison, the best-matching scanner for the same product and layer can be identified for a given product, layer, and scanner.

[0075] More data-driven methods will now be described in connection with Figure 10 . The method uses an encoder-decoder network, where the encoder EN encodes the input data x into a reduced space or latent space representation LS, and the decoder DE decodes the latent space representation back to the original data or an approximation of the original data x' (assuming sufficient training). Then, matching can be performed within the latent space LS; for example, the latent space can include vector representations and the matching can be performed by means of (n-dimensional) vector comparison.

[0076] Typically, models are trained on historical scanner datasets for multiple scanner platforms. By inputting data from multiple scanners (characterized by scanner ID), the model allows the similarity between scanners to be evaluated based on their positions in the latent space. Thus, tools can be ranked; for example, according to their similarity (proximity in the latent space) to a tool of interest, such as a matching tool or a tool to be replaced).

[0077] The method can also be used to determine (match) corrections based on selecting a reference value in the latent space (e.g., the average of the tool data represented therein), determining the vector displacement of the tool of interest relative to this reference value, and decoding this vector displacement into a correction for the tool of interest (or each tool), the correction being intended to remove these differences such that they perform more similarly (i.e., all show performance similar to a reference tool).

[0078] In a particular embodiment, first monitoring data is obtained from cyclic monitoring; for example, by means of a monitoring wafer of the type described Figure 5 The data can include overlaps, or other parameters of interest measurements performed on the monitoring wafer for baseline monitoring and stability control. Thus, the evaluated lithography matching performance can include overlap matching performance, and the monitoring data can include an overlap measurement grid (e.g., described as a wafer map or fingerprint). The monitoring data can be obtained by measuring one or more monitoring substrates periodically processed on a corresponding lithography apparatus or other tool. For example, the monitoring data can include one or both of inter-field data corresponding to multiple lithography exposure fields and intra-field data corresponding to a lithography exposure field.

[0079] Optionally, the monitoring data can include other scanner background data: such as alignment data, leveling data, temperature data, etc. The transformation in the latent space can transform each scanner into a reference or average scanner. Then, the output can include a set of machine matching overlap corrections (e.g., including corrections for each scanner).

[0080] Knowledge of existing machine matching methods and related models can be included in the encoder network (e.g., averaging some or all of the parameters and functions of previous runs). The differences between tools can be studied by projecting the latent space vectors back onto the measurement / machine parameters.

[0081] The measurement data is mapped to a vector in the latent space such that basic mathematical operations (e.g., addition, subtraction) can be performed on the vectors in the latent space. Thus, a specific reference state can be subtracted from (or added to) the data set. Moreover, other operations can be performed based on the nature of the data set, for example, subtracting reference data (reference state) related to a first type of scanner and adding reference data related to a second type of scanner. The trained network can capture unknown error sources and adapt to new scanner capabilities, and provide easier calibration for cross-platform matching.

[0082] In practice, when performing machine matching, it is challenging to distinguish between the part that can be fixed by APC / scanner calibration and the scanner-to-scanner differences that cause increased overlap and / or focus. This is because, although the statistical nature of the scanner-to-scanner differences seems to be within a small / limited range (in terms of mean and standard deviation), the impact of non-linear effects on the fingerprint differences can be significant.

[0083] Therefore, in a third main embodiment, a non-linear data-driven machine matching method is proposed. The method includes identifying the latent structure of the detection data related to the scanner (e.g., obtained from the monitored wafers as already described) by using non-linear dimensionality reduction techniques such as clustering and manifold learning techniques. Once clustered, a first set or cluster that shares similarities but does not have the same shape is identified within the monitored data. Then, these first sets each have their main fingerprints removed, as these fingerprints can be corrected, for example, using the APC correction loop described above. What remains is the processed monitored data (fingerprints) related to the nano-scale effect characteristics specific to each individual machine / chuck / track, etc. Thus, this transformed monitored data can be used to reveal the ideal / calibrated performance of the scanner. By performing a second non-linear dimensionality reduction (e.g., using clustering and manifold learning techniques) on this processed monitored data, multiple second sets (final data sets) can be obtained. Each of these second sets or data sets can determine a proposed match for the machine. The fact that the method is data-driven means that there is no need to make assumptions, and the determined match depends only on the measured data and performance.

[0084] Figure 11It is a flowchart describing such a method of matching (e.g., lithography) machines in a way that minimizes nanoscale differences between machines. Known or standard modeling techniques (e.g., using 6 parameters or higher order or any other alignment model) are used to model 1110 a monitoring data set 1100 (e.g., overlap, focus, or other parameters in the data of interest of a monitoring wafer) to obtain a modeled data set (e.g., fingerprint data). In step 1120, a first clustering and manifold learning step is performed on the fingerprint data, and in step 1130, the common data of each cluster or group is removed (e.g., the average value of each group is removed). In step 1140, a second clustering and manifold learning step is performed on the processed data (the common content has been removed). In step 1150, machines that are grouped together in the latent space defined by the previous steps are identified as matching machines (or their components, such as tracks / chucks, etc.). A pattern classification or feature extraction step 1160 (e.g., principal component analysis, other component analysis, or any pattern recognition and feature extraction algorithm) can be performed on the latent representation to identify and classify patterns and trends in the clusters. Each cluster can represent multiple machines with similar behavior, and this behavior can stem from several independent root causes. This last step 1160 can be used to find and identify these root causes / failure modes of the behavior observed within a cluster (e.g., heat-induced patterns, wafer load-induced patterns).

[0085] Figure 12 The clustering and manifold learning steps are shown in a simple 2D example. Figure 12 (a) is an example of fingerprint data, and Figure 12 (b) shows the result of the clustering step, thus showing three main clusters or groups (each circled in the figure). Figure 12 (c) is a manifold representation of the data that can be used to identify the "continuous" structure of the latent process. Then this data can be hierarchically processed to obtain Figure 12 (d) the representation, which describes the order of the data within the clusters.

[0086] The basic method of this embodiment can also be used for production monitoring via monitoring wafers. Figure 13 Conceptually shows the basic concept. By Figure 11 The foregoing steps described can be used to calculate / identify the latent structure of monitoring data related to multiple machines. Figure 13 (a) shows the result of this method, where each point represents a monitoring wafer. This shows a snapshot of how the machines perform in terms of the shape of the monitoring wafers. Figure 13(b) is a separate snapshot of the cluster of interest and includes the monitoring wafers for a particular machine / chuck / track; this snapshot can then be used as a reference to inspect future wafers. If everything is under control, any new / future wafers with the same origin should be classified / identified as members of the current cluster and manifold. Such future wafers are represented by the Figure 13 gray dots in (c). On the other hand, when a new cluster is formed, as Figure 13 represented by the black dots in (c). This indicates that there are significant changes in, for example, the monitoring of wafer production, and markings may accordingly appear when this occurs.

[0087] Note that the teachings herein (for all embodiments) can be extended to any type of processing tool where there may be a matching requirement (e.g., where replacement may also be unavailable) and / or which is subject to drift relative to a reference and for which correction is to be made. In addition to scanners (or steppers or any other lithographic exposure tool), such tools can include any metrology tool, polishing tool, etching tool / chamber, deposition tool, etc.

[0088] The methods described herein can be used to build a product that (1) tunes a scanner together with an active control loop, (2) optimizes the routing of wafers during production, and / or (3) performs scanner-to-process equipment matching.

[0089] It should be noted that although the description herein generally relates to a (single) latent (or reduced feature) space, this should not be considered limiting. The principles described herein can be applied with and / or to any number of latent spaces. For example, the systems, methods, (metrology) devices, non-transitory computer-readable media, etc. described herein can be configured such that a matching metric and / or correction can be determined based on one or more data associated with multiple scanners and represented in multiple (e.g., at least two) latent spaces.

[0090] Multiple latent spaces can be used serially (e.g., for analyzing a dataset and / or making a first match prediction, then a second match prediction, etc.), in parallel (e.g., for analyzing a dataset and / or making match predictions simultaneously), and / or otherwise. Advantageously, the individual latent spaces associated with a suitable model can be more stable than a single latent space. For example, different latent spaces can focus on specific properties of the dataset, e.g., one for retrieving a first match metric related to the overlapping properties of a scanner of interest, another for scanner classification based on the aberrations of the projection optics of the scanner of interest, etc. A combined latent space can be configured to capture all possibilities, while in the case of different latent spaces, each latent space can be configured (e.g., trained) to focus on a specific topic and / or aspect of the dataset. The individual latent spaces may be simpler but better at capturing information (e.g., when set accordingly).

[0091] In some embodiments, one or more latent spaces can include at least two latent spaces, multiple latent spaces, and / or other numbers of latent spaces, where a single latent space corresponds to different mechanisms of a model for defining the latent space. Different mechanisms of the model can include an encoding mechanism (e.g., Figure 10 the EN shown), a decoding mechanism (e.g., Figure 10 the DE shown), a match metric determination mechanism, and a scanner correction determination mechanism (e.g., determining one or more corrections to improve the match quality between scanners). In some embodiments, different mechanisms can correspond to different operations performed by one or more models for determining parameters of interest such as match metrics or corrections. By way of non-limiting example, in some embodiments, multiple latent spaces can be used in parallel, e.g., one for image encoding and / or decoding, another for predicting match metrics, and yet another for correction settings (e.g., predicting or recommending scanner set points) etc. The individual latent spaces corresponding to different mechanisms can be more stable than a single latent space associated with multiple mechanisms.

[0092] In some embodiments, the individual latent spaces can be associated with including as input (e.g., as Figure 10is associated with different independent parameters within the (multiple) datasets of the input 'x' depicted therein. Each latent space corresponding to a different independent parameter can also be more robust than a single latent space associated with multiple parameters. For example, in some embodiments, the present system and method can include or utilize a first latent space for matching overlaps between scanners, and a second separate latent space for processing interference that affects the imaging properties (affecting the dimensional properties of the patterns generated by the scanner of interest). The first latent space can be configured (e.g., trained) to perform overlap matching or characterization, and independently of the first latent space, the second latent space can be configured (e.g., trained) to process imaging differences caused by tool-specific properties. It should be noted that this is merely one possible example and is not intended to be limiting. Many other possible examples can be envisioned.

[0093] Figure 14 is a block diagram showing a computer system 1400 that can assist in implementing the methods and processes disclosed herein. The computer system 1400 includes a bus 1402 or other communication mechanism for communicating information, and a processor 1404 (or processors 1404 and 1405) coupled to the bus 1402 for processing information. The computer system 1400 also includes a main memory 1406 coupled to the bus 1402 for storing instructions and information to be executed by the processor 1404, such as random access memory (RAM) or other dynamic storage devices. The main memory 1406 can also be used to store temporary variables or other intermediate information during the execution of instructions to be executed by the processor 1404. The computer system 1400 further includes a read-only memory (ROM) 1408 or other static storage device coupled to the bus 1402 for storing static information and instructions for the processor 1404. A storage device 1410, such as a magnetic disk or optical disk, is provided and coupled to the bus 1402 for storing information and instructions.

[0094] The computer system 1400 can be coupled via the bus 1402 to a display 1412 for displaying information to a computer user, such as a cathode ray tube (CRT), flat panel display, or touch panel display. An input device 1414 including alphanumeric keys and other keys is coupled to the bus 1402 for communicating information and command selections to the processor 1404. Another type of user input device is a cursor control 1416 for communicating direction information and command selections to the processor 1404 and for controlling the movement of a cursor on the display 1412, such as a mouse, trackball, or cursor direction keys. Such input devices typically have two degrees of freedom in two axes (i.e., a first axis (e.g., x) and a second axis (e.g., y)), which allows the device to specify a position in a plane. A touch panel (screen) display can also be used as an input device.

[0095] One or more methods described herein can be performed by a computer system 1400 in response to one or more sequences of one or more instructions contained in a main memory 1406 being executed by a processor 1404. These instructions can be read from another computer-readable medium, such as a storage device 1410, into the main memory 1406. Execution of the instruction sequences contained in the main memory 1406 causes the processor 1404 to perform the process steps described herein. One or more processors in a multiprocessing arrangement can also be employed to execute the instruction sequences contained in the main memory 1406. In alternative embodiments, hardwired circuitry can be used in place of or in combination with software instructions. Accordingly, the description herein is not limited to any specific combination of hardware circuitry and software.

[0096] As used herein, the term "computer-readable medium" refers to any medium that participates in providing instructions to a processor 1404 for execution. Such a medium can take many forms, including but not limited to non-volatile media, volatile media, and transmission media. Non-volatile media includes, for example, optical or magnetic disks, such as storage device 1410. Volatile media includes dynamic memory, such as main memory 1406. Transmission media includes coaxial cables, copper wire, and fiber optics, including the wires that comprise a bus 1402. Transmission media can also take the form of acoustic or light waves, such as those generated during radio frequency (RF) and in infrared (IR) data communications. Common forms of computer-readable media include, for example, a floppy disk, a flexible disk, a hard disk, a magnetic tape, any other magnetic medium, a CD-ROM, a DVD, any other optical medium, punch cards, paper tape, any other physical medium with patterns of holes, a RAM, a PROM, and an EPROM, a FLASH-EPROM, any other memory chip or cartridge, a carrier wave as described below, or any other medium readable by a computer.

[0097] Various forms of computer-readable media can be involved in carrying one or more sequences of one or more instructions to a processor 1404 for execution. For example, the instructions can initially be carried on a magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A modem local to the computer system 1400 can receive the data on the telephone line and convert the data into an infrared signal using an infrared transmitter. An infrared detector coupled to the bus 1402 can receive the data carried in the infrared signal and place the data on the bus 1402. The bus 1402 carries the data to the main memory 1406, from which the processor 1404 retrieves and executes the instructions. The instructions received by the main memory 1406 can optionally be stored on the storage device 1410 either before or after execution by the processor 1404.

[0098] The computer system 1400 also preferably includes a communication interface 1418 coupled to the bus 1402. The communication interface 1418 provides a two-way data communication coupling with a network link 1420 connected to a local area network 1422. For example, the communication interface 1418 can 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 1418 can be a Local Area Network (LAN) card to provide a data communication connection to a compatible LAN. A wireless link can also be implemented. In any such implementation, the communication interface 1418 transmits and receives electrical, electromagnetic, or optical signals that carry a digital data stream representing various types of information.

[0099] The network link 1420 generally provides data communication to other data devices through one or more networks. For example, the network link 1420 can provide a connection through the local area network 1422 to a main computer 1424 or to data equipment operated by an Internet Service Provider (ISP) 1426. The ISP 1426 in turn provides data communication services through the global packet data communication network (now commonly referred to as the “Internet” 1428). Both the local area network 1422 and the Internet 1428 use electrical, electromagnetic, or optical signals that carry a digital data stream. Signals passing through various networks and signals on the network link 1420 and through the communication interface 1418 (which carry digital data to and from the computer system 1400) are example forms of carrier waves that convey information.

[0100] The computer system 1400 can send messages and receive data including program code over the network, network link 1420, and communication interface 1418. In the Internet example, a server 1430 can transmit the requested code for an application through the Internet 1428, ISP 1426, local area network 1422, and communication interface 1418. For example, one such downloaded application can provide for one or more of the techniques described herein. The received code can be executed by the processor 1404 upon receipt, and / or stored in the storage device 1410, or other non-volatile storage device for later execution. In this manner, the computer system 1400 can obtain application program code in the form of a carrier wave.

[0101] Embodiments can be implemented in a lithographic apparatus such as described with reference to Figure 1 a lithographic apparatus including:

[0102] an illumination system configured to provide a projection beam of radiation;

[0103] A support structure configured to support a patterning device configured to pattern a projection beam according to a desired pattern;

[0104] A substrate table configured to hold a substrate;

[0105] A projection system configured to project the patterned beam onto a target portion of the substrate, and

[0106] A processing unit configured to perform any of the methods described herein.

[0107] Embodiments may be implemented in any apparatus such as those represented in the lithography cell referred to in the reference Figure 2 described.

[0108] Embodiments may be implemented in a computer program product comprising machine-readable instructions for causing a general purpose data processing apparatus to perform the steps of the described methods.

[0109] Although specific reference may be made herein to the use of a lithographic apparatus in the manufacture of ICs, it should be understood that the lithographic apparatus described herein may have other applications. Possible other applications include the manufacture of integrated optical systems, guidance and detection patterns for magnetic domain memories, flat panel displays, liquid crystal displays (LCDs), thin film magnetic heads, and the like.

[0110] Although embodiments of the invention may be specifically referred to herein in the context of an inspection or lithographic apparatus, embodiments of the invention may be used in other apparatuses. Embodiments of the invention may form part of a mask inspection apparatus, a lithographic apparatus, or any apparatus for measuring or processing an object such as a wafer (or other substrate) or a mask (or other patterning device). It is also noted that the term metrology apparatus or metrology system encompasses the term inspection apparatus or inspection system or may be replaced by the term inspection apparatus or inspection system. The metrology or inspection apparatus disclosed herein may be used to detect defects on or in a substrate and / or defects in structures on a substrate. In such embodiments, the characteristics of the structure on the substrate may be related to, for example, defects in the structure, missing specific portions of the structure, or the presence of undesired structures on the substrate.

[0111] Although specific reference is made to "metrology apparatus / tool / system" or "inspection apparatus / tool / system", these terms may refer to the same or similar types of tools, apparatuses or systems. For example, an inspection or metrology apparatus comprising an embodiment of the invention may be used to determine the characteristics of an entity system (such as a structure on a substrate or a wafer). For example, an inspection apparatus or metrology apparatus comprising an embodiment of the invention may be used to detect defects in a substrate or defects in a structure on a substrate or a wafer. In such embodiments, the characteristics of the entity structure may be related to defects in the structure, missing specific portions of the structure, or the presence of undesired structures on the substrate or wafer.

[0112] Although the above has specifically referred to the use of embodiments of the present invention in the context of optical lithography, it should be understood that, where the context permits, the present invention is not limited to optical lithography and can be used in other applications (such as imprint lithography).

[0113] Other embodiments are disclosed in the list of numbered items below.

[0114] 1. A method for determining matching performance between tools for semiconductor manufacturing, the method comprising:

[0115] Obtaining a plurality of data sets associated with a plurality of tools;

[0116] Obtaining a representation of the data sets in a reduced space to obtain a reduced data set, the reduced space having a reduced dimension; and

[0117] Determining a matching metric and / or a matching correction based on the reduced data set characterized in the reduced space.

[0118] 2. The method according to item 1, wherein each data set is associated with a different respective tool.

[0119] 3. The method according to item 1 or 2, wherein the data sets are associated with changes over time of one or more tools and / or manufacturing parameters.

[0120] 4. The method according to any one of the preceding items, wherein the data sets describe parameters of a substrate during the entire manufacturing process of one or more tools.

[0121] 5. The method according to any one of the preceding items, wherein the representation includes at least one model configured to represent the data sets in the reduced space, the at least one model includes one or more functional models based on known physics associated with a particular manufacturing step or process and associated tools, and the method comprises: determining one or more functional indicators based on the one or more functional models and the plurality of data sets.

[0122] 6. The method according to item 5, wherein the one or more functional indicators describe deviations of parameter values from a nominal behavior derived from the known physics.

[0123] 7. The method according to item 5 or 6, wherein each of the one or more functional indicators is trained using one or more of: statistical techniques, optimization, regression, or machine learning techniques.

[0124] 8. The method according to any one of clauses 5 to 7, comprising: combining and / or aggregating the function indicators by tool and / or by process to obtain a tool function fingerprint including a model, the function of the model defining the performance of the tool.

[0125] 9. The method according to clause 8, wherein machine learning techniques are applied to the function indicators to identify the tool function fingerprint.

[0126] 10. The method according to any one of clauses 5 to 9, comprising determining which function indicators are more relevant to the matching metric.

[0127] 11. The method according to any one of clauses 5 to 10, comprising grading the function indicators or the tool function fingerprint according to the matching metric.

[0128] 12. The method according to any one of clauses 5 to 11, wherein the grading comprises: grading the function indicators or the tool function fingerprint according to the similarity to a tool of interest or other reference.

[0129] 13. The method according to any one of clauses 5 to 12, comprising applying a clustering algorithm to the function indicators or the tool function fingerprint to determine the matching metric.

[0130] 14. The method according to any one of clauses 5 to 13, comprising: applying a decision model that outputs a value for each of one or more class indicators based on parameter data to parameter data related to one or more tools to be matched, each of the one or more class indicators indicating the quality of a manufacturing process; and

[0131] deciding or verifying whether the machine is well-matched based on the class indicators.

[0132] 15. The method according to clause 14, comprising using a decision model trained for a first tool to absolutely predict the performance based on the parameter data of a second tool, so as to evaluate whether the first tool and the second tool are well-matched.

[0133] 16. The method according to clause 14 or 15, wherein each of the one or more class indicators is derived from the one or more function indicators by classifying the function indicators according to one or more thresholds applied to and / or learned for the one or more function indicators.

[0134] 17. The method according to clause 1 or 2, wherein the representation comprises an encoder-decoder network model, the encoder-decoder network model being operable to encode the data set into the reduced spatial representation and decode back from the reduced spatial representation to the data set.

[0135] 18. The method according to item 17, wherein the reduced space representation is a latent space, the latent space includes a vector representation, and the matching metric is based on vector comparison.

[0136] 19. The method according to item 18, comprising:

[0137] selecting a reference within the latent space;

[0138] determining a vector displacement of one or more of the plurality of tools relative to the reference; and

[0139] decoding the vector displacement into a correction for one or more of the plurality of tools, each correction causing the corresponding tool to perform more similarly to the reference.

[0140] 20. The method according to any one of items 17 to 19, comprising: grading the tools according to the proximity of the tools to a tool of interest or other reference in the latent space.

[0141] 21. The method according to any one of items 17 to 20, comprising: subtracting reference data related to a first type of tool and adding reference data related to a second type of tool within the latent space to match the first type of tool with the second type of tool.

[0142] 22. The method according to any one of items 17 to 21, comprising: training the model on historical scanner datasets for a plurality of tools and a plurality of tool types.

[0143] 23. The method according to item 1 or 2, wherein the step of obtaining a representation of the dataset in the reduced space comprises: performing one or more non-linear dimensionality reduction techniques on the dataset.

[0144] 24. The method according to item 23, wherein the one or more non-linear dimensionality reduction techniques comprise: performing clustering and manifold learning on the dataset to group the dataset into data groups; and

[0145] determining the matched tools as tools belonging to a common data group.

[0146] 25. The method according to item 24, comprising:

[0147] performing a first clustering and manifold learning step to obtain a first group;

[0148] removing the common and / or dominant data patterns of each first group to obtain a processed dataset; and

[0149] Performing a second clustering and manifold learning step on the processed data set to obtain the data groups.

[0150] 26. The method according to item 24 or 25 further includes: performing a pattern classification and / or feature classification step on one or more data groups to identify root causes or failure modes.

[0151] 27. The method according to any one of items 24 to 26 includes performing production monitoring based on the reduced space; the method includes,

[0152] Obtaining one or more other said data sets related to the actual production process;

[0153] For the corresponding said data groups, referring to the one or more other said data sets in the reduced space.

[0154] 28. The method according to item 27 includes: if the reference indicates a significant change between the one or more other said data sets and the corresponding said data groups, marking potential problems.

[0155] 29. The method according to any one of items 17 to 28, wherein each of the data sets includes: monitoring data from cyclic monitoring for stability control of the multiple tools.

[0156] 30. The method according to item 29, wherein the monitoring data includes overlapping or focused measurement grids.

[0157] 31. The method according to item 29 or 30 obtains the monitoring data by measuring one or more monitoring substrates periodically processed on the corresponding tools.

[0158] 32. The method according to item 29, 30 or 31, wherein the monitoring data includes other tool environments, such as one or more of alignment data, leveling data, temperature data.

[0159] 33. The method according to any one of the foregoing items, wherein the multiple tools include one or more of the following: lithography exposure tools, metrology tools, polishing tools, etching tools / chambers, and deposition tools.

[0160] 34. A semiconductor manufacturing process includes a method for determining lithography matching performance according to any one of the foregoing items.

[0161] 35. A computer program product includes machine-readable instructions for causing a general-purpose data processing device to perform the steps of the method according to any one of items 1 to 33.

[0162] 36. A processing unit and a storage device, comprising the computer program product according to item 35.

[0163] 37. A lithographic apparatus, comprising:

[0164] An illumination system configured to provide a projection beam of radiation;

[0165] A support structure configured to support a patterning device, the patterning device being configured to pattern the projection beam according to a desired pattern;

[0166] A substrate table configured to hold a substrate;

[0167] A projection system configured to project the patterned beam onto a target portion of the substrate; and

[0168] The processing unit according to item 36.

[0169] 38. A lithographic cell, comprising the lithographic apparatus according to item 37.

[0170] 39. A non - transitory computer - readable medium having instructions thereon, which when executed by a computer cause the computer to:

[0171] Obtain a plurality of data sets related to a plurality of tools used in a semiconductor manufacturing process; [[ID=S29]]

[0172] Obtain a representation of the data sets in a reduced space to obtain a reduced data set, the reduced space having a reduced dimension; and

[0173] Determine a matching metric and / or matching correction based on the reduced data set characterized in the reduced space.

[0174] 40. The medium according to item 39, wherein the reduced space comprises one or more latent spaces.

[0175] 41. The medium according to item 40, wherein the one or more latent spaces comprise at least two latent spaces.

[0176] 42. The medium according to item 40 or 41, wherein the one or more latent spaces comprise a plurality of latent spaces, and each of the plurality of latent spaces corresponds to a different mechanism of a model for defining the one or more latent spaces.

[0177] 43. The medium according to item 42, wherein the different mechanisms of the model comprise an encoding mechanism and a decoding mechanism.

[0178] 44. The medium according to item 43, wherein different mechanisms of the model further include a matching metric determination mechanism and / or a tool calibration determination mechanism.

[0179] 45. The medium according to any one of items 40 to 44, wherein the one or more latent spaces include at least two latent spaces associated with different independent parameters included in the plurality of data sets.

[0180] 46. The medium according to item 45, wherein the different independent parameters include overlap-related parameters and imaging-related parameters.

[0181] 47. The method according to any one of items 1 to 33, wherein the reduced space includes one or more latent spaces.

[0182] 48. The method according to item 47, wherein the one or more latent spaces include at least two latent spaces.

[0183] 49. The method according to item 47 or 48, wherein the one or more latent spaces include a plurality of latent spaces, and each of the plurality of latent spaces corresponds to a different mechanism of the model for defining the one or more latent spaces.

[0184] 50. The method according to item 49, wherein the different mechanisms of the model include an encoding mechanism and a decoding mechanism.

[0185] 51. The method according to item 50, wherein the different mechanisms of the model further include a matching metric determination mechanism and / or a tool calibration determination mechanism.

[0186] 52. The method according to any one of items 47 to 51, wherein the one or more latent spaces include at least two latent spaces associated with different independent parameters included in the plurality of data sets.

[0187] 53. The method according to item 52, wherein the different independent parameters include overlap-related parameters and imaging-related parameters.

[0188] Although specific embodiments of the present invention have been described above, it should be understood that the present invention can be implemented in forms different from the above. The above description is intended to be illustrative rather than restrictive. Therefore, those of ordinary skill in the art can understand that the described invention can be modified without departing from the scope of protection of the following claims.

Claims

1. A method for determining matching performance between tools for semiconductor manufacturing, the method comprising: Obtaining a plurality of data sets associated with a plurality of tools; Obtaining a representation of the data sets in a reduced space to obtain a reduced data set, the reduced space having a reduced dimension, wherein the obtaining comprises one of the following: Performing one or more non-linear dimensionality reduction techniques on the data sets; or Using an encoder-decoder network model to encode the data sets into the reduced space representation and decoding back from the reduced space representation to the data sets; and Determining a matching metric and / or matching correction based on the reduced data set characterized in the reduced space.

2. The method according to claim 1, wherein, Each data set is associated with a different respective tool.

3. The method according to claim 1, wherein The data sets are associated with changes over time of one or more tools and / or manufacturing parameters.

4. The method according to claim 1, wherein, The reduced space representation is a latent space, the latent space includes vector representations, and the matching metric is based on vector comparison.

5. The method according to claim 4, further comprising: Selecting a reference within the latent space; Determining a vector displacement of one or more of the plurality of tools relative to the reference; and And Decoding the vector displacement into a correction for one or more of the plurality of tools, each correction causing its respective tool to perform more similarly to the reference.

6. The method according to claim 4 or 5, further comprising grading the tools according to their proximity in the latent space to a tool of interest or other reference.

7. The method according to claim 4 or 5, further comprising subtracting reference data associated with a first type of tool within the latent space and adding reference data associated with a second type of tool to match the first type of tool with the second type of tool.

8. The method according to claim 4, further comprising training the model for historical scanner data sets of a plurality of tools and a plurality of tool types.

9. The method according to claim 1 or 2 or 3, wherein, The one or more non-linear dimensionality reduction techniques include: performing clustering and manifold learning on the data sets to group the data sets into data groups; and determining the matching tools as tools belonging to a common data group.

10. The method according to claim 9, further comprising: Performing a first clustering and manifold learning step to obtain a first group; Removing common and / or dominant data patterns of each first group to obtain a processed data set; and And Performing a second clustering and manifold learning step on the processed data set to obtain the data groups.

11. A non-transitory computer-readable medium having instructions thereon that, when executed by a computer, cause the computer to: Obtain a plurality of data sets associated with a plurality of tools used in a semiconductor manufacturing process; Obtain a representation of the data sets in a reduced space to obtain a reduced data set, the reduced space having a reduced dimension, wherein the obtaining comprises one of the following: Performing one or more non-linear dimensionality reduction techniques on the data sets; or Encode the dataset into the reduced spatial representation using an encoder-decoder network model and decode back from the reduced spatial representation to the dataset; and Determine a matching metric and / or a matching correction based on the reduced dataset characterized in the reduced space.

12. The medium according to claim 11, wherein, The representation is a latent space, the latent space includes a vector representation, and the matching metric is based on vector comparison.

13. The medium according to claim 11 or 12, wherein, The reduced space includes a plurality of latent spaces, where each of the plurality of latent spaces corresponds to a different mechanism of the model for defining the reduced space.

14. The medium according to claim 13, wherein, The different mechanisms of the model further include a matching metric determination mechanism and / or a tool correction determination mechanism.

15. The medium according to claim 14, wherein The one or more latent spaces include at least two latent spaces associated with different independent parameters included in the plurality of datasets.

Images