Composite data for device metrology

Abstract

A machine learning model using composite metrology data uses measured metrology data from a device under test to determine at least one parameter of the device. The composite metrology data is generated by merging measured metrology data from a reference device with synthetic metrology data calculated from a model of the reference device. The composite metrology data may further be generated based on synthetic metrology data calculated from a model for the modified reference device. The modified reference device may be generated using a change in at least one parameter of the model to extend a parameter space of the training range.

Description

[0001] Cross-references to related patent applications

[0002] This application claims priority to and the benefit of U.S. Provisional Application No. 63 / 412,339, entitled “ENHANCED MACHINE LEARNING RECIPE,” filed on September 30, 2022, U.S. Provisional Application No. 63 / 498,474, entitled “COMPOSITE DATA FOR OPTICAL METROLOGY,” filed on April 26, 2023, and U.S. Non-Provisional Application No. 18 / 339,982, entitled “COMPOSITE DATA FOR DEVICE METROLOGY,” filed on June 22, 2023, all of which are assigned to the assignee of this application and are incorporated herein by reference in their entirety. Technical Field

[0003] The subject matter described herein relates generally to metrology, and more particularly to training and using a machine learning model to characterize at least one parameter of a device structure. Background Art

[0004] Semiconductor and other similar industries often use optical metrology, such as optical metrology or X-ray metrology, to perform non-contact evaluation of samples during processing. For example, with optical metrology, the sample under test is illuminated with light, such as at a single wavelength or multiple wavelengths. After the light interacts with the sample, the resulting light is detected and analyzed to determine at least one characteristic of the sample.

[0005] Various types of metrology for process improvement, monitoring, and control, such as optical critical dimensions (OCD) of device structures, often employ modeling of the structures being measured. For example, a model can be generated based on the material and nominal parameters of the structure, such as film thickness, line and space widths, etc. Shrinking critical dimensions and smaller error margins combined with increasingly complex structures, such as 3D NAND, are challenging current modeling capabilities.

[0006] Machine learning-based OCD metrology solutions and other similar metrology solutions can be useful when matched with reference metrology data, but only in scenarios where the online process conditions are similar to those present in the process conditions used for model training. However, robustness is a challenge for machine learning-based metrology solutions because there is typically only a limited amount of reference metrology data available for training machine learning recipes. For example, process changes, especially in the early development stages, may undermine the practicality of reference metrology data for training machine learning recipes. This effect is more significant when the raw spectral data collected from samples after modifying the manufacturing process shows a certain degree of deviation from the data previously collected and used for model training. A possible solution to this problem is to obtain additional reference metrology data and retrain the machine learning model after the process modification. However, obtaining updated reference metrology data for retraining the machine learning model is inefficient because the measurement of the reference metrology data may be too expensive and the machine learning model retraining may be time-consuming. In addition, obtaining additional reference metrology data and retraining the machine learning model in response to process modifications does not take into account any further process modifications and must be performed iteratively. Therefore, improvements are expected. Summary of the invention

[0007] The metrology of the device structure may be performed using a machine learning model that is trained using composite metrology data. The composite metrology data may be generated, for example, by merging measured metrology data from a reference device with synthetic metrology data calculated from a model of the reference device. The composite metrology data may be further generated based on synthetic metrology data calculated from a model for a modified reference device. The modified reference device is generated using a change in at least one parameter of the model to expand the parameter space of the training range. The composite metrology data may be generated, for example, by modifying synthetic metrology data calculated from a model for the modified reference device based on a change (e.g., a mismatch or spectral shift) between measured metrology data from the reference device and synthetic metrology data calculated from a model for the reference device. In another specific implementation, the composite metrology data may be generated by modifying measured metrology data from the reference device based on a change (e.g., a difference between synthetic metrology data calculated from a model of the reference device and synthetic metrology data calculated from a model for the modified reference device). The metrology of the device structure may be performed using a machine learning model that is trained using a training data set that includes at least the composite metrology data.

[0008] In one specific implementation, a method for characterizing a device on a sample includes: obtaining measured metrology data from the device; and determining at least one parameter of the device based on the measured metrology data using a machine learning model using composite metrology data. Each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device.

[0009] In one specific implementation, a metrology system configured to support characterization of a device on a sample includes: a radiation source configured to generate radiation to be incident on the device on the sample; at least one detector configured to detect radiation from the device generated in response to the radiation incident on the device; and at least one processor coupled to the at least one detector. The at least one processor is configured to obtain measured metrology data from the device, and determine at least one parameter of the device based on the measured metrology data using a machine learning model using composite metrology data. Each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device.

[0010] In one specific implementation, a system configured to support characterization of a device on a sample includes: means for obtaining measured metrology data from the device; and means for determining at least one parameter of the device based on the measured metrology data using a machine learning model using composite metrology data. Each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device.

[0011] In one specific implementation, a method for characterizing a device on a sample includes obtaining measured metrology data for a reference device for the device. The method also includes generating a first set of synthetic metrology data for a first model of the reference device, and generating a second set of synthetic metrology data for a second model of a modified reference device that is changed relative to the first model. Composite metrology data for the modified reference device is generated. Each composite metrology data is generated by combining the measured metrology data, the first synthetic metrology data, and the second synthetic metrology data, and at least the composite metrology data is stored as a training data set.

[0012] In one specific implementation, a computer system configured to support characterization of a device on a sample includes at least one processor configured to obtain measured metrology data for a reference device for the device. The at least one processor is also configured to generate a first set of synthetic metrology data for a first model of the reference device, and to generate a second set of synthetic metrology data for a second model of the modified reference device that is changed relative to the first model. The at least one processor is also configured to generate composite metrology data for the modified reference device. Each composite metrology data is generated by merging the measured metrology data, the first synthetic metrology data, and the second synthetic metrology data, and at least the composite metrology data is stored as a training data set.

[0013] In one specific implementation, a system configured to support characterization of a device on a sample includes means for obtaining measured metrology data for a reference device for the device. The system also includes means for generating a first set of composite metrology data for a first model of the reference device and generating a second set of composite metrology data for a second model of a modified reference device that is changed relative to the first model. Composite metrology data for the modified reference device is generated. Each composite metrology data is generated by combining the measured metrology data, the first composite metrology data, and the second composite metrology data, and at least the composite metrology data is stored as a training data set.

[0014] In one specific implementation, a method for characterizing a device on a sample includes obtaining measured optical metrology data for a reference device for the device and generating composite optical metrology data, wherein each composite optical metrology data is generated by merging the measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device. The method also includes training a machine learning model using a training data set including at least the composite optical metrology data to characterize the device using the measured optical metrology data from the device.

[0015] In one specific implementation, a computer system configured to support characterization of a device on a sample includes: at least one memory configured to store measured optical metrology data and composite optical metrology data; and at least one processor coupled to the at least one memory. The at least one processor is configured to obtain measured optical metrology data for a reference device for the device, and to generate composite optical metrology data, wherein each composite optical metrology data is generated by merging the measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device. The at least one processor is also configured to train a machine learning model using a training data set including at least the composite optical metrology data to characterize the device using the measured optical metrology data from the device.

[0016] In one specific implementation, a computer system for characterizing a device on a sample includes: means for obtaining measured optical metrology data for a reference device for the device; and means for generating composite optical metrology data, wherein each composite optical metrology data is generated by combining the measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device. The computer system also includes means for training a machine learning model using a training data set including at least the composite optical metrology data to characterize the device using the measured optical metrology data from the device. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic diagram of a metrology apparatus is shown that can be used to generate metrology data to produce reference data for training a machine learning model and / or to generate experimental data to characterize parameters of a device under test structure using a trained machine learning model, as described herein.

[0018] Figure 2 A workflow is shown for offline recipe creation and online inference based on a training dataset including composite metrology data as discussed herein.

[0019] Figure 3 is a diagram showing the parameter space for metrology using machine learning.

[0020] Figure 4A A workflow for generating composite metrology data based on merging measured metrology data and synthetic metrology data to be included in training data is shown.

[0021] Figure 4BA workflow for generating composite metrology data by combining synthetic metrology data calculated for a modified reference device with the changes between measured metrology data from a reference device and the synthetic metrology data calculated for the reference device is shown.

[0022] Figure 4C Another workflow for generating composite metrology data by combining synthetic metrology data calculated for a modified reference device with changes between measured metrology data from a reference device and synthetic metrology data calculated for the reference device is shown.

[0023] Figure 4D A workflow for generating composite metrology data by combining measured metrology data from a reference device with synthetic metrology data calculated for the reference device and the variation between synthetic metrology data calculated for a modified reference device is shown.

[0024] Figure 4E Another workflow for generating composite metrology data by combining measured metrology data from a reference device with synthetic metrology data calculated for the reference device and the variation between synthetic metrology data calculated for a modified reference device is shown.

[0025] Figure 5 A workflow for generating composite reference data based on merging parameter values ​​from a reference device and parameter values ​​from a model of the reference device and a modified reference device is shown.

[0026] Fig. 6A and Figure 6B Graphs showing a comparison of composite spectral data and measured spectral data for different Miller matrix elements are shown.

[0027] Fig. 7A and Figure 7B is a graph showing the training and testing of a machine learning model trained using only the measured spectra.

[0028] Fig. 8A and Figure 8B is a graph showing the training and testing of a machine learning model trained using composite metrology spectra and measured spectra.

[0029] Fig. 9A and Fig. 9B An example of a device structure having non-critical parameters that are strongly correlated with the critical parameters to be characterized of the device structure is shown.

[0030] Fig. 10A Graphs showing sample and modeled spectra for a number of Miller matrix elements are shown.

[0031] Fig. 10BA graph showing signals from off-diagonal Miller matrix components corresponding to asymmetric parameters is shown.

[0032] Fig.11 An illustrative flow chart depicting example operations for supporting characterization of a device on a sample according to implementations described herein is shown.

[0033] Fig.12 An illustrative flow chart depicting example operations for supporting characterization of a device on a sample according to implementations described herein is shown.

[0034] Fig.13 An illustrative flow chart depicting example operations for supporting characterization of a device on a sample according to implementations described herein is shown. DETAILED DESCRIPTION

[0035] During the manufacture of semiconductor devices and similar equipment, it is often necessary to monitor the manufacturing process by non-destructively measuring these devices. One type of metrology that can be used to non-destructively measure samples during processing is optical metrology, which can use a single wavelength or multiple wavelengths and can include, for example, ellipsometry, reflectometry, Fourier transform infrared spectroscopy (FTIR), etc. Other types of metrology can also be used, including X-ray metrology, photoacoustic metrology, electron beam (E-beam) metrology, etc.

[0036] Optical metrology (such as thin film metrology and optical critical dimension (OCD) metrology) and other types of metrology sometimes use physical modeling of device structures (e.g., sometimes referred to herein as OCD modeling or using an OCD model, or simply modeling or using a model). Simulated metrology data (e.g., simulated optical data) can be calculated for the physical model using rigorous coupled wave analysis (RCWA), finite difference time domain (FDTD), or finite element method (FEM), or other similar techniques. Variable parameters in the physical model, such as layer thickness, line width, spacer width, sidewall angle, material properties, etc., can be adjusted, and simulated data calculated for each change. The data calculated from the device structure can be compared with the simulated data for each parameter change (e.g., in a nonlinear regression process) until a good fit is achieved, at which point the values ​​of the fitted parameters are determined to be accurate representations of the parameters of the device structure. However, techniques using physical modeling have a high computational cost due to the calculations required to simulate the metrology data. As device structures become more complex, such techniques become less useful due to slow solution times and limitations on model accuracy.

[0037] Machine learning is another technique that can be used for the measurement of various parameters of equipment structures for process improvement, monitoring and control. Unlike modeling, machine learning does not use physical models. Instead, training data is obtained from reference samples and used to train machine learning models. The training data may include, for example, spectral signals and values ​​of structural parameters of interest from reference samples. The machine learning model is automatically "trained" based on the training data to find relevant spectral features and learn the inherent relationships and connections between input features and output features in order to make decisions and predictions on new data.

[0038] One of the challenges for metrology using machine learning is the robustness of machine learning training. For example, there is typically only a limited amount of reference data available for training a machine learning model, which inherently limits the predictions that can be generated by the machine learning model. For example, training data can be obtained based on measured data from at least one reference device. However, the measured data is limited to the number of available reference devices, and therefore, there are practical limits on the variation of parameters from the reference devices that can be present in the training data acquired in this manner. In addition, manufacturing process modifications may occur over time, for example, when developing device structures, which may reduce the relevance of training data obtained from earlier measurements. Although additional training data can be obtained by generating and measuring additional reference devices based on process modifications and retraining machine learning models using newly acquired training data, this process is inefficient because the measurement of reference metrology data is expensive and retraining is time-consuming. In addition, the process of obtaining new training data and retraining the machine learning model may need to be performed iteratively for each process modification.

[0039] Another challenge is that some key parameters to be measured (such as stacking) may have low sensitivity. Therefore, problems with training data may occur due to signal size and / or system noise of key parameters in the training data set. In addition, non-critical parameters (e.g., with stronger signals) may be highly correlated with key parameters. For example, stacking destroys the structural symmetry in the device, resulting in spectral responses in the non-diagonal components of the Miller matrix; however, the amplitudes of these non-diagonal components are low. In addition, when there is structural asymmetry, non-critical parameters such as tilt can be highly correlated with the stacking. Therefore, it may be challenging to measure the stacking using conventional spectral fitting with only an OCD model. Although machine learning can be used for measurement, the robustness of the machine learning recipe is closely related to the experimental design (DOE) and the quality and quantity of references for the training data set. For example, common failure modes are related to process variations. When process variations are not included in the DOE and the affected parameters have a strong correlation with the stacking, the machine learning recipe will perform poorly by attributing the spectral response of such process variations to the stacking, resulting in inaccurate predictions and requiring recipe rework.

[0040] Training data can be augmented using synthetic (simulated) data calculated from a physical model. For example, physical modeling techniques (such as OCD modeling) can be used to generate synthetic metrology data for changes in at least one parameter in the physical model, which can be used to increase the training sample parameter space to obtain better process variation coverage. However, synthetic metrology data has limited practicality as training data. For example, synthetic training data cannot be easily directly adopted for use in applications that utilize key parameters with low sensitivity, such as stacking measurements. For example, in applications with relatively small key parameter signals, physical models need to be extremely accurate to generate useful synthetic metrology data. In addition, the measured data contains system noise, which can change with the system, wafer, environment, and measurement time. It is difficult to determine whether the synthetic metrology data generated by the physical model that simulates the key parameters is accurate or overfits the system noise. In addition, the synthetic metrology data does not include system noise. Therefore, when the synthetic metrology data and the measured data are collected together separately as training data, the machine learning model may not extract the correct measurement information from both and may treat one as an outlier.

[0041] As discussed herein, in order to overcome the limitation of using only measured data or a collection of measured data and separate synthetic metrology data as training data, composite metrology data may be generated and used as training data. Composite metrology data is a mixture of measured metrology data and synthetic metrology data. Therefore, instead of collecting only measured metrology data and synthetic metrology data separately into a training data set, the measured metrology data and the synthetic metrology data are merged together to form composite metrology data. Composite metrology data may be included in a training data set. In some specific implementations, composite metrology data and measured metrology data may be included in a training data set. For reference purposes, the metrology data used to generate composite metrology data is sometimes described herein as optical data or, in particular, spectral data, but it should be understood that other types of metrology data, such as X-ray data, photoacoustic data, and electron beam data, may be used.

[0042] Composite metrology data is generated using measured metrology data from a reference device, a first set of composite metrology data from a first model of the reference device (e.g., an accurate model after being fitted to the measured metrology data), and a second set of composite metrology data from a modified second model of the reference device (e.g., at least one parameter of the first model is changed to produce the modified second model of the reference device). For example, in some implementations, the second set of composite metrology data may be modified based on changes between the measured metrology data and the first set of composite metrology data, while in other implementations, the measured metrology data may be modified based on changes between the first set of composite metrology data and the second set of metrology data.

[0043] In some implementations, at least one parameter of the first model that is changed to produce the second model of the modified reference device can include a key parameter, which will increase the parameter space for the training data set. Composite reference data associated with the composite metrology data can be generated, for example, by merging the values ​​of the reference parameter, the fitted parameter, and the changed parameter, and used as a label in the training data set.

[0044] In some implementations, at least one parameter of the first model that is changed to generate the second model of the modified reference device can be a non-critical parameter. The non-critical parameter can be, for example, a parameter that is strongly correlated with the critical parameter. Thus, composite metrology data can be generated based on the measured metrology data and the synthetic metrology data generated from the changes in the highly correlated non-critical parameters. The composite metrology data can be associated with reference values ​​for the critical parameters that are labels in the training data set.

[0045] Therefore, by using composite metrology data, machine learning recipes are able to tolerate a wide range of process variations without the need to prepare actual wafers with extensive process coverage for machine learning recipe creation.

[0046] By way of example, Figure 1 A schematic diagram of a metrology apparatus 100 is shown, which can be used to generate metrology data to produce reference data for training a machine learning model and / or to produce experimental data to characterize parameters of a device structure under test using a trained machine learning model, as described herein. The metrology apparatus 100 is shown as an optical metrology apparatus, but other types of metrology apparatuses may be used, such as an X-ray metrology apparatus, a photoacoustic metrology apparatus, or an electron beam metrology apparatus. The optical metrology apparatus 100 may be configured to perform, for example, spectroscopic reflectometry, spectroscopic ellipsometry (including Miller matrix ellipsometry), spectroscopic scatterometry, stack scatterometry, interferometry, or FTIR measurement of a sample 101. The sample 101 may, for example, include at least one device structure to be measured. It should be understood that the optical metrology apparatus 100 is shown as an example of a metrology apparatus, and other metrology apparatuses may be used if desired, including normal incidence apparatuses, non-polarized apparatuses, etc.

[0047] The metrology device 100 includes a source 110 that generates radiation that is incident on a sample. Figure 1The metrology device 100 in FIG. 1 is shown as an optical metrology device 100, and thus the source is a light source 110 that produces light 102, but other types of metrology devices may use sources that produce other types of radiation (e.g., X-rays or electron beams). The light 102 produced by the light source 110 may include a range of wavelengths, i.e., a continuous range or multiple discrete wavelengths, such as UV visible light with wavelengths such as between 200 nm and 1000 nm; or may be a single wavelength. The optical metrology device 100 includes focusing optics 120 and 130 that focus and receive the light and direct the light so as to be obliquely incident on the top surface of the sample 101. The optics 120, 130 may be refractive, reflective, or a combination thereof, and may be an objective lens.

[0048] The reflected light may be focused by lens 114 and received by detector 150. Detector 150 may be a conventional charge coupled device (CCD), photodiode array, CMOS or similar type of detector. If broadband light is used, detector 150 may be, for example, a spectrometer, and detector 150 may generate a spectral signal as a function of wavelength. A spectrometer may be used to disperse the full spectrum of received light into spectral components across the detector pixel array. One or more polarization elements may be located in the beam path of optical metrology apparatus 100. For example, optical metrology apparatus 100 may include one or both (or neither) of one or more polarization elements 104 in the beam path before sample 101 and a polarization element (analyzer) 112 in the beam path after sample 101, and may include one or more additional elements 105a and 105b, such as compensators or photoelastic modulators, which may be before, after, or both before and after sample 101. By using a spectroscopic ellipsometer using a dual rotating compensator between polarization elements 104 and 112 and the sample, a full Miller matrix may be measured.

[0049] The optical metrology apparatus 100 further includes one or more computing systems 160 configured to perform measurement of at least one parameter of the sample 101 using the methods described herein. The one or more computing systems 160 are coupled to the detector 150 to receive metrology data acquired by the detector 150 during measurement of the structure of the sample 101. The acquisition of data may be to generate reference data from one or more reference devices for training a machine learning model and / or to generate experimental data from the device under test to characterize at least one parameter of the device. The one or more computing systems 160 may be, for example, a workstation, a personal computer, a central processing unit, or other suitable computer system, or multiple systems. The one or more computing systems 160 may be configured to perform optical metrology based on spectral processing or based on processing any other desired type of optical metrology data or other metrology data (e.g., according to the methods described herein).

[0050] It should be understood that the one or more computing systems 160 may be a single computer system or multiple separate or linked computer systems, which are interchangeably referred to herein as computing system 160, at least one computing system 160, or one or more computing systems 160. The computing system 160 may be included in the optical metrology device 100, or connected to the optical metrology device or otherwise associated with the optical metrology device. Different subsystems of the optical metrology device 100 may each include a computing system that is configured to perform steps associated with the associated subsystem. For example, the computing system 160 may control the positioning of the sample 101, for example, by controlling the movement of a stage 109 coupled to the chuck. The stage 109 may, for example, be capable of horizontal movement in Cartesian (i.e., X and Y) coordinates or polar (i.e., R and θ) coordinates, or some combination of the two. The stage may also be capable of vertical movement along the Z coordinate. The computing system 160 may further control the operation of the chuck 108 to hold or release the sample 101. The computing system 160 may further control or monitor the rotation of one or more polarization elements 104, 112 or elements 105a and 105b, which may be compensators or photoelastic modulators, among others.

[0051] The computing system 160 may be communicatively coupled to the detector 150 in any manner known in the art. For example, one or more computing systems 160 may be coupled to a separate computing system associated with the detector 150. The computing system 160 may be configured to receive and / or obtain metrology data or information from one or more subsystems of the optical metrology device 100, such as the detector 150, and the controller polarization elements 104, 112 and elements 105a, 105b, etc., via a transmission medium that may include wired and / or wireless portions. Thus, the transmission medium may serve as a data link between the computing system 160 and other subsystems of the optical metrology device 100.

[0052] The computing system 160 includes at least one processor 162 and a memory 164, and a user interface (UI) 168 communicatively coupled via a bus 161. The memory 164 or other non-transitory computer-usable storage medium includes computer-readable program code 166 embodied therein and can be used by the computing system 160 to cause one or more computing systems 160 to control the optical metrology device 100 and perform functions including generating composite metrology data, using composite metrology data to train a machine learning model, or using a machine learning model trained with composite metrology data to characterize parameters of the device structure, as described herein. For example, as shown, the memory 164 may include instructions for causing the processor 162 to perform both physical modeling and machine learning (ML), as discussed herein. In view of the present disclosure, a person of ordinary skill in the art may implement the data structures and software codes for automatically implementing one or more actions described in this detailed description, and these data structures and software codes are stored on a computer-usable storage medium such as the memory 164, which may be any device or medium that can store code and / or data for use by a computer system such as the computing system 160. Computer usable storage media may be, but are not limited to, read-only memory, random access memory, magnetic and optical storage devices such as disk drives, magnetic tape, etc. In addition, the functions described herein may be embodied in whole or in part within the circuitry of an application specific integrated circuit (ASIC) or a programmable logic device (PLD), and these functions may be embodied in a computer-understandable descriptor language that may be used to create an ASIC or PLD that operates as described herein.

[0053] The computing system 160 may be configured, for example, to obtain optical metrology data from a plurality of measurement sites on one or more samples (e.g., calibration samples or training samples). Each measurement site includes a device structure, and the optical metrology data may include a spectral signal or other types of optical metrology data. The computing system 160 may be configured to generate a metrology solution for optical measurement of device structures as discussed herein, including using modeling and machine learning to generate composite metrology data for training a machine learning model. In some implementations, different computing systems and / or different optical metrology devices may be used to obtain reference data from training samples, generate composite metrology data, use composite metrology data to train a machine learning model, or use a machine learning model trained with composite metrology data to characterize at least one parameter of the device structure under test. The reference data, the training data set including the composite metrology data, or the trained machine learning model may be provided to the computing system 160, for example, via a computer-readable program code 166 on a non-transitory computer-usable storage medium such as a memory 164.

[0054] Reference data, training data sets including composite metrology data, or trained machine learning models may be stored in memory 164. Further results of the analysis of the data (e.g., parameters to characterize the structure of the device under test) may be reported, such as stored in memory 164 associated with sample 101 and / or indicated to a user via UI 168, an alarm, or other output device. In addition, results from the analysis may be reported and fed forward or fed back to processing equipment to adjust appropriate manufacturing steps to compensate for any detected differences in the manufacturing process. For example, computing system 160 may include communication port 169, which may be any type of communication connection, such as a communication connection to the Internet or any other computer network. Communication port 169 may be used to receive instructions for programming computing system 160 to perform any one or more of the functions described herein and / or to output, for example, a signal with measurement results and / or instructions to another system such as an external process tool in a feedforward or feedback process to adjust process parameters associated with the manufacturing process steps of the sample based on the measurement results.

[0055] By way of example, Figure 2 A workflow 200 for offline recipe creation and online inference based on a training data set as discussed herein is shown. As shown, during a learning phase 210, a training set 212 of metrology data is provided. The training set 212 of metrology data may, for example, include composite metrology data generated based on a combination of theoretically calculated metrology data (synthetic metrology data) and experimentally measured metrology data (measured metrology data) obtained from one or more reference structures using an optical metrology device 100, as discussed herein. For example, each composite metrology data is a combination of measured metrology data and synthetic metrology data. The synthetic metrology data may cover a wider range of process condition variations than the measured metrology data. The training set 212 of metrology data may also include experimentally measured data, such as metrology data measured from one or more reference structures. At least a portion of the training set 212 of metrology data may be labeled based on key parameter reference values.

[0056] As shown, training 214 is performed based on training set 212 of metrology data. For example, training can include feature extraction, regression, classification, and other techniques to generate machine learning model 216. For example, training 214 can use any desired machine learning algorithm that can use composite metrology data as at least a portion of training set 212 of metrology data. For example, the machine learning algorithm can be a supervised learning algorithm or an unsupervised learning algorithm.

[0057] As discussed herein, a machine learning model 216 trained using a training data set including composite metrology data may be used for measurement of at least one parameter of a device under test. By using composite metrology data to train the machine learning model 216, the machine learning model 216 may be better able to tolerate a variety of process variations than only measured metrology data, only pure synthetic metrology data, or a combination of measured metrology data and pure synthetic metrology data in the training set 212 of metrology data. In some implementations, at least one parameter that may be measured using the machine learning model 216 may include a critical parameter (such as stackup) having low sensitivity because the composite metrology data reduces or eliminates issues attributable to signal size of the critical parameter and system noise in the training data set.

[0058] During the inference phase 220, for example using Figure 1 The optical metrology apparatus 100 is shown to obtain test data 222 (e.g., optical metrology data) from a device structure under test. The test data 222 may, for example, be the same type of metrology data used in the training set 212 of metrology data. Using the test data 222 as input data, a trained machine learning model 216 may be used to infer or predict (224) and output 226 the value of a desired feature, such as a parameter of interest of the device structure under test. Because the machine learning model 216 is trained on composite metrology data that is exposed to a wider range of process condition variations, if the composite metrology data better represents the experimental metrology data after a process modification, the model predictions may have better metrology performance in the event of a process modification.

[0059] Figure 3 300 is a diagram illustrating a parameter space for metrology using machine learning. The parameter space shown in diagram 300 includes two parameters, parameter 1 along the X-axis and parameter 2 along the Y-axis. As shown using the black circles in diagram 300, a reference target having parameter variations is measured to produce a plurality of training samples 312 that define a training range 310 in the parameter space. Measurements of a device under test having parameters within the training range 310 will be relatively accurate, but the accuracy decreases when the parameters are outside the training range 310. For example, as shown using the white circles, a test sample 322 generated by measuring a device having parameters outside the training range 310 will result in reduced accuracy. For example, a device having parameters outside the training range 310 may be produced due to manufacturing process variations that occur after the initial training samples 312 are obtained.

[0060] In order to expand the training range, additional training data can be obtained. As shown in the gray circle, a composite training sample 332 can be obtained, thereby generating an extended training range 330. As shown, the extended training range 330 includes the test sample 322, and is therefore sufficient to generate an acceptable accuracy for the test sample 322. Each composite training sample 332 is generated based on a combination of measured metrology data and synthetic metrology data. Synthetic metrology data can be generated, for example, by including some disturbances to reference data (e.g., training sample 312). The synthetic metrology data is then combined with the reference data (e.g., training sample 312) to generate a composite training sample 332 with parameters outside the training range 310. Therefore, the parameter space defined by the composite training sample is extended from the initial training range 310.

[0061] The machine learning model may be trained using composite data, i.e., a training data set that includes both the training range 310 and the extended training range 330, and is therefore exposed to a wider variation in process conditions than if trained only on the training range 310. Therefore, in the event of a process modification, predictions from the trained machine learning model may have better metrological performance, for example, if the composite metrological data is a better representation of the data measured after the process modification.

[0062] Figure 4A A workflow 400 is shown for generating composite metrology data based on merging measured metrology data and synthetic metrology data to be included in training data. Composite metrology data is generated based on a combination of measured metrology data and synthetic metrology data. For reference purposes, metrology data discussed herein (e.g., composite metrology data, measured metrology data, and synthetic metrology data) are sometimes described as optical metrology data and, in particular, spectral data, but it should be understood that other types of metrology data may be used.

[0063] As shown at block 410, measured metrology data is obtained from a reference device having a first set of parameters that may represent process conditions A1 (experimental spectrum A1). The reference device may, for example, be generated using the same manufacturing process as the device under test (process conditions A1), but may have at least one parameter intentionally changed from its nominal value. The measured metrology data from block 410 may, for example, be obtained by Figure 1 The optical metrology apparatus 100 shown is obtained from a reference device.

[0064] At box 420, the measured metrology data (experimental spectrum A1) is fitted to the physical model. For example, a first model (e.g., an OCD model) may be generated based on a value of at least one parameter for a reference device, which value may be known, for example, based on a CD-SEM or other similar type of measurement of a reference structure or other similar reference structure. The calculated metrology data is generated for the physical model using a modeling technique (such as RCWA, FDTD, FEM, etc.), which is fitted to the measured metrology data. For example, various parameters of the physical model may be adjusted, and the calculated metrology data generated for each change is compared to the measured metrology data (e.g., in a nonlinear regression process) until a good fit is achieved. Once a good fit is achieved, the values ​​of the fitted parameters for the physical model are considered to be accurate representations of the parameters of the reference device.

[0065] At block 422 , calculated metrology data corresponding to the best fit to the measured metrology data from block 420 is generated as a first set of synthetic metrology data for the reference device, where the first set of parameters represents process condition A1 (eg, calculated spectrum A1 ).

[0066] In addition, as shown at block 430, a second model with changes relative to the first model is generated to simulate the modified reference device. For example, at least one parameter of the first model from block 420 may be changed to generate the second model. The change to at least one parameter for the second model may, for example, be greater than any change in the change of the parameter found in the reference device representing process condition A1. In some implementations, at least one parameter changed for the second model may be a geometric parameter and an optical constant that is changed to mimic the change due to the expected modified process condition B1, which may be obtained from user input or through process simulation as shown at block 425.

[0067] For example, OCD model parameters that are susceptible to changes due to changes in process conditions may be varied by perturbations that may be defined for parameter P as follows:

[0068] P Perturbed =P OCD_Fit + random_perturbation(P OCD_Fit ) Equation 1

[0069] Where P Perturbed is the perturbed parameter P, P OCD_Fit is the result for the parameter P after fitting the OCD model to the experimental spectrum A1 (at block 420). OCD_Fit The random perturbation of the function can be defined as follows:

[0070] random_perturbation(P OCD_Fit )=B+factor*std_dev(P OCD_Fit )*rand_num Equation 2

[0071] Where B corresponds to the bias term incorporating the experimental design (DOE) conditions, factor is the multiplicative factor controlling the variance of the random disturbance and can be, for example, between 0.1 and 10, std_dev is the standard deviation and rand_num is a random number that can range, for example, between -0.5 and 0.5, but any desired range can be used, such as between -1 and 1, between 0.1 and 10, etc.

[0072] The one or more parameters that are changed can be critical parameters, non-critical parameters, or both critical parameters and non-critical parameters. In some implementations, the critical parameters of the model can remain static, and only the non-critical parameters in the physical model can be changed. For example, if the critical parameter is stackup, the same stackup value can be used in the first model and the second model, but other non-critical parameters can be changed. Non-critical parameters can be strongly correlated with critical parameters. For example, in some device structures, a bit line tilt parameter in the structure can be strongly correlated with the stackup, such as a change in tilt produces a change in optical metrology data similar to a change in the stackup. Other examples of parameters that can be related to the stackup can be layer thickness or critical dimension.

[0073] At box 432, using a modeling technique (such as RCWA, FDTD, FEM, etc.), the calculated metrology data can be generated for the second model as a second set of synthetic metrology data for the modified reference device having a second set of parameters, which can represent process condition B1 (e.g., calculated spectrum B1).

[0074] As shown, based on the combination of the measured metrology data (experimental spectrum A1) from box 410, the first set of synthetic metrology data (calculated spectrum A1) from box 422, and the second set of synthetic metrology data (calculated spectrum B1) from box 432, composite metrology data (composite spectrum) can be generated. By way of example, in some specific implementations, the change between the measured metrology data (experimental spectrum A1) and the first set of synthetic metrology data (calculated spectrum A1) can be determined, and the change can then be combined with the second set of synthetic metrology data (calculated spectrum B1) to generate composite metrology data (composite spectrum). The change can be, for example, a mismatch between the measured metrology data and the synthetic metrology data, or it can be a spectral transformation between the measured metrology data and the synthetic metrology data. In other specific implementations, the change between the first set of synthetic metrology data (calculated spectrum A1) and the second set of synthetic metrology data (calculated spectrum B1) can be determined, and the change can then be combined with the measured metrology data (experimental spectrum A1) to generate composite metrology data (composite spectrum). The change between the first set of composite meter data and the second set of composite meter data may be, for example, the difference between the composite meter data.

[0075] Thus, the resulting composite metrology data at block 450 is similar to the measured metrology data from block 410, but includes changes in the parameter space to expand the training range, e.g., Figure 3 As shown in the expanded training range 330 in .

[0076] The method may be performed for each individual measured metrology data (e.g., measured metrology data obtained from different reference devices with different parameter values). Figure 4A The process of generating composite metrology data shown in FIG. 1 can be repeated for different changes in parameters to generate multiple composite metrology data. The composite metrology data and the measured metrology data in some specific implementations can be combined to form a training set 212 of metrology data (e.g., Figure 2 shown).

[0077] Thus, the composite metrology data is based on measured metrology data, but also includes simulated changes in metrology data introduced by changes in at least one parameter. Figure 2 ) to expand the parameter space, thereby increasing practicality. In addition, the synthetic information combined with the measured metrology data can be used to help the machine learning model break the metrology data correlation of non-critical parameters, but without creating any new critical parameter data. Therefore, the resulting machine learning model 216 trained using the composite metrology data will be robust and able to tolerate a variety of process changes without requiring a large amount of measured reference data.

[0078] Figure 4BA workflow 402 for a specific implementation of generating composite metrology data based on a merging of measured metrology data and synthetic metrology data to produce training data is shown. The workflow 402 is similar to Figure 4A As shown in the workflow 400, similar designated elements are the same.

[0079] Similar to Figure 4A In the illustrated workflow 400, at box 410, measured metrology data is obtained from a reference device having a first set of parameters, which may represent process condition A1 (experimental spectrum A1). At box 420, the measured metrology data (experimental spectrum A1) is fitted to a physical model, and at box 422, calculated metrology data corresponding to the best fit to the measured metrology data is generated as a first set of synthetic metrology data for the reference device, wherein the first set of parameters represents process condition A1 (e.g., calculated spectrum A1).

[0080] At box 430, a second model simulating the modified reference device is generated (in which one or more parameters are changed relative to the parameters used in the first model from box 420), and at box 432, calculated metrology data is generated for the second model as a second set of synthetic metrology data for the modified reference device having a second set of parameters, which second set of parameters may represent process condition B1 (e.g., calculated spectrum B1).

[0081] like Figure 4B As shown at block 440, the variation between the measured metrology data (experimental spectrum A1) at block 410 and the first set of synthetic metrology data (calculated spectrum A1) from block 422 may be determined as a mismatch between the measured metrology data (experimental spectrum A1) and the first set of synthetic metrology data (calculated spectrum A1). For example, the mismatch spectrum may be determined as:

[0082] Mismatch spectrum = experimental spectrum A1 - OCD best fit spectrum (process condition A1) Equation 3 where the OCD best fit spectrum is the calculated spectrum A1.

[0083] The mismatch from block 430 , which is based on the measured metrology data from block 410 and the first set of composite metrology data from block 422 , may be added to the second set of composite metrology data from block 432 to produce composite metrology data representing process condition B1 at block 450 .

[0084] Figure 4C A workflow 404 for a specific implementation of generating composite metrology data based on a combined combination of measured metrology data and synthetic metrology data to produce training data is shown. The workflow 404 is similar to Figure 4A As shown in the workflow 400, similar designated elements are the same.

[0085] Similar to Figure 4A In the illustrated workflow 400, at box 410, measured metrology data is obtained from a reference device having a first set of parameters, which may represent process condition A1 (experimental spectrum A1). At box 420, the measured metrology data (experimental spectrum A1) is fitted to a physical model, and at box 422, calculated metrology data corresponding to the best fit to the measured metrology data is generated as a first set of synthetic metrology data for the reference device, wherein the first set of parameters represents process condition A1 (e.g., calculated spectrum A1).

[0086] At box 430, a second model simulating the modified reference device is generated (in which one or more parameters are changed relative to the parameters used in the first model from box 420), and at box 432, calculated metrology data is generated for the second model as a second set of synthetic metrology data for the modified reference device having a second set of parameters, which second set of parameters may represent process condition B1 (e.g., calculated spectrum B1).

[0087] like Figure 4C As shown at box 442, the change between the measured metrology data (experimental spectrum A1) at box 410 and the first set of synthetic metrology data (calculated spectrum A1) from box 422 can be determined as a spectral transformation between the measured metrology data (experimental spectrum A1) and the first set of synthetic metrology data (calculated spectrum A1). For example, a mathematical model is generated to convert the first set of synthetic metrology data (calculated spectrum A1) to the measured metrology data (experimental spectrum A1) for process condition A1.

[0088] The transformation determined in block 442 (which is based on the measured metrology data from block 410 and the first set of composite metrology data from block 422 ) may then be applied to the second set of composite metrology data from block 432 to produce composite metrology data representative of process condition B1 at block 450 .

[0089] Figure 4D A workflow 406 for a specific implementation of generating composite metrology data based on combining measured metrology data and synthetic metrology data to produce training data is shown. The workflow 406 is similar to Figure 4A As shown in the workflow 400, similar designated elements are the same.

[0090] Similar to Figure 4AIn the illustrated workflow 400, at box 410, measured metrology data is obtained from a reference device having a first set of parameters, which may represent process condition A1 (experimental spectrum A1). At box 420, the measured metrology data (experimental spectrum A1) is fitted to a physical model, and at box 422, calculated metrology data corresponding to the best fit to the measured metrology data is generated as a first set of synthetic metrology data for the reference device, wherein the first set of parameters represents process condition A1 (e.g., calculated spectrum A1).

[0091] At box 430, a second model simulating the modified reference device is generated (in which one or more parameters are changed relative to the parameters used in the first model from box 420), and at box 432, calculated metrology data is generated for the second model as a second set of synthetic metrology data for the modified reference device having a second set of parameters, which second set of parameters may represent process condition B1 (e.g., calculated spectrum B1).

[0092] like Figure 4D As shown at block 444, the change between the first set of composite metrology data (the calculated spectrum A1) from block 422 and the second set of composite metrology data (the calculated spectrum B1) from block 432 may be determined as a spectral difference. For example, the spectral difference may be determined as:

[0093] Spectral difference = OCD best fit spectrum (process condition A1) - OCD spectrum (process condition B1) Equation 4 where the OCD best fit spectrum is the calculated spectrum A1, and the OCD spectrum is the calculated spectrum B1.

[0094] The spectral differences from block 444 (which are based on the first set of composite metrology data from block 422 and the second set of composite metrology data from block 432 ) may be added to the measured metrology data from block 410 to produce composite metrology data representative of process condition B1 at block 450 .

[0095] Figure 4E Shows Figure 4D A more detailed view of a workflow 406 for a specific implementation of generating composite metrology data based on merging measured metrology data and synthetic metrology data to produce training data is shown, wherein spectral differences between a first set of synthetic metrology data and a second set of synthetic metrology data are used.

[0096] like Figure 4E As shown, a first set of synthetic metrology data (calculated metrology data) is obtained at block 422. The calculated metrology data from block 422 may be obtained, for example, based on fitting the measured metrology data from block 410 to a first physical model (e.g., Figure 4D420 in FIG. 4 ). A physical model (e.g., an OCD model) may be generated based on values ​​of key parameters (and any other parameters) for a reference structure, which values ​​may be known, for example, based on CD-SEM or other similar types of measurements of the reference structure. Calculated metrology data is generated for the physical model and fit to the measured data in a nonlinear regression process until a good fit is achieved, at which point the values ​​of the fitted parameters for the physical model are determined to be accurate representations of the parameters of the reference structure.

[0097] As shown, at blocks 432a, 432b, 432c, and 432d (sometimes collectively referred to as block 432), a second physical model (e.g., Figure 4D A second set of synthetic metrology data (the calculated metrology data) is generated based on changes in one or more non-critical parameters in the first physical model (shown in box 430 in FIG. 1 ), i.e., the same reference value is used for the critical parameter used in the first model. For example, if the critical parameter is stackup, the same stackup value is used in the first physical model in box 422 to generate the calculated metrology data, and the same stackup value is used in the second physical model in box 432 to generate the calculated optical metrology data. The calculated metrology data in box 432 may be generated for the second physical model using the same modeling technique (e.g., RCWA, FDTD, FEM, etc.) used to generate the calculated metrology data in box 422. In some specific implementations, the non-critical parameters that are changed in box 432 to generate the calculated metrology data may be strongly correlated with the critical parameters. As in the examples discussed above, in some device structures, a bit line tilt parameter in the structure may be strongly correlated with the stackup, for example, a change in tilt produces a change in metrology data similar to a change in stackup. Other examples of parameters that may be related to stackup may be layer thickness or critical dimension. For example, Figure 4E Calculated metrology data determined for four changes in non-critical parameters having values ​​of -0.2, -0.1, +0.1, and +0.2, respectively, are shown from blocks 432a, 432b, 432c, and 432d. It should be understood that additional or fewer changes may be used, and that the values ​​of the changes may be different and may depend on the type of non-critical parameter being changed and the number of changes.

[0098] The differences between the first set of composite metrology data from the first model (from block 422) and the second set of composite metrology data from block 432 are calculated to generate differences (I, II, III, and IV) at blocks 444a, 444b, 444c, and 444d (sometimes collectively referred to as blocks 444). The differences determined at block 444 represent changes in the metrology data due to changes in non-critical parameters (where the values ​​of the critical parameters are fixed).

[0099] As shown, the measured metrology data from block 410 may be modified based on the differences from block 444 to generate composite metrology data for each change in a non-critical parameter at blocks 450a, 450b, 450c, and 450d (sometimes collectively referred to as block 450). For example, each of the differences from block 444 may be added to the measured metrology data from block 410 that was used to fit to produce the calculated metrology data from block 422 and that has the same reference value for the critical parameter as used in the first set of composite metrology data from block 422 and the second set of composite metrology data from block 432. Thus, the resulting composite metrology data in block 450 is similar to the measured metrology data from block 410 in that it has the same reference value for the critical parameter and system noise, but includes changes in metrology data due to changes in the non-critical parameters.

[0100] Figure 4D and Figure 4E The process of generating composite metrology data shown can be performed for each individual measured metrology data (e.g., having different key parameter values) to generate multiple composite metrology data for each measured metrology data. The composite metrology data and, in some implementations, the associated measured metrology data can be combined to form a training set 212 of metrology data (e.g., Figure 2 shown).

[0101] Thus, the composite metrology data is based on the measured metrology data, but includes simulated differences in the metrology data introduced by changes in non-critical parameters, which in some implementations may be highly correlated with the critical parameters. Therefore, the composite metrology having the same critical parameter reference values ​​as the measured metrology data is used to train the machine learning model 216 (e.g., Figure 2 ). Composite metrology data that combines synthetic information with measured information can be used to help the machine learning model break metrology data dependencies for non-critical parameters, but without creating any new critical parameter data. Therefore, the physical model does not need to be fitted for the critical parameters, which reduces or eliminates problems associated with signal sensitivity associated with the critical parameters and the presence of systematic noise in the reference data rather than the synthetic metrology data. Therefore, the resulting machine learning model 216 trained using the composite data will be robust and able to tolerate a variety of process variations without requiring a large amount of measured reference data.

[0102] Figure 5 A workflow 500 for a specific implementation of generating composite reference data based on merging parameter values ​​from a reference device and parameter values ​​from a model of the reference device and a modified reference device is shown. The workflow 500 generates a composite reference data that can be used with training data (e.g., Figure 2The composite reference data included in the training set 212 of the metrological data in the composite metrological data together with the composite reference data, for example, if the reference data are Figure 4A , Figure 4B and Figure 4C The key parameters are changed in the second model as discussed in workflows 400, 402, and 404 in FIG. If the key parameters are not changed in the second model, the composite reference data generated by workflow 500 may not be included in the training data, such as reference data. Figure 4D and Figure 4E As discussed in workflow 406 .

[0103] At block 510, similar to Figure 4A At block 410 in FIG. 1 , the measured metrology data is obtained from a reference device having a first set of reference parameters that may represent process conditions A1 (experimental spectrum A1). Reference parameters from the reference device (reference A1) are shown at block 512. The values ​​for the reference parameters may be known, such as based on other similar types of measurements of a CD-SEM or reference structure or other similar reference structure.

[0104] At box 520, the measured metrology data (experimental spectrum A1) is fitted to the physical model and can be Figure 4B The same process performed at block 420 in FIG. 5 is performed, for example, by fitting the measured metrology data to the calculated metrology data and using a nonlinear regression process to adjust the parameters of the physical model until a good fit is achieved. Once a good fit is achieved, the values ​​of the fitted parameters for the physical model are considered to be an accurate representation of the parameters of the reference device. The resulting parameter values ​​from the fitting process in block 520 are generated as a first set of key parameter values ​​for the first model of the reference device at block 522 (fitted key parameter values ​​for A1).

[0105] As shown at block 530, a second model having changes to the key parameters relative to the first model is generated to simulate a modified reference plant by changing the values ​​of one or more key parameters of the first model from block 520, for example based on the application of a change representing an estimated process variation for the plant, which may be obtained from user input or obtained through a process simulation as shown at block 525, which may represent process condition B1. It should be understood that non-key parameters may also be changed relative to the first model representing possible process variations for the plant. The generation of the second model at block 530 may be performed at Figure 4B The same process performed at block 430 in FIG. At block 532, parameter values ​​for the second model are generated as a second set of key parameter values ​​for the second model (key parameter values ​​for the change of B1).

[0106] At block 540, key parameter deviations are determined based on the first set of key parameter values ​​from block 522 and the second set of key parameter values ​​from block 532. For example, parameter deviations are differences between expected parameter values ​​and actual parameter values. Thus, at block 540, key parameter deviations may be determined as differences between the first set of key parameter values ​​from block 522 and the second set of key parameter values ​​from block 532. Then, at block 550, the key parameter deviations are combined with the reference parameters from block 512 to generate a composite reference. The composite reference data may be used as a label for process B1 in the training data set.

[0107] By way of example, Fig. 6A and Figure 6B is a graph showing a comparison of composite spectral data and measured spectral data for a plurality of Miller matrix elements (M33 and M34, respectively). The "experimental training" curve is an experimental spectrum generated from a reference under process condition A, and the "experimental test" curve is an experimental spectrum generated from a reference under process condition B. The "composite" curve is the composite spectral data generated after applying a change representing a change between process conditions A and B, as discussed herein.

[0108] By way of example, Fig. 7A is a graph showing the performance during training of a machine learning model using only measured reference optical metrology data (spectra) and showing training and testing (validation) of the trained model. Figure 7B Shown from Fig. 7A Blind testing of the machine learning model. As shown in the figure, the blind test resulted in an accuracy of 0.763 (R 2 ), where the slope is 0.609 and the root mean square error (RMSE) is 0.53.

[0109] on the contrary, Fig. 8A is a graph showing the performance during training of a machine learning model using composite optical metrology data (spectra) and measured reference optical metrology data (spectra) and showing training and testing (validation) of the trained model. Figure 8B Shown from Fig. 8A As shown in the figure, blind testing leads to a Figure 7B The higher accuracy shown, where the accuracy (R 2 ) is 0.91, the slope is 0.991, and the root mean square error (RMSE) is 0.33.

[0110] As discussed above, composite metrology data can be used for measurements of critical parameters with low sensitivity, such as stackup, because composite metrology data reduces or eliminates the problem of signal magnitude due to critical parameters and systematic noise in the training data set. For example, stackup destroys structural symmetry, resulting in spectral responses in the non-diagonal components of the Miller matrix; however, the amplitude of those non-diagonal components is low. In addition, when structural asymmetry is present, several parameters can be highly correlated with stackup.

[0111] By way of example, Fig. 9A An example of a device structure 900 having a critical parameter with low sensitivity that can be measured using metrology techniques such as those discussed herein is shown. The device structure 900 is shown as a simplified DRAM gate including a plurality of bit lines 902 aligned with an underlying structure 904. For example, Fig. 9A The process expectations for the stack and the tilt for the bit line 902 are shown in dashed lines. By way of example, Fig. 9B An example of a device structure 950 similar to the device structure 900 but with a non-zero overlay (OVL) error and a non-zero tilt of a bit line 952 is shown. The overlay error OVL, for example, can be a key parameter for characterizing the device structure via optical metrology. The tilt or other parameters of the bit line 952, such as a critical dimension or thickness, can be non-critical parameters but can be strongly correlated with the overlay error OVL. In other words, changes in non-critical parameters, such as tilt, can strongly affect the same optical metrology data used to determine the overlay error OVL, making it difficult to measure the overlay.

[0112] By way of example, Fig. 10A is a graph showing sample and modeled spectra for a number of Miller matrix elements (i.e., M12, M13, M14, M22, M23, M24, M33, M34, and M44) for which a good fit has been achieved using conventional physical modeling techniques to measure thickness and OCD parameters. In contrast, Fig. 10B Asymmetric signals from off-diagonal Miller matrix components corresponding to stacking and tilt are shown. Fig. 10B As can be seen from the figure, the sensitivity of the signal to stacking and tilt is significantly lower than what can be fitted using physical modeling techniques, such as Fig. 10A shown.

[0113] As discussed in this paper, by using composite measurement data, a training dataset can be used to train Figure 2In the machine learning model 216 shown, the training data set includes composite metrology data for measurements of key parameters with low sensitivity (such as stackup) because the composite metrology data reduces or eliminates the problems of signal magnitude and system noise due to the key parameters in the training data set. Therefore, by using composite metrology data, the machine learning model 216 is able to better tolerate a variety of process variations than if pure synthetic metrology data were used with the measured data in the training set 212 of metrology data.

[0114] Fig.11 An illustrative flow chart describing example operations 1100 for supporting characterization of a device on a sample according to some implementations is shown. In some implementations, the example operations 1100 may be performed by a metrology device (such as a computer having one or more processors (e.g., such as a Figure 1 The metering device 100) is executed by the processor 162 in the computing system 160.

[0115] The one or more processors may obtain measured metrology data from a device (1102). For example, the means for obtaining measured metrology data from a device may be a metrology device 100 and a Figure 1 The measured metrology data may be, for example, Figure 2 Test data 222 is shown.

[0116] The one or more processors may determine at least one parameter of the device based on the measured metrology data using a machine learning model using composite metrology data. Each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device (1104), such as for example Figure 2 The learning phase 210 and the inference phase 220 are shown as well as FIG. 4A to FIG. 4E A component for determining at least one parameter of a device based on measured metrology data using a machine learning model using composite metrology data may be, for example, a metrology device 100, the metrology device including a device configured to generate training data by Figure 1 The computer readable program code 166 shown in FIG. 1 is a computing system 160 of a processor 162 executing a machine learning model, wherein each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data for a first model of the reference device. In some optional implementations, as shown in dashed lines, each composite metrology data may include a combination of metrology data and the first synthetic metrology data, and also includes second synthetic metrology data (1106) for a second model of the modified reference device that is changed relative to the first model, such as for example, for Figure 2 The learning phase 210 and the inference phase 220 are shown as well as FIG. 4A to FIG. 4E The generation of training data shown in workflows 400, 402, 404, and 406 in FIG.

[0117] The one or more processors may provide (e.g., report) at least one parameter of the device to characterize the device on the sample. For example, the means for providing at least one parameter of the device to characterize the device on the sample may be the metering device 100, and the Figure 1 The computing system 160 is shown interacting with a processor 162 and a memory 164 as well as a UI 168 .

[0118] In some implementations, the second model of the modified reference device has at least one parameter that is changed relative to the first model, e.g. FIG. 4A to FIG. 4E As discussed in block 430 of .

[0119] In some implementations, the machine learning model can further use composite reference parameters for the modified reference device based on a combination of the reference parameters of the reference device, a first set of key parameter values ​​generated for the first model, and a second set of key parameter values ​​generated for the second model, for example, as in Figure 5 As discussed at blocks 540, 512 and 550 of .

[0120] In some implementations, the first model can be generated by fitting metrology data measured from a reference device to synthetic optical metrology data for the first model, e.g., as in FIG. 4A to FIG. 4E As discussed at box 420 in .

[0121] In some implementations, the modified second model of the reference device can be generated by changing at least one parameter of the first model, for example, as in FIG. 4A to FIG. 4E As discussed at box 430 in .

[0122] In some implementations, second synthetic metrology data for a second model of a modified reference device can be generated based on changes between metrology data measured from the reference device and first synthetic metrology data for the first model, e.g., as described in the reference device. Figure 4A Discussed and in FIG. 4B to FIG. 4C The variation between the metrology data measured from the reference device and the first set of synthetic metrology data for the first model may be, for example, a mismatch between the metrology data and the first synthetic metrology data (e.g., Figure 4B 440 in ) or a spectral transformation between the metrology data and the first synthetic metrology data (as discussed in Figure 4C (as shown in box 442 in FIG. ).

[0123] In some implementations, composite metrology data can be generated by modifying metrology data measured from a reference device with a determined difference between first composite metrology data for a first model of the modified reference device and second composite metrology data for a second model, e.g., as described in the reference Figure 4A Discussed and in FIG. 4D to FIG. 4E As shown at boxes 444 and 450 in.

[0124] In some implementations, the composite metrology data can include multiple sets of composite metrology data for a corresponding plurality of modified reference devices.

[0125] In some implementations, the measured metrology data includes a measured spectrum, and the composite metrology data includes a composite spectrum.

[0126] Fig.12 An illustrative flow chart describing example operations 1200 for supporting characterization of a device on a sample according to some implementations is shown. In some implementations, the example operations 1200 may be performed by a metrology device (such as a computer having at least one processor (e.g., such as a Figure 1 The metering device 100) is executed by the processor 162 in the computing system 160.

[0127] The at least one processor may obtain measured metrology data for a reference device for the device (1202). For example, the means for obtaining measured metrology data for the reference device may be the metrology device 100 and the reference device 100. Figure 1 The measured metrology data may be, for example, reference data having known values ​​of key parameters such as stacking. The measured stacking metrology data may be, for example, from FIG. 4A to FIG. 4E Block 410 and Figure 5 The measured metrology data of block 510 is shown.

[0128] The at least one processor may generate a first set of synthetic metrology data for a first model of a reference device (1204). For example, the first set of synthetic metrology data may be FIG. 4A to FIG. 4E The means for generating a first set of synthetic metrology data for a first model of a reference device may be a metrology device 100, the metrology device comprising Figure 1 Computer readable program code 166 is shown configuring processor 162 of computing system 160 .

[0129] The at least one processor may generate a second set of composite metrology data for a second model of the modified reference device that is changed relative to the first model (1206). For example, the second set of composite metrology data may be FIG. 4A to FIG. 4E A means for generating a second set of synthetic metrology data for a second model of a modified reference device that is changed relative to the first model may be a metrology device 100, the metrology device comprising a Figure 1 Computer readable program code 166 is shown configuring processor 162 of computing system 160 .

[0130] The at least one processor may generate composite metrology data for the modified reference device, each composite metrology data being generated by combining the measured metrology data, the first composite metrology data, and the second composite metrology data (1208). For example, the composite metrology data may be FIG. 4A to FIG. 4E The generated composite metrology data shown in block 450 of FIG. A component for generating composite metrology data for a modified reference device may be a metrology device 100 including a device having a Figure 1 The computer readable program code 166 shown in FIG. 1 configures the processor 162 of the computing system 160 , each composite metrology data is generated by combining the measured metrology data, the first composite metrology data, and the second composite metrology data.

[0131] The at least one processor may store (1210) at least the composite metrology data as a training data set in, for example, Figure 1 The training data set may be, for example, Figure 2 Shown and referenced FIG. 4A to FIG. 4E A training set 212 of the metrology data in question. A means for storing at least the composite metrology data as a training data set may be a metrology device 100 comprising a processor 162 configured by a computer readable program code 166 and Figure 1 The computing system 160 is shown with memory 164 in the computing system 160 .

[0132] In some optional implementations, the at least one processor may further train a machine learning model with a training data set including at least composite metrology data to characterize the device using measured metrology data from the device (1212), such as by using Figure 2 The training set 212 of metrology data shown is used to train 214 the machine learning model 216 shown.

[0133] In some implementations, at least one processor can generate the first model by fitting the measured metrology data to a first set of synthetic metrology data for the first model, e.g., as in FIG. 4A to FIG. 4E As discussed at box 420 in .

[0134] In some implementations, at least one processor may generate a modified second model of the reference device by changing at least one parameter of the first model, for example, as in FIG. 4A to FIG. 4E As discussed at box 430 in .

[0135] In some implementations, at least one processor can generate composite metrology data for a modified reference device by determining a change between the measured metrology data and the first set of composite metrology data, e.g., as a reference Figure 4A The discussed Figure 4B and Figure 4C The at least one processor may further modify the second set of composite metering data using the changes between the measured metering data and the first set of composite metering data, as shown in blocks 440 and 442 of FIG. Figure 4A The discussed Figure 4B and Figure 4C The variation between the measured metrology data and the first set of composite metrology data may be, for example, a mismatch between the measured metrology data and the first set of composite metrology data (e.g., Figure 4B 440 in ) or a spectral transformation between the measured metrology data and the first set of synthetic metrology data (as discussed in Figure 4C (as shown in box 442 in FIG. ).

[0136] In some implementations, at least one processor may generate composite metrology data for a modified reference device by determining a difference between a first set of composite metrology data and a second set of composite metrology data, for example, as shown in FIG. Figure 4A The discussed Figure 4D and Figure 4E The at least one processor may further modify the measured metering data using the difference between the first set of composite metering data and the second set of composite metering data, as shown in block 444 of FIG. Figure 4A The discussed Figure 4D and Figure 4E As shown at boxes 444 and 450 in.

[0137] In some implementations, the at least one processor may further generate multiple sets of composite metrology data for the corresponding multiple modified reference devices, and store the multiple sets of composite metrology data as the training data set.

[0138] The measured metrology data may be, for example, a measured spectrum, and the first set of synthetic metrology data and the second set of synthetic metrology data may be synthetic spectra.

[0139] In some implementations, at least one processor may further generate a first set of key parameter values ​​for a first model of a reference device, for example, as in Figure 5 The at least one processor may further generate a second set of key parameter values ​​for a second model, for example, as discussed in block 522 of Figure 5 The at least one processor may generate composite reference parameters for the modified reference device by combining the reference parameters, the first set of key parameter values, and the second set of key parameter values, for example, as discussed at blocks 512, 540, and 550. The at least one processor may store the composite reference parameters along with the composite metrology data as a training data set, for example, in a Figure 1 For example, the composite reference parameters for the modified reference device may be determined by determining the key parameter deviation between the first set of key parameter values ​​and the second set of key parameter values ​​(e.g., as in Figure 5 and modifying the reference parameter with the key parameter deviation (e.g., as discussed in block 540 of Figure 5 (as discussed in box 550 of ).

[0140] Fig.13 An illustrative flow chart describing example operations 1300 for supporting characterization of a device on a sample according to some implementations is shown. In some implementations, the example operations 1300 may be performed by a metrology device (such as a computer having at least one processor (e.g., such as a Figure 1 The metering device 100) is executed by the processor 162 in the computing system 160.

[0141] The at least one processor may obtain measured metrology data for a reference device for the device (1302). For example, the means for obtaining measured metrology data for the reference device may be the metrology device 100 and the reference device 100. Figure 1 The measured metrology data may be, for example, reference data having known values ​​of key parameters such as stacking. The measured stacking metrology data may be, for example, from FIG. 4A to FIG. 4E Block 410 and Figure 5 The measured metrology data of block 510 is shown.

[0142] The at least one processor may generate composite metrology data, each composite metrology data being generated by combining the measured metrology data for the reference device with first synthetic metrology data of the first model for the reference device (1304). For example, the composite metrology data may be FIG. 4A to FIG. 4E The generated composite metering data shown in block 450 of FIG. In some implementations, the first set of composite metering data may be FIG. 4A to FIG. 4EA component for generating composite metrological data may be a metrological device 100, the metrological device comprising a Figure 1 The computer system 160 of the processor 162 configured with the computer readable program code 166 shown, each composite metrology data is generated by combining the measured metrology data for the reference device with the first synthetic metrology data for the first model of the reference device. In some optional specific implementations, as shown in dashed lines, each composite metrology data may include a combination of the metrology data and the first synthetic metrology data, and also includes second synthetic metrology data (1306) for a second model of the modified reference device that is changed relative to the first model, and the composite metrology data may be for the modified reference device, such as for example Figure 2 The learning phase 210 and the inference phase 220 are shown as well as FIG. 4A to FIG. 4E In some implementations, the second set of synthetic measurement data may be FIG. 4A to FIG. 4E The calculated metering data shown in box 432.

[0143] The at least one processor trains a machine learning model with a training data set including at least composite metrology data to characterize the device using measured metrology data from the device (1308), such as by using Figure 2 A method for training a machine learning model using a training data set including at least composite metrology data to characterize a device using measured metrology data from the device may be, for example, a metrology device 100, the metrology device including a plurality of components ... Figure 1 Computer readable program code 166 is shown configuring processor 162 of computing system 160 .

[0144] In some implementations, at least one processor can generate the first model by fitting the measured metrology data to a first set of synthetic metrology data for the first model, e.g., as in FIG. 4A to FIG. 4E As discussed at box 420 in .

[0145] In some implementations, at least one processor may generate a modified second model of the reference device by changing at least one parameter of the first model, for example, as in FIG. 4A to FIG. 4E As discussed at box 430 in .

[0146] In some implementations, at least one processor can generate composite metrology data for a modified reference device by determining a change between the measured metrology data and the first set of composite metrology data, e.g., as a reference Figure 4A The discussed Figure 4B and Figure 4C The at least one processor may further modify the second set of composite metering data using the changes between the measured metering data and the first set of composite metering data, as shown in blocks 440 and 442 of FIG. Figure 4A The discussed Figure 4B and Figure 4C The variation between the measured metrology data and the first set of composite metrology data may be, for example, a mismatch between the measured metrology data and the first set of composite metrology data (e.g., Figure 4B 440 in ) or a spectral transformation between the measured metrology data and the first set of synthetic metrology data (as discussed in Figure 4C (as shown in box 442 in FIG. ).

[0147] In some implementations, at least one processor may generate composite metrology data for a modified reference device by determining a difference between a first set of composite metrology data and a second set of composite metrology data, for example, as shown in FIG. Figure 4A The discussed Figure 4D and Figure 4E The at least one processor may further modify the measured metering data using the difference between the first set of composite metering data and the second set of composite metering data, as shown in block 444 of FIG. Figure 4A The discussed Figure 4D and Figure 4E As shown at boxes 444 and 450 in.

[0148] In some implementations, the at least one processor may further generate multiple sets of composite metrology data for the corresponding multiple modified reference devices, and store the multiple sets of composite metrology data as the training data set.

[0149] The measured metrology data may be, for example, a measured spectrum, and the first set of synthetic metrology data and the second set of synthetic metrology data may be synthetic spectra.

[0150] In some implementations, at least one processor may further generate a first set of key parameter values ​​for a first model of a reference device, for example, as in Figure 5 The at least one processor may further generate a second set of key parameter values ​​for a second model (if used), e.g., as in Figure 5The at least one processor may generate composite reference parameters for the modified reference device by combining the reference parameters, the first set of key parameter values, and the second set of key parameter values ​​(if used), e.g., as discussed at blocks 512, 540, and 550. The at least one processor may store the composite reference parameters along with the composite metrology data as a training data set, e.g., in a computer system. Figure 1 For example, the composite reference parameters for the modified reference device may be determined by determining the key parameter deviation between the first set of key parameter values ​​and the second set of key parameter values ​​(e.g., as in Figure 5 and modifying the reference parameter with the key parameter deviation (e.g., as discussed in block 540 of Figure 5 (as discussed in box 550 of ).

[0151] The above description is intended to be illustrative rather than limiting. For example, the above examples (or one or more aspects thereof) may be used in combination with each other. Other specific implementations may be used, such as those of ordinary skill in the art after reading the above description. In addition, various features may be grouped together, and less than all features of a specific disclosed specific implementation may be used. Therefore, the following aspects are thus incorporated into the above description as examples or specific implementations, wherein each aspect is independently used as a separate specific implementation, and it is contemplated that such specific implementations may be combined with each other in various combinations or permutations. Therefore, the spirit and scope of the appended claims should not be limited to the foregoing description.

Claims

1. A method for characterizing a device on a sample, the method comprising: obtaining measured metrological data from the device; as well as At least one parameter of the device is determined based on the measured metrology data using a machine learning model using composite metrology data, wherein each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device. 2 . The method of claim 1 , wherein the first model is generated by fitting metrology data measured from the reference device to synthetic metrology data for the first model. 3 . The method of claim 1 , wherein each composite metrology data comprises the merger of the metrology data and the first synthetic metrology data and further comprises second synthetic metrology data for a second model of a modified reference device that is changed relative to the first model. 4 . The method of claim 3 , wherein the second model of the modified reference device has at least one parameter that is changed relative to the first model.

5. A method according to claim 3, wherein the machine learning model further uses composite reference parameters for the modified reference device based on a combination of reference parameters of the reference device, a first set of key parameter values ​​generated for the first model, and a second set of key parameter values ​​generated for the second model.

6. The method of claim 3, wherein the second model of the modified reference device is generated by changing at least one parameter of the first model. 7 . The method of claim 3 , wherein second synthetic metrology data for the second model of the modified reference device is generated based on a change between metrology data measured from the reference device and first synthetic metrology data for the first model.

8. The method of claim 3, wherein the composite metrology data is generated by modifying metrology data measured from the reference device using determined differences between first composite metrology data for the first model and second composite metrology data for the second model for the modified reference device.

9. The method of claim 3, wherein the composite metrology data comprises multiple sets of composite metrology data for a corresponding plurality of modified reference devices.

10. The method of claim 1, wherein the measured metrology data comprises a measured spectrum and the composite metrology data comprises a composite spectrum.

11. A metrology system configured to support characterization of a device on a sample, the metrology system comprising: a source configured to generate radiation to be incident on the device on the sample; at least one detector configured to detect radiation from the device generated in response to the radiation incident on the device; and at least one processor coupled to the at least one detector, wherein the at least one processor is configured to: Obtaining measured metrology data from the device; and At least one parameter of the device is determined based on the measured metrology data using a machine learning model using composite metrology data, wherein each composite metrology data includes a combination of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device.

12. The metrology system of claim 11, wherein the first model is generated by fitting metrology data measured from the reference device to synthetic metrology data for the first model.

13. The metrology system of claim 11, wherein each composite metrology data comprises the merger of the metrology data and the first synthetic metrology data and further comprises second synthetic metrology data for a second model of a modified reference device that is changed relative to the first model.

14. The metrology system of claim 13, wherein the second model of the modified reference device has at least one parameter that is changed relative to the first model.

15. The metrology system of claim 13, wherein the machine learning model further uses composite reference parameters for the modified reference device based on a combination of reference parameters of the reference device, a first set of key parameter values ​​generated for the first model, and a second set of key parameter values ​​generated for the second model.

16. The metrology system of claim 13, wherein the second model of the modified reference device is generated by changing at least one parameter of the first model.

17. The metrology system of claim 13, wherein second synthetic metrology data for the second model of the modified reference device is generated based on a change between metrology data measured from the reference device and first synthetic metrology data for the first model.

18. The metrology system of claim 13, wherein the composite metrology data is generated by modifying metrology data measured from the reference device using determined differences between first composite metrology data for the first model and second composite metrology data for the second model for the modified reference device.

19. The metrology system of claim 13, wherein the composite metrology data comprises multiple sets of composite metrology data for a corresponding plurality of modified reference devices.

20. The metrology system of claim 11, wherein the measured metrology data comprises a measured spectrum and the composite metrology data comprises a composite spectrum.

21. A metrology system configured to support characterization of a device on a sample, the metrology system comprising: means for obtaining measured metrological data from said device; and Means for determining at least one parameter of the device based on the measured metrology data using a machine learning model using composite metrology data, wherein each composite metrology data includes a merger of metrology data measured for a reference device and first synthetic metrology data of a first model for the reference device.

22. The metrology system of claim 21, wherein each composite metrology data comprises the merger of the metrology data and the first composite metrology data and further comprises second composite metrology data for a second model of a modified reference device that is changed relative to the first model.

23. A method for characterizing a device on a sample, the method comprising: obtaining optical metrology data of a measurand for a reference device for the device; generating composite optical metrology data, wherein each composite optical metrology data is generated by combining measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device; as well as A machine learning model is trained using a training data set including at least the composite optical metrology data to characterize the device using measured optical metrology data from the device.

24. The method of claim 23, further comprising generating the first model by fitting the measured optical metrology data to a first set of synthetic optical metrology data for the first model.

25. The method of claim 23, wherein each composite metrology data comprises a merger of the metrology data and the first synthetic metrology data, and further comprises second synthetic metrology data for a second model of a modified reference device that is changed relative to the first model.

26. The method of claim 25, further comprising generating the second model of the modified reference device by changing at least one parameter of the first model.

27. The method of claim 25, wherein generating the composite optical metrology data for the modified reference device comprises: determining a change between the measured optical metrology data and the first set of synthetic optical metrology data; as well as The change between the measured optical metrology data and the first set of synthetic optical metrology data is utilized to modify a second set of synthetic optical metrology data.

28. The method of claim 27, wherein the variation between the measured optical metrology data and the first set of synthetic optical metrology data comprises a mismatch between the measured optical metrology data and the first set of synthetic optical metrology data or a spectral shift between the measured optical metrology data and the first set of synthetic optical metrology data.

29. The method of claim 25, wherein generating the composite optical metrology data for the modified reference device comprises: determining a difference between the first set of synthetic optical metrology data and the second set of synthetic optical metrology data; as well as The measured optical metrology data is modified using the difference between the first set of composite optical metrology data and the second set of composite optical metrology data.

30. The method of claim 23, further comprising: A plurality of sets of composite optical metrology data are generated for a corresponding plurality of modified reference devices, wherein the training data set includes the plurality of sets of composite optical metrology data.

31. The method of claim 23, wherein the measured optical metrology data comprises a measured optical spectrum, and wherein the first group comprises composite optical metrology data.

32. The method of claim 25, wherein the reference device has reference parameters, the method further comprising: generating a first set of key parameter values ​​for the first model of the reference device; generating a second set of key parameter values ​​for the second model; as well as Composite reference parameters for the modified reference device are generated by combining the reference parameters, the first set of key parameter values, and the second set of key parameter values, wherein the training data set includes the composite reference parameters and the composite optical metrology data.

33. The method of claim 32, wherein generating the composite reference parameters for the modified reference device comprises: determining a key parameter deviation between the first set of key parameter values ​​and the second set of key parameter values; as well as The reference parameter is modified using the key parameter deviation.

34. A computer system configured to support characterizing a device on a sample, the computer system comprising: at least one memory configured to store measured optical metrology data and composite optical metrology data; and at least one processor coupled to the at least one memory, wherein the at least one processor is configured to: obtaining optical metrology data of a measurand for a reference device for the device; generating composite optical metrology data, wherein each composite optical metrology data is generated by combining measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device; as well as A machine learning model is trained using a training data set including at least the composite optical metrology data to characterize the device using measured optical metrology data from the device.

35. The computer system of claim 34, wherein the at least one processor is further configured to generate the first model by fitting the measured optical metrology data to the first set of synthetic optical metrology data for the first model.

36. The computer system of claim 34, wherein each composite metrology data comprises a merger of the metrology data and the first composite metrology data, and further comprises second composite metrology data for a second model of a modified reference device that is changed relative to the first model.

37. The computer system of claim 36, wherein the at least one processor is further configured to generate the second model of the modified reference device by changing at least one parameter of the first model.

38. The computer system of claim 36, wherein the at least one processor is configured to generate the composite optical metrology data for the modified reference device by being configured to: determining a change between the measured optical metrology data and the first set of synthetic optical metrology data; and The change between the measured optical metrology data and the first set of synthetic optical metrology data is utilized to modify a second set of synthetic optical metrology data.

39. The computer system of claim 38, wherein the variation between the measured optical metrology data and the first set of synthetic optical metrology data comprises a mismatch between the measured optical metrology data and the first set of synthetic optical metrology data or a spectral shift between the measured optical metrology data and the first set of synthetic optical metrology data.

40. The computer system of claim 36, wherein the at least one processor is configured to generate the composite optical metrology data for the modified reference device by being configured to: determining a difference between the first set of composite optical metrology data and the second set of composite optical metrology data; and The measured optical metrology data is modified using the difference between the first set of composite optical metrology data and the second set of composite optical metrology data.

41. The computer system of claim 34, wherein the at least one processor is further configured to: A plurality of sets of composite optical metrology data are generated for a corresponding plurality of modified reference devices, wherein the training data set includes the plurality of sets of composite optical metrology data.

42. The computer system of claim 34, wherein the measured optical metrology data comprises a measured optical spectrum, and wherein the first group comprises composite optical metrology data.

43. The computer system of claim 36, wherein the reference device has reference parameters, wherein the at least one processor is further configured to: generating a first set of key parameter values ​​for the first model of the reference device; generating a second set of key parameter values ​​for the second model; as well as Composite reference parameters for the modified reference device are generated by combining the reference parameters, the first set of key parameter values, and the second set of key parameter values, wherein the training data set includes the composite reference parameters and the composite optical metrology data.

44. The computer system of claim 43, wherein the at least one processor is configured to generate the composite reference parameters for the modified reference device by being configured to: determining a key parameter deviation between the first set of key parameter values ​​and the second set of key parameter values; and The reference parameter is modified using the key parameter deviation.

45. A computer system for characterizing a device on a sample, the computer system comprising: means for obtaining measured optical metrology data for a reference device for said device; means for generating composite optical metrology data, wherein each composite optical metrology data is generated by combining measured optical metrology data for the reference device and first synthetic optical metrology data for a first model of the reference device; and for training a machine learning model using a training data set including at least the composite optical metrology data to characterize components of the device using measured optical metrology data from the device.

46. ​​The computer system of claim 45, wherein each composite metrology data comprises a merger of the metrology data and the first composite metrology data, and further comprises second composite metrology data for a second model of a modified reference device that is changed relative to the first model.

Images