Metrology system and method based on scattering measurement

Abstract

The invention relates to a scattering measurement-based metrology system and method. The measurement model is trained from raw scatterometry data collected by a scatterometry-based overlay metrology system from an experimentally designed DOE wafer. Each measurement site includes one or more metrology targets fabricated with programmed overlay variations and known process variations. Each measurement site is measured with a known metrology system variation. In this manner, the measurement model is trained to separate the actual overlay from process variations and metrology system variations that affect overlay measurements. Thus, estimation of the actual overlay from the training measurement model is robust to process variations and metrology system variations. The measurement model is trained based on scatter measurement data collected from the same metrology system used to perform measurements. Thus, the measurement model is insensitive to system errors, asymmetry, and the like.

Description

[0001] Related Application

[0002] This application is a divisional application of the application patent application with application number 201880025696.X, filed on March 1, 2018, entitled “Metrology system and method based on scatterometry”.

[0003] Cross Reference to Related Applications

[0004] This patent application claims priority under 35 U.S.C. §119 to U.S. Provisional Patent Application No. 62 / 465,163, entitled “Method and System for Process-Robust Overlay Metrology Using Optical Scatterometry,” filed March 1, 2017, the subject matter of which is incorporated herein by reference in its entirety. TECHNICAL FIELD

[0005] The described embodiments relate to metrology systems and methods, and more particularly, the described embodiments relate to methods and systems that improve overlay measurements. BACKGROUND

[0006] Semiconductor devices such as logic and memory devices are typically fabricated by a sequence of processing steps applied to a sample. Various features and multiple structural levels of the semiconductor devices are formed by these processing steps. For example, photolithography is one semiconductor manufacturing process that involves generating patterns on a semiconductor wafer. Additional examples of semiconductor manufacturing processes include, but are not limited to, chemical-mechanical polishing, etch, deposition, and ion implantation. Multiple semiconductor devices can be fabricated on a single semiconductor wafer and are then separated into individual semiconductor devices.

[0007] Metrology processes are used in various steps during the semiconductor manufacturing process to detect defects on wafers to facilitate higher yields. Optical metrology techniques offer the possibility of high throughput and no risk of sample destruction. A large number of optical metrology based techniques, including scatterometry and reflectometry implementations and associated analysis algorithms, are commonly used to characterize critical dimensions, film thickness, composition, overlay, and other parameters of nanoscale structures.

[0008] Semiconductor devices are typically fabricated by depositing a series of layers onto a substrate. Some or all of these layers contain various patterned structures. The relative positions of structures within and between specific layers are crucial to the performance of the electronic device. Overlay refers to the relative positions of overlying or interleaved structures on the same or different layers of a wafer. Overlay error refers to the deviation from the nominal (i.e., desired) relative positions of the overlying or interleaved structures. The larger the overlay error, the more structural misalignments there are. If the overlay error is too large, the performance of the fabricated electronic device will be compromised.

[0009] Overlay errors are typically evaluated based on measurements of specific target structures formed on a wafer at various locations using photolithography tools. Optical metrology techniques are commonly employed to perform overlay measurements. In some instances, image-based overlay (IBO) metrology is used. IBO measurements involve imaging a specific target based on reflected light. The target structure can take many forms, such as a box-in-box structure or a rod-in-rod structure. In one example, a box is formed on one layer of the wafer, and a second, smaller box is formed on another layer. Local overlay errors are measured by comparing the alignment between the centers of the two boxes. Such measurements are performed at the wafer locations where the target structure is available. Overlay is measured by processing each image to estimate the distances between target features printed on different layers from the measurement images.

[0010] Unfortunately, these specific target structures often do not conform to the design rules of the particular semiconductor manufacturing processes used to produce electronic devices. This leads to errors in estimating the superimposed errors associated with the actual device structure manufactured according to the applicable design rules. For example, IMO metrology typically requires patterns containing thick lines with critical dimensions far exceeding the critical dimensions of the design rules to be successfully resolved by an optical microscope.

[0011] In some other instances, the superposition is estimated using light scattered or diffracted from the superimposed targets. Superposition metrology based on scattering measurements (often called Scatter Measurement Superposition (SCOL) or Diffraction-Based Superposition (DBO)) primarily relies on differential measurements of optical signals corresponding to diffracted optical signals from two different targets, each with a programmed superposition offset. Unknown superposition errors are then extracted based on these differential measurements.

[0012] Most existing methods based on scattering measurements characterize superposition error by relying on metrics sensitive to structural asymmetry. For example, existing angle-resolved scattering superposition (SCOL) methods characterize superposition based on the measurement asymmetry between the +1 and -1 diffraction orders. However, relying solely on asymmetry as an indicator of superposition error is problematic because process-induced variations (which include both symmetric and asymmetric variations) significantly affect superposition measurements. For instance, asymmetric process variations such as sidewall angle asymmetry, spectral profile asymmetry, or beam illumination asymmetry couple into the superposition of measurement signals, creating asymmetry. This leads to inaccurate measurements of superposition error. In other instances, symmetric process variations, such as film thickness variations, couple into the superposition of measurement signals, creating asymmetry.

[0013] Typically, robustness to process variations (i.e., variations in non-superimposed parameters affecting measurement asymmetry) in SCOL and DBO is addressed by selecting specific illumination wavelengths insensitive to process variations and optimizing the target design to reduce sensitivity to these variations. Unfortunately, both approaches are limited in their effectiveness. For example, selecting a specific illumination wavelength can lead to small superimposed measurement errors, but only within a small range across the entire process window. This makes measurements unreliable and inconsistent, requiring frequent re-evaluation of the illumination wavelength. Target design optimization is very time-consuming and requires accurate models of the structure, material dispersion, and optical system. Validating the accuracy of the model is also very challenging because masks and targets are often unavailable during formulation development. Furthermore, while target optimization reduces measurement sensitivity to process variations, it does not completely address robustness across the entire process variation window.

[0014] Figures 1A to 1C This illustrates a pupil image of a measurement diffraction order associated with SCOL measurements performed at three different illumination wavelengths. Figure 1A Image 10 depicting the measured diffraction order at an illumination wavelength of 523 nm. Figure 1B Image 11 depicting the measured diffraction order at an illumination wavelength of 579 nm. Figure 1C Image 12 depicting the measured diffraction order at an illumination wavelength of 668 nm.

[0015] like Figure 1B As explained, measurement image 11 is distorted by an arc passing through the middle of the image due to resonance at the wavelength. The total intensity of the measured diffraction order is affected by the resonant arc and the resulting superposition estimate. Images 10 and 12 do not exhibit this distortion, and the resulting superposition estimates associated with these illumination wavelengths are more accurate.

[0016] Figure 2 A plot depicting the inaccuracies of superimposed measurements via the SCOL system within the wavelength range. Figure 13Plot line 14 depicts three different wavelength ranges where superposition inaccuracies spike to unacceptable levels. Points 15 to 17 are respectively... Figures 1A to 1C The illumination wavelengths described herein correspond to those in the diagram. Figure 2 This is commonly referred to as the inaccuracy landscape. This plot is used to analyze superimposed inaccuracies and robustness to process variations.

[0017] Current formulation development methods for specific superimposed measurement applications aim to avoid wavelength sub-regions where inaccuracy spikes. However, the presence of symmetrical and asymmetrical process variations causes a shift in the inaccuracy profile, complicating the selection of suitable illumination wavelengths.

[0018] Figure 3 Plot 20 depicts several inaccuracy states associated with different values ​​of spectral profile asymmetry. Plot line 21 depicts inaccuracy without spectral profile asymmetry. Plot line 22 depicts inaccuracy with 2 nm spectral profile asymmetry. Plot line 23 depicts inaccuracy with 4 nm spectral profile asymmetry. Plot line 24 depicts inaccuracy with 8 nm spectral profile asymmetry. (The text repeats itself here.) Figure 3 As illustrated, the amplitude of the induced inaccuracy in superposition measurements increases with increasing structural asymmetry (e.g., due to printing errors). In this example, the increase in inaccuracy is linearly proportional to the amplitude of the spectral profile asymmetry.

[0019] Figure 4 Plot 30 depicts several inaccuracy scenarios associated with different values ​​of height variation (symmetric process variation) of the superimposed structure. Plot line 31 depicts inaccuracies without height variation. Plot line 32 depicts inaccuracies with a height variation of +6 nanometers. Plot line 33 depicts inaccuracies with a height variation of -6 nanometers. (The text repeats itself here.) Figure 4 As explained in the paper, the wavelength of the inaccuracy state shifts as the symmetry process changes.

[0020] Figure 3 and 4 This indicates that superposition inaccuracies depend on symmetrical and asymmetrical variations. Superposition errors are amplified by asymmetrical process variations and shifted by symmetrical process variations. If symmetrical process variations exist, the peak of superposition inaccuracy is not fixed at a specific wavelength. Therefore, selecting a suitable illumination wavelength based on a specific inaccuracy profile cannot sufficiently prove the existence of asymmetrical and symmetrical process variations. This risk can be mitigated by selecting illumination wavelengths based on several inaccuracy profiles covering the range of symmetrical and asymmetrical process variations; however, in some cases, the variations are too large for any illumination wavelength to result in sufficiently accurate superposition measurements. Therefore, in some situations, it is impossible to generate a superposition measurement scheme based on the selection of illumination wavelength.

[0021] Additional description of the inaccuracy situation is presented in the following literature by Bringholz, Barak, et al., “Accuracy in optical overlay metrology,” Proc. of SPIE, Vol. 9778, 9778H-1-19, March 24, 2016, which is incorporated herein by reference in its entirety.

[0022] Conventional SCOL and DBO techniques require four different targets (e.g., a metrology target with four different cells) to measure superposition in two directions (i.e., two cells are associated with each different direction). This increases the move-acquire-measure (MAM) time and the target area on the wafer.

[0023] Furthermore, the stacking accuracy of conventional SCOL and DBO techniques is significantly affected by variations in the optical system and aberrations. This makes accurate stacking measurements and sufficiently accurate tool matching difficult to achieve.

[0024] The increasing demands for smaller resolutions and larger wafer area in future overlay metrology applications pose challenges to metrology. Therefore, there is a need for improved methods and systems for overlay measurements. Summary of the Invention

[0025] This paper describes a method and system for robust superposition error measurement based on a trained measurement model. The measurement model is trained from raw scattering measurement data collected from a Design of Experiments (DOE) wafer by a superposition metrology system based on scattering measurements. Each measurement site includes one or more metrological targets manufactured with programmed superposition variations and known process variations. Each measurement site is measured with known metrological system variations. In this way, the measurement model is trained to decouple the actual superposition from the process variations and metrological system variations that affect the superposition measurement. Therefore, the actual superposition is estimated to be robust to process variations and metrological system variations by the trained measurement model.

[0026] By generating a measurement model using only raw scattering measurement data, the errors and approximations associated with traditional model-based metrology methods are reduced. Furthermore, the measurement model is insensitive to systematic errors, asymmetries, etc., because it is trained on scattering measurement data collected from a specific metrology system and used to perform measurements based on scattering measurement data collected from the same metrology system.

[0027] According to one embodiment of this disclosure, a scattering measurement-based metrology system is provided. The scattering measurement-based metrology system includes: an illumination source configured to provide illumination light to a plurality of superimposed metrology targets; a detector configured to detect an image of the amount of light scattered from each of the plurality of superimposed metrology targets in response to the illumination light; and a computing system. The plurality of superimposed metrology targets are manufactured with different known programmed superposition values ​​and different known values ​​of at least one manufacturing process variable. The image is detected by a plurality of pixels of the detector, and the detected optical signal associated with a measurement of each of the superimposed metrology targets includes a training data amount. The computing system is configured to estimate the actual superposition value associated with each of the plurality of superimposed metrology targets; and to train a measurement model based on the estimated value of the actual superposition and the training data amount. The estimation of the actual superposition value associated with each of the plurality of superimposed metrology targets is based on a plurality of measurements performed by the scattering measurement-based metrology system.

[0028] According to another embodiment of this disclosure, a scattering measurement-based metrology system is provided. The scattering measurement-based metrology system includes: an illumination source configured to provide illumination light at multiple different illumination wavelengths to each of a plurality of superimposed metrology targets; a detector configured to detect an image of the amount of light scattered from each superimposed metrology target at each different illumination wavelength onto a plurality of pixels of the detector; and a computing system. The scattered light amount includes +1 and -1 diffraction orders. The computing system is configured to estimate candidate values ​​for actual superposition based on the difference between the +1 diffraction order and the -1 diffraction order for each different illumination wavelength; determine a peak value for actual superposition from the candidate values ​​for actual superposition associated with the different illumination wavelengths; determine a wavelength difference relative to the wavelength associated with the peak value that minimizes superposition variation within the wafer, the determined wavelength difference being a distance from the peak value; and estimate a zero-error superposition (ZEO) value for actual superposition based on measurements of superposition at the distance from the peak value.

[0029] According to another embodiment of this disclosure, a metrology method is provided. The metrology method includes: illuminating each of a plurality of superimposed metrology targets with illumination light generated by an optical illumination source of a metrology system based on scattering measurements; collecting the amount of light scattered from each of the plurality of superimposed metrology targets in response to the illumination light; detecting an image of the amount of light collected from each of the plurality of superimposed metrology targets by a plurality of pixels of an optical detector of the metrology system based on scattering measurements, wherein the detection optical signal associated with a measurement of each of the superimposed metrology targets includes a training data amount; estimating an actual superposition value associated with each of the plurality of superimposed metrology targets, wherein the estimation of the actual superposition value associated with each of the plurality of superimposed metrology targets is based on a plurality of measurements performed by the metrology system based on scattering measurements; and training a measurement model based on the estimated value of the actual superposition and the training data amount. The plurality of superimposed metrology targets are manufactured with different known programmed superposition values ​​and different known values ​​of at least one manufacturing process variable.

[0030] In one aspect, a set of optimally programmed superposition variations is created, varying according to the location on one or more DOE wafers. In some instances, cross-wafer variations of one or more process parameters and cross-wafer variations of one or more structural parameters, varying according to the location on the DOE wafers (which induce variations in superposition measurements), are also created.

[0031] In another respect, DOE includes a range of different values ​​of one or more system parameter values ​​associated with (several) scattering measurement systems used to measure the DOE metrological target to generate training data.

[0032] In another approach, the actual superposition value associated with each manufacturing metrology target is estimated based on a series of measurements for that target, where the scattering measurement superposition system itself employs the Zero Error Superposition (ZEO) estimation method. The ZEO method is used to more accurately estimate the superposition from SCOL superposition measurements performed at several different illumination wavelengths.

[0033] In another embodiment, each measurement site comprises a single-cell metrological target having an array of grating structures that is periodic in at least one direction. Therefore, the scattering measurement of the single-cell metrological target is sensitive to superposition in at least one direction. In some embodiments, the single-cell metrological target comprises an array of grating structures that is periodic in at least two directions. Therefore, the measurement of the single-cell target is sensitive to superposition in both directions. Generally, the superimposed structures being measured can be located on the same or different layers of the substrate.

[0034] In another aspect, a single unit metrological target comprises an array of grating structures having two or more different periodicities in at least one direction. For example, the top grating and the bottom grating may be segmented by smaller pitch gratings or device-like structures.

[0035] In another aspect, the training measurement model is used as a measurement model for measuring other targets with unknown superposition values.

[0036] In another respect, the methods and systems described herein are not limited to measuring superposition errors. Generally, the scattering-based measurement techniques described herein can be applied to the measurement of other processes, structures, dispersion parameters, or any combination of these parameters. By way of non-limiting examples, profile geometry parameters (e.g., critical dimensions), process parameters (e.g., focal length and dosage), dispersion parameters, pitch wander, or any combination of parameters can be measured. A set of training objectives with programmed variations for each parameter of interest must be provided. The measurement model is then trained based on scattering measurement data collected at the measurement sites, which contains the range of programmed variations for each parameter of interest.

[0037] In another aspect, methods and systems for training measurement models include optimization algorithms to automate any or all components required to achieve the training of the measurement model.

[0038] In another aspect, scattering measurement data from multiple different targets are collected for model building, training, and measurement. Using measurement data associated with multiple targets with different structures but formed by the same process conditions and programmed superposition values ​​increases the information embedded in the model and reduces the superposition correlation with changes in process or other parameters.

[0039] In another aspect, measurement data derived from measurements performed at multiple values ​​of one or more measurement system parameters are collected for model building, training, and measurement. By way of non-limiting examples, measurements performed at multiple illumination wavelengths, polarizations, etc., are used to train the measurement model, and measurements are performed using the trained measurement model described herein.

[0040] In another aspect, measurement data derived from measurements performed using a combination of multiple different measurement techniques is collected for model building, training, and measurement. Using measurement data associated with multiple different measurement techniques increases the information content of the combined signal set and reduces the superimposed correlation with variations in process or other parameters.

[0041] In another approach, the results of the measurement model described herein can be used to provide active feedback to process tools (e.g., lithography tools, etching tools, deposition tools, etc.). For example, the value of the superposition error determined using the methods described herein can be transmitted to the lithography tool to adjust the lithography system to achieve the desired output. Similarly, etching parameters (e.g., etching time, diffusion rate, etc.) or deposition parameters (e.g., time, concentration, etc.) can be included in the measurement model to provide active feedback to the etching tool or deposition tool, respectively.

[0042] The foregoing is an overview and is therefore necessarily simplified, generalized, and omits details; thus, those skilled in the art will understand that the overview is for illustrative purposes only and is in no way limiting. Other aspects, inventive features, and advantages of the apparatus and / or process described herein will become apparent from the non-limiting detailed description presented herein. Attached Figure Description

[0043] Figures 1A to 1C This illustrates a pupil image of a measurement diffraction order associated with SCOL measurements performed at three different illumination wavelengths.

[0044] Figure 2 A plot depicting the inaccuracies of superimposed measurements via the SCOL system within the wavelength range.

[0045] Figure 3 A plot depicting several inaccuracy states, each associated with different values ​​of spectral profile asymmetry.

[0046] Figure 4 A plot depicting several inaccuracy states, each associated with different values ​​of symmetric process variation.

[0047] Figure 5 This describes a system 100 for measuring superposition errors according to the exemplary method presented herein.

[0048] Figure 6 Description suitable for the metering system of the present invention (e.g. Figure 5 The method for implementing the measurement system 100 described herein.

[0049] Figure 7 Description suitable for the metering system of the present invention (e.g. Figure 5 The method for implementing the measurement system 100 described herein.

[0050] Figure 8 A description of a single unit superimposed measurement target in one embodiment.

[0051] Figure 9 A description of a single-cell superimposed metering target in another embodiment.

[0052] Figure 10 A plot depicting several superimposed error measurements that vary according to wavelength.

[0053] Figure 11 The description includes a plot of the interpolated R-curve that varies with wavelength.

[0054] Figure 12 This section describes the plotting of several R-curves that vary depending on the wavelength.

[0055] Figure 13 Plot the superimposed estimation curves and R-curves associated with a series of measurements performed by the SCOL metrology system at different wavelengths.

[0056] Figure 14 A plot showing the relationship between each point and a set of points superimposed on measurements evaluated at different distances from the peak position within the wafer. Detailed Implementation

[0057] Detailed references will now be made to the background examples and some embodiments of the present invention, which are illustrated in the accompanying drawings.

[0058] This paper describes a method and system for robust superposition error measurement based on a trained measurement model. A measurement model is trained from raw scattering measurement data collected from a Design of Experiments (DOE) wafer by a superposition metrology system based on scattering measurements. Each measurement site contains one or more metrological targets manufactured with programmed superposition variations and known process variations. Each measurement site is measured by the scattering-based superposition metrology system against known metrological system variations (i.e., measurement system parameter values) to which it is sensitive. In this way, the measurement model is trained to decouple the actual superposition from the process variations and metrological system variations that affect the superposition measurement. Therefore, the actual superposition estimated by the trained measurement model is robust to process variations and metrological system variations. Consequently, the estimation of the actual superposition is consistent and accurate across the entire illumination wavelength range without requiring target design optimization.

[0059] The measurement model described in this paper reduces the errors and approximations associated with conventional model-based metrology methods by generating the model using only raw scattering measurement data. Furthermore, the measurement model is insensitive to systematic errors, asymmetries, etc., because it is trained on scattering measurement data collected from a specific metrology system and used to perform measurements based on scattering measurement data collected from the same metrology system.

[0060] Figure 5 This describes a system 100 used to measure superposition errors according to the exemplary method presented herein. For example... Figure 5As shown, system 100 can be used to perform superimposed scattering measurements (SCOL) of one or more structures of sample 107. In this respect, system 100 may include a scatterometer equipped with an illuminator 101 and an imaging detector 111. The illuminator 101 of system 100 is configured to generate and direct illumination of a selected wavelength or wavelength range (e.g., 100 nm to 2500 nm) onto structures disposed on the surface of sample 107. The imaging detector 111 is further configured to receive illumination scattered from the surface of sample 107. In some embodiments, illuminator 101 includes at least one laser. In some of these embodiments, the laser is capable of selecting the illumination wavelength based on a command signal received from computing system 130.

[0061] exist Figure 5In the depicted embodiments, system 100 includes a wavelength selection device 102, a beam shaping device 103, and a polarization control device 104 in the illumination path. Wavelength selection device 102 includes one or more optical elements (e.g., optical filters, etc.) configured to selectively allow a desired illumination wavelength or wavelength range to pass through and reject other illumination wavelengths. In some embodiments, wavelength selection device 102 is controlled by computing system 130. In these embodiments, computing system 130 is configured to transmit control commands indicating the desired wavelength or wavelength range to wavelength selection device 102. In response, wavelength selection device 102 selectively allows the desired wavelength or wavelength range to pass through. Beam shaping device 103 includes one or more optical elements configured to shape the illumination beam provided to the surface of sample 107. In some embodiments, beam shaping device 103 is controlled by computing system 130. In these embodiments, computing system 130 is configured to transmit control commands indicating the desired beam shape to beam shaping device 103. In response, beam shaping device 103 selectively reshapes the illumination beam to achieve the desired illumination beam shape provided to the surface of sample 107. In some embodiments, illumination light emitted from illuminator 101 is polarized by polarization control device 104 to generate a polarized illumination beam provided to sample 107. In some embodiments, polarization control device 104 is controlled by computing system 130. In these embodiments, computing system 130 is configured to transmit control commands indicating the desired polarization to polarization control device 104. In response, polarization control device 104 selectively polarizes the illumination light to achieve the desired polarization state. Illumination beam 112 is directed to beam splitter element 105, which directs illumination beam 112 toward objective lens 106 and onto the surface of sample 107. Radiation scattered by the measured metrological target 108 disposed on sample 107 is collected by objective lens 106 and passed through beam splitter element 105. In some embodiments, collected light 113 is analyzed by polarization analyzer device 109 to generate an analytical collected beam provided to imaging detector 111. In some embodiments, the polarization analyzer device 109 is controlled by a computing system 130. Collected light 113 is directed to a focusing optics 110, which images the collected light 113 onto an imaging detector 111. The imaging detector 111 is positioned in a pupil plane that aligns with the surface of the sample 107. The measurement pupil image 120 is transmitted to the computing system 130 for analysis of the measured target.

[0062] In another embodiment, system 100 may include one or more computing systems 130 for performing superimposed measurements based on a trained measurement model developed according to the methods described herein. The one or more computing systems 130 may be communicatively coupled to imaging detector 111. In one aspect, the one or more computing systems 130 are configured to receive measurement data 120 associated with measurements of a metrological target placed on sample 107.

[0063] It should be understood that the various steps described herein can be implemented by a single computer system 130 or, alternatively, multiple computer systems 130. Furthermore, different subsystems of system 100 (e.g., imaging detector 111) may include computer systems suitable for implementing at least a portion of the steps described herein. Therefore, the above description should not be construed as limiting the invention, but rather as illustrative only. Additionally, one or more computing systems 130 may be configured to perform any other(s) steps of any of the method embodiments described herein.

[0064] Additionally, computer system 130 may be communicatively coupled to imaging detector 111 in any manner known in the art. For example, one or more computing systems 130 may be coupled to a computing system associated with imaging detector 111. In another instance, imaging detector 111 may be directly controlled by a single computer system coupled to computer system 130.

[0065] The computer system 130 of the overlay metering system 100 can be configured to receive and / or acquire data or information from subsystems of the system (e.g., imaging detector 111 and the like) via a transmission medium that may include wired and / or wireless portions. In this way, the transmission medium can serve as a data link between the computer system 130 and other subsystems of the system 100.

[0066] The computer system 130 of the superimposed measurement system 100 can be configured to receive and / or acquire data or information (e.g., measurement results, modeling inputs, modeling results, etc.) from other systems via a transmission medium that may include wired and / or wireless components. In this manner, the transmission medium can act as a data link between the computer system 130 and other systems (e.g., on-board memory of the measurement system 100, external memory, reference measurement sources, or other external systems). For example, the computer system 130 can be configured to receive measurement data from a storage medium (i.e., memory 132 or external memory) via a data link. For example, measurement results obtained using the imaging detector 111 can be stored in a permanent or semi-permanent memory device (e.g., memory 132 or external memory). In this respect, measurement results can be imported from on-board memory or external memory systems. Furthermore, the computer system 130 can transmit data to other systems via the transmission medium. For example, the parameter model or superimposed parameters 121 determined by the computer system 130 can be transmitted and stored in external memory. In this respect, measurement results can be exported to another system.

[0067] The computing system 130 may include (but is not limited to) a personal computer system, a mainframe computer system, a workstation, a graphics computer, a parallel processor, or any other device known in the art. Generally, the term "computing system" may be broadly defined to cover any device having one or more processors that execute instructions from memory media.

[0068] The program instructions 134 for implementing methods such as those described herein can be transmitted via a transmission medium such as a wire, cable, or wireless transmission link. For example, as... Figure 5 As described, program instructions 134 stored in memory 132 are transferred to processor 131 via bus 133. Program instructions 134 are stored in computer-readable media (e.g., memory 132). Exemplary computer-readable media include read-only memory, random access memory, magnetic disk or optical disk or magnetic tape.

[0069] Figure 6 Description suitable for the metering system of the present invention (e.g. Figure 5 The method 200 is implemented by the metrology system 100 described herein. It should be understood that, in one aspect, the data processing block of method 200 may be implemented via a pre-programmed algorithm executed by one or more processors of the computing system 130 or any other general-purpose computing system. It should be understood herein that specific structural aspects of the metrology system 100 are not intended to be limiting, but should be interpreted only as illustrative.

[0070] In block 201, each of the plurality of superimposed metrological targets is illuminated with illumination light generated by an optical illumination source (e.g., illuminator 101) of a superimposed metrological system based on scattering measurements. The plurality of metrological targets are manufactured with different known programmed superposition values ​​and different known values ​​of at least one manufacturing process variable.

[0071] Generally, the metrological targets used for training are fabricated on one or more Design of Experiments (DOE) wafers. Each DOE wafer is exposed with a known programming overlay. The programming overlay can have different values ​​within a domain of a DOE wafer, between different domains of a DOE wafer, or between different DOE wafers. Typically, the range of programming overlay variations is designed to be the same as or greater than the expected variation of the overlay (i.e., the overlay process window).

[0072] In some embodiments, the superimposed variations are organized into a Design of Experiments (DOE) pattern on the surface of a semiconductor wafer (e.g., a DOE wafer). In this way, measurement sites query different locations on the wafer surface corresponding to different superimposed values.

[0073] In some embodiments, the programmed overlay variations are designed to minimize their correlation with any anticipated process variations or structural parameter variations. Additionally, the programmed overlay variations are designed to minimize their correlation with variations in other overlay structures (e.g., overlays in other directions, deeper overlays, etc.).

[0074] In one aspect, a set of optimal programming overlay variations and their locations on one or more DOE wafers are determined by, for example, a computing system 130. In one instance, the computing system 130 is configured to define a set of programming overlay values ​​spanning a desired range of overlay variations. For example, a set of programming overlay values ​​may include a range of programming overlay values ​​from -10 nm to 10 nm with a resolution of 1 nm in the x-direction (i.e., programming overlay X = {-10 nm, -9 nm, …, 0 nm … , 9 nm, 10 nm}) and a range of programming overlay values ​​from -10 nm to 10 nm with a resolution of 1 nm in the y-direction (i.e., programming overlay Y = {-10 nm, -9 nm, … , 0 nm … , 9 nm, 10 nm}).

[0075] In this example, the computing system 130 is further configured to determine cross-wafer variations of one or more process parameters and one or more structural parameters (which induce variations in superimposed measurements). In this way, the process variations and structural parameter variations are determined to vary based on their positions (e.g., {x, y} coordinates) on the DOE wafer.

[0076] The computing system 130 is further configured to assign a position to each of the programmed overlay values ​​(e.g., mapping the values ​​of {programmed overlay X, programmed overlay Y} to different {x, y} wafer coordinates). The assignment of positions is optimized to minimize the correlation between programming overlay variations and process variations.

[0077] The computing system 130 is further configured to transmit the programmed overlay values ​​and their assigned wafer coordinates as an overlay DOE (i.e., scanner recipe) to the photolithography tool. The photolithography tool then prints the programmed overlay values ​​onto the desired locations on the DOE wafer(s).

[0078] Generally, DOE wafer sets are manufactured with known values ​​for at least one manufacturing process variable. In some embodiments, one or more layers of a wafer including metrological targets are manufactured with programmed values ​​for lithographic focal length and dosage. However, the robustness of the training measurement model is improved by manufacturing DOE wafer sets with known values ​​for a larger number of manufacturing process variables. Ideally, DOE wafer sets are manufactured with known values ​​across a process window of all process and structural parameters. However, practical constraints on manufacturing and measurement times impose limitations on the number of process and structural variables that will be considered within the DOE set.

[0079] In one example, the DOE wafer set contains a wide range of programmed values ​​for the critical dimension (CD) of the bottom grating structure. The bottom grating structure is repeatedly fabricated with different lithographic dose values ​​during the photolithography step used to form the bottom grating pattern.

[0080] In another example, the DOE wafer set includes a range of programmed values ​​for the focal length, dose, superposition in the x-direction, and superposition in the y-direction of the grating stack structure, as well as a range of programmed values ​​for the focal length and dose of the previous grating layer.

[0081] In another example, a DOE wafer set includes a range of process parameter values ​​for each process step (e.g., deposition, etching, chemical mechanical polishing, etc.).

[0082] In another aspect, DOE encompasses a range of different values ​​for one or more system parameters associated with the scattering measurement system or the system used to measure the DOE metrological target to generate training data. Therefore, the trained measurement model is robust to variations in the measurement system (e.g., variations in the optical system). In one instance, the optical scattering measurement signal associated with the measurement of each metrological target comprises signals collected under multiple different system conditions. For example, the collected optical signals may include signals from multiple wavelengths and multiple polarizations.

[0083] In box 202, by (for example) Figure 5 The light-collecting device of the scattering measurement superposition measurement system 100 described herein collects the amount of light scattered from each of the multiple superposition measurement targets in response to illumination light.

[0084] In box 203, an image is generated by multiple pixels of an optical detector of a metrology system based on scattering measurements detecting the amount of light collected from each of a plurality of superimposed targets. The detection optical signals associated with the measurements of each of the metrology targets in the DOE wafer set measured at each different measurement system configuration include the amount of training data.

[0085] exist Figure 5 In the depicted embodiments, the optically sensitive region of the imaging detector 111 is located at or near the pupil image plane of the scattering-based metrology system 100. Therefore, the system 100 collects a pupil image of the light diffracted from each metrology target. In some embodiments, the system 100 collects a pupil image of the light diffracted at a diffraction order of +1 / -1 from each metrology target.

[0086] In box 204, the actual superposition value associated with each of the multiple superposition measurement targets is estimated.

[0087] In some instances, the actual superposition value associated with each measurement target is estimated as the known programmed superposition value associated with each measurement target.

[0088] In some other instances, the actual superposition value associated with each metrological target is estimated by measuring each target using a trusted reference metrology system. For example, any of the following can be used as a reference metrology system: critical-size scanning electron microscopy (CD-SEM), X-ray enhanced SEM, optical critical-size tools, transmission electron microscopy (TEM), and small-angle X-ray scattering (SAXS) tools. In one instance, the actual superposition value is estimated by performing SCOL or IBO measurements at multiple azimuth angles and subtracting tool-induced displacement (TIS) errors.

[0089] In another aspect, the actual superposition value associated with each manufacturing metrology target is estimated based on a series of measurements for each manufacturing metrology target, wherein the scattering measurement superposition measurement system itself employs the zero-error superposition (ZEO) estimation method described herein.

[0090] The ZEO method is used to more accurately estimate superposition from SCOL superposition measurements performed at several different illumination wavelengths. An objective lens is used to obtain the zero-error superposition position (i.e., a specific distance from the peak) of the reference SCOL superposition measurement curve. The distance between the zero-error superposition position and the peak of the SCOL superposition curve remains very close, regardless of significant variations in process and structural parameter values. As described above, SCOL superposition curves are typically scaled, shifted, or both depending on process and structural parameter values. Therefore, the shape of the estimated SCOL superposition curve remains largely unchanged across the wavelength (regardless of variations in process and structural parameter values), and the distance from the zero-error superposition position to the peak remains nearly constant. Specifically, asymmetric variations cause shape scaling, symmetric variations (e.g., process variations) cause shape shifts in the wavelength direction (i.e., the horizontal direction), and variations in programmed superposition cause shifts in the estimated superposition direction (i.e., the vertical axis). However, regardless of these variations, the distance from the zero-error superposition to the peak remains almost constant.

[0091] Figure 10 A plot 150 depicts several superimposed error measurements that vary according to wavelength. Each plotted line represents a measurement of a metrological target manufactured with different values ​​of one or more process variables and asymmetric structural variables. For example... Figure 10 As explained in the paper, the shape of the SCOL superposition estimation curve remains intact, but the curves are scaled and shifted relative to each other.

[0092] In one example, a scattering measurement superposition system is used to measure the optical response of each metrological target at several different illumination wavelengths. In this context, the scattering measurement superposition system illuminates each of the superimposed metrological targets with several different illumination wavelengths, collects the amount of light scattered from each of the multiple superimposed metrological targets in response to each different illumination wavelength at the +1 / -1 diffraction order, and detects an image of the amount of light collected from each superimposed metrological target at each different illumination wavelength.

[0093] The computing system (e.g., computing system 130) estimates the candidate values ​​for the actual superposition based on the difference between the +1 diffraction order and the -1 diffraction order for each different illumination wavelength.

[0094] In some embodiments, candidate values ​​are interpolated onto wavelengths to achieve smaller wavelength intervals. This provides smoother curves, more accurate peaks, and more accurate estimation of zero-error superposition locations. In some instances, linear, polynomial, or spline interpolation methods are used to achieve high smoothness and flexibility. In one instance, cubic spline interpolation is used.

[0095] In some embodiments, a wavelength-dependent superposition estimate is used for ZEO analysis. However, in some other embodiments, a wavelength-dependent R value is used for ZEO analysis. In some embodiments, an R curve is advantageous because it is generally smoother. The R curve is calculated using equation (1), where brackets (<>) denote the inner product over pixels in the mask, I o Let G represent the apodization function used in a charge-coupled device (CCD) camera as an imaging detector, and let G be the SCOL difference signal representing the measurement asymmetry.

[0096] (1)

[0097] Figure 11 The illustration includes a plot 160 showing the interpolated R-curve that varies according to wavelength. Similarly, Figure 12 The illustration includes a plot 170 containing several R-curves that vary according to wavelength. Each plotted line represents a measurement of a metrological target manufactured with different values ​​of one or more process variables. For example... Figure 12 As explained in the text, the shape of the R curve remains intact, but the curves are shifted relative to each other.

[0098] The computational system is further configured to determine the actual superimposed peak from the candidate values. Multiple ZEO points exist across the entire illumination wavelength range; however, not all ZEO points are easily detected. Peaks in the superimposed estimation curve, or R-curve, are more readily detected. Once a peak is detected, a ZEO location search is performed within the region of interest surrounding the selected peak.

[0099] Figure 13 A plot 180 depicts the superimposed estimation curve 182 and the R-curve 181 associated with a series of measurements performed by the SCOL metrology system at different wavelengths. (See also...) Figure 13 As explained, the superimposed estimated curve, or R-curve, typically contains two or three peaks, each located within a specific wavelength range. In some instances, the illumination wavelength range is subdivided into several non-overlapping intervals, each associated with only one peak. Within each interval, the maximum point of the curve and its associated illumination wavelength are identified. Figure 13 As depicted, the superimposed estimation curve 182 is almost symmetrical near the middle peak of the R curve. (As shown in the figure...) Figure 13 As explained in the paper, the superposition estimation curve 182 intersects the actual superposition value at almost the same distance D from the peak position (the middle peak of the R curve).

[0100] The computational system was further configured to determine the wavelength difference relative to the peak value associated with minimizing intra-wafer superposition variation. Figure 13 In the illustrated example, the calculation system identifies the distance D between the peak of the R curve 181 and the SCOL overlay estimate associated with ZEO.

[0101] SCOL superposition estimation of a specific measurement site measured at a specific wavelength i Described by equation (2), where

[0102] (2)

[0103] O0 is a programmed overlay value associated with the measurement site, w i It is a change within the chip, and e i This is the SCOL measurement error. It allows for the estimation of variations within the wafer as having zero mean and variance. The Gaussian distribution, i.e., w i ~N(0, Similarly, the measurement error of SCOL can be estimated as having zero mean and variance. The Gaussian distribution, i.e., e i ~N(0, Furthermore, it is assumed that the intra-wafer variation is uncorrelated with the SCOL measurement error. In one instance, the ZEO location where the error is minimized is determined by minimizing the root mean square error of the SCOL measurement error, as described by equation (3), where

[0104] (3)

[0105] E[] represents the expected value. When O0 is unavailable, the ZEO location where the error is minimized is determined by minimizing the root mean square error of the SCOL measurement, as described by equation (4).

[0106] (4)

[0107] because Since it is unknown and constant, minimizing equation (4) simplifies to Var[S] 2 The computer system is thus configured to find the optimal illumination wavelength (i.e., the illumination wavelength associated with the peak, separated by a distance D) that minimizes the intra-wafer variation in the SCOL superposition measurement. This leads to a large error because the slope around the identified peak is very steep. To mitigate this risk, the intra-wafer variance is evaluated and averaged at two symmetrical locations relative to the peak (i.e., at a distance + / -d from the peak), as illustrated by equation (5), where

[0108] (5)

[0109] f(d) is the variation of the average value at a symmetrical position located at a distance d from the peak. The solution to equation (5) is given by equation (6), where...

[0110] (6)

[0111] d * It is among them that makes Var[S 2 Minimize the estimated value of d. ZEO estimate O * It is described by equation (7).

[0112] (7)

[0113] In some instances, the computing system is configured to: calculate the intra-wafer variation over a set of discrete distances from the peak, select the distance where the intra-wafer variation is minimal, and determine the average SCOL measurement at the wavelength position at the optimal distance + / - from the peak as the ZEO value. Figure 14 A plot 190 illustrates a set of points associated with the intra-wafer variation superimposed on measurements evaluated according to equation (6) at different distances d from the peak position. (See also...) Figure 14 As described, the minimum variance is located at a distance of approximately 30 nanometers from the peak position. In this example, based on a distance d of approximately 30 nanometers... * The ZEO value is estimated using equation (7) at the location.

[0114] Although equations (5 to 7) evaluate the ZEO value based on the average of two points symmetrically located about the peak position, in some other instances, the ZEO value can be determined based on two points located at different distances (e.g., -d1 and +d2) from the peak position. More generally, the SCOL measurement range within the distance range r from the peak can be used to find the ZEO value. In this example, the superposition estimation is described by equations (8) and (9), where c i These are weighting coefficients.

[0115] (8)

[0116] in

[0117] (9)

[0118] In this scheme, the optimization problem can be expressed by equation (10), where c i The value is analyzed to minimize S(d) i The variance of the sum of weighted values ​​of ) is the ZEO value. The ZEO value can be expressed by equation (11).

[0119] (10)

[0120] (11)

[0121] In some instances, the SCOL pattern is flatter around the peak. In these instances, a thorough search along a single direction from the peak can be performed to find the ZEO location. Additionally, if the SCOL patterns associated with each region of interest exhibit very similar patterns, then the ZEO value associated with each peak can be appropriately calculated, and the results can then be averaged to obtain a final estimate of the ZEO value.

[0122] In box 205, the measurement model is trained based on the estimated values ​​of the actual overlays and the amount of training data. The measurement model is structured to receive measurement data generated by the metrology system at one or more measurement sites and directly determine the overlay associated with each measurement target.

[0123] In this regard, the training measurement model establishes a functional relationship between the optical scattering measurement signal and the superimposed value. In some embodiments, the relationship is described by a complex function, wherein the function coefficients are estimated based on the training signal (i.e., the collected optical scattering measurement signal) and the estimated value of the actual superposition.

[0124] In some instances, the measurement model is trained based on key features extracted from the scattering measurement training data and estimates of the actual superposition. In these instances, multiple key features of the training data are determined based on a transformation of the amount of training data that reduces its size. Several key features are extracted from the scattering measurement training data based on mathematical transformations. The transformation reduces the size of the scattering measurement data and maps the original signal to a new set of reduced signals. The transformation is determined based on changes in the superposition of the scattering measurement training data. Each measurement signal is considered as a variation of the original signal across different superposition measurements within the set of scattering measurement training data. The transformation can be applied to all measurement signals or a subset of measurement signals. In some instances, signals undergoing analysis are randomly selected. In other instances, signals undergoing analysis are selected due to their relatively high sensitivity to changes in superposition. For example, signals insensitive to changes in superposition can be ignored. By way of non-restriction, the transformation can be implemented using any of the following: principal component analysis (PCA) models, kernel PCA models, nonlinear PCA models, independent component analysis (ICA) models or other dimensionality reduction methods using a dictionary, discrete cosine transform (DCT) models, fast Fourier transform (FFT) models, wavelet models, etc.

[0125] In some embodiments, the measurement model is implemented as a neural network model. In one instance, the number of nodes in the neural network is selected based on several features extracted from the training data. In other instances, the measurement model may be implemented as a linear model, a multinomial model, a response surface model, a support vector machine model, a decision tree model, a random forest model, or other types of models. In some instances, the measurement model may be implemented as a combination of models. In some instances, the selected model is trained based on key features (reduced signal set) and known superimposed variations. The model is trained such that its output fits the superimposed variation of all measurement signals in the superimposed variation space defined by the DOE measurement target set.

[0126] Various metrological targets can be considered within the scope of this invention. In some embodiments, the metrological target is based on a conventional linear / spatial target. In some other embodiments, the metrological target is a device-like structure. In some other embodiments, the metrological target is the actual device itself, and therefore no specific metrological target is employed. Regardless of the type of metrological target employed, a set of training targets with known programming offsets must be provided to train the measurement model. Once the model has been trained, it can be used to perform measurements on structures with unknown superpositions.

[0127] Training targets can be provided on a separate training wafer or a production wafer. In some instances, metrology targets are located within the dicing lines of the production wafer. In other instances, metrology targets are located in the active die region.

[0128] In some embodiments, measurements for model training are performed in the cutting line region and subsequent measurements are performed in the periodic region of the actual device.

[0129] In some embodiments, multiple different target offsets in orthogonal directions are used in each die. This can advantageously minimize the impact of the underlying layer on measurement accuracy.

[0130] In one aspect, each measurement site comprises a single-cell metrology target having an array of grating structures periodically arranged in at least one direction. Therefore, the scattering measurement of the single-cell metrology target is sensitive to superposition in at least one direction. In some embodiments, the single-cell metrology target comprises an array of grating structures periodically arranged in at least two directions. Therefore, the measurement of the single-cell target is sensitive to superposition in both directions. Generally, the measurement superposition structure can be located on the same layer or different layers of the substrate.

[0131] Figure 8A description of a single-cell superimposed metrology target 140 in one embodiment is provided. The metrology target 140 comprises a top grating 142 and at least one of a bottom grating 141. For different superposition values, the +1 and -1 diffraction orders captured on the pupil image by the imaging detector 111 are sensitive to superposition in both directions (e.g., ΔX in the x-direction and ΔY in the y-direction). In one embodiment, the nominal offset between gratings 141 and 142 is approximately 150 nanometers in both the x and y directions. As described above, a typical target is an SCOL or DBO target with two or more cells per superposition direction. Figure 8 As depicted, measurement target 140 is a single-cell target with sensitivity in two orthogonal directions. A significant reduction in move-acquisition-measurement (MAM) time is achieved by collecting signals from one cell instead of four.

[0132] In another aspect, a single unit metering target comprises an array of grating structures having two or more different periodicities in at least one direction. For example, the gratings at the top layer 142 and the bottom layer 141 may be segmented by smaller (e.g., design rules) pitch gratings or device-like structures.

[0133] Figure 9 A description of a single-cell superimposed metrology target 145 in one embodiment is provided. The metrology target 145 includes a grating in a top layer 142 and a grating in a bottom layer 141. For different superposition values, the +1 and -1 diffraction orders captured by the imaging detector 111 on the pupil image are sensitive to superposition in both directions (e.g., ΔX in the x-direction and ΔY in the y-direction). Furthermore, the metrology target 145 includes multi-pitch patterns in both directions. The multi-pitch patterns provide additional patterning on the pupil image and increase the pupil image area sensitive to superposition variations. Figure 9 As depicted, the multi-pitch pattern comprises a periodicity P1 in two directions and another periodicity P2 in two directions. In one embodiment, P1 is approximately 600 nanometers and P2 is approximately 750 nanometers.

[0134] As described above, the measurement methods and systems described herein are not constrained by any specific objective. In general, the methods and systems described herein can be used for any objective that exhibits sensitivity to superposition when measured by an available measurement system.

[0135] In another aspect, the trained model is used as a measurement model for measuring other targets with unknown superposition values. Figure 7 Description suitable for the metering system of the present invention (e.g. Figure 5Method 210 is implemented by the metrology system 100 described herein. It should be understood that, in one aspect, the data processing block of method 210 may be implemented via a pre-programmed algorithm executed by one or more processors of the computing system 130 or any other general-purpose computing system. It should be understood herein that specific structural aspects of the metrology system 100 are not intended to be limiting, but should be interpreted only as illustrative.

[0136] In block 211, at least one superimposed metrology target is illuminated using illumination light generated by an optical illumination source (e.g., illuminator 101) of a superimposed metrology system based on scattering measurements (which generates training data for training a measurement model). The metrology targets have unknown superposition errors. Typically, the metrology targets are mounted on a wafer different from the wafers used to generate the training data. Measurements are typically performed in the same or similar process steps as the training wafers. Additionally, the metrology targets mounted on the wafer of interest are typically the same or similar to the metrology targets mounted on the training wafers. However, in some embodiments, the metrology targets measured for model training and the metrology targets measured for superposition estimation are mounted on the same wafer.

[0137] In block 212, an imaging detector 111 collects the amount of light scattered from at least one superimposed metering target in response to illumination light.

[0138] In box 213, an image of the amount of light collected from at least one superimposed target is detected by multiple pixels of an optical detector of a metrology system based on scattering measurements.

[0139] In box 214, the actual superposition value associated with at least one superposition target is estimated based on the detected image of at least one superposition target and a trained measurement model. In one example, the computation system 130 uses a functional relationship established through the trained measurement model to estimate the actual superposition value based on the measured scattering signal.

[0140] In block 215, the estimated superposition is stored in memory. For example, the superposition value may be stored on the board of the measurement system 100 (e.g., stored in memory 132) or may be transmitted to an external memory device (e.g., via output signal 121).

[0141] In some embodiments, the computing system 130 is configured to determine multiple key features of a detected image of at least one superimposed target based on a transformation that reduces the image size. In these embodiments, the estimation of the actual superimposed value associated with at least one superimposed target is based on multiple key features and a trained measurement model. In some embodiments, the transformation is the same transformation used to reduce the size of the corresponding training data described in reference method 200. Preferably, the same analysis used in method 200 to extract features from the scattering measurement data is used to extract features. In this way, the size reduction of the acquired data is performed by the same transformation used to reduce the size of the training data.

[0142] In another respect, the methods and systems described herein are not limited to measuring superposition errors. Generally, the aforementioned measurement techniques based on scattering measurements can be applied to the measurement of other process, structural, dispersion parameters, or any combination of these parameters. By way of non-limiting examples, the aforementioned techniques can be used to measure profile geometry parameters (e.g., critical dimensions), process parameters (e.g., focal length and dosage), dispersion parameters, pitch wander, or any combination of parameters, as well as superposition errors. A set of training objectives with programmed variations for each parameter of interest must be provided. The measurement model is then trained based on scattering measurement data collected at the measurement sites (which includes the range of programmed variations for each parameter of interest), as described herein with reference to superposition. By way of non-limiting examples, the parameters of interest can be process parameters, structural parameters, dispersion parameters, and layout parameters.

[0143] In some instances, different known values ​​of at least one additional parameter of interest are used as the measurement targets for model training. Therefore, the training of the measurement model is also based on different known values ​​of at least one additional parameter of interest and the amount of training data. In some embodiments, reference metrics such as CD-SEM, XSEM, OCD, etc., are used to obtain reference values ​​associated with each parameter of interest.

[0144] In this manner, the measurement model trained as described herein can be used to estimate superimposed values ​​and additional parameters such as critical dimensions, edge placement errors, lithographic focal length, lithographic dose, and other shape and film parameters of interest. In some embodiments, the superimposed values ​​and estimates of the parameters of interest can be generated based on measurement data collected from a single unit target, device structure, or any other target or group of targets that exhibit measurement sensitivity to the parameters of interest. In this manner, estimates of multiple parameters of interest can be obtained simultaneously based on measurements of one or more units.

[0145] In another aspect, methods and systems for training measurement models include optimization algorithms to automate any or all components required to achieve the training of the measurement model.

[0146] In some instances, the optimization algorithm is configured to maximize the performance of the measurement (defined by the cost function) by optimizing any or all of the following parameters: the type of feature extraction model (i.e., the transformation), the parameters of the selected feature extraction model, the type of measurement model, and the parameters of the selected measurement model. The optimization algorithm may incorporate user-defined heuristics and may be a combination of nested optimizations (e.g., combinatorial and sequential optimization).

[0147] In another approach, scattering measurement data from multiple different targets are collected for model building, training, and measurement. Using measurement data associated with multiple targets of different structures but formed by the same process conditions and programmed superposition values ​​increases the information embedded in the model and reduces superposition dependence on process or other parameter variations. The additional information embedded in the model allows the superposition-related information to be decoupled from information associated with other parameters (e.g., film thickness, CD, etc.) that can similarly affect the measurement signal. In these instances, training data using images containing multiple different targets at one or more measurement sites achieves more accurate superposition estimation. In some instances, a mixture of isolated and dense line / spatial targets is used to decouple superposition from underlying effects.

[0148] In another instance, signals from multiple targets can be processed to reduce sensitivity to process variations and increase sensitivity to parameters of interest. In some instances, signals from different targets are subtracted from each other. In others, signals from different targets are fitted to a model, and residual signals are used to construct, train, and apply the measurement model described herein. In one instance, signals from two different targets are subtracted to eliminate or significantly reduce the effects of process noise in each measurement. Generally, various mathematical calculations can be applied between signals from different targets to determine signals that reduce sensitivity to process variations and increase sensitivity to parameters of interest.

[0149] In another aspect, measurement data derived from measurements performed at multiple values ​​of one or more measurement system parameters are collected for model building, training, and measurement. By way of non-limiting examples, measurements performed at multiple illumination wavelengths, polarizations, etc., are used to train the measurement model, and the trained measurement model is used to perform measurements, as described herein.

[0150] In another aspect, measurement data derived from measurements performed using a combination of multiple different measurement techniques is collected for model building, training, and measurement. Using measurement data associated with multiple different measurement techniques increases the information content of the combined signal set and reduces the superimposed correlation with variations in process or other parameters. Measurement data can be derived from measurements performed using any combination of multiple different measurement techniques. In this way, different measurement sites can be measured by multiple different measurement techniques (e.g., scattering measurements, imaging, and other signal information sources) to increase the measurement information that can be used to estimate parameters of interest.

[0151] Generally, since the data processed by the feature extraction model and the measurement model used for training and measurement is in vector form, any measurement technique or a combination of two or more measurement techniques can be considered within the scope of this invention. Because the techniques described herein perform vector computation of the data, each collected signal is processed independently. Furthermore, data from multiple different measurements can be cascaded, regardless of whether the data is two-dimensional, one-dimensional, or even single-point data.

[0152] Exemplary measurement techniques that can provide data for analysis according to the metrological techniques described herein include (but are not limited to) imaging reflectometers, imaging spectroreflectometers, polarization spectroreflectometers, scanning reflectometer systems, systems with two or more reflectometers capable of parallel data acquisition, systems with two or more spectroreflectometers capable of parallel data acquisition, systems with two or more polarization spectroreflectometers capable of parallel data acquisition, systems with two or more polarization spectroreflectometers capable of serial data acquisition without moving the wafer stage or any optical elements or reflectometer stage, imaging spectrometers, and imaging systems with wavelength filters. Imaging systems include systems with long-pass wavelength filters, imaging systems with short-pass wavelength filters, imaging systems without wavelength filters, interferometric imaging systems, imaging ellipsometers, imaging spectral ellipsometers, scanning ellipsometer systems, systems with two or more ellipsometers capable of parallel data acquisition, systems with two or more ellipsometers capable of serial data acquisition without moving the wafer stage or any optical components or ellipsometer stage, Michelson interferometers, Mach-Zehnder interferometers, Sagnac interferometers, scanning incident angle systems, and scanning azimuth angle systems. Furthermore, generally, measurement data collected by different measurement techniques and analyzed according to the methods described herein can be collected by multiple tools rather than by a single tool integrating multiple technologies.

[0153] In another aspect, signals from multiple metrologies can be processed to reduce sensitivity to process variations and increase sensitivity to parameters of interest. In some instances, signals from targets derived from different metrologies are subtracted from each other. In other instances, signals from targets derived from different metrologies are fitted to a model, and residual signals are used to construct, train, and apply the measurement model described herein. In one instance, signals from targets derived from two different metrologies are subtracted to eliminate or significantly reduce the effects of process noise in each measurement. Generally, various mathematical calculations can be applied between signals from different metrologies to determine signals that reduce sensitivity to process variations and increase sensitivity to parameters of interest.

[0154] Generally speaking, signals from multiple targets measured by various metrology techniques increase the information content of the combined signal group and reduce the superposition correlation with changes in process or structural parameters.

[0155] In some instances, the model building, training, and measurement methods described herein were implemented using elements of the SpectraShape® optical critical size metrology system purchased from KLA-Tencor in Milpitas, California. This method created models ready for immediate use after the system collected DOE wafer spectra.

[0156] In some other instances, the model building and training methods described herein are implemented offline by a computing system, for example, implementing AcuShape® software purchased from KLA-Tencor in Milpitas, California. The resulting trained model can be incorporated as an element of an AcuShape® library accessible by a metrology system performing the measurements.

[0157] In another aspect, the metrology system (e.g., metrology system 100) used to perform the superposition measurements described herein includes an infrared optical measurement system. In these embodiments, metrology system 100 includes an infrared light source (e.g., an arc lamp, an electrodeless lamp, a laser continuous plasma (LSP) source, or a supercontinuum source). Infrared supercontinuum laser sources are superior to conventional lamp sources due to their higher achievable power and brightness in the infrared region of the spectrum. In some instances, the power provided by a supercontinuum laser is sufficient to measure superposition structures with opaque film layers.

[0158] A potential problem with superimposed measurements is the lack of sufficient light penetration to the bottom grating. In many instances, a non-transparent (e.g., opaque) film layer exists between the top and bottom gratings. Examples of such opaque films include amorphous carbon and tungsten silicide (WSi). xThe layers consist of tungsten, titanium nitride, amorphous silicon, and other metallic and non-metallic layers. Illumination light limited to wavelengths in the visible range and below (e.g., between 250 nm and 700 nm) typically cannot penetrate to the bottom grating. However, illumination light in the infrared spectrum and above (e.g., greater than 700 nm) typically penetrates the opaque layers more effectively.

[0159] An effective target design or measurement structure propagates a non-zero diffraction order between the first and second patterns, such that the relative positions of the two patterns affect the intensity of the outgoing diffracted beam detected in the far field.

[0160] The operational definition of "opaque" in the ultraviolet and visible range is that the prediction accuracy of SCOL in the wavelength range of 250 nm to 700 nm is much worse than the required accuracy. This is attributed to the attenuation of the propagation diffraction order carrying the relative positional information between the first and second patterns. When absorption is significantly less, measuring the SCOL signal with an illumination wavelength greater than 700 nm (e.g., 800 nm to 1650 nm) improves SCOL accuracy. In embodiments where illumination light with a wavelength greater than 700 nm is used, the design pitch of the measurement target is selected such that a usable SCOL signal exists.

[0161] In some embodiments described herein, training data is generated by measuring the pixels of pupil images of multiple metrological targets measured by an SCOL metrology system. However, in general, training data can also be generated by measuring the pixels of field images of multiple metrological targets measured by a diffraction-based superposition (DBO) metrology system, signals generated by an imaging IBO metrology system, signals generated by an X-ray metrology system, signals generated by a spectroscopic ellipsometry (SE) system, and so on.

[0162] In another aspect, the results of the measurement model described herein can be used to provide active feedback to process tools (e.g., lithography tools, etching tools, deposition tools, etc.). For example, the value of the superposition error determined using the methods described herein can be transmitted to the lithography tool to adjust the lithography system to achieve the desired output. Similarly, etching parameters (e.g., etching time, diffusion rate, etc.) or deposition parameters (e.g., time, concentration, etc.) can be included in the measurement model to provide active feedback to the etching tool or deposition tool, respectively.

[0163] Generally, the systems and methods described herein can be implemented as part of a process that prepares a measurement model for offline or on-device measurements. Additionally, two measurement models and any reparameterized measurement models can describe one or more target structures and measurement sites.

[0164] As described herein, the term "critical dimension" includes any critical dimension of a structure (e.g., bottom critical dimension, middle critical dimension, top critical dimension, sidewall angle, grating height, etc.), any critical dimension between two or more structures (e.g., the distance between two structures), and displacement between two or more structures (e.g., the superposition displacement between superimposed grating structures, etc.). Structures may include three-dimensional structures, patterned structures, superimposed structures, etc.

[0165] As described herein, the terms “critical size application” or “critical size measurement application” encompass any critical size measurement.

[0166] As described herein, the term "metrology system" includes at least part of any system used to characterize samples in any aspect, encompassing measurement applications such as critical size metrology, superposition metrology, focal length / dosage metrology, and composition metrology. However, such technical terms do not limit the scope of the term "metrology system" as described herein. Additionally, metrology system 100 may be configured for measuring patterned wafers and / or unpatterned wafers. The metrology system may be configured as an LED inspection tool, edge inspection tool, backside inspection tool, macroscopic inspection tool, or multi-mode inspection tool (involving data from one or more platforms simultaneously), and any other metrology or inspection tool that benefits from calibrating system parameters based on critical size data.

[0167] This document describes various embodiments of semiconductor processing systems (e.g., inspection systems or lithography systems) that can be used to process samples. The term "sample" is used herein to refer to a wafer, a photomask, or any other sample that can be processed by components known in the art (e.g., for printing or inspecting defects).

[0168] As used herein, the term "wafer" generally refers to a substrate formed of semiconductor or non-semiconductor materials. Examples include (but are not limited to) single-crystal silicon, gallium arsenide, and indium phosphide. Such substrates are typically found and / or processed in semiconductor manufacturing plants. In some cases, a wafer may consist only of a substrate (i.e., a bare wafer). Alternatively, a wafer may contain one or more layers of different materials formed on the substrate. The one or more layers formed on the wafer may be "patterned" or "unpatterned." For example, a wafer may contain multiple bare wafers with repeatable patterned features.

[0169] A "photomask" can be a photomask at any stage of the photomask assembly process, or a finished photomask that may or may not be released for use in a semiconductor manufacturing plant. A photomask or "mask" is generally defined as a substantially transmissive substrate having substantially opaque areas formed thereon and arranged in a pattern. The substrate may comprise, for example, a glass material, such as amorphous SiO2. A photomask can be placed over a wafer coated with resist during the exposure step of a photolithography process, such that the pattern on the photomask can be transferred to the resist.

[0170] One or more layers formed on a wafer may be patterned or unpatterned. For example, a wafer may comprise multiple dies, each having repeatable pattern features. The formation and processing of such material layers can ultimately result in a completed device. Many different types of devices can be formed on a wafer, and the term wafer, as used herein, is intended to encompass wafers on which any type of device known in the art is manufactured.

[0171] In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functionality may be stored as one or more instructions or code on or transmitted via a computer-readable medium. A computer-readable medium includes both computer storage media and communication media, including any media that facilitates the transfer of a computer program from one location to another. Storage media may be any available media accessible by a general-purpose computer or a special-purpose computer. For example (and without limitation), this computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage or other magnetic storage devices, or any other media that can be used to carry or store desired program code elements in the form of instructions or data structures and is accessible by a general-purpose computer or a special-purpose computer or a general-purpose or special-purpose processor. Furthermore, any connection may be appropriately referred to as computer-readable media. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technology (such as infrared, radio, and microwave), then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technology (such as infrared, radio, and microwave) is included in the definition of media. As used herein, disks and optical discs include optical discs (CDs), laser discs, optical discs, digital versatile discs (DVDs), floppy disks, and Blu-ray discs, wherein disks typically magnetically reproduce data while optical discs optically reproduce data using lasers. The above combinations should also be included within the scope of computer-readable media.

[0172] Although certain specific embodiments have been described above for guiding purposes, the teachings of this patent document are generally applicable and not limited to the specific embodiments described above. Therefore, various modifications, adaptations, and combinations of various features of the described embodiments can be practiced without departing from the scope of the invention as set forth in the claims.

Claims

1. A metrological system based on scattering measurement, comprising: A lighting source configured to provide illumination light to a plurality of superimposed metering targets, wherein the plurality of superimposed metering targets are manufactured with different known programmed superimposed values ​​and different known values ​​of at least one manufacturing process variable; A detector configured to detect an image of the amount of light scattered from each of the plurality of superimposed metrological targets in response to the amount of illumination light, the image being detected by a plurality of pixels of the detector, the detection optical signal associated with the measurement of each of the superimposed metrological targets including an amount of training data; and The computing system is configured to: Estimate the actual superposition value associated with each of the plurality of superimposed metrological targets, wherein the estimate of the actual superposition value associated with each of the plurality of superimposed metrological targets is based on a plurality of measurements performed by the scattering-based metrological system; and The measurement model is trained based on the estimated values ​​of the actual superposition and the amount of training data. The estimation of the actual superimposed value involves: Each of the plurality of superimposed measurement targets is illuminated with a plurality of different illumination wavelengths; The amount of light scattered from each of the plurality of superimposed metering targets is collected in response to each different illumination wavelength; wherein the amount of light includes +1 and -1 diffraction orders; An image of the amount of light collected from each superimposed metrology target at each different illumination wavelength onto multiple pixels of the optical detector of the metrology system based on scattering measurement; The candidate values ​​for the actual superposition are estimated based on the difference between the +1 diffraction order and the -1 diffraction order for each different illumination wavelength; The peak value of the actual superposition is determined from the candidate values ​​of the actual superposition associated with the different illumination wavelengths; Determine the wavelength difference relative to the wavelength associated with the peak value that minimizes the superposition variation within the wafer; the determined wavelength difference is the distance from the peak value; and The zero-error superposition ZEO value of the actual superposition is estimated based on the superposition measurement at the distance from the peak.

2. The metrology system based on scattering measurement according to claim 1, wherein the detector is located at or near the pupil image plane of the metrology system based on scattering measurement.

3. The metrology system based on scattering measurement according to claim 1, wherein the calculation system is further configured to: Multiple key features of the training data are determined based on a transformation of the training data volume by reducing the size of the training data, wherein the training of the measurement model is based on the multiple key features extracted from the training data volume and the estimated values ​​that are actually superimposed.

4. The metrology system based on scattering measurement according to claim 3, wherein the transformation of the training data involves any of the following models: principal component analysis (PCA), independent component analysis (ICA), kernel PCA, nonlinear PCA, fast Fourier transform (FFT), discrete cosine transform (DCT), and wavelet model.

5. The metrology system based on scattering measurement according to claim 1, wherein the measurement model is any of the following: linear model, polynomial model, neural network model, support vector machine model, decision tree model, and random forest model.

6. The metrology system based on scattering measurement according to claim 1, wherein the training data includes measurement data associated with the plurality of superimposed metrology targets, formed by the same process conditions and programmed superposition values.

7. The metrology system based on scattering measurements according to claim 1, wherein the amount of training data includes scattering measurements obtained by multiple different metrology techniques.

8. The metrology system based on scattering measurement according to claim 1, wherein the plurality of superimposed metrology targets are further generated with different known values ​​of at least one additional parameter of interest, and wherein the training of the measurement model is further based on the different known values ​​of the at least one additional parameter of interest and the amount of training data.

9. The metrology system based on scattering measurement according to claim 8, wherein the at least one additional parameter of interest includes any of process parameters, structural parameters, dispersion parameters, and layout parameters.

10. The metrology system based on scattering measurement according to claim 1, wherein the illumination source is further configured to illuminate at least one superimposed metrology target having an unknown superposition error, wherein the detector is further configured to detect an image of the amount of light scattered from the at least one superimposed metrology target, and wherein the computing system is further configured to estimate the actual superposition value associated with the at least one superimposed metrology target based on the detected image of the at least one superimposed metrology target and the trained measurement model.

11. The metrology system based on scattering measurement according to claim 10, wherein the computing system is further configured to determine a plurality of principal features of the image of the at least one superimposed metrology target based on a transformation that reduces the size of the image, wherein the estimate of the actual superimposed value associated with the at least one superimposed metrology target is based on the plurality of principal features and the trained measurement model.

12. The metrology system based on scattering measurement according to claim 10, wherein the values ​​of the photolithography process variables are adjusted based on the estimated values ​​that are actually superimposed.

13. The metrology system based on scattering measurement according to claim 1, wherein the plurality of superimposed metrological targets are disposed on a plurality of semiconductor wafers, wherein each of the plurality of semiconductor wafers is manufactured with different values ​​of at least one semiconductor manufacturing process variable.

14. The metrology system based on scattering measurement according to claim 1, wherein the plurality of superimposed metrology targets are disposed on a plurality of domains of one or more semiconductor wafers.

15. The scattering-based metrology system of claim 1, wherein the image is provided at multiple different values ​​of one or more measurement system parameters, and the amount of light scattered from each of the multiple superimposed metrology targets is detected.

16. The metrology system based on scattering measurement according to claim 1, wherein at least one of the superimposed metrology targets is a single-cell metrology target having a grating structure array with periodicity in at least one direction.

17. The metrology system based on scattering measurement according to claim 1, wherein at least one of the superimposed metrology targets is a single-cell metrology target having a grating structure array with periodicity in at least two directions.

18. The metrology system based on scattering measurement according to claim 1, wherein at least one of the superimposed metrology targets is a single-unit metrology target having two or more grating structure arrays with different periodicities in at least one direction.

19. The metrology system based on scattering measurement according to claim 1, wherein the estimate of the actual superposition of the values ​​associated with each of the plurality of superposition metrology targets is based on the programmed superposition value.

20. The metrology system based on scattering measurements according to claim 1, wherein the estimate of the actual superimposed value associated with each of the plurality of superimposed metrological targets is based on the measurement of each of the plurality of superimposed metrological targets by a reference metrology system.

21. The metrology system based on scattering measurement according to claim 20, wherein the reference metrology system is a scanning electron microscope (SEM).

22. A metrological system based on scattering measurement, comprising: A lighting source configured to provide illumination light at multiple different illumination wavelengths to each of multiple superimposed metering targets; A detector configured to detect an image of the amount of light scattered onto a plurality of pixels of the detector at each different illumination wavelength from each superimposed metering target, wherein the amount of scattered light includes +1 and -1 diffraction orders; and The computing system is configured to: The candidate values ​​for the actual superposition are estimated based on the difference between the +1 diffraction order and the -1 diffraction order for each different illumination wavelength; The peak value of the actual superposition is determined from the candidate values ​​of the actual superposition associated with the different illumination wavelengths; The wavelength difference is determined relative to the wavelength associated with the peak value that minimizes the superposition variation within the wafer; the determined wavelength difference is the distance from the peak value. and The zero-error superposition ZEO value of the actual superposition is estimated based on the superposition measurement at the distance from the peak.

23. A measurement method, comprising: Each of a plurality of superimposed metrological targets is illuminated with illumination light generated by an optical illumination source of a metrological system based on scattering measurements according to any one of claims 1-23, wherein the plurality of superimposed metrological targets are manufactured with different known programmed superposition values ​​and different known values ​​of at least one manufacturing process variable; In response to the illumination light, the amount of light scattered from each of the plurality of superimposed metering targets is collected; An image of the amount of light collected from each of a plurality of superimposed metrological targets by multiple pixels of an optical detector of a metrology system based on scattering measurements, wherein the detection optical signal associated with the measurement of each of the superimposed metrological targets includes the amount of training data; Estimate the actual superposition value associated with each of the plurality of superimposed metrological targets, wherein the estimate of the actual superposition value associated with each of the plurality of superimposed metrological targets is based on a plurality of measurements performed by the scattering-based metrological system; and The measurement model is trained based on the estimated values ​​of the actual superposition and the amount of training data. The estimation of the actual superimposed value involves: Each of the plurality of superimposed measurement targets is illuminated with a plurality of different illumination wavelengths; The amount of light scattered from each of the plurality of superimposed metering targets is collected in response to each different illumination wavelength; wherein the amount of light includes +1 and -1 diffraction orders; An image of the amount of light collected from each superimposed metrology target at each different illumination wavelength onto multiple pixels of the optical detector of the metrology system based on scattering measurement; The candidate values ​​for the actual superposition are estimated based on the difference between the +1 diffraction order and the -1 diffraction order for each different illumination wavelength; The peak value of the actual superposition is determined from the candidate values ​​of the actual superposition associated with the different illumination wavelengths; Determine the wavelength difference relative to the wavelength associated with the peak value that minimizes the superposition variation within the wafer; the determined wavelength difference is the distance from the peak value; and The zero-error superposition ZEO value of the actual superposition is estimated based on the superposition measurement at the distance from the peak.

24. The method of claim 23, wherein the detector is located at or near the pupil image plane of the scattering measurement-based metrology system.

25. The method of claim 23, wherein the plurality of superimposed measurement targets are also generated with different known values ​​of at least one additional concern parameter, and wherein the training of the measurement model is further based on the different known values ​​of the at least one additional concern parameter and the amount of training data.

26. The method of claim 23, further comprising: At least one superimposed measurement target is illuminated using illumination light generated by the optical illumination source of the scattering-based measurement system, the at least one superimposed measurement target having an unknown superposition error; In response to the illumination light, the amount of light scattered from the at least one superimposed metering target is collected; An image of the amount of light collected from the at least one superimposed metering target, detected by multiple pixels of the optical detector of the scattering measurement-based metrology system; The actual superposition value associated with the at least one superposition measurement target is estimated based on the detected image of the at least one superposition measurement target and the trained measurement model; and The superimposed value is stored in the memory.

27. The method of claim 26, further comprising: The values ​​of the lithography process variables are adjusted based on the estimated values ​​obtained from the actual superposition.

28. The method of claim 23, wherein at least one of the superimposed metrological targets is a single-cell metrological target having a grating structure array with periodicity in at least one direction.

29. The method of claim 23, wherein at least one of the superimposed metrological targets is a single-cell metrological target having a grating structure array with periodicity in at least two directions.

30. The method of claim 23, wherein at least one of the superimposed metrological targets is a single-cell metrological target having an array of grating structures having two or more different periodicities in at least one direction.

Images