Measurement in the presence of CMOS underarray (CuA) structures using model-less machine learning.

Abstract

The system may include a controller comprising one or more processors, the one or more processors being configured to execute program instructions causing the one or more processors to perform a measurement recipe, the measurement recipe being performed by: receiving optical measurement data of a training sample comprising a complementary metal-oxide-semiconductor underarray (CuA) device, wherein the CuA device comprises a CMOS structure disposed beneath a periodic memory array structure; receiving reference data of the training sample, wherein the reference data comprises measured values ​​of the geometric parameters of the CuA device; training a machine learning model using the optical measurement data and the reference data of the training sample; receiving optical measurement data of a test sample comprising a CuA device; and determining one or more measured values ​​of the geometric parameters of the CuA device on the test sample by using the machine learning model with the optical measurement data of the test sample.

Description

Technical Field

[0001] The present disclosure generally relates to optical measurement, and particularly to optical measurement of a memory structure including an embedded CMOS structure.

Background Art

[0002] One approach to meeting the requirements for improving the performance of memory devices (such as 3D memory devices) while maintaining or reducing the physical size is to fabricate a CMOS circuit (such as a logic circuit) under the memory array structure. This approach is generally referred to as complementary metal oxide semiconductor (CMOS) under array (CuA) technology.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Patent Document 5

Patent Document 6

Patent Document 7

Patent Document 8

Patent Document 9

Patent Document 10

Patent Document 11

[0004] [Non-Patent Document 1] Germer et al., "Intercomparison between optical and x-ray scatterometry measurements of FinFET structures," Proceedings of SPIE, Vol. 8681, p. 86810Q, 2013. [Non-Patent Document 2] Kline et al., "X-ray scattering critical dimensional metrology using a compact x-ray source for next generation semiconductor devices," Journal of Micro / Nanolithography, MEMS, and MOEMS, Vol. 16, No. 1, 2017. [Overview of the project] [Problems that the invention aims to solve]

[0005] However, CuA technology presents unique challenges for measurement systems used in process control, because the underlying CMOS circuitry can affect the measurements of the memory array structure. Therefore, the development of systems and methods to address this challenge is required. [Means for solving the problem]

[0006] Methods according to one or more exemplary embodiments of the present disclosure are disclosed. In some embodiments, the method includes the step of generating optical measurement data for one or more training samples, the CuA devices comprising CMOS structures disposed beneath a periodic memory array structure. In some embodiments, the method includes the step of generating reference data for the one or more training samples, the reference data comprising measurements of the geometric parameters of the CuA devices. In some embodiments, the method includes the step of training a machine learning model using the optical measurement data and the reference data of the one or more training samples. In some embodiments, the method includes the step of generating optical measurement data for one or more test samples, the CuA devices. In some embodiments, the method includes the step of determining one or more measurements of the geometric parameters of the CuA devices on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples.

[0007] This disclosure discloses a system according to one or more exemplary embodiments of the present disclosure. In some embodiments, the system includes a controller comprising one or more processors, the one or more processors being configured to execute program instructions causing the one or more processors to perform a measurement recipe, the measurement recipe being performed by: receiving optical measurement data of one or more training samples comprising complementary metal-oxide-semiconductor underarray (CuA) devices, wherein the CuA devices comprise CMOS structures disposed beneath a periodic memory array structure; receiving reference data of the one or more training samples, wherein the reference data comprises measured values ​​of the geometric parameters of the CuA devices; training a machine learning model using the optical measurement data of the one or more training samples and the reference data; receiving optical measurement data of one or more test samples comprising CuA devices; and determining one or more measured values ​​of the geometric parameters of the CuA devices on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples.

[0008] This disclosure discloses a system according to one or more exemplary embodiments of the present disclosure. In some embodiments, the system includes an optical characterization system. In some embodiments, the system includes a reference characterization system. In some embodiments, the system includes a controller comprising one or more processors, the one or more processors being configured to execute program instructions causing the one or more processors to perform a measurement recipe, the measurement recipe being performed by: receiving optical measurement data of one or more training samples comprising complementary metal-oxide-semiconductor underarray (CuA) devices from the optical characterization system, wherein the CuA devices comprising CMOS structures disposed beneath a periodic memory array structure; receiving reference data of the one or more training samples, wherein the reference data comprising measured values ​​of the geometric parameters of the CuA devices from the reference characterization system; training a machine learning model using the optical measurement data of the one or more training samples and the reference data; receiving optical measurement data of one or more test samples comprising CuA devices from the optical characterization system; and determining one or more measured values ​​of the geometric parameters of the CuA devices on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples.

[0009] The above summary and the following detailed description are for illustrative and illustrative purposes only and should not be understood as necessarily limiting the invention described in the claims. The accompanying drawings are incorporated herein and constitute part of this specification and illustrate embodiments of the invention, and together with the above summary, are useful in explaining the principles of the invention.

[0010] Those skilled in the art will gain a deeper understanding of the numerous advantages of this disclosure by referring to the attached drawings.

Brief Description of the Drawings

[0011] [Figure 1A] It is a block diagram of a measurement system according to one or some embodiments of the present disclosure. [Figure 1B] It is a schematic diagram of a characteristic evaluation subsystem configured as an optical characteristic evaluation subsystem according to one or some embodiments of the present disclosure. [Figure 1C] It is a schematic diagram of a characteristic evaluation subsystem configured as an X-ray characteristic evaluation subsystem according to one or some embodiments of the present disclosure. [Figure 1D] It is a schematic diagram of a characteristic evaluation subsystem configured as a particle beam characteristic evaluation subsystem according to one or some embodiments of the present disclosure. [Figure 2] It is a schematic diagram of a CuA device according to one or some embodiments of the present disclosure. [Figure 3] It is a flowchart showing steps executed in a method for evaluating the characteristics of a CuA device according to one or some embodiments of the present disclosure.

Modes for Carrying Out the Invention

[0012] Next, the subject matter of the present disclosure shown in the accompanying drawings will be referred to in detail. In the present disclosure, specific embodiments and their detailed features are specifically shown and described. The embodiments described herein should be construed as exemplary and not restrictive. Those skilled in the art will readily understand that various changes / modifications can be made to the form and details without departing from the spirit and scope of the present disclosure.

[0013] Each embodiment of this disclosure relates to a system and method for optical measurement of complementary metal-oxide-semiconductor (CMOS) underarray (CuA) devices, which is based on an optical measurement data (such as full-loop data) of a completed CuA device and a machine learning model trained on reference data generated by another tool that provides ground truth measurements of parameters of interest.

[0014] CuA structures (such as CuA memory structures) generally may include logic circuits (such as CMOS logic circuits) that are physically located (e.g., disposed) beneath memory array structures (such as three-dimensional (3D) memory stacks or 3D NAND structures). In this specification, the term “CuA structure” may encompass various designs of logic / memory array structures. Thus, this disclosure is not limited to any particular CuA configuration.

[0015] Optical measurement is typically used for semiconductor process control. This is because optical measurement can provide relatively high measurement throughput and is typically non-destructive. In optical measurement, light is shone onto a sample, and measurements are generated based on the light emitted from the sample in response. In optical measurement of subsurface features, the light typically needs to travel through at least the upper part of the sample to reach the feature of interest below the surface. Therefore, optical measurement systems typically utilize light of wavelengths selected to travel through the structure of interest with relatively little absorption.

[0016] However, in the case of CuA structures, incident light can interact with both the memory array structure and the embedded CMOS structure, which can impair the ability to generate isolated measurements for the memory array structure. This specification assumes that conventional optical measurement techniques may not be sufficient for accurately characterizing CuA structures, especially as dimensions decrease and device complexity increases. For example, some techniques can utilize wavelengths of light that are within the transmission range for the memory array structure of interest and at least partially absorbed by the underlying logic circuit. As an example, some logic circuits can utilize a polysilicon layer. A polysilicon layer absorbs light with wavelengths longer than approximately 450 nanometers (nm). In this case, isolated measurements of the memory array structure can be generated by performing optical measurements at wavelengths shorter than approximately 450 nm. However, such techniques may limit CuA designs to specific ones incorporating the aforementioned absorbing materials, the sensitivity obtained may be limited for deep structures, and furthermore, the values ​​obtained may be limited in broadband optical measurement methods that utilize data from multiple wavelengths. As another example, some techniques utilize supervised training of artificial neural networks using labeled optical measurement data generated by additional measurement methods. However, these techniques may have various limitations. For example, these limitations include, but are not limited to, the need for fast sampling of ground truth reference data, the considerable time required to generate a sufficient number of labels for training, limited performance for deeply embedded structures, low sensitivity to process changes, and lack of general applicability to CuA structures outside the training dataset.

[0017] In some embodiments of this disclosure, a machine learning model is trained to generate measurements of a CuA device or a portion thereof based on training data, the training data including optical measurement data of a completed CuA device and reference data providing ground truth measurements of parameters of interest. Thus, isolated measurements of a periodic memory array structure can be generated, although the optical measurement data may be affected by the presence of the underlying CMOS structure. Specifically, the machine learning model can identify patterns between the optical measurement data and ground truth measurements of a completed CuA device.

[0018] Next, with reference to Figures 1A to 3, a system and method for characterizing CuA structures according to one or more embodiments of the present disclosure will be described in detail.

[0019] Figure 1A is a block diagram of a measurement system 100 according to one or more embodiments of the present disclosure. In some embodiments, the measurement system 100 includes a characterization subsystem 102 for generating measurement data of a sample 104 using optical techniques, and a controller 106 for generating one or more measurements based on the measurement data.

[0020] The characterization subsystem 102 may include any components or combinations thereof suitable for generating measurement data for the sample 104.

[0021] In some embodiments, the characterization subsystem 102 includes an optical characterization subsystem 102 for generating measurement data based on the interaction between the sample 104 and light. For example, the characterization subsystem 102 may include, but is not limited to, a spectroscopic ellipsometer (SE), an SE with multiple illumination angles, an SE that measures Müller matrix elements (e.g., using a rotational compensator), a single-wavelength ellipsometer, a beam profile ellipsometer (angle-resolved ellipsometer), a beam profile reflectometer (angle-resolved reflectometer), a broadband reflectometer (spectroscopic reflectometer), a single-wavelength reflectometer, an angle-resolved reflectometer, an imaging system, a scatorometer (such as a speckle analyzer), or any combination thereof.

[0022] In some embodiments, the characterization subsystem 102 includes an X-ray characterization subsystem 102 for generating measurement data based on the interaction between the sample 104 and X-rays. For example, the characterization subsystem 102 may include, but is not limited to, a small-angle X-ray scattering (SAXS) system or an X-ray reflection scatterometry (SXR) system.

[0023] In some embodiments, the characterization subsystem 102 includes a particle beam characterization subsystem 102 for generating measurement data based on the interaction between the sample 104 and the particle beam. For example, the particle beam may be, but is not limited to, an electron beam (e beam), an ion beam, or a neutral particle beam.

[0024] In some embodiments, the characterization subsystem 102 provides multiple types of measurements. In some embodiments, the measurement system 100 includes multiple measurement subsystems 102, each subsystem providing a different combination of one or more measurements. Furthermore, the measurement system 100 can be provided as a single tool or as multiple tools. An outline of a single tool providing multiple measurement configurations is described in Patent Document 1, which is incorporated herein by reference in its entirety. An outline of multiple tool and structure analysis is described in Patent Document 2, which is incorporated herein by reference in its entirety.

[0025] Furthermore, Patent Documents 3, 4, 5, 6, 7, and 8 are incorporated herein by reference in their entirety.

[0026] In some embodiments, the controller 106 includes one or more processors 108, each processor 108 configured to execute a set of program instructions held in memory 110 or a memory device, which can cause the processor 108 to perform various operations.

[0027] One or more processors 108 of the controller 106 may include any processor or processing element known in the art. In this disclosure, the terms “processor” or “processing element” may be broadly defined to include any device that includes one or more processing elements or logic elements (e.g., one or more microprocessor devices, one or more application-specific integrated circuit (ASIC) devices, one or more field-programmable gate arrays (FPGAs), one or more digital signal processors (DSPs)). In this sense, one or more processors 108 may include any device configured to execute algorithms and / or instructions (such as program instructions stored in memory). In some embodiments, one or more processors 108 may be embodied as a desktop computer, a mainframe computer system, a workstation, an image computer, a parallel processor, a network computer, or any other computer system configured to execute a program configured to operate the characterization subsystem 102 or to operate in conjunction with the characterization subsystem 102, as described throughout this disclosure. Furthermore, various subsystems of the measurement system 100 may include processors or logic elements suitable for performing at least some of the steps described in this disclosure. Therefore, the above description should not be construed as limiting the embodiments of the present disclosure, but merely as illustrative. Furthermore, the steps described throughout the present disclosure may be performed by a single controller or by multiple controllers. In addition, the controller 106 may include one or more controllers housed in one common housing or multiple housings. Thus, any controller or combination thereof can be packaged separately as modules suitable for integration into the measurement system 100.

[0028] The memory 110 may include any storage medium known in the art that is suitable for storing program instructions that can be executed by one or more associated processors 108. For example, the memory 110 may include a non-temporary storage medium. Another example of the memory 110 may include, but is not limited to, read-only memory (ROM), random access memory (RAM), magnetic or optical memory devices (such as disks), magnetic tape, solid-state drives, etc. The memory 110 may be housed together with one or more processors 108 in a common controller housing. In some embodiments, the memory 110 may be located remotely from the physical location of one or more processors 108 and the controller 106. For example, one or more processors 108 of the controller 106 may access remote memory (such as a server) that is accessible via a network (such as the internet or an intranet).

[0029] The controller 106 may be coupled to communicate with any or any combination thereof of components of the measurement system 100. In some embodiments, the controller 106 may receive data (such as measurement data) from one or more components of the measurement system 100. In some embodiments, the controller 106 may control one or more components of the measurement system 100 by drive signals. More generally, the controller 106 may perform any of the steps described herein.

[0030] In some embodiments, the controller 106 generates one or more measurements of the sample 104 based at least in part on measurement data generated by the characterization subsystem 102. Measurement of parameters of interest may include several algorithms that can be performed by the controller 106. For example, the optical interaction between the incident beam and the sample 104 can be modeled using an EM (electromagnetic) solver, and algorithms that can be used for this include, but are not limited to, exact coupled-wave analysis (RCWA), finite element method (FEM), method of moments, surface integral method, volume integral method, or finite difference time-domain (FDTD) method. Modeling (parameterization, etc.) of the sample 104 can be done using a geometric engine, a process modeling engine, or a combination of both. An outline of the use of process modeling is described in Patent Document 9, which is incorporated herein by reference in its entirety. Geometric engines are implemented, for example, in KLA Corporation's AcuShape® software.

[0031] The controller 106 may analyze the collected measurement data using any suitable combination of data fitting and / or optimization techniques, which include, but are not limited to, libraries, fast low-dimensional models, regression analysis, statistical techniques (see, for example, Patent Document 10), machine learning algorithms (e.g., neural networks, support vector machines (SVM), principal component analysis (PCA), independent component analysis (ICA), local linear embedding (LLE), and more generally, dimensionality reduction techniques), sparse representation techniques, Fourier transform techniques, wavelet transform techniques, or Kalman filtering. An outline of statistical model-based measurement is described in Patent Document 10, which is incorporated herein by reference in its entirety. The controller 106 may also analyze the collected measurement data using algorithms that do not involve modeling, optimization, and / or fitting. An outline of characterization of patterned wafers is described in Patent Document 11, which is incorporated herein by reference in its entirety. In some embodiments, the controller 106 utilizes one or more algorithms to improve consistency between tools of the same or different types (e.g., between individual instances or configurations of a single characterization subsystem 102).

[0032] The controller 106 may be designed to provide efficient performance through any appropriate technique. For example, such techniques include, but are not limited to, parallelization, computational distribution, load balancing, multi-service support, and dynamic load optimization. Furthermore, the controller 106 can perform any step using any type of configuration, or a combination thereof. For example, such configurations include, but are not limited to, dedicated hardware (such as an FPGA), software, or firmware.

[0033] The controller 106 may further generate any type of measurement of sample 104 (or a portion thereof) based at least in part on measurement data from the characterization subsystem 102. In some embodiments, the controller 106 generates measured measurements. For example, measured measurements may include, but are not limited to, overlay measurements, minimum linewidth (CD) measurements, shape measurements (such as height measurements, slope measurements, and sidewall angle measurements), stress measurements, composition measurements, bandgap measurements, electrical property measurements, or process state measurements (such as focus and / or exposure levels, resist state, partial pressure, temperature, and focusing model). In some embodiments, the controller 106 generates inspection measurements, which perform at least one of the identification or classification of one or more defects on sample 104.

[0034] The measurement system 100 and any of its components (such as the characterization subsystem 102 or the controller 106) may be configured to perform a recipe (such as a measurement recipe). The recipe may define various configuration parameters and / or steps to be performed in a single measurement or a series of measurements.

[0035] As an example, the recipe may include various aspects of the design of sample 104 (e.g., the design of the CuA device 202 on sample 104), including, but not limited to, the layout of features on one or more sample layers, the size of the features, or the pitch of the features. As another example, the recipe may include illumination parameters, including, but not limited to, the wavelength of illumination, the illumination pupil distribution (e.g., the angular distribution of illumination and the illumination intensity at each associated angle), the polarization of the incident illumination, the spatial distribution of illumination, or the height of the sample. As yet another example, the recipe may include focusing parameters, including, but not limited to, the focusing pupil distribution (e.g., the desired angular distribution of light from the sample used for measurement and the filtered intensity at each associated angle), the setting of the focusing aperture for selecting the part of the sample of interest, the polarization of the focused light, or a wavelength filter. As yet another example, the recipe may include various processing steps. For example, the controller 106 may perform these processing steps to generate measurements based on the measurement data generated according to the above recipe.

[0036] Referring now to Figure 2, which is a schematic diagram of a CuA device 202 according to one or more embodiments of the present disclosure. The CuA device 202 may include a memory array structure 204 and various CMOS structures 206 (such as logic structures) located beneath (e.g., disposed beneath) the memory array structure 204.

[0037] The memory array structure 204 may include any number or type of structures suitable for forming a memory array. For example, the memory array structure 204 may include, but is not limited to, a 3D NAND structure. The 3D NAND structure is formed of feature objects 208 having a certain pattern, which are contained within the multilayer stack 210. Furthermore, such a memory array structure 204 is typically a periodic structure having periodicity along one or more dimensions.

[0038] The CMOS structure 206 may contain any number or type of structures fabricated beneath the memory array structure 204. For example, the CMOS structure 206 may be suitable for controlling and / or powering the memory array structure 204, but this is not required. Thus, the combination of the CMOS structure 206 and the memory array structure 204 can form a memory device (such as a 3D memory device). Furthermore, the CMOS structure 206 typically has a spatially varied distribution, and therefore the number and / or design of its constituent features may not have periodicity across the entire CuA device 202. Thus, the CMOS structure 206 can generally be described as non-periodic. However, it should be noted that the CMOS structure 206 may exhibit local periodicity in some regions.

[0039] Furthermore, the memory array structure 204 and / or the CMOS structure 206 can generally have any design, and therefore the term CuA device 202 as herein is not limited to a specific design. As an example, the CuA device 202 may include a layer interposed between the memory array structure 204 and the CMOS structure 206, for example, a source layer 212 (such as a polysilicon source layer), but is not limited to this. As another example, although not shown, the CuA device 202 may include a layer interposed between the CMOS structure 206 and the substrate 214.

[0040] Next, with reference to Figure 3, techniques for characterizing the CuA device 202 or a portion thereof, according to one or more embodiments of the present disclosure, will be described in detail.

[0041] At various stages of the manufacturing process, it may be desirable to generate measurements of the structures constituting the CuA device 202 (such as the CMOS structure 206 and / or the memory array structure 204). Such measurements may include, but are not limited to, measurement measurements and defect measurements (such as inspection measurements). Measurement measurements may include, but are not limited to, overlay measurements between structures manufactured with different lithography exposures, minimum linewidth (CD) measurements of one or more features, heights of one or more features, or inclinations of one or more features. Inspection measurements may include, but are not limited to, identifying and / or characterizing defects in the manufacturing process (e.g., unwanted or missing features, features with incorrect shape or position). Furthermore, such measurements can be used for a variety of purposes, including, but are not limited to, process control, deployment, or evaluation of the performance of the manufactured CuA device 202.

[0042] Measurements may be generated after any process step for manufacturing the CuA device 202. For example, measurements may be generated after manufacturing the CMOS structure 206 and / or after manufacturing the memory array structure 204 to form the complete CuA device 202. For convenience of explanation, measurements of the complete CuA device 202, including both the memory array structure 204 and the underlying CMOS structure 206, are referred to herein as “full-loop” measurements.

[0043] This specification assumes that measurements at a given process step can generally provide information about any features already fabricated on the sample 104, depending on the interaction between the illumination beam 114 and the sample 104. Thus, it can be difficult to isolate and measure newly fabricated features. For example, full-loop measurements can generally provide information about, or be influenced by, both the memory array structure 204 and the underlying CMOS structure 206, which can limit or impair the ability to generate isolated measurements of the memory array structure 204.

[0044] In some embodiments, measurements of various test structures may be generated to assist in generating isolated measurements of specific features. For example, measurements of a test structure that includes a memory array structure 204 but does not include a corresponding embedded CMOS structure 206 are referred to herein as “short-loop” measurements.

[0045] Furthermore, measurements at any process step can generally be generated using any suitable technique. These techniques include, but are not limited to, optical, X-ray, and particle-based techniques. However, various measurement techniques can have various trade-offs. For example, optical measurement techniques can generally provide non-destructive measurements with high measurement throughput, but their resolution may be limited or restricted to certain types of structures (such as periodic structures). This type of structure is determined based on the corresponding analysis or modeling step. Therefore, optical measurements are typically used during execution when throughput is particularly critical. Another example is X-ray and / or particle-based techniques, which may offer higher resolution than some optical techniques, but have the drawback of relatively low throughput and / or destructive measurements. Consequently, these techniques are typically used for reference measurements.

[0046] However, this specification assumes that generating all types of measurements in every measurement step may not be feasible or desirable depending on the application. In such cases, depending on the available data, various techniques can be used to generate measurements of a specific structure (such as isolated measurements of the memory array structure 204).

[0047] Figure 3 is a flowchart showing the steps performed in a method 300 for characterizing a CuA device 202 according to one or more embodiments of this disclosure. It should be understood that embodiments and techniques for realizing them described herein in the context of the measurement system 100 also apply to method 300. For example, any steps related to method 300 may be performed by the controller 106 and / or the characterization subsystem 102 of the measurement system 100. However, it should also be noted that method 300 is not limited to the configuration of the measurement system 100.

[0048] This specification assumes that Method 300 may be suitable for applications where it is desirable to generate isolated measurements of the memory array structure 204 without using short-loop measurements, but is not limited thereto. One of the goals of this technique is to build a machine learning model to correlate optical measurement data (such as data of the CuA device 202 generated by the optical characterization subsystem 102, such as full-loop data) with measurements of interest of the CuA device 202. For example, measurements of interest may include, but are not limited to, overlay measurements, minimum linewidth (CD) measurements, shape measurements (such as height measurements, slope measurements, and sidewall angle measurements), stress measurements, composition measurements, bandgap measurements, electrical property measurements, process state measurements (such as focus and / or exposure state, resist state, partial pressure, temperature, and focusing model), defect identification, or defect classification. Such a machine learning model may be trained using optical measurement data of the CuA device 202 on a training sample (such as full-loop data on the training sample) and reference data of the same training sample. Reference data can be generated using appropriate tools, such as, but not limited to, the X-ray characterization subsystem 102 and the particle beam characterization subsystem 102. Furthermore, the reference data may provide direct ground truth measurements of the parameters of interest. Once the machine learning model is trained, it can be used to generate measurements of the CuA device 202 (e.g., one under construction) on the test specimen based on the optical measurement data of that specimen.

[0049] In some embodiments, Method 300 includes step 302 of generating optical measurement data for one or more training samples containing a CuA device 202. The optical measurement data may include, but is not limited to, any type of data generated by any type of optical system, such as the optical characterization subsystem 102 shown in Figure 1B. For example, the optical measurement data may include, but is not limited to, spectral data from an ellipsometer, reflectometer, or scatometer (which collect, for example, ellipsometry data, reflectometry data, or scatometry data, respectively).

[0050] Furthermore, in some embodiments, the one or more training samples include known variability in the CuA device 202 (such as known variability related to process bias). Thus, the impact of such variability can be identified and incorporated into the model of Method 300, described later. Such a procedure is called Design of Experiments (DOE) and can improve the robustness of Method 300.

[0051] In some embodiments, Method 300 includes a step 304 to generate reference data for one or more training samples, the reference data including measurements of geometric parameters of the CuA device 202. Geometric parameters may include, but are not limited to, any physical properties of the structure in question, such as CD, feature height, feature inclination (e.g., sidewall angle), film thickness, overlay between features produced by different lithographic exposures, or composition.

[0052] The above reference data can be generated using any suitable technique. In some embodiments, the reference data is generated using a high-resolution characterization subsystem 102, which may include, but is not limited to, an X-ray characterization subsystem 102 or a particle beam characterization subsystem 102. Non-limiting examples include, but are not limited to, transmission electron microscope (TEM) data, scanning electron microscope (SEM) data (such as minimum linewidth SEM (CD-SEM) data or electron beam SEM (EB-SEM) data), SAXS data (such as transmission SAXS (T-SAXS) data or CD-SAXS data), X-ray photoelectron spectroscopy (XPS) data, or X-ray diffraction (XRD) data. CD-SAXS data may be particularly suitable for use as reference data, but this is not a limitation of this disclosure.

[0053] Thus, accurate measurements of various geometric parameters related to the CuA device 202 can be obtained directly from the reference data or through analysis of the reference data.

[0054] In some embodiments, method 300 includes step 306 of training a machine learning model using optical measurement data (such as that in step 302) and reference data (such as that in step 304).

[0055] Thus, the machine learning model can determine the relationship between the parameters of physical geometric interest on the CuA device 202 (or a part thereof) provided from the reference data and the optical measurement data, which is a type of data that may be generated during manufacturing by the characterization subsystem 102. As previously stated herein, the optical measurement data (such as full-loop data) of the completed CuA device 202 may be affected by both the memory array structure 204 and the underlying CMOS structure 206. In applications where short-loop data (e.g., optical measurement data of the memory array structure 204 without the underlying CMOS structure 206) is unavailable for some reason, it may be difficult or impossible to provide isolated measurements of the memory array structure 204 using conventional techniques. However, it is assumed herein that a machine learning model trained using reference data that provides full-loop optical measurement data and ground truth measurements (such as those of the memory array structure 204) in step 306 can generate isolated measurements of the memory array structure 204 despite the influence of the underlying CMOS structure 206.

[0056] The above machine learning model may include any suitable type of machine learning technique known in the art, and may include any combination of learning techniques. Learning techniques may include, but are not limited to, supervised learning, unsupervised learning, or reinforcement learning. In some embodiments, the machine learning model is a neural network model.

[0057] In some embodiments, method 300 includes step 308 of generating optical measurement data for one or more test specimens containing a CuA device 202. In some embodiments, method 300 includes step 310 of determining one or more measurements of the geometric parameters of the CuA device 202 on the one or more test specimens by using the machine learning model with the optical measurement data for the one or more test specimens.

[0058] Once the above machine learning model has been trained, it can be used to generate measurements of the CuA device 202 or a part thereof (such as the memory array structure 204 or the CMOS structure 206) based on optical measurement data of additional test specimens. The above machine learning model can provide any type of measurement, including, but not limited to, overlay measurements, minimum linewidth (CD) measurements, shape measurements (such as height measurements, tilt measurements, and sidewall angle measurements), stress measurements, composition measurements, bandgap measurements, electrical property measurements, process state measurements (such as focus and / or exposure state, resist state, partial pressure, temperature, and focusing model), defect identification, or defect classification.

[0059] Next, with reference to Figures 1B to 1D, various non-limiting configurations of the characterization subsystem 102 according to one or more embodiments of the present disclosure will be described in detail.

[0060] In some embodiments, the characterization subsystem 102 is an optical measurement subsystem that generates measurement data based on the interaction between the sample 104 and light. Figure 1B is a schematic diagram of the characterization subsystem 102 configured as an optical characterization subsystem 102 according to one or more embodiments of the present disclosure. For example, the characterization subsystem 102 may include, but is not limited to, a spectroscopic ellipsometer (SE), an SE with multiple illumination angles, an SE that measures Müller matrix elements (e.g., using a rotational compensator), a single-wavelength ellipsometer, a beam profile ellipsometer (e.g., an angle-resolved ellipsometer), a beam profile reflectometer (e.g., an angle-resolved reflectometer), a broadband reflectometer (e.g., a spectroscopic reflectometer), a single-wavelength reflectometer, an angle-resolved reflectometer, an imaging system, a scatorometer (e.g., a speckle analyzer), or any combination thereof.

[0061] In some embodiments, the characterization subsystem 102 includes an illumination source 112 configured to produce at least one illumination beam 114. The illumination from the illumination source 112 may include one or more selected wavelengths of light, including but not limited to ultraviolet (UV), visible, or infrared (IR) radiation. For example, the characterization subsystem 102 may include one or more apertures in the illumination pupil plane. These apertures are for splitting the illumination from the illumination source 112 into one or more illumination beams 114 or illumination lobes. In this regard, the characterization subsystem 102 may provide two-pole illumination, four-pole illumination, and so on. Furthermore, the spatial profiles of one or more illumination beams 114 on the sample 104 can be controlled by apertures in the field of view to have any selected spatial profile.

[0062] The illumination source 112 may include any type of illumination source suitable for providing at least one illumination beam 114. In some embodiments, the illumination source 112 is a laser source. For example, the illumination source 112 may include, but is not limited to, one or more narrowband laser sources, broadband laser sources, supercontinuum laser sources, white light laser sources, etc. In some embodiments, the illumination source 112 includes a laser sustained plasma (LSP) source. For example, the illumination source 112 may include, but is not limited to, an LSP lamp, an LSP bulb, or an LSP chamber suitable for housing one or more elements that can emit broadband illumination when excited to a plasma state by a laser source. In some embodiments, the illumination source 112 includes a lamp light source. In some embodiments, the illumination source 112 may include, but is not limited to, an arc lamp, a discharge lamp, an electrodeless lamp, etc.

[0063] The illumination source 112 may provide one or more illumination beams 114 using free-space technology and / or optical fibers.

[0064] In some embodiments, the characterization subsystem 102 directs the illumination beam 114 to the sample 104 via an illumination path 118 by at least one illumination lens 116 (such as an objective lens). The illumination path 118 may include one or more optical components suitable for modifying and / or adjusting the illumination beam 114, and for directing the illumination beam 114 to the sample 104. In some embodiments, the illumination path 118 includes one or more illumination path optical elements 120 for shaping or otherwise controlling the illumination beam 114. For example, the illumination path optical elements 120 may include, but are not limited to, one or more field diaphragms, one or more pupil diaphragms, one or more polarizers, one or more filters, one or more beam splitters, one or more diffusers, one or more homogenizers, one or more apodizers, one or more beam shapers, or one or more mirrors (such as stationary mirrors, translational mirrors, scanning mirrors, etc.).

[0065] The characterization subsystem 102 may position the sample 104 for measurement using any suitable technique. In some embodiments, as shown in Figure 1B, the characterization subsystem 102 includes a sample stage 122 for positioning the sample 104 relative to the illumination beam 114, and the sample stage 122 includes one or more actuators (e.g., linear actuators, tip / tilt actuators, rotary actuators, etc.). In some embodiments, although not explicitly shown, the characterization subsystem 102 includes beam scanning optical elements (e.g., galvanometer mirrors, scanning prisms, etc.) for positioning and / or scanning one or more illumination beams 114.

[0066] In some embodiments, the characterization subsystem 102 includes at least one focusing lens 124. The focusing lens 124 is for capturing light or other radiation (referred to herein as focused light 126) emitted from the sample 104 and directing it through a focusing path 130 to one or more detectors 128. The focusing path 130 may include one or more optical elements suitable for modifying and / or adjusting the focused light 126 from the sample 104. In some embodiments, the focusing path 130 includes one or more focusing path optical elements 132 for shaping or otherwise controlling the focused light 126. For example, the focusing path optical elements 132 may include, but are not limited to, one or more field diaphragms, one or more pupil diaphragms, one or more polarizers, one or more filters, one or more beam splitters, one or more diffusers, one or more homogenizers, one or more apodizers, one or more beam shapers, or one or more mirrors (such as stationary mirrors, translational mirrors, scanning mirrors, etc.).

[0067] The characterization subsystem 102 may generally include any number or type of detectors 128. For example, the characterization subsystem 102 may include at least one single-pixel detector 128, which may be, for example, a photodiode, an avalanche photodiode, or a single-photon detector. Another example is that the characterization subsystem 102 may include at least one multi-pixel detector 128, which may be, for example, a charge-coupled device (CCD) or complementary metal-oxide-semiconductor (CMOS) device, a line detector, or a time-delay integral (TDI) detector.

[0068] The detector 128 can be positioned at any selected location within the focusing path 130. In some embodiments, the characterization subsystem 102 includes the detector 128 for generating an image of the sample 104 at the field of view (e.g., a plane conjugate to the sample 104). In some embodiments, the characterization subsystem 102 includes the detector 128 for generating an image of the pupil at the pupil (e.g., a diffraction plane). In this regard, the pupil image may correspond to the angular distribution of light from the sample 104 on the detector 128. For example, the diffraction order related to the diffraction of the illumination beam 114 from the sample 104 (e.g., an overlay target on the sample 104) may be imaged or otherwise observed at the pupil. In a general sense, the detector 128 can capture any combination of reflected (or transmitted), scattered, or diffracted light from the sample 104.

[0069] The illumination path 118 and focusing path 130 of the characterization subsystem 102 can be oriented in various configurations. For example, as shown in Figure 1B, the illumination path 118 and focusing path 130 may include non-overlapping optical paths. In some embodiments, although not explicitly shown, the characterization subsystem 102 may include a beam splitter, which is oriented so that a common objective lens can direct the illumination beam 114 toward the sample 104 while simultaneously capturing the focused light 126.

[0070] Figure 1C is a schematic diagram of a characterization subsystem 102 configured as an X-ray characterization subsystem 102 according to one or more embodiments of the present disclosure. Such a characterization subsystem 102 may, but is not limited to, a small-angle X-ray scatometer (SAXR) or a soft X-ray reflectometer (SXR). An overview of X-ray characterization systems and related measurement techniques is described in Patent Documents 12, 13, 14, 15, 16, 17, Non-Patent Document 1, Non-Patent Document 2, and 18, all of which are incorporated herein by reference.

[0071] In some embodiments, the illumination source 112 is an X-ray source configured to generate an X-ray illumination beam 114 having any particle energy (such as soft X-rays or hard X-rays). The characterization subsystem 102 may include any combination of components suitable for capturing the associated focusing signal 134, which may include, but is not limited to, X-ray emission, light emission, or particle emission.

[0072] For example, the characterization subsystem 102 may include an X-ray illumination lens 116 suitable for parallelizing or focusing the X-ray illumination beam 114, and a focusing path lens (not shown) suitable for focusing, parallelizing, and / or focusing the focusing signal 134 from the sample 104. Furthermore, the characterization subsystem 102 may include various illumination path optical elements (not shown) and / or focusing path optical elements (not shown), for example, X-ray parallelizing mirrors, specular X-ray optical elements (such as oblique incidence elliptical mirrors), polycapillary optical elements (such as hollow tube X-ray waveguides), multilayer optical elements or multilayer systems, or any combination thereof. In some embodiments, the characterization subsystem 102 includes an X-ray detector 128, for example, an X-ray monochromator (such as a crystalline monochromator like a Loxley-Tanner-Bowen monochromator), an X-ray aperture, an X-ray beam stop, or a diffractive optical element (such as a zone plate).

[0073] Figure 1D is a schematic diagram of a characterization subsystem 102 configured as a particle beam characterization subsystem 102 according to one or more embodiments of the present disclosure.

[0074] In one embodiment, the illumination source 112 includes a particle source (such as an electron beam source or an ion beam source), and therefore the illumination beam 114 includes a particle beam (such as an electron beam or a particle beam). The illumination source 112 may include any particle source known in the art that is suitable for generating the particle illumination beam 114. For example, the illumination source 112 may include, but is not limited to, an electron gun or an ion gun. In another embodiment, the illumination source 112 is configured to supply a particle beam having an adjustable energy. As an example, the illumination source 112 including an electron source may provide an acceleration voltage in the range of 0.1 kilovolts (kV) to 30 kV, but is not limited to that. As another example, the illumination source 112 including an ion source may provide an ion beam having an energy in the range of 1 kiloelectron volt (keV) to 50 keV, but is not required.

[0075] In another embodiment, the illumination path 118 includes one or more particle focusing elements (such as an illumination lens 116 and a focusing lens 124). For example, the one or more particle focusing elements may include, but are not limited to, a single particle focusing element or one or more particle focusing elements forming a composite system. In another embodiment, the one or more particle focusing elements include an illumination lens 116 configured to direct the particle illumination beam 114 toward the sample 104. Furthermore, the one or more particle focusing elements may include, but are not limited to, any type of electron lens known in the art, including electrostatic lenses, magnetic lenses, single-potential lenses, or dual-potential lenses.

[0076] In another embodiment, the characterization subsystem 102 includes one or more particle detectors 128 for imaging or otherwise detecting particles emanating from the sample 104. For example, the detector 128 may include an electron collector (such as a secondary electron collector or a backscatter electron detector). Alternatively, the detector 128 may include a photon detector (e.g., a photodetector, an X-ray detector, or a scintillation element coupled to a photomultiplier tube (PMT) detector) for detecting electrons and / or photons from the sample surface.

[0077] The subject matter described herein may, in some cases, represent various components that are contained within or connected to other components. It should be understood that such configurations are merely illustrative, and in practice, many other configurations can be implemented to achieve the same function. Conceptually, any arrangement of components to achieve the same function is effectively “associated” in such a way that the desired function is achieved. Therefore, any two components combined herein to achieve a particular function, regardless of each configuration or intermediate component, can be considered “associated” with one another in such a way that the desired function is achieved. Similarly, any two such associated components can be considered “connected” or “joined” with one another to achieve the desired function, and any two components that can be associated in such a way can be considered “joinable” with one another to achieve the desired function. Specific examples of joinability include, but are not limited to, components that are physically interactable and / or interact physically, and / or interact wirelessly and / or interact wirelessly, and / or interact logically and / or interact logically.

[0078] The present disclosure and its many associated advantages are understood from the above description, and it is clear that various modifications can be made to the form, structure, and arrangement of the above components without departing from the subject matter disclosed herein or sacrificing any of its important advantages. The forms described herein are for illustrative purposes only, and the following claims are intended to encompass such modifications. Furthermore, the present invention should be understood to be defined by the appended claims.

Claims

1. It is a system, The system includes a controller which includes one or more processors, the one or more processors being configured to execute program instructions that cause the one or more processors to perform a measurement recipe, and the measurement recipe is A step of receiving optical measurement data of one or more training samples comprising a complementary metal-oxide-semiconductor underarray (CuA) device, wherein the CuA device comprises a CMOS structure disposed beneath a periodic memory array structure; A step of receiving reference data of one or more training samples, wherein the reference data includes measured values ​​of the geometric parameters of the CuA device, The steps include training a machine learning model using the optical measurement data and reference data of one or more training samples, The steps include receiving optical measurement data of one or more test samples containing a CuA device, The steps include determining one or more measured values ​​of the geometric parameters of the CuA device on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples, A system implemented by [the organization / organization].

2. The aforementioned machine learning model, Neural Network The system according to claim 1, including the following:

3. The system according to claim 1, wherein the machine learning model utilizes at least one of unsupervised learning, supervised learning, or reinforcement learning.

4. The optical measurement data of the one or more training samples and the one or more test samples are At least one of the following: ellipsometric data, reflectometric data, or scantometric data. The system according to claim 1, including the following:

5. The reference data of the one or more training samples is At least one of the following: transmission electron microscope (TEM) data, scanning electron microscope (SEM) data, SAXS data, X-ray photoelectron spectroscopy (XPS) data, or X-ray diffraction (XRD) data. The system according to claim 1, including the following:

6. The reference data of the one or more training samples is Minimum linewidth small-angle X-ray spectroscopy (CD-SAXS) data The system according to claim 1, including the following:

7. The aforementioned 1 or more measurement values ​​are Measurement values The system according to claim 1, including the following:

8. The measured value is, At least one of the following: overlay measurement, minimum linewidth (CD) measurement, feature height measurement, or slope measurement. The system according to claim 7, including the system described in claim 7.

9. The aforementioned 1 or more measurement values ​​are Test measurement values The system according to claim 1, including the following:

10. The aforementioned test measurement values ​​are At least one of the following: identification or classification of defects in the second structure. The system according to claim 9, including the system described in claim 9.

11. It is a system, Optical properties evaluation system, Reference characteristic evaluation system, A controller comprising one or more processors, wherein the one or more processors are configured to execute program instructions causing the one or more processors to perform a measurement recipe, and the measurement recipe is The steps include receiving optical measurement data of one or more training samples, each containing a complementary metal-oxide-semiconductor underarray (CuA) device, from an optical characterization system, wherein the CuA device includes a CMOS structure disposed beneath a periodic memory array structure; A step of receiving reference data of one or more training samples, wherein the reference data includes measured values ​​of the geometric parameters of the CuA device from the reference characteristic evaluation system. The steps include training a machine learning model using the optical measurement data and reference data of one or more training samples, The steps include receiving optical measurement data of one or more test samples containing a CuA device from the optical property evaluation system, The steps include determining one or more measured values ​​of the geometric parameters of the CuA device on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples, A system implemented by [the organization / organization].

12. The optical characteristics evaluation system At least one of the following: ellipsometer, reflectometer, or scatometer The system according to claim 11, including the following:

13. The aforementioned reference characteristic evaluation system At least one of either an X-ray characterization system or a particle beam characterization system. The system according to claim 11, including the following:

14. It is a method, A step of generating optical measurement data of one or more training samples comprising complementary metal-oxide-semiconductor (CMOS) underarray (CuA) devices, wherein the CuA devices comprise a CMOS structure disposed beneath a periodic memory array structure; A step of generating reference data for one or more training samples, wherein the reference data includes measured values ​​of the geometric parameters of the CuA device. The steps include training a machine learning model using the optical measurement data and reference data of one or more training samples, A step of generating optical measurement data for one or more test samples containing a CuA device, The steps include determining one or more measured values ​​of the geometric parameters of the CuA device on the one or more test samples by using the machine learning model with the optical measurement data of the one or more test samples, Methods that include...

15. The aforementioned machine learning model, Neural Network The method according to claim 14, including the method described in claim 14.

16. The method according to claim 14, wherein the machine learning model utilizes at least one of unsupervised learning, supervised learning, or reinforcement learning.

17. The optical measurement data of the one or more training samples and the one or more test samples are At least one of the following: ellipsometric data, reflectometric data, or scantometric data. The method according to claim 14, including the method described in claim 14.

18. The reference data of the one or more training samples is At least one of the following: transmission electron microscope (TEM) data, scanning electron microscope (SEM) data, small-angle X-ray spectroscopy (SAXS) data, X-ray photoelectron spectroscopy (XPS) data, or X-ray diffraction (XRD) data. The method according to claim 14, including the method described in claim 14.

19. The reference data of the one or more training samples is Minimum linewidth small-angle X-ray spectroscopy (CD-SAXS) data The method according to claim 14, including the method described in claim 14.

20. The aforementioned 1 or more measurement values ​​are Measurement values The method according to claim 14, including the method described in claim 14.

21. The measured value is, At least one of the following: overlay measurement, minimum linewidth (CD) measurement, shape measurement, stress measurement, composition measurement, bandgap measurement, electrical property measurement, or process state measurement. The method according to claim 20, including the method described in claim 20.

22. The aforementioned 1 or more measurement values ​​are Test measurement values The method according to claim 14, including the method described in claim 14.

23. The aforementioned test measurement values ​​are At least one of the following: identification or classification of defects in the second structure. The method according to claim 22, including the method described in claim 22.

Images