Full wafer metrology upsampling

Abstract

A system and method for OCD metrology is provided that includes receiving training data for training an OCD machine learning (ML) model, the training data measured from a plurality of wafers and including pairs of corresponding input and tag data sets obtained from each respective wafer. Each pair of input data sets includes a plurality of scatterometry data sets measured at a plurality of respective positions defined by the first map. Each pair of label datasets includes one or more critical dimension (CD) parameters for a respective location defined by a second graph that includes at least one location that is not in the first graph. In accordance with the first map, the OCD ML model is applied to a new set of scatterometry datasets measured from the position of the new wafer to generate predicted CD parameters for the position of the second map on the new wafer.

Description

Technical Field

[0001] The present invention relates generally to the field of optical inspection of integrated circuit wafer patterns, and in particular to algorithms for measuring wafer pattern parameters. Background Art

[0002] Integrated circuits (ICs) are produced on semiconductor wafers by multiple steps of depositing, altering, and removing thin layers, which are built up into stacked structures on the wafer. These stacked structures (also called "stacks", "features") can be formed in repeating patterns that, like diffraction gratings, have optical properties. Modern optical metrology methods for measuring critical dimensions (CDs) and material properties of these patterns exploit these optical properties, for example, by applying rigorous coupled wave analysis (RCWA) to scatterometry data to determine the CD and material properties at a given measurement location.

[0003] Hereinafter, CD and material properties are also referred to as "parameters of interest" (POI), or simply "parameters". These parameters may include the height, width and spacing of the stack. As described by Dixit et al. in "Sensitivity analysis and line edge roughness determination of 28-nmpitch silicon fins using Mueller matrix spectroscopic ellipsometry-based optical critical dimension metrology" J. Micro / Nanolith. MEMS MOEMS. 14 (3), 031208 (2015), which is incorporated herein by reference, pattern parameters may also include: sidewall angle (SWA), spacer width, spacer pull-down, epitaxial proximity, footing / undercut, two-dimensional overfill / underfill (HKMG), three-dimensional profile (FinFET) and line edge roughness (LER).

[0004] Optical critical dimension (OCD) metrology employs scatterometry methods to measure scatterometry data, i.e., reflected optical radiation indicative of an optical property of a pattern. A collection of scatterometry data, which may also be referred to as a scatterometry signature, may include data points of reflected irradiance versus angle of incidence of the radiation (which may be a zero-order measurement). Alternatively or in addition, the scatterometry data may include a spectrogram, which is a measure of the intensity of the reflected radiation over a range of wavelengths or frequencies. Additional types of scatterometry data known in the art may also be applied to OCD metrology.

[0005] U.S. Patent 6,476,920, “Method and apparatus for measurements of patterned structures,” by Scheiner and Machavariani, incorporated herein by reference, describes the development of an “optical model” (also referred to as a “physical model”). An optical model is a function (i.e., a collection of algorithms) that defines the relationship between reflected radiation and the physical structure of a wafer. That is, an optical model is a theoretical model of how light reflects from a pattern with known parameters. Thus, such an optical model can be applied to generate estimates of scatterometry data from a known set of pattern parameters that will be measured during a metrology session, such as on a production wafer during HVM. The optical model can also be designed to perform the inverse (or “inverse”) function of estimating the pattern parameters based on the measured scatterometry data.

[0006] Optical models are often used for OCD metrology during IC production to determine whether wafer patterns are manufactured with the correct parameters based on scatterometry. Pattern parameters on a given wafer can be measured to determine how much the parameters vary from design specifications, which can specify allowable deviations from the mean.

[0007] Machine learning (ML) techniques can also be used to estimate pattern parameters based on scatterometry data. For example, as described in PCT patent application WO 2019 / 239380 by Rothstein et al., which is incorporated herein by reference, a machine learning model can be trained to identify the correspondence between measured scatterometry data and reference parameters.

[0008] Exemplary scatterometry tools for measuring (acquiring) scatterometry data (e.g., spectrograms) may include spectroscopic ellipsometers (SE), spectroscopic reflectometers (SR), polarization spectroscopic reflectometers, and other optical critical dimension (OCD) metrology tools. Such tools are incorporated into currently available OCD metrology systems. One such OCD metrology system is the NOVA prism OCD metrology tool, commercially available from Nova Measuring Instruments Ltd. of Rehovot, Israel, which can measure pattern parameters at a designated test site or "in-mold". Additional methods for measuring critical dimension (CD) include interferometry, X-ray Raman spectroscopy (XRS), X-ray diffraction (XRD), and pump-probe tools, among others. Some examples of such tools are disclosed in U.S. Patents US10,161,885, US10,054,423, US9,184,102 and US10,119,925 and International Pending Patent Application Publication WO2018 / 211505, all of which are assigned to the applicant and are incorporated herein by reference in their entirety. Scattering measurement tools configured as an integrated metrology (IM) platform enable advanced process control (APC) to monitor and control wafer-to-wafer variations of complex high-end CMP and etch applications during high-volume manufacturing (HVM) with high productivity and reliability required for the most advanced logic and memory technology nodes. Integrated metrology systems include Nova i550, i570 and Astera, commercially available from Nova Measurement Instruments Ltd. of Rehovot, Israel, which are integrated with processing equipment such as CMP polishers.

[0009] High-precision methods for measuring pattern parameters that do not rely on the above optical models include wafer measurement using equipment such as CD scanning electron microscopes (CD-SEMs), atomic force microscopes (AFMs), cross-sectional tunneling electron microscopes (TEMs), or X-ray metrology tools. These methods are generally more expensive and time-consuming than optical modeling methods.

[0010] Regardless of whether the optical model and the ML model are applied to the scatterometry data, the process of obtaining the scatterometry data is time consuming. On the other hand, sparse measurements of the scatterometry data may not be sufficient to indicate problems with wafer production. Embodiments of the present invention as disclosed below address these shortcomings of sparse scatterometry data measurements. Summary of the invention

[0011] Embodiments of the present invention provide systems and methods for generating machine learning models for optical critical dimension (OCD) monitoring, including "upsampling" to improve OCD resolution. Advanced wafer manufacturing can benefit from more extensive critical dimension (CD) parameter sampling, but extensive sampling conflicts with measurement cycle time goals and overall metrology tool costs. Therefore, manufacturers typically use "sparse" sampling schemes that reduce monitoring delays. The systems and methods provided herein provide CD parameters of interest by upsampling that avoids monitoring time delays. The result is improved OCD monitoring with greater accuracy and robustness, without corresponding time delays, thereby improving process control. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to better understand the various embodiments of the present invention and to show how to implement the present invention, reference is made to the accompanying drawings by way of example. The structural details of the present invention are shown to provide a basic understanding of the present invention, and the description in conjunction with the accompanying drawings makes it apparent to those skilled in the art how to embody several forms of the present invention in practice. In the accompanying drawings:

[0013] Figure 1 is a block diagram of a system for generating a machine learning model for OCD metrology by sampling from a sparse wafer map according to an embodiment of the present invention;

[0014] Figure 2 is a schematic diagram of applying a machine learning model to upsample a sparse wafer map to a denser wafer map according to an embodiment of the present invention;

[0015] Figure 3 is a flow chart depicting a process for generating a machine learning model for OCD metrology by sampling from a sparse wafer map in accordance with an embodiment of the present invention;

[0016] Figure 4 is a schematic diagram depicting the use of input and output (i.e., target) data in generating a machine learning model for sampling from a sparse wafer map according to an embodiment of the present invention;

[0017] Figure 5 is a schematic diagram of a neural network acting as a machine learning model for sampling from a sparse wafer map according to an embodiment of the present invention;

[0018] FIG. 6A to FIG. 6C is a graph validating a machine learning model for sampling from a sparse wafer map according to an embodiment of the present invention; and

[0019] Figure 7 is a graph indicating a “trust” parameter that may be associated with input data sampled from a sparse wafer map in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0020] Embodiments of the present invention provide systems and methods for generating machine learning (ML) models for optical critical dimension (OCD) monitoring by training the ML model with scatterometry data for upsampling from "sparse" wafer maps, i.e., for predicting OCD parameters that are not measured directly from a sparse set of wafer parameter measurements.

[0021] Although advanced OCD metrology equipment can provide fast and non-destructive measurement of critical dimensions, manufacturers typically only perform optical metrology measurements from a sparse subset (typically -25%) of the die on a wafer to meet cycle time targets. Wafer sampling plans may be carefully designed to acquire measurements from all wafer quadrants and known radii of interest, but useful details indicating process variations may still be missed. These details can include process-specific within-wafer (WiW) profiles, topographic variations within a die region of interest, or process chamber-specific WiW profiles. In addition, a larger number of measurements are typically required while supporting process changes, qualifying new equipment, and for offset detection and recovery. Upsampling measurements acquired from a sparse wafer map, as provided by the methods and systems of the present invention, achieves the advantage of acquiring more wafer measurements without affecting manufacturing throughput.

[0022] Figure 1 is a schematic diagram of a system 10 for generating a machine learning model for OCD metrology according to an embodiment of the present invention.

[0023] The system 10 can be operated within a production line (not shown) for producing and monitoring one or more wafers 12. As shown, the wafer 12 includes patterns 14 (also referred to herein as "structures"). These patterns have critical dimensions (CDs) or "parameters", as shown in the pattern magnification 14a, which parameters can include height ("h"), width ("w"), and spacing ("p"), as well as other parameters described in the above background technology. Typically, a single wafer includes multiple dies that are designed to have the same pattern (i.e., the same pattern design). For each pattern ("point of interest") in each die, a set of multiple parameters can be measured.

[0024] Manufacturing variations result in slight variations in the parameters of the measurement locations between wafers and across a single wafer. These variations result in variations in the measured scatterometry data. During a typical monitoring period, a scatterometry data set is measured at the measurement locations defined by the wafer map, typically for the same pattern in each of multiple dies. The optical model is then applied to the scatterometry data set to determine a set of one or more POIs (i.e., CD parameters). The scatterometry data set is also written herein as a vector and the set of one or more CD parameters at a given measurement location is written as a vector The set of all scatterometry data sets measured from a given wafer is written as the set S and The set of vectors (ie, CD parameters measured at each of a plurality of common measurement locations in a plurality of corresponding dies of a wafer) is referred to as the set P.

[0025] The system 10 includes a light source 20 that generates a light beam 22 of a predetermined wavelength range. During the monitoring process, the light beam 22 is reflected from the wafer pattern 14 (indicated as reflected or "scattered" light 24) toward a spectrophotometric detector 26. In some configurations, the light source and the spectrophotometric detector (e.g., an ellipsometer or a spectrophotometer) are included in an OCD metrology system 30. The construction and operation of the metrology system 30 can be of any known kind, for example, such as the types disclosed in U.S. Patents US 5,517,312, US 6,657,736, and US 7,169,015 and International Pending Patent Application Publication WO2018 / 211505, all of which are assigned to the applicant and are incorporated herein by reference in their entirety. Typically, the metrology system 30 includes additional components not shown, such as a light guide optical device, which may include a beam deflector having an objective lens, a beam splitter, and a reflector. Additional components of such a system may include an imaging lens, a polarizing filter, a variable aperture diaphragm, and a motor. The operation of such elements is typically automated by a computer controller, which may include I / O devices and may also be configured to perform data processing tasks, such as generating scatterometry data 32 .

[0026] The scatterometry data 32 generated by the metrology system 30 may include a spectrogram 34, which may be represented in vector form, whose data points are measures of reflected light intensity "E" at different light wavelengths. The scatterometry data may also or alternatively be a mapping of reflected irradiance versus angle of incidence. In typical OCD metrology, the range of light measured may cover the visible spectrum, and may also include wavelengths in the ultraviolet and infrared regions. A typical spectrogram output for OCD metrology may have 245 data points, covering a wavelength range of 200 nm to 970 nm.

[0027] As described above, a scatterometry data set is measured from each point of a predefined map of the wafer. All measurement locations of the map are copies of a given pattern and ideally will have the same CD parameters. In actual manufacturing, processing conditions on the wafer surface vary, and variations between scatterometry data sets at different measurement locations indicate different CD parameters for the measurement locations.

[0028] In an embodiment of the present invention, a scatterometry dataset from a "sparse" map of measurement locations on a wafer is applied to train an ML model 40 to improve the OCD measurement resolution, i.e., to predict parameter values ​​for a larger number of measurement locations on the wafer based on the sparse measurements. Hereinafter, the process of predicting the value of a higher resolution map from a sparse map is also referred to as "upsampling". It should be understood that a "sparse map" is defined as a map with fewer measurement locations than a denser target map. Hereinafter, a denser map is also referred to herein as a "full wafer map" (FWM), but it should also be understood that a map referred to herein as a FWM may cover less than 100% of the dies of a wafer (and conversely, may also include more measurement locations than dies). In a typical metrology practice for HVM, such as using an integrated metrology platform, a sparse map may be a typical high throughput map indicating the measurement locations of approximately 20% to 30% of the dies in a wafer design, while a FWM may cover 80% to 100%. However, it should be understood that the present invention is applicable to a wide range of sparse and dense maps, where the difference between the maps may be a single measurement location.

[0029] The ML model can be based on ML tools known in the art, such as neural networks, random forests, or any type of ML regression algorithm. It can include known methods to avoid overfitting, such as regularization, model ensembles, or intelligent feature extraction and selection. Reference parameters 44 from a denser wafer map can be used as target labels for ML training. The reference parameters can be obtained from measurement locations of one or more wafers by the same OCD spectrum used to acquire the scatterometry data set 36, or by other means known in the art (such as CD-SEM, AFM, TEM, X-ray metrology). After training, the ML model 40 is used to predict pattern parameters based on the scatterometry data set. The predicted pattern parameters can then be applied, for example, in monitoring of wafer production.

[0030] The ML model 40 may operate independently of the metrology system 30 or may be integrated with the metrology system.

[0031] Figure 2202 indicates white points at locations measured by spectroscopy. Dense map 206 shows sparse measurements as white points and extrapolated measurements obtained by the ML model as black points. The backgrounds of both figures indicate shaded contour maps for given CD parameters. The contours are generated by interpolation between measured points for the sparse map and by interpolation between measured points and predicted points for the dense map. Interpolation is performed by simple radial basis functions and cubic spline interpolation. As shown, the contour map is significantly more detailed (i.e., has a higher resolution) for the dense map. The high-resolution data provided by the ML model reveals enhanced visibility of POI variability. Greater visibility can help detect process and hardware performance early and help successfully achieve key deliverables.

[0032] Figure 3 3 is a flow chart depicting a computer-implemented process 300 for generating a machine learning model for OCD metrology by sampling from a sparse wafer map in accordance with an embodiment of the present invention. The process 300 includes two phases, namely, a training phase 310 for generating the ML model 40 as described above and a production phase 320 implemented by the ML model 40.

[0033] The first step 312 of the ML training phase includes generating (ie, measuring) a scatterometry dataset (eg, a spectrogram) for each measurement location of a sparse wafer map of a plurality of wafers. The scatterometry dataset measured from each wafer is used as input data for training.

[0034] The second step 316 of training includes acquiring (e.g., measuring and / or calculating) corresponding labeled training data (i.e., target data) for each wafer, the labeled training data being values ​​of one or more CD parameters for each measurement location of a second, typically denser map (e.g., a full wafer map (FWM)) for the plurality of wafers. It should be understood that in some implementations, the second map differs from the first map only in that it includes at least one location that is not in the first map, and the first map is typically, but not necessarily, a subset of the second map. In some implementations, the CD parameters of the second map are obtained from a scatterometry dataset, and assuming that the first map is a subset of the second map, the scatterometry dataset associated with the first map is used as the input training dataset.

[0035] At step 318 , the ML model is then trained using the input obtained from each wafer and the corresponding labeled dataset.

[0036] Once the ML model is generated, it is applied in the production stage 320. At step 322, a new set of scatterometry data sets are measured from the sparse set of measurement locations on the new wafer. The ML model is then applied to the new set of scatterometry data sets to determine CD parameter values ​​for the measurement locations on the FWM of the new wafer, for example, during HVM using an integrated metrology platform, at step 324. The generated ML model can be deployed on an integrated metrology tool for APC during HVM.

[0037] Figure 4 is a graphical view of the steps for generating the ML model 40. A set of n wafers (shown as W1 to W n ), each wafer has a sparse map and a dense map (FWM). The scatterometry data set is measured from the measurement locations of the sparse map, indicated in the figure as both a collection of spectral maps and an equation of the following form:

[0038]

[0039] Each lowercase The vector represents a scatterometry data set from a single measurement location, and the capital S represents the set of all scatterometry data sets of the sparse map of the wafer. A sparse set is indicated as having l number of data sets.

[0040] The CD parameters for all measured positions of the dense map are given by equations of the following form:

[0041]

[0042] The CD parameters for each measurement position are represented by lowercase The vector indicates that the dense map is indicated as having m measurement locations, where m>l. As shown in the figure, the ML model 40 is trained from n pairs of corresponding S and P.

[0043] Figure 5 is a schematic diagram of a neural network 500 that acts as a machine learning model for sampling from a sparse wafer map. A set S of scatterometry data sets from a sparse wafer map is fed to input nodes 520. The total number of input nodes 520 is equal to the number of data points of the set S, which is equal to the number of measurement locations l multiplied by the number of data points d for each scatterometry data set (e.g., 245 data points, as described above). For a target set P of k predicted CD parameters at each predicted location of a dense wafer map, this number l×d of input nodes is transmitted to m×k output nodes 530. One or more hidden layers 540 connect the input nodes and the output nodes.

[0044] FIG. 6A to FIG. 6Cis a graph of a machine learning model for sampling from a sparse wafer map according to an embodiment of the present invention. CD measurement data are often merged for visual analysis. Such merging can take the form of radial profiles, contour plots, and various statistical measures, such as wafer average and WiW standard deviation. Fig. 6A It is shown that the radial graph with ML predicted CD values ​​is closer to the radial graph of actual CD values ​​than the radial graph of sparse CD values ​​(interpolation is performed between measured positions to determine radial values).

[0045] Figure 6B and Figure 6C shows the die level measurements ( Figure 6B ) and wafer-level measurements ( Figure 6C ) between the actual FWM CD measurements and the ML predicted values. It can be seen that the correlations each show a nearly linear fit with a slope close to unity.

[0046] Figure 7 is a graph indicating a “trust” score that can be associated with input data sampled from a sparse wafer map. The trust score is a measure of confidence in the predicted result based on the “closeness” (i.e., the inverse of the “distance”) of the input data to the range spanned by the input data of the training set. When the input data (i.e., the set S of scatterometry data sets) has a trust score greater than, for example, 90%, the result can be considered reliable. The trust score can be based on, for example, a value function distance between a function and a set of functions, as described in International Patent Application WO2021140515 entitled “Detecting outliers and anomalies for OCD metrology machine learning”, the description of which is incorporated herein by reference.

[0047] The table in the figure presents the trust scores determined from online process of record (POR) wafers (indicated as "Process") compared to the trust scores of experimental wafers that underwent stack changes (Experimental). The online process data shows that the trust scores are well above 90% for most sites. (The few points below 0.90 (not shown) are edge die locations, which exhibit greater stack variability than the locations included in the training set.) The experimental data shows that the trust scores are mainly below 90%, indicating that the process stack of these wafers is significantly different from the process stack that produced the training set. These results indicate that the ML prediction solution is sensitive to online process perturbations. Therefore, the trust score is a quality metric that can alert the user to wafer anomalies or serve as an indicator when additional ML training data is needed.

[0048] In short, it is important to design the sampling scheme to adequately capture all radial variations and spatial variability modes to ensure that the measurements are truly representative of the process. In addition, since different process tools have different spatial variability signatures during the process flow, the sampling scheme at different process steps should be modified accordingly. Due to cycle time constraints, sparse measurements are often employed, but as a result, defects may only be exposed at end-of-line (EOL) testing. As disclosed herein, upsampling simulates full-wafer sampling, enhancing visibility of potential problems and thereby improving process control.

[0049] It should be understood that the processing elements shown or described herein are preferably implemented by one or more computers in computer hardware and / or computer software contained in non-transitory computer-readable media in accordance with conventional techniques, such as using computer processors, memories, I / O devices, and network interfaces coupled via a computer bus or alternative connection means.

[0050] Unless otherwise specified, the term "processor" or "device" is intended to include any processing device, such as, for example, a processing device including a CPU (central processing unit) and / or other processing circuitry (e.g., a GPU), and may refer to more than one processing device. Various elements associated with a processing device may be shared by other processing devices.

[0051] The term "memory" as used herein is intended to include memory associated with a processor or CPU, such as, for example, RAM, ROM, fixed memory devices (e.g., hard drive), removable memory devices (e.g., floppy disk, tape), flash memory, etc. Such memory may be considered a computer-readable storage medium.

[0052] Additionally, the phrase “input / output device” or “I / O device” may include one or more input devices (e.g., keyboard, mouse, scanner, HUD, etc.) for inputting data to the processing unit, and / or one or more output devices (e.g., speakers, displays, printers, HUD, AR, VR, etc.) for presenting results associated with the processing unit.

[0053] Embodiments of the present invention may include systems, methods, and / or computer program products. A computer program product may include (one or more) computer-readable storage media having computer-readable program instructions thereon for causing a processor to perform aspects of the present invention.

[0054] Computer readable storage medium can be a tangible device that can retain and store instructions for use by instruction execution devices. Computer readable storage medium can be, for example, but not limited to, electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of computer readable storage medium includes the following: portable computer floppy disk, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disk read-only memory (CD-ROM), digital versatile disk (DVD), blue light, magnetic tape, holographic memory, memory stick, floppy disk, mechanical encoding device (such as a punch card or a convex structure in a groove with instructions recorded thereon), and any suitable combination of the foregoing. Computer readable storage medium as used herein should not be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (e.g., a light pulse by a fiber optic cable), or an electrical signal transmitted by a wire.

[0055] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to a corresponding computing / processing device, or downloaded to an external computer or external storage device via a network (e.g., the Internet, a local area network, a wide area network, and / or a wireless network). A network adapter card or network interface in each computing / processing device can receive computer-readable program instructions from a network and forward the computer-readable program instructions for storage in a computer-readable storage medium in the corresponding computing / processing device.

[0056] The computer-readable program instructions for performing the operation of the present invention can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages ​​(including object-oriented programming languages ​​such as Java, Smalltalk, C++, etc., and conventional process programming languages ​​such as "C" programming language or similar programming languages). The computer-readable program instructions can be executed completely on the user's computer, partly on the user's computer as an independent software package, partly on the user's computer and partly on a remote computer or completely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any type of network (including a local area network (LAN) or a wide area network (WAN)), or can be connected to an external computer (for example, by using the Internet of an Internet service provider). In some embodiments, the electronic circuit system including, for example, a programmable logic circuit system, a field programmable gate array (FPGA) or a programmable logic array (PLA) can execute the computer-readable program instructions by utilizing the state information of the computer-readable program instructions to personalize the electronic circuit system, thereby performing various aspects of the present invention.

[0057] Where aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention, it will be understood that each block of the flowchart illustrations and / or block diagrams and combinations of blocks in the flowchart illustrations and / or block diagrams can be implemented by computer-readable program instructions.

[0058] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device create a device for implementing the functions / actions specified in one or more boxes of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, which can guide the computer, programmable data processing device, and / or other device to operate in a specific manner, so that the computer-readable storage medium having the instructions stored therein includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes of the flowchart and / or block diagram.

[0059] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing device, or other device to cause a series of operational steps to be performed on the computer, other programmable device, or other device, thereby producing a computer-implemented process, so that the instructions executed on the computer, other programmable device, or other device implement the functions / actions specified in one or more boxes of the flowchart and / or block diagram.

[0060] Any flow chart and block diagram included herein illustrate the architecture, function and operation of the possible implementation of the system, method and computer program product according to various embodiments of the present invention.In this respect, each square block in the flow chart or block diagram can represent a module, segment or part of an instruction, which can include one or more executable instructions for realizing a specified logical function.In some alternative implementations, the function marked in the square block can appear in the order shown in this article.For example, in fact, the two square blocks shown continuously can be performed substantially simultaneously, or these square blocks can sometimes be performed in reverse order, depending on the functionality involved.It will also be noted that the combination of the square blocks in each square block of the block diagram and / or the flow chart illustration and the block diagram and / or the flow chart illustration can be realized by a system based on special hardware that performs a specified function or action or performs a combination of special hardware and computer instructions.

[0061] The description of various embodiments of the present invention has been presented for the purpose of illustration, but is not intended to be exhaustive or limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terms used herein are selected to best explain the principles of the embodiments, practical applications, or technical improvements over technologies found on the market, or to enable those of ordinary skill in the art to understand the embodiments disclosed herein.

[0062] Example

[0063] Examples of the present invention may include the following configurations.

[0064] Example 1 is a method for optical critical dimension (OCD) metrology, which includes the following steps.

[0065] 1) receiving training data for training an OCD machine learning (ML) model, the training data being measured from a plurality of wafers and comprising a plurality of pairs of corresponding input and label data sets obtained from each respective wafer. The input data set of each pair comprises a plurality of scatterometry data sets measured at a plurality of respective positions defined by a first map of each wafer. The label data set of each pair comprises one or more critical dimension (CD) parameters of a plurality of respective positions defined by a second map of each wafer, the second map comprising at least one position not in the first map.

[0066] 2) Apply the input and label pairs of the training dataset to generate the OCD ML model.

[0067] 3) Based on the first graph, applying the OCD ML model to a new set of scatterometry data sets measured from locations on the new wafer to generate predicted CD parameters for the locations on the new wafer in the second graph.

[0068] Example 2 is a method including the steps of Example 1, further defined as the first graph being a subset of the second graph.

[0069] Example 3 is a method including the steps of Example 2, further defined as receiving training data comprising: receiving a scatterometry data set for the first map and an additional scatterometry data set for each second map position not included in the first map for each wafer, and determining a label data set of CD parameters for each wafer from all of the scatterometry data sets for the second map.

[0070] Example 4 is a method including the steps of any one of Examples 1 to 3, further comprising: measuring a distance of a new set of scatterometry data sets to a series of scatterometry data sets of an input training data set to verify a confidence score of a prediction of a CD parameter of a location in the second map on the new wafer.

[0071] Example 5 is a method including the steps of any of Examples 1 to 4, further comprising: calculating a radial plot of predicted CD parameters for new wafers to identify process control issues and issue user alerts accordingly.

[0072] Example 6 is a method including the steps of any one of Examples 1 to 5, further defined as one or more CD parameters of the label data set being obtained by applying the optical model to a scatterometry data set obtained at locations of the second map.

[0073] Example 7 is a method including the steps of any one of Examples 1 to 5, further defined as, the one or more CD parameters of the label data set are obtained by applying the second ML model to the scatterometry data set obtained at the locations of the second map.

[0074] Example 8 is a method including the steps of any one of Examples 1 to 5, further defined as one or more CD parameters of the label data set are obtained from one or more of a CD scanning electron microscope (CD-SEM), an atomic force microscope (AFM), a cross-sectional tunneling electron microscope (TEM), or an X-ray metrology tool.

[0075] Example 9 is a method including the steps of any of Examples 1 to 8, further defined by each scatterometry data set being a spectrogram.

[0076] Example 10 is a method including the steps of any one of Examples 1 to 9, further defined as the scatterometry data set being measured by one or more instruments comprising a spectroscopic ellipsometer (SE), a spectroscopic reflectometer (SR), and a polarization spectroscopic reflectometer.

[0077] Example 11 is a method including the steps of any one of Examples 1 to 10, further defined wherein the OCD ML model is one of a neural network or a random forest algorithm.

[0078] Example 12 is a system for optical critical dimension (OCD) metrology, comprising a processor having a non-transitory memory, the memory comprising instructions that, when executed by the processor, cause the processor to implement the steps of any one of Examples 1 to 11.

[0079] Example 13 is a non-transitory machine-accessible storage medium having instructions stored thereon, which, when executed by a machine, causes the machine to implement the steps of any one of Examples 1 to 11.

[0080] Validation of the trained (upsampled) model can be done via time-based triggering on real full-wafer maps collected on the same tool (integrated metrology) or a standalone system, alternatively the trigger can be based on the performance of predicted WiW / intra-area variations and / or model confidence metrics.

Claims

1. A method for optical critical dimension (OCD) metrology, comprising: receiving training data for training an OCD machine learning (ML) model, the training data being measured from a plurality of wafers and comprising a plurality of pairs of corresponding input and label data sets obtained from each respective wafer, wherein the input data set of each pair comprises a plurality of scatterometry data sets measured at a plurality of corresponding locations defined by the first map of each wafer, and wherein the label data set of each pair includes one or more critical dimension (CD) parameters for a plurality of corresponding locations defined by a second map of each wafer, the second map including at least one location not in the first map; Applying the input and label pairs of the training dataset to generate the OCD ML model; and Based on the first map, the OCD ML model is applied to a new set of scatterometry data sets measured from locations on a new wafer to generate predicted CD parameters for the locations on the new wafer of the second map.

2. The method according to claim 1, wherein: The first graph is a subset of the second graph.

3. The method according to claim 2, wherein: Receiving the training data includes: receiving the scatterometry data set for the first map and an additional scatterometry data set for each second map position not included in the first map for each wafer, and determining the label data set of CD parameters for each wafer from all of the scatterometry data sets of the second map.

4. The method according to claim 1, further comprising: The distance of the new set of scatterometry data sets to a series of scatterometry data sets of the input training data set is measured to verify the confidence score of the prediction of the CD parameter of the location of the second map on the new wafer.

5. The method according to claim 1, further comprising: A radial plot of the predicted CD parameters for the new wafer is calculated to identify process control issues and issue user alerts accordingly.

6. The method according to claim 1, wherein: The one or more CD parameters of the label data set are obtained by applying an optical model to a scatterometry data set obtained at locations of the second map.

7. The method according to claim 1, wherein: The one or more CD parameters of the label dataset are obtained by applying a second ML model to a scatterometry dataset obtained at locations of the second map.

8. The method according to claim 1, wherein: The one or more CD parameters of the label data set are obtained from one or more of a CD scanning electron microscope CD-SEM, an atomic force microscope AFM, a cross-sectional tunneling electron microscope TEM, or an X-ray metrology tool.

9. The method according to claim 1, wherein: Each scatterometry data set is a spectroscopic graph.

10. The method according to claim 1, wherein: The scatterometry data set is measured by one or more instruments including a spectroscopic ellipsometer SE, a spectroscopic reflectometer SR and a polarimetric spectroscopic reflectometer.

11. The method according to claim 1, wherein: The OCD ML model is one of a neural network or a random forest algorithm.

12. A system for optical critical dimension (OCD) metrology, comprising a processor having a non-transitory memory, the memory comprising instructions which, when executed by the processor, cause the processor to perform the following steps: receiving training data for training an OCD machine learning (ML) model, the training data being measured from a plurality of wafers and comprising a plurality of pairs of corresponding input and label data sets obtained from each respective wafer, in, The input data set of each pair comprises a plurality of scatterometry data sets measured at a plurality of corresponding locations defined by a first map of each wafer, and wherein the label data set of each pair includes one or more critical dimension (CD) parameters for a plurality of corresponding locations defined by a second map of each wafer, the second map including at least one location not in the first map; Applying the input and label pairs of the training dataset to generate the OCD ML model; and Based on the first map, the OCD ML model is applied to a new set of scatterometry data sets measured from locations on a new wafer to generate predicted CD parameters for the locations on the new wafer of the second map.

13. A non-transitory machine-accessible storage medium storing instructions which, when executed by a machine, cause the machine to perform the following steps: receiving training data for training an optical critical dimension (OCD) machine learning model, the training data being measured from a plurality of wafers and comprising a plurality of pairs of corresponding input and label data sets obtained from each respective wafer, in, The input data set of each pair comprises a plurality of scatterometry data sets measured at a plurality of corresponding locations defined by a first map of each wafer, and wherein the label data set of each pair includes one or more critical dimension (CD) parameters for a plurality of corresponding locations defined by a second map of each wafer, the second map including at least one location not in the first map; Applying the input and label pairs of the training dataset to generate the OCD ML model; and Based on the first map, the OCD ML model is applied to a new set of scatterometry data sets measured from locations on a new wafer to generate predicted CD parameters for the locations on the new wafer of the second map.

Images