System and method for rendering SEM images and predicting defect imaging conditions for substrates using 3D designs

Abstract

A system for characterizing a specimen is disclosed. In one embodiment, the system includes a characterization subsystem configured to acquire one or more images of the specimen, and a controller communicatively coupled to the characterization subsystem. The controller may be configured to receive training images of one or more features of the specimen from the characterization subsystem, receive training three-dimensional (3D) design images corresponding to the one or more features of the specimen, generate a deep learning predictive model based on the training images and the training 3D design images, receive product 3D design images of the one or more features of the specimen, generate simulated images of the one or more features of the specimen using the deep learning predictive model based on the product 3D design images, and determine one or more properties of the specimen based on the one or more simulated images.

Description

[Technical field]

[0001] The present invention relates generally to the field of specimen characterization and metrology, and more particularly to a system and method for simulating defects in a specimen using machine learning techniques. [Background technology]

[0002] (CROSS REFERENCE TO RELATED APPLICATIONS) This application claims priority to Indian Provisional Patent Application No. 201941030221, filed on July 26, 2019, titled “RENDERING SEM IMAGE AND PREDICTING DEFECT IMAGING CONDITION USING 3D DESIGN” and listing Arpit Yati and Chandrashekaran Gurumurthy as inventors, the disclosures of which are incorporated herein by reference in their entireties.

[0003] The demand for electronic logic and memory devices with smaller footprints and functionality presents a wide range of manufacturing challenges beyond the scope of fabrication at the desired scale. In the context of semiconductor manufacturing, predicting and removing defects in semiconductor devices is a critical step in improving throughput and yield. One of the main challenges in identifying defects is the inability to accurately predict the location, type, and structure of the defects.

[0004] Conventional techniques used to predict the relative location, type, and structure of defects involve the use of simulation. For example, in situations where a process engineer knows the approximate location, type, and general structure of a probable defect, a Monte Carlo simulation may be used to simulate a scanning electron microscope (SEM) image of the defect. Similarly, optical simulations that solve Maxwell's equations may be used to predict the imaging conditions required to predict and identify defect signals. The simulated SEM images and identified imaging conditions may then be used to more effectively identify defects, improving throughput and yield. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] US Patent Application Publication No. 2017 / 0148226 [Patent Document 2] US Patent Application Publication No. 2019 / 0228522 Summary of the Invention [Problem to be solved by the invention]

[0006] These conventional techniques utilizing Monte Carlo simulation suffer from several drawbacks. First, these conventional simulation techniques are computationally expensive and time consuming. In particular, running a Monte Carlo simulation to generate SEM images based on design images / data can be very time consuming. Thus, these techniques may only be able to predict / identify a few defects per day, which greatly impedes the throughput of the semiconductor manufacturing process. In addition, Monte Carlo-based simulation techniques cannot render design images / data based on SEM images and cannot determine material properties of specimens based on simulated SEM images.

[0007] It would therefore be desirable to provide a system and method that obviates one or more of the shortcomings of previous approaches identified above. [Means for solving the problem]

[0008] A system for characterizing a specimen is disclosed. In one embodiment, the system includes a characterization subsystem configured to acquire one or more images of the specimen. In another embodiment, the system includes a controller communicatively coupled to the characterization subsystem. The controller may be configured to receive one or more training images of one or more features of the specimen from the characterization subsystem, receive one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen, generate a deep learning predictive model based on the one or more training images and the one or more training 3D design images, receive one or more product 3D design images of the one or more features of the specimen, generate one or more simulated images of the one or more features of the specimen using the deep learning predictive model based on the one or more product 3D design images, and determine one or more properties of the specimen based on the one or more simulated images.

[0009] A system for characterizing a specimen is disclosed. In one embodiment, the system includes a controller configured to: receive one or more training images of one or more features of the specimen, receive one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen, generate a deep learning predictive model based on at least the one or more training images and the one or more training 3D design images, receive one or more product 3D design images of the one or more features of the specimen, generate one or more simulated images of the one or more features of the specimen based on the one or more product 3D design images using the deep learning predictive model, and determine one or more properties of the specimen based on the one or more simulated images.

[0010] A system for characterizing a specimen is disclosed. In one embodiment, the system includes a characterization subsystem configured to acquire one or more images of the specimen, and a controller communicatively coupled to the characterization subsystem. The controller may be configured to receive one or more training images of one or more features of the specimen from the characterization subsystem, receive one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen, generate a deep learning predictive model based on the one or more training images and the one or more training 3D design images, receive one or more product images of the one or more features of the specimen, generate one or more simulated 3D design images of the one or more features of the specimen based on the one or more product images using the deep learning predictive model, and determine one or more properties of the specimen based on the one or more simulated 3D design images.

[0011] A system for characterizing a specimen is disclosed. In one embodiment, the system includes a controller configured to: receive one or more training images of one or more features of the specimen, receive one or more training 3D design images corresponding to the one or more features of the specimen, generate a deep learning predictive model based on at least the one or more training images and the one or more training 3D design images, receive one or more product images of the one or more features of the specimen, generate one or more simulated 3D design images of the one or more features of the specimen based on the one or more product images using the deep learning predictive model, and determine one or more properties of the specimen based on the one or more simulated 3D design images.

[0012] A method for characterizing a specimen is disclosed. In one embodiment, the method includes acquiring one or more training images of one or more features of the specimen with a characterization subsystem, receiving one or more training three-dimensional (3D) design images of the one or more features of the specimen, generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images, receiving one or more product 3D design images of the one or more features of the specimen, generating one or more simulated images of the one or more features of the specimen using the deep learning predictive model based on the one or more product 3D design images, and determining one or more properties of the specimen based on the one or more simulated images.

[0013] A method for characterizing a specimen is disclosed. In one embodiment, the method includes acquiring one or more training images of one or more features of the specimen with a characterization subsystem, receiving one or more training three-dimensional (3D) design images of the one or more features of the specimen, generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images, receiving one or more product images of the one or more features of the specimen, generating one or more simulated design images of the one or more features of the specimen based on the one or more product images using the deep learning predictive model, and determining one or more properties of the specimen based on the one or more simulated design images.

[0014] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not necessarily restrictive of the invention as claimed. The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and, together with the general description, serve to explain the principles of the invention.

[0015] Many advantages of the present disclosure may be better understood, by those skilled in the art, by reference to the following accompanying drawings. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 shows a flow chart for generating a simulated SEM image of a specimen using Monte Carlo simulation. [Diagram 2] 1 illustrates a system that utilizes machine learning techniques to generate simulated images of a specimen to characterize the specimen, in accordance with one or more embodiments of the present disclosure. [Figure 3A] 1 illustrates a system that utilizes machine learning techniques to generate simulated images of a specimen to characterize the specimen, in accordance with one or more embodiments of the present disclosure. [Figure 3B] 1 illustrates a system that utilizes machine learning techniques to generate simulated images of a specimen to characterize the specimen, in accordance with one or more embodiments of the present disclosure. [Figure 4A] 1 illustrates a flowchart for training a deep learning predictive model in accordance with one or more embodiments of the present disclosure. [Figure 4B] 1 illustrates a flowchart of a trained deep learning predictive model configured to generate a simulated image of a specimen based on a product 3D design image of the specimen, in accordance with one or more embodiments of the present disclosure. [Figure 4C] 1 illustrates a flowchart of a trained deep learning predictive model configured to generate a simulated three-dimensional design image of a specimen based on a product image of the specimen, in accordance with one or more embodiments of the present disclosure. [Diagram 5] 1 illustrates a flowchart for training a deep learning predictive model including a conditional generative adversarial network (CGAN), in accordance with one or more embodiments of the present disclosure. [Figure 6] 1 illustrates a flowchart for adaptively modifying a three-dimensional design image used to create a specimen in a production process that utilizes a deep learning predictive model, in accordance with one or more embodiments of the present disclosure. [Figure 7] 1 illustrates a flowchart of a method for generating a simulated image based on a three-dimensional design image utilizing a deep learning predictive model, in accordance with one or more embodiments of the present disclosure. [Figure 8]1 illustrates a flowchart of a method for generating simulated three-dimensional design images based on scanning electron microscope (SEM) images utilizing a deep learning predictive model in accordance with one or more embodiments of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] The present disclosure has been particularly shown and described with respect to certain embodiments and certain features thereof. The embodiments described herein are to be construed as illustrative and not limiting. It will be readily apparent to those skilled in the art that various changes and modifications in form and detail may be made therein without departing from the spirit and scope of the present disclosure.

[0018] Reference will now be made in detail to the disclosed subject matter, which is illustrated in the accompanying drawings.

[0019] Embodiments of the present disclosure relate to systems and methods for generating simulated scanning electron microscope (SEM) images and / or simulated optical images based on three-dimensional (3D) design images utilizing machine learning techniques (such as deep learning predictive models). Additional embodiments of the present disclosure relate to systems and methods for generating simulated three-dimensional (3D) design images based on scanning electron microscope (SEM) images utilizing machine learning techniques (such as deep learning predictive models). Further embodiments of the present disclosure relate to adaptively modifying a 3D design image of a specimen utilizing simulated SEM images and / or simulated 3D design images generated using machine learning techniques.

[0020] It is contemplated herein that embodiments of the present disclosure may enable faster and more efficient identification of specimen defects throughout the fabrication process, thereby improving throughput and yield. Additionally, embodiments of the present disclosure may enable adaptive modification of 3D design data to improve specimen fabrication throughout the fabrication process. It is further contemplated that the systems and methods of the present disclosure may enable generation of 3D design images / files from SEM images, as well as prediction of material properties utilizing the generated simulated SEM images, which is not possible with conventional modeling techniques.

[0021] 1 illustrates a flowchart 100 for generating a simulated SEM image of a specimen using Monte Carlo simulation. The flowchart 100 illustrated in FIG. 1 is provided as an example of a conventional simulation technique utilizing Monte Carlo simulation. It is contemplated herein that a brief description of the flowchart 100 may provide a baseline against which the attendant advantages of the present disclosure may be compared.

[0022] To predict the relative locations, types, and structures of defects in a specimen (e.g., a semiconductor wafer), a conventional technique involves inputting a three-dimensional (3D) design image 102 (e.g., a 3D design file 102) into one or more processors and running a Monte Carlo simulation 104. The Monte Carlo simulation 104 may then generate one or more simulated scanning electron microscope (SEM) images 106 that are intended to mimic the relative locations, types, and structures of defects in the specimen.

[0023] These conventional techniques utilizing Monte Carlo simulation 104 suffer from several drawbacks. First, these conventional simulation techniques are computationally expensive and time consuming. In particular, running the Monte Carlo simulation 104 to generate simulated SEM images 106 based on the 3D design images 102 can be very time consuming. Thus, these techniques may only be able to predict / identify a few defects per day, which greatly impedes the throughput of the semiconductor manufacturing process. Furthermore, Monte Carlo-based simulation techniques cannot render the 3D design images 102 based on the acquired SEM images, and cannot predict the material properties of the specimen based on the simulated SEM images 106.

[0024] Accordingly, embodiments of the present disclosure are directed to systems and methods that obviate one or more of the shortcomings of previous approaches identified above.

[0025] 2 illustrates a system 200 for generating simulated images of a specimen utilizing machine learning techniques to characterize the specimen, in accordance with one or more embodiments of the present disclosure. System 200 may include, but is not limited to, one or more characterization subsystems 202. System 200 may further include, but is not limited to, a controller 204 including one or more processors 206 and memory 208, and a user interface 210.

[0026] The characterization subsystem 202 may include any characterization subsystem 202 known in the art, including, but not limited to, an optical-based characterization system, a charged particle-based characterization system, etc. For example, the characterization subsystem 202 may include a scanning electron microscope (SEM) characterization system. In one embodiment, the controller 204 is communicatively coupled to the one or more characterization subsystems 202. In this regard, the one or more processors 206 of the controller 204 may be configured to generate one or more control signals that are configured to adjust one or more characteristics of the characterization subsystem 202.

[0027] 3A illustrates a system 200 that utilizes machine learning techniques to generate simulated images of a specimen to characterize the specimen, in accordance with one or more embodiments of the present disclosure. In particular, FIG. 3A illustrates system 200 that includes an optical characterization subsystem 202a.

[0028] The optical characterization subsystem 202a may include any optically based characterization system known in the art, including but not limited to an image-based metrology tool. For example, the characterization subsystem 202a may include an optical critical dimension metrology tool. The optical characterization subsystem 202a may include, but is not limited to, an illumination source 212, an illumination arm 211, a collection arm 213, and a detector assembly 226.

[0029] In one embodiment, the optical characterization subsystem 202a is configured to inspect and / or measure the specimen 220 disposed on the stage assembly 222. The illumination source 212 may include any illumination source known in the art for generating the illumination 201, including but not limited to a broadband radiation source. In another embodiment, the optical characterization subsystem 202a may include an illumination arm 211 configured to direct the illumination 201 toward the specimen 220. It is noted that the illumination source 212 of the optical characterization subsystem 202a may be configured in any orientation known in the art, including but not limited to a dark field orientation, a bright field orientation, etc.

[0030] The specimen 220 may include any specimen known in the art, including, but not limited to, a wafer, a reticle, a photomask, and the like. In one embodiment, the specimen 220 is disposed on a stage assembly 222 to facilitate movement of the specimen 220. In another embodiment, the stage assembly 222 is an actuatable stage. For example, the stage assembly 222 may include, but is not limited to, one or more translation stages adapted to selectively translate the specimen 220 along one or more linear directions (e.g., an x-direction, a y-direction, and / or a z-direction). As another example, the stage assembly 222 may include, but is not limited to, one or more rotation stages adapted to selectively rotate the specimen 220 along a rotational direction. As another example, the stage assembly 222 may include, but is not limited to, a rotational stage and a translation stage adapted to selectively translate the specimen 220 along a linear direction and / or rotate the specimen 220 along a rotational direction. It is noted herein that the system 200 may operate in any scanning mode known in the art.

[0031] The illumination arm 211 may include any number and type of optical components known in the art. In one embodiment, the illumination arm 211 includes one or more optical elements 214, a beam splitter 216, and an objective lens 218. In this regard, the illumination arm 211 may be configured to focus illumination 201 from an illumination source 212 onto a surface of a specimen 220. The one or more optical elements 214 may include any optical element known in the art, including, but not limited to, one or more mirrors, one or more lenses, one or more polarizers, one or more beam splitters, etc.

[0032] In another embodiment, the optical characterization subsystem 202a includes a collection arm 213 configured to collect illumination reflected or scattered from the specimen 220. In another embodiment, the collection arm 213 may direct and / or focus the reflected and scattered light via one or more optical elements 224 to one or more sensors of a detector assembly 226. The one or more optical elements 224 may include any optical element known in the art, such as, but not limited to, one or more mirrors, one or more lenses, one or more polarizers, one or more beam splitters, etc. It is noted that the detector assembly 226 may include any sensor and detector assembly known in the art for detecting illumination reflected or scattered from the specimen 220.

[0033] In another embodiment, the detector assembly 226 of the optical characterization subsystem 202 is configured to collect metrology data of the specimen 220 based on illumination reflected or scattered from the specimen 220. In another embodiment, the detector assembly 226 is configured to transmit the collected / acquired images and / or metrology data to the controller 204.

[0034] As previously described herein, the controller 204 of the system 200 may include one or more processors 206 and a memory 208. The memory 208 may include program instructions configured to cause the one or more processors 206 to perform various steps of the present disclosure. In one embodiment, the program instructions are configured to cause the one or more processors 206 to adjust one or more properties of the optical characterization subsystem 202 to perform one or more measurements of the specimen 220.

[0035] In additional and / or alternative embodiments, the characterization subsystem 202 may include a charged particle based characterization subsystem 202. For example, the characterization subsystem 202 may include a SEM characterization subsystem, as shown in FIG.

[0036] 3B illustrates a system for utilizing machine learning techniques to generate simulated images of a specimen to characterize the specimen, in accordance with one or more embodiments of the present disclosure. In particular, FIG. 3B illustrates a system 200 that includes a SEM characterization subsystem 202b.

[0037] In one embodiment, the SEM characterization subsystem 202b is configured to perform one or more measurements on the specimen 220. In this regard, the SEM characterization subsystem 202b may be configured to acquire one or more images of the specimen 220. The SEM characterization subsystem 202b may include, but is not limited to, an electron beam source 228, one or more electron optical elements 230, one or more electron optical elements 232, and an electron detector assembly 234 including one or more electron sensors 236.

[0038] In one embodiment, the electron beam source 228 is configured to direct one or more electron beams 229 to the specimen 220. The electron beam source 228 may form an electron optical column. In another embodiment, the electron beam source 228 includes one or more additional and / or alternative electron optical elements 230 that are configured to focus and / or direct the one or more electron beams 229 to the surface of the specimen 220. In another embodiment, the SEM characterization subsystem 202b includes one or more electron optical elements 232 that are configured to collect secondary and / or backscattered electrons 231 emitted from the surface of the specimen 220 in response to the one or more electron beams 229. It is noted herein that the one or more electron optical elements 230 and the one or more electron optical elements 232 may include any electron optical element configured to direct, focus, and / or collect electrons, including, but not limited to, one or more deflectors, one or more electron optical lenses, one or more condenser lenses (e.g., magnetic condenser lenses), one or more objective lenses (e.g., magnetic condenser lenses), etc.

[0039] It is noted that the electron-optical assembly of the SEM characterization subsystem 202b is not limited to the electron-optical elements shown in Figure 3B, which are provided for illustrative purposes only. It is further noted that the system 200 may include any number and type of electron-optical elements necessary to direct / focus one or more electron beams 229 onto the specimen 220 and, in response, collect and image emitted secondary and / or backscattered electrons 231 onto the electron detector assembly 234.

[0040] For example, the system 200 may include one or more electron beam scanning elements (not shown) including, but not limited to, one or more electromagnetic scanning coils or electrostatic deflectors suitable for controlling the position of the one or more electron beams 229 relative to the surface of the specimen 220. Additionally, the one or more scanning elements may be utilized to scan the one or more electron beams 229 across the specimen 220 in a selected pattern.

[0041] In another embodiment, the secondary and / or backscattered electrons 231 are directed to one or more sensors 236 of the electron detector assembly 234. The electron detector assembly 234 of the SEM characterization subsystem 202 may include any electron detector assembly known in the art suitable for detecting backscattered and / or secondary electrons 231 emitted from the surface of the specimen 220. In one embodiment, the electron detector assembly 234 includes an electron detector array. In this regard, the electron detector assembly 234 may include an array of electron detecting portions. Further, each electron detecting portion of the detector array of the electron detector assembly 234 may be positioned to detect an electron signal from the specimen 220 associated with one of the one or more electron beams 229 incident thereon. In this regard, each channel of the electron detector assembly 234 may correspond to an electron beam 229 of the one or more electron beams 229. The electron detector assembly 234 may include any type of electron detector known in the art. For example, the electronic detector assembly 234 may include a microchannel plate (MCP), PIN or pn junction detector array, such as, but not limited to, a diode array or an avalanche photodiode (APD). As another example, the electronic detector assembly 234 may include a fast scintillator / PMT detector.

[0042] 3B illustrates the SEM characterization subsystem 202b as including an electron detector assembly 234 that includes only a secondary electron detector assembly, which should not be considered a limitation of the present disclosure. In this regard, it is noted that the electron detector assembly 234 may include, but is not limited to, a secondary electron detector, a backscattered electron detector, and / or a primary electron detector (e.g., an in-column electron detector). In another embodiment, the SEM characterization subsystem 202 may include multiple electron detector assemblies 234. For example, the system 200 may include a secondary electron detector assembly 234a, a backscattered electron detector assembly 234b, and an in-column electron detector assembly 234c.

[0043] In one embodiment, the one or more processors 206 of the controller 204 are configured to analyze the output of the detector assembly 226 / electron detector assembly 234. In one embodiment, the set of program instructions is configured to cause the one or more processors 206 to analyze one or more characteristics of the specimen 220 based on an image received from the detector assembly 226 / electron detector assembly 234. In another embodiment, the set of program instructions is configured to cause the one or more processors 206 to modify one or more characteristics of the system 200 to maintain focus on the specimen 220 and / or the detector assembly 226 / electron detector assembly 234. For example, the one or more processors 206 may be configured to adjust one or more characteristics of the illumination source 212 / electron beam source 228 and / or other elements of the system 200 to focus the illumination 201 and / or the one or more electron beams 229 on the surface of the specimen 220. As another example, the one or more processors 206 may be configured to adjust one or more elements of the system 200 to collect illumination and / or secondary electrons 231 from a surface of the specimen 220 and focus the collected illumination onto the detector assembly 226 / electron detector assembly 234. As another example, the one or more processors 206 may be configured to adjust one or more focusing voltages applied to one or more electrostatic deflectors of the electron beam source 228 to independently adjust the position or alignment of one or more electron beams 229 to scan the electron beams 229 across the specimen 220.

[0044] In one embodiment, the one or more processors 206 may be communicatively coupled to memory 208, where the one or more processors 206 are configured to execute a set of program instructions stored in memory 208, the set of program instructions configured to cause the one or more processors 206 to perform various functions and steps of the present disclosure.

[0045] 2 and 3A-3B, system 200 includes a user interface 210 communicatively coupled to controller 204. In another embodiment, user interface 210 includes a user input device and a display. The user input device of user interface 210 may be configured to receive one or more input commands from a user, where the one or more input commands are configured to input data into system 200 and / or adjust one or more characteristics of system 200. In another embodiment, the display of user interface 210 may be configured to display system 200 data to a user.

[0046] As previously described herein, the one or more processors 206 of the controller 204 may be communicatively coupled to the memory 208, where the one or more processors 206 are configured to execute a set of program instructions stored in the memory 208, the set of program instructions configured to cause the one or more processors 206 to perform various functions and steps of the present disclosure. In this regard, the controller 204 may be configured to: receive one or more training images 225 of one or more features of the specimen 220 from the characterization subsystem 202, receive one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen 220, generate a deep learning predictive model based on the one or more training images 225 and the one or more training 3D design images, receive one or more 3D product design images of the one or more features of the specimen 220, generate one or more simulated images of the one or more features of the specimen 220 using the deep learning predictive model based on the one or more 3D product design images, and determine one or more properties of the specimen 220 based on the one or more simulated images. Each of these steps / functions of the controller 204 is described in further detail herein.

[0047] 4A illustrates a flowchart 400 for training a deep learning predictive model in accordance with one or more embodiments of the present disclosure. In this regard, flowchart 400 may be considered a conceptual flowchart illustrating steps performed by / within one or more processors 206 of controller 204.

[0048] In one embodiment, the controller 204 is configured to receive from the characterization subsystem 202 one or more training images 225 of one or more features of the specimens 220. For purposes of this disclosure, the term "training images" may be considered images that are used as inputs to train a machine learning classifier.

[0049] For example, as shown in FIG. 3A, the controller 204 may be configured to receive one or more optical training images 225 of one or more features of the specimen 220 from the optical characterization subsystem 202a. As another example, as shown in FIG. 3B, the controller 204 may be configured to receive one or more SEM training images 225 of one or more features of the specimen 220 from the SEM characterization subsystem 202b. In this regard, the training images 225 shown in FIG. 4A may include optical training images 225, SEM training images 225, etc. In additional and / or alternative embodiments, the controller 204 may be configured to receive the one or more training images 225 from a source other than the one or more characterization subsystems 202. For example, the controller 204 may be configured to receive the one or more training images 225 of the features of the specimen 220 from an external storage device and / or the memory 208. In another embodiment, the controller 204 may be further configured to store the received training images 225 in the memory 208.

[0050] The one or more features of the specimen may include any feature that may be of interest throughout the fabrication / characterization process of the specimen 220, including, but not limited to, patterned structures, defects, high aspect ratio (HAR) structures, critical dimension (CD) structures, etc. The controller 204 may be configured to store one or more training images 225 in the memory 208.

[0051] In another embodiment, the controller 204 is configured to receive one or more training three-dimensional (3D) design images 404 corresponding to one or more features of the specimen 220. The term "training 3D design image 404" and similar terms may be considered any data, file, or image related to the design of the specimen 220. In this regard, the term "3D design image" may be used interchangeably with the terms "3D design data" and "3D design file." As used herein, it is contemplated that the training 3D design images 404 may include material information and / or specimen topography information, which may enable the generation of a more sophisticated deep learning predictive model 402, which is configured to determine the material and / or topography properties of the specimen 220 in a subsequent step. In some embodiments, the one or more training 3D design images 404 are stored in the memory 208, such that the controller 204 is configured to receive and / or retrieve the one or more training 3D design images 404 from the memory 208.

[0052] In another embodiment, the controller 204 is configured to generate the deep learning predictive model 402 based on the one or more training images 225 and the one or more training 3D design images 404. In this regard, the one or more training images 225 and the one or more training 3D design images 404 may be used as inputs to train the deep learning predictive model 402. The deep learning predictive model 402 may include any type of machine learning algorithm / classifier and / or deep learning technique or classifier known in the art, including, but not limited to, conditional generative adversarial networks (CGANs), convolutional neural networks (CNNs) (e.g., GoogleNet, AlexNet, etc.), ensemble learning classifiers, random forest classifiers, artificial neural networks (ANNs), etc. In an embodiment, the controller 204 may be configured to store the generated deep learning predictive model 402 in the memory 208.

[0053] It is contemplated herein that the controller 204 may be configured to generate the deep learning predictive model 402 via supervised learning and / or unsupervised learning. Training / generating the deep learning predictive model 402 may include training the deep learning predictive model 402 to generate simulated images 408 based on the product 3D design images 406, as shown in FIG. 4B. Conversely, training / generating the deep learning predictive model 402 may include training the deep learning predictive model 402 to generate simulated 3D design images 410 based on the product images 235, as shown in FIG. 4C. In this regard, the deep learning predictive model 402 may include any algorithm, classifier, or predictive model configured to generate the simulated images 408 and / or the simulated 3D design images 410, as will be discussed in more detail herein.

[0054] FIG. 4B illustrates a flowchart 401 of a trained deep learning predictive model 402 configured to generate a simulated image 408 of a specimen 220 based on a product 3D design image 406 of the specimen 220 in accordance with one or more embodiments of the present disclosure.

[0055] In one embodiment, after the deep learning predictive model 402 is trained, the controller 204 may be configured to receive one or more product 3D design images 406 of one or more features of the specimen 220. As used herein, the term "product" (e.g., "product" 3D design image 406, "product" image 435) may be used to refer to design images / data and / or images that are not used to train the deep learning predictive model 402. In particular, the product 3D design image 406 may refer to design data / images of the product specimen 220 of a fabrication / characterization process for which a simulated image 408 is desired. For example, the product 3D design image 406 may include a design image of the specimen 220 to be fabricated in a fabrication process. As previously described, the one or more product 3D design images 406 may be stored in the memory 208 such that the controller 204 is configured to receive and / or retrieve one or more product 3D design images 410 from the memory 208.

[0056] In an embodiment, the one or more features of the specimen 220 may include any feature of interest throughout the fabrication / characterization process of the specimen 220, including, but not limited to, patterned structures, defects, high aspect ratio (HAR) structures, critical dimension (CD) structures, etc. For example, when designing the specimen 220 to be fabricated, a process engineer may know the relative location, type, and rough structure of defects that are likely to be created during the fabrication of the specimen 220. To predict what optical and / or SEM images of likely defects will look like to more efficiently identify the defects after fabrication, the process engineer may generate manufacturing 3D design images 406 of likely defects at various depths. These product 3D design images 406 may then be input into the deep learning predictive model 402.

[0057] In another embodiment, the controller 204 is configured to generate one or more simulated images 408 of one or more features of the specimen 220 based on one or more 3D product design images using the deep learning predictive model 402. For example, continuing with the same example above, one or more product 3D design images 406 of likely defects (e.g., features) in the future specimen 220 may be input into the deep learning predictive model 402. The controller 204 may be configured to receive these product 3D design images 406 of the likely defects and generate one or more simulated images 408 of the likely defects (e.g., features) using the deep learning predictive model 402. In this regard, a user may be able to visualize the intensity of various likely defects in the simulated images 408. In an embodiment, the deep learning predictive model 402 may be configured to generate simulated SEM images (e.g., simulated SEM images 408) of the likely defects and / or simulated optical images (e.g., simulated optical images 408) of the likely defects. The controller 204 may be configured to store one or more simulated images 408 in the memory 208 .

[0058] In an embodiment, the simulated image 408 (e.g., simulated SEM image 408, simulated optical image 408) generated by the deep learning predictive model 402 may be used as a reference image for subsequent characterization, inspection, and detection. For example, by generating the simulated image 408 of the defect, a user and / or the system 200 may be configured to recognize / identify a likely characteristic of the defect (e.g., defect intensity, defect size, etc.). When the specimen 220 is subsequently inspected, the system 200 may be configured to identify the defect more quickly and efficiently based on the known likely characteristics of the defect identified in the simulated image 408. Furthermore, by using the simulated image 408 as a reference image, the controller 204 may be configured to identify defect imaging conditions of the characterization subsystem 202 that are optimized for characterizing the defect in a subsequent inspection. The controller 204 may generate one or more control signals configured to selectively adjust one or more characteristics of the characterization subsystem 202 to achieve the identified defect imaging conditions.

[0059] In another embodiment, the controller 204 is configured to determine one or more characteristics of the specimen 220 based on the one or more simulated images 408. Continuing with the same example above, the controller 204 may be configured to determine one or more characteristics of a likely defect (e.g., feature) in the future specimen 220 based on the generated simulated SEM image 408 and / or the generated simulated optical image 408. The properties of the specimen 220 and / or the features of the specimen 220 that may be determined based on the simulated images 408 may include any property of the specimen 220 known in the art, including, but not limited to, the location of the defect in the specimen 220, the type of defect in the specimen 220, the material of the specimen 220, and topography measurements of the surface and / or patterns of the specimen 220, etc.

[0060] As previously discussed, conventional modeling techniques based on Monte Carlo simulations may not be able to determine material properties of a specimen based on simulated images generated by the Monte Carlo simulations. In comparison, the machine learning techniques of the present disclosure may enable material property determination based on simulated images 408 and / or simulated 3D design images 410 generated by the deep learning predictive model 402.

[0061] In another embodiment, the controller 204 may be configured to adaptively modify one or more characteristics of one or more product 3D design images 406 of the specimen 220 to generate one or more modified product 3D design images. For example, if one or more product 3D design images 406 include design data of likely defects in the specimen 220 to be fabricated, the controller 204 may utilize information collected from the simulated images 408 to adaptively modify the product 3D design images 406 to generate one or more modified product 3D design images. By modifying the 3D design image / data of the specimen 220 prior to fabrication, various features and characteristics of the specimen 220 may be modified during fabrication. For example, by modifying the 3D design image / data of the specimen 220 prior to fabrication, the system 200 may reduce the number of defects in the specimen 220, reduce the size of defects in the specimen 220, modify characteristics of defects in the specimen 220 to make them easier to identify, etc.

[0062] In some embodiments, the controller 204 may be further configured to generate control signals configured to cause the fabrication tools to fabricate the specimen 220 based on the modified product 3D design image. For example, with reference to FIGS. 3A-3B, the system 200 may further include one or more fabrication tools communicatively coupled to the controller 204. The one or more fabrication tools may include any fabrication tool known in the art configured to fabricate the specimen 220, including, but not limited to, lithography tools, etching tools, deposition tools, polishing tools, etc. Continuing with the same example, the controller 204 may be configured to generate one or more control signals configured to cause the fabrication tools to fabricate the specimen 220 based on the modified product 3D design image stored in the memory 208.

[0063] It is noted herein that the generation of the simulated SEM image 408 may enable improved SEM-to-design alignment. In particular, by generating the simulated SEM image 408 based on the product 3D design image 406, the controller 204 may be configured to more efficiently orient and align the simulated SEM image 408 with the product 3D design image 406 than would otherwise be possible with a non-simulated SEM image. In this regard, embodiments of the present disclosure may enable mapping of characterizations of features of the specimen 220 identified in the simulated SEM image 408 to their respective locations within the product 3D design image. Additionally, embodiments of the present disclosure may enable more accurate critical dimension (CD) overlay.

[0064] In additional and / or alternative embodiments, the controller 204 may be configured to perform a Monte Carlo simulation to evaluate the accuracy of the deep learning predictive model 402 and / or to validate the resulting simulated SEM images 408 against the unavailability of non-simulated SEM images. For example, as shown in FIG. 4B, the controller 204 may be configured to generate one or more simulated SEM images 408 from one or more product 3D design images 406 via the deep learning predictive model 402. The controller 204 may then be configured to perform one or more Monte Carlo simulations on one or more product 3D design images corresponding to one or more features of the specimen 220 to generate one or more comparison images (e.g., comparative SEM images). The controller 204 may then be configured to determine an accuracy metric of the deep learning predictive model 402 by comparing one or more simulated SEM images 408 generated by the deep learning predictive model 402 to one or more comparative SEM images generated by the Monte Carlo simulation.

[0065] In addition to generating simulated images 408 based on product 3D design images 406, as shown in FIG 4B, the deep learning predictive model 402 can also be trained to generate simulated 3D design images 410 based on product images 235. This can be further understood with reference to FIG 4C.

[0066] FIG. 4C illustrates a flowchart 403 of a trained deep learning predictive model 402 configured to generate a simulated 3D design image 410 based on a product image 235 of a specimen 220 in accordance with one or more embodiments of the present disclosure.

[0067] In one embodiment, the controller 204 is configured to receive one or more product images 235 of one or more features of the specimen 220 from the characterization subsystem 202. For example, as shown in FIG. 3A, the controller 204 may be configured to receive one or more optical product images 235 of one or more features of the specimen 220 from the optical characterization subsystem 202a. As another example, as shown in FIG. 3B, the controller 204 may be configured to receive one or more SEM product images 235 of one or more features of the specimen 220 from the SEM characterization subsystem 202b. In this regard, the product images 235 shown in FIG. 4C may include optical product images 225, SEM product images 225, etc. In additional and / or alternative embodiments, the controller 204 may be configured to receive one or more product images 235 from a source other than the one or more characterization subsystems 202. For example, the controller 204 may be configured to receive one or more product images 235 of one or more features of the specimen 220 from an external storage device. In another embodiment, the controller 204 may be further configured to store the received training images 225 in the memory 208 .

[0068] In another embodiment, the controller 204 is configured to generate one or more simulated 3D design images 410 of one or more features of the specimen 220 based on one or more product images 235 using the deep learning predictive model 402. For example, the one or more product images 235 may include images of defects in the specimen 220. The controller 204 may be configured to receive these product images 235 of the defects and generate one or more simulated 3D design images 410 of the defects (e.g., features) using the deep learning predictive model 402. In this regard, design data / images may be created for product images 235 that do not have associated design data. As previously discussed herein, conventional modeling techniques based on Monte Carlo simulations are unable to recreate design data / images based on optical and / or SEM images. In this regard, the system 200 of the present disclosure may enable the recreation of 3D design data from images, which is not feasible with currently available technology.

[0069] 4B , the controller 204 may be further configured to determine one or more characteristics of the specimen 220 based on the simulated 3D design image 410. In addition, the simulated 3D design image 410 may also be used to adaptively modify one or more characteristics of the system 200. Further, the controller 204 may be configured to generate one or more control signals, which are configured to cause one or more fabrication tools to fabricate the specimen 220 based on the generated simulated 3D design image 410.

[0070] As previously described herein, the deep learning predictive model 402 may include any type of machine learning algorithm / classifier and / or deep learning technique or classifier known in the art, including, but not limited to, conditional generative adversarial networks (CGANs), convolutional neural networks (CNNs) (e.g., GoogleNet, AlexNet, etc.), ensemble learning classifiers, random forest classifiers, artificial neural networks (ANNs), etc. Deep learning predictive models 402, including CGANs, may be further understood with reference to FIG.

[0071] FIG. 5 illustrates a flowchart for training a deep learning predictive model 402 including a conditional generative adversarial network (CGAN) in accordance with one or more embodiments of the present disclosure.

[0072] In an embodiment, the CGAN (e.g., deep learning predictive model 402) may include, but is not limited to, a generator 412 and a discriminator 414. In an embodiment, the generator 412 may be configured to generate simulated training images 416 (e.g., “fake” images). Conversely, the discriminator 414 may be configured to receive the training 3D design images 404 and either the training images 235 (e.g., “real” images of the specimen 220) or the simulated training images 416 (e.g., “fake” images) and determine whether an image associated with the training 3D design image 404 is real or fake. The “real” or “fake” determination may be referred to as the “discriminator output.” In this regard, training the CGAN (e.g., deep learning predictive model 402) may include training the discriminator 414 to be more accurate at identifying fake images and training the generator 412 to generate more realistic “fake” images to “fool” the discriminator 414. As the discriminator 414 gets better at discerning between real images (e.g., training images 235) and false images (e.g., simulated training images 416), the generator 412 should get better at generating more realistic simulated training images 416. After the generator 412 is sufficiently trained, the simulated training images 416 may be output from the deep learning predictive model 402 as the simulated images 408 shown in FIG.

[0073] Thus, in some embodiments, the generator 412 of the CGAN may be configured to generate one or more simulated images 408 based on one or more training 3D design images 404 of one or more features of the specimens 220. The controller 204 may then be configured to train a discriminator 414 of the CGAN based on the one or more training 3D design images 404, the one or more training images 235, and the one or more simulated training images 408. Training the discriminator 414 may include training the discriminator 414 to generate more accurate discriminator outputs (e.g., “real”, “fake”). As the discriminator 414 becomes more efficient at identifying “fake” simulated training images 416, the generator 412 may also become more efficient at generating more realistic simulated training images 416. In this regard, the generator 412 may be said to be “trained” based on the discriminator outputs. Thus, in some embodiments, by training the discriminator 414 to generate a more accurate discriminator output, the controller 204 may be further configured to train a generator 412 of the CGAN based on the one or more discriminator outputs of the discriminator 414. Upon training the generator 412, the generator 412 may be configured to generate one or more simulated images 408 of one or more features of the specimen 220 based on the one or more product 3D design images 406, as shown in FIG.

[0074] FIG. 6 illustrates a flowchart 600 for adaptively modifying three-dimensional design images / files used to create specimens in a production process that utilizes deep learning predictive models within the specimen production process, in accordance with one or more embodiments of the present disclosure.

[0075] In step 602, 3D design images / files are created. For example, the 3D design images may be created by the controller 204 with or without input from a user. These 3D design images may be stored in the memory 208.

[0076] In step 604, one or more specimens are fabricated. The one or more specimens may be fabricated based on the 3D design image / file generated in step 602. It is noted herein that fabrication of the one or more specimens may be performed by any fabrication process or fabrication tool known in the art, including, but not limited to, lithography tools, etching tools, deposition tools, polishing tools.

[0077] In step 606, the one or more prepared specimens are inspected / reviewed. The one or more specimens may be inspected / reviewed via any characterization tool known in the art, including, but not limited to, an optical characterization tool (e.g., optical characterization subsystem 202a), a SEM characterization tool (e.g., SEM characterization subsystem 202b).

[0078] In step 602, the deep learning predictive model is trained. For example, the controller 204 may be configured to receive 3D design images (e.g., training 3D design images 404) and one or more training images 235 from an inspection / review process and train the deep learning predictive model 402.

[0079] After training the deep learning predictive model 402, the controller 204 may be configured to adaptively modify the 3D design image and / or the fabrication process. For example, once the deep learning predictive model 402 is trained, the deep learning predictive model 402 may be configured to determine one or more characteristics of the specimen 220 based on the simulated 3D design image 410. The controller 204 may be configured to adaptively modify one or more characteristics of the one or more product 3D design images 406 of the specimen 220 to generate one or more modified product 3D design images. For example, if the one or more product 3D design images 406 include design data of a likely defect in the specimen 220 to be fabricated, the controller 204 may utilize information gathered from the simulated images 208 to adaptively modify the product 3D design images 406 to generate one or more modified product 3D design images. By modifying the 3D design image / data of the specimen 220 prior to fabrication, various features and characteristics of the specimen 220 may be modified during fabrication. For example, by modifying the 3D design image / data of the specimen 220 prior to fabrication, the system 200 can reduce the number of defects in the specimen 220, reduce the size of defects in the specimen 220, modify the characteristics of defects in the specimen 220 to make them easier to identify, etc.

[0080] Additionally, the simulated 3D design image 410 may also be used to adaptively modify one or more characteristics of the system 200. For example, after training the deep learning predictive model 402, the controller 204 may be further configured to generate a control signal, which is configured to cause a fabrication tool to fabricate the specimen 220 based on the modified product 3D design image and / or the simulated 3D design image 410.

[0081] It is noted herein that one or more components of system 200 may be communicatively coupled to various other components of system 200 in any manner known in the art. For example, one or more processors 206 may be communicatively coupled to each other or to another component via a wired (e.g., copper wire, fiber optic cable, etc.) or wireless connection (e.g., RF coupling, IR coupling, WiMax, Bluetooth, 3G, 4G, 4GLTE, 5G, etc.). As another example, controller 204 may be communicatively coupled to one or more components of characterization subsystem 202 via any wired or wireless connection known in the art.

[0082] In one embodiment, the one or more processors 206 may include any one or more processing elements known in the art. In this sense, the one or more processors 206 may include any microprocessor-type device configured to execute software algorithms and / or instructions. In one embodiment, the one or more processors 206 may comprise a desktop computer, a mainframe computer system, a workstation, an image computer, a parallel processor, or another computer system (e.g., a networked computer) configured to execute programs as described throughout this disclosure, the programs being configured to operate the system 200. It should be appreciated that the steps described throughout this disclosure may be performed by a single computer system, or alternatively, multiple computer systems. Furthermore, it should be appreciated that the steps described throughout this disclosure may be performed by any one or more of the one or more processors 206. In general, the term "processor" may be broadly defined to encompass any device having one or more processing elements, which executes program instructions from memory 208. Additionally, different subsystems of system 200 (e.g., illumination source 212, electron beam source 228, detector assembly 226, electron detector assembly 234, controller 204, user interface 210, etc.) may include processors or logic elements suitable for performing at least a portion of the steps described throughout this disclosure. Thus, the above description should not be construed as a limitation of the present disclosure, but merely as an example.

[0083] The memory 208 may include any storage medium known in the art suitable for storing program instructions executable by the associated one or more processors 206 and data received from the characterization subsystem 202. For example, the memory 208 may include a non-transitory memory medium. For example, the memory 208 may include, but is not limited to, a read-only memory (ROM), a random access memory (RAM), a magnetic or optical memory device (e.g., disk), a magnetic tape, a solid-state drive, and the like. It is further noted that the memory 208 may be housed in a common controller housing with the one or more processors 206. In alternative embodiments, the memory 208 may be located remotely with respect to the physical location of the processors 206, the controller 204, and the like. In another embodiment, the memory 208 maintains program instructions for causing the one or more processors 206 to perform various steps described throughout this disclosure.

[0084] In one embodiment, the user interface 210 is communicatively coupled to the controller 204. In one embodiment, the user interface 210 may include, but is not limited to, one or more desktops, tablets, smartphones, smart watches, etc. In another embodiment, the user interface 210 includes a display that is used to display data of the system 200 to a user. The display of the user interface 210 may include any display known in the art. For example, the display may include, but is not limited to, a liquid crystal display (LCD), an organic light emitting diode (OLED) based display, or a CRT display. Those skilled in the art should recognize that any display device that can be integrated with the user interface 210 is suitable for implementation in the present disclosure. In another embodiment, a user may input selections and / or commands responsive to data displayed to the user via a user input device of the user interface 210.

[0085] 7 illustrates a flowchart of a method 700 for generating a simulated image 408 based on a three-dimensional design image 406 utilizing a deep learning predictive model 402, in accordance with one or more embodiments of the present disclosure. It is noted herein that the steps of method 700 may be implemented in whole or in part by system 200. However, it is further recognized that method 700 is not limited to system 200 in that additional or alternative system-level embodiments may perform all or a portion of the steps of method 700.

[0086] In step 702, one or more training images of one or more features of the specimen are acquired using a characterization subsystem. For example, as shown in FIG. 3A, the controller 204 may be configured to receive one or more optical training images 225 of one or more features of the specimen 220 from the optical characterization subsystem 202a. As another example, as shown in FIG. 3B, the controller 204 may be configured to receive one or more SEM training images 225 of one or more features of the specimen 220 from the SEM characterization subsystem 202b.

[0087] One or more training three-dimensional (3D) design images of one or more features of the specimen are received at step 704. For example, the one or more training 3D design images 404 may be stored in memory 208 such that controller 204 is configured to receive and / or retrieve the one or more training 3D design images 404 from memory 208.

[0088] In step 706, a deep learning predictive model is generated based on the one or more training images and the one or more training 3D design images. For example, the one or more training images 225 and the one or more training 3D design images 404 may be used as inputs to train the deep learning predictive model 402. The deep learning predictive model 402 may include any type of machine learning algorithm / classifier and / or deep learning technique or classifier known in the art, including, but not limited to, conditional generative adversarial networks (CGANs), convolutional neural networks (CNNs) (e.g., GoogleNet, AlexNet, etc.), ensemble learning classifiers, random forest classifiers, artificial neural networks (ANNs), etc.

[0089] At step 708, one or more product 3D design images of one or more features of the specimen are received. For example, as shown in FIG. 3A, the controller 204 can be configured to receive one or more product 3D design images 406 of one or more features of the specimen 220.

[0090] At step 710, one or more simulated 3D design images of one or more features of the specimen are generated based on the one or more product 3D design images and using the deep learning predictive model. For example, the controller 204 can be configured to receive the one or more product 3D design images 406 of the likely defects and generate one or more simulated images 408 of the likely defects (e.g., features) using the deep learning predictive model 402.

[0091] At step 712, one or more characteristics of the specimen are determined based on the one or more simulated images. For example, continuing with the same example, the controller 204 may be configured to determine one or more characteristics of a likely defect (e.g., feature) in the future specimen 220 based on the generated simulated SEM image 408 and / or the generated simulated optical image 408. The characteristics of the specimen 220 and / or the features of the specimen 220 that may be determined based on the simulated image 408 may include any characteristic of a specimen known in the art, including, but not limited to, a location of a defect in the specimen 220, a type of defect in the specimen 220, a material of the specimen 220, and topographical measurements of a surface and / or pattern of the specimen 220.

[0092] 8 illustrates a flowchart of a method 700 for generating a simulated image 408 based on a three-dimensional design image 406 utilizing a deep learning predictive model 402, in accordance with one or more embodiments of the present disclosure. It is noted herein that the steps of method 700 may be implemented in whole or in part by system 200. However, it is further recognized that method 700 is not limited to system 200 in that additional or alternative system-level embodiments may perform all or a portion of the steps of method 700.

[0093] In step 802, one or more training images of one or more features of the specimen are acquired using a characterization subsystem. For example, as shown in FIG. 3A, the controller 204 may be configured to receive one or more optical training images 225 of one or more features of the specimen 220 from the optical characterization subsystem 202a. As another example, as shown in FIG. 3B, the controller 204 may be configured to receive one or more SEM training images 225 of one or more features of the specimen 220 from the SEM characterization subsystem 202b.

[0094] One or more training three-dimensional (3D) design images of one or more features of the specimen are received at step 804. For example, the one or more training 3D design images 404 may be stored in memory 208 such that controller 204 is configured to receive and / or retrieve the one or more training 3D design images 404 from memory 208.

[0095] In step 806, a deep learning predictive model is generated based on the one or more training images and the one or more training 3D design images. For example, the one or more training images 225 and the one or more training 3D design images 404 may be used as inputs to train the deep learning predictive model 402. The deep learning predictive model 402 may include any type of machine learning algorithm / classifier and / or deep learning technique or classifier known in the art, including, but not limited to, conditional generative adversarial networks (CGANs), convolutional neural networks (CNNs) (e.g., GoogleNet, AlexNet, etc.), ensemble learning classifiers, random forest classifiers, artificial neural networks (ANNs), etc.

[0096] At step 808, one or more product images of one or more features of the specimen are received. For example, as shown in FIGURE 3A, the controller 204 may be configured to receive one or more optical product images 235 of one or more features of the specimen 220 from the optical characterization subsystem 202a. According to another example, as shown in FIGURE 3B, the controller 204 may be configured to receive one or more SEM product images 235 of one or more features of the specimen 220 from the SEM characterization subsystem 202b.

[0097] At step 810, one or more simulated 3D design images of one or more features of the specimen are generated based on the one or more product images and using the deep learning predictive model. For example, the controller 204 may be configured to receive one or more product images 235 of the likely defects and generate one or more simulated 3D design images 410 of the likely defects (e.g., features) using the deep learning predictive model 402.

[0098] At step 812, one or more characteristics of the specimen are determined based on the one or more simulated 3D design images. For example, continuing with the same example, the controller 204 may be configured to determine one or more characteristics of likely defects (e.g., features) of the future specimen 220 based on the generated simulated SEM image 408 and / or the generated simulated optical image 408. The characteristics of the specimen 220 and / or the features of the specimen 220 that may be determined based on the simulated images 408 may include any characteristics of a specimen known in the art, including, but not limited to, a location of a defect in the specimen 220, a type of defect in the specimen 220, a material of the specimen 220, and topographical measurements of a surface and / or pattern of the specimen 220.

[0099] Those skilled in the art will recognize that the components (e.g., operations), devices, objects, and their accompanying descriptions described herein are used as examples for conceptual clarity, and that various configuration modifications are contemplated. Thus, as used herein, the specific examples described and the accompanying discussion are intended to be representative of their more general classification. In general, the use of a specific example is intended to represent that classification, and should not be considered as a limitation to not including the specific components (e.g., operations), devices, and objects.

[0100] Those skilled in the art will appreciate that there are a variety of vehicles (e.g., hardware, software, and / or firmware) in which the processes and / or systems and / or other techniques described herein may be implemented, and that the preferred vehicle will vary depending on the context in which the processes and / or systems and / or other techniques are deployed. For example, if the implementer determines that speed and accuracy are paramount, the implementer may select a primarily hardware and / or firmware vehicle; alternatively, if flexibility is paramount, the implementer may select a primarily software implementation, or again alternatively, the implementer may select some combination of hardware, software, and / or firmware. Thus, there are several possible vehicles in which the processes and / or devices and / or other techniques described herein may be implemented, and no vehicle is inherently better than another in that any vehicle utilized is a choice depending on the context in which the vehicle is deployed and the particular interests of the implementer (e.g., speed, flexibility, or predictability), any of which may be subject to change.

[0101] The preceding description is presented to enable one skilled in the art to make and use the invention as provided in the context of a particular application and its requirements. As used herein, directional terms such as "top", "bottom", "over", "under", "upper", "upward", "lower", "down" and "downward" are intended to provide relative positions for purposes of description and are not intended to specify an absolute frame of reference. Various modifications to the described embodiments will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments. Thus, the present invention is not intended to be limited to the specific embodiments shown and described, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0102] With respect to the use of substantially any plural and / or singular term herein, those of skill in the art may interpret from the plural to the singular and / or from the singular to the plural as appropriate to the context and / or application. The various singular / plural permutations are not expressly set forth herein for the sake of clarity.

[0103] All methods described herein may include storing results of one or more steps of the method embodiments in memory. The results may include any of the results described herein and may be stored in any manner known in the art. The memory may include any of the memories described herein, or any other suitable storage medium known in the art. After storing the results, the results may be accessed in the memory and used by any of the method or system embodiments described herein, formatted for display to a user, used by another software module, method, or system, etc. Furthermore, the results may be stored "permanently," "semi-permanently," "temporarily," or for a period of time. For example, the memory may be a random access memory (RAM), and the results may not necessarily remain in memory indefinitely.

[0104] It is further contemplated that each of the method embodiments described above may include any other step(s) of any other method(s) described herein. In addition, each of the method embodiments described above may be performed by any of the systems described herein.

[0105] The subject matter described herein may depict different components contained within or connected to another component. It should be understood that such depicted architectures are merely exemplary, and that in fact many other architectures that achieve the same functionality may be implemented. In a conceptual sense, any arrangement of components to achieve the same functionality is effectively "associated" to achieve the desired functionality. Thus, any two components herein that are combined to achieve a particular functionality may be considered to be "associated" with one another, such that the desired functionality is achieved without regard to the architecture or intermediate components. Similarly, any two components so associated may also be considered to be "connected" or "coupled" with one another to achieve the desired functionality, and any two components that may be so associated may be considered to be "couplable" with one another to achieve the desired functionality. Specific examples of components that may be coupled include, but are not limited to, components that are physically interactable and / or physically interacting, and / or components that are wirelessly interactable and / or wirelessly interacting, and / or components that are logically interactable and / or logically interacting.

[0106] It should further be understood that the present invention is defined by the appended claims. In general, it will be understood by those skilled in the art that the terms used herein, particularly in the appended claims (e.g., the body of the appended claims), are generally intended as "open" terms (e.g., the term "including" should be interpreted as "including but not limited to," the term "having" should be interpreted as "having at least," the term "includes" should be interpreted as "including but not limited to," etc.). If a specific number of introduced claim recitations are intended, such intent will be explicitly set forth in the claims, and it will be further understood by those skilled in the art that, in the absence of such recitation, no such intent exists. For example, to aid in understanding, the following appended claims may include the use of the introductory phrases "at least one" and "one or more" to introduce the claim recitations. However, the use of such phrases should not be construed to mean that the introduction of a claim recitation with the indefinite article "a" or "an" limits the scope of any particular claim containing the claim recitation so introduced to inventions containing only one such recitation, even if the same claim includes the introductory phrase "one or more" or "at least one" and the indefinite article "a" or "an" (e.g., "a" and / or "an" should generally be interpreted to mean "at least one" or "one or more"), and the same applies to the use of definite articles used to introduce claim recitations. In addition, even if a specific number of introduced claim recitations is explicitly recited, one of ordinary skill in the art will recognize that such a recitation should generally be interpreted to mean at least the recited number (e.g., a recitation of only "two recitations" without any other modifier generally means at least two recitations, or more than two recitations).Further, in instances where a phrase similar to "at least one of A, B, and C, etc." is used, such a configuration is generally intended to have the meaning as one of ordinary skill in the art would understand the phrase (e.g., "a system having at least one of A, B, and C" includes, but is not limited to, systems having only A, only B, only C, A and B together, A and C together, B and C together, and / or A, B and C together, etc.). In instances where a phrase similar to "at least one of A, B, or C, etc." is used, such a configuration is generally intended to have the meaning as one of ordinary skill in the art would understand the phrase (e.g., "a system having at least one of A, B, or C" includes, but is not limited to, systems having only A, only B, only C, A and B together, A and C together, B and C together, and / or A, B and C together, etc.). It will be further understood by those skilled in the art that virtually any disjunction and / or phrase expressing two or more alternative terms, whether in the description, claims or drawings, should be understood to contemplate the possibility of including one of the terms, either of the terms, or both terms. For example, the phrase "A or B" will be understood to include the possibilities of "A" or "B" or "A and B."

[0107] It will be apparent that the present disclosure and many of its attendant advantages will be understood from the foregoing description, and that various changes in the form, construction and arrangement of the elements may be made without departing from the disclosed subject matter or sacrificing all of its important advantages. The forms described are merely illustrative, and it is the intent of the following claims to encompass and include such modifications. It is to be understood, further, that the invention is defined by the appended claims.

Claims

1. 1. A system for characterizing a specimen, comprising: a characterization subsystem configured to acquire one or more images of the specimen; a controller communicatively coupled to the characterization subsystem, the controller comprising one or more processors configured to execute a set of program instructions stored in a memory, the set of program instructions configured to cause the one or more processors to: receiving one or more training images of one or more features of a specimen from the characterization subsystem, the specimen comprising a semiconductor wafer; receiving one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen; generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images; receiving one or more product 3D design images of one or more features of the specimen; generating at least one of one or more simulated scanning electron microscope images or one or more optical images of the one or more features of the specimen using the deep learning predictive model based on the one or more product 3D design images; and a controller configured to cause one or more material properties of the specimen to be determined based on at least one of the one or more simulated scanning electron microscope images or the one or more optical images. system.

2. 10. The system of claim 1, wherein at least one of the one or more training images of the one or more features of the specimen comprises an optical image of the specimen.

3. 10. The system of claim 1, wherein at least one of the one or more training images of the one or more features of the specimen comprises a scanning electron microscope (SEM) image of the specimen.

4. The system of claim 1 , wherein the deep learning predictive model comprises a conditional generative adversarial network (CGAN).

5. The one or more processors: generating one or more simulated training images based on the one or more training 3D design images using a generator of the CGAN; training a discriminator of the CGAN based on the one or more training images, the one or more training 3D design images, and the one or more simulated training images; The system of claim 4 further configured to:

6. The system of claim 5 , wherein the one or more processors are further configured to train the generator of the CGAN based on one or more discriminator outputs of the discriminator.

7. generating the one or more simulated scanning electron microscope images or the one or more optical images of the one or more features of the specimen based on the one or more product 3D design images using the deep learning predictive model, generating, using a generator of the CGAN, the one or more simulated scanning electron microscope images or the one or more optical images of the one or more features of the specimen based on the one or more product 3D design images; The system of claim 5 , comprising:

8. 10. The system of claim 1, wherein the one or more characteristics of the specimen further include at least one of a topography measurement of a surface of the specimen and a location of a defect in the specimen or a type of defect in the specimen.

9. The one or more processors: adaptively modifying one or more characteristics of the one or more product 3D design images of the specimen to generate one or more modified product 3D design images. The system of claim 1 further configured to:

10. one or more fabrication tools communicatively coupled to the controller, the one or more processors of the controller comprising: generating one or more control signals configured to cause the one or more fabrication tools to fabricate a product specimen based on the one or more modified product 3D design images; The system of claim 9 , further comprising one or more fabrication tools further configured to:

11. The one or more processors: performing one or more Monte Carlo simulations on the one or more product 3D design images corresponding to the one or more features of the specimen to generate one or more comparison images; determining an accuracy metric of the deep learning predictive model by comparing the one or more simulated images generated by the deep learning predictive model to the one or more comparison images; The system of claim 1 further configured to:

12. 1. A system for characterizing a specimen, comprising:

1. A controller including one or more processors configured to execute a set of program instructions stored in a memory, the set of program instructions configuring the one or more processors to: receiving one or more training images of one or more features of a specimen, the specimen comprising a semiconductor wafer; receiving one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen; generating a deep learning predictive model based on at least the one or more training images and the one or more training 3D design images; receiving one or more product 3D design images of one or more features of the specimen; generating at least one of one or more simulated scanning electron microscope images or one or more optical images of the one or more features of the specimen using the deep learning predictive model based on the one or more product 3D design images; and a controller configured to cause one or more material properties of the specimen to be determined based on at least one of the one or more simulated scanning electron microscope images or one or more optical images. system.

13. The system of claim 12 , wherein the one or more training images are received from a characterization subsystem configured to acquire one or more images of a specimen.

14. The one or more processors: adaptively modifying one or more characteristics of the one or more product 3D design images of the specimen to generate one or more modified product 3D design images. The system of claim 12 further configured to:

15. one or more fabrication tools communicatively coupled to the controller, the one or more processors of the controller comprising:

15. The system of claim 14, further comprising: one or more fabrication tools further configured to generate one or more control signals configured to cause the one or more fabrication tools to fabricate a product specimen based on the one or more modified product 3D design images.

16. The system of claim 12 , wherein the deep learning predictive model comprises a conditional generative adversarial network (CGAN).

17. 1. A system for characterizing a specimen, comprising: a characterization subsystem configured to acquire one or more images of a specimen including a semiconductor wafer; a controller communicatively coupled to the characterization subsystem, the controller comprising one or more processors configured to execute a set of program instructions stored in a memory, the set of program instructions configured to cause the one or more processors to: receiving one or more training images of one or more features of a specimen from the characterization subsystem; receiving one or more training three-dimensional (3D) design images corresponding to the one or more features of the specimen; generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images; receiving one or more product images of one or more features of the specimen, the one or more product images including at least one of one or more scanning electron microscope images or one or more optical images; generating one or more simulated 3D design images of the one or more features of the specimen using the deep learning predictive model based on the one or more product images; and a controller configured to cause one or more material properties of the specimen to be determined based on the one or more simulated 3D design images. system.

18. 1. A system for characterizing a specimen, comprising:

1. A controller including one or more processors configured to execute a set of program instructions stored in a memory, the set of program instructions configuring the one or more processors to: receiving one or more training images of one or more features of a specimen, the specimen comprising a semiconductor wafer; receiving one or more training 3D design images corresponding to the one or more features of the specimen; generating a deep learning predictive model based on at least the one or more training images and the one or more training 3D design images; receiving one or more product images of one or more features of the specimen, the one or more product images including at least one of one or more scanning electron microscope images or one or more optical images; generating one or more simulated 3D design images of the one or more features of the specimen using the deep learning predictive model based on the one or more product images; and a controller configured to cause one or more material properties of the specimen to be determined based on the one or more simulated 3D design images. system.

19. 1. A method for characterizing a specimen, comprising: acquiring, with a characterization subsystem, one or more training images of one or more features of a specimen comprising a semiconductor wafer; receiving one or more training three-dimensional (3D) design images of the one or more features of the specimen; generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images; Receiving one or more product 3D design images of one or more features of the specimen; generating at least one of one or more simulated scanning electron microscope images or one or more optical images of the one or more features of the specimen using the deep learning predictive model based on the one or more product 3D design images; determining one or more material properties of the specimen based on at least one of the one or more simulated scanning electron microscope images or the one or more optical images. method.

20. 1. A method for characterizing a specimen, comprising: acquiring, with a characterization subsystem, one or more training images of one or more features of a specimen comprising a semiconductor wafer; receiving one or more training three-dimensional (3D) design images of the one or more features of the specimen; generating a deep learning predictive model based on the one or more training images and the one or more training 3D design images; receiving one or more product images of the one or more features of the specimen, the one or more product images including at least one of one or more scanning electron microscope images or one or more optical images; generating one or more simulated design images of the one or more features of the specimen using the deep learning predictive model based on the one or more product images; determining one or more properties of the specimen based on the one or more simulated design images, the one or more properties including a material property of the specimen; method.

Images