Depth Reconstruction of 3D Images of Samples in Charged Particle Systems

Abstract

Depth reconstruction of 3D images of samples in charged particle systems. Methods and systems for high-resolution reconstruction of imaged 3D samples using a slice and view process, where the depth of electron interaction of the imaging beam is greater than the slice thickness. Data obtained through such slice and view processes is enhanced with a depth blur reduction algorithm configured to reduce depth blur caused by portions of first data and second data respectively generated by electron interactions outside a first layer and a second layer to produce enhanced first data and enhanced second data. Subsequently, a high-resolution 3D reconstruction of the sample is generated using the enhanced first data and the enhanced second data. In some embodiments, the depth blur reduction algorithm may be selected from a set of such algorithms individually configured for certain microscope conditions, sample conditions, or combinations thereof.

Description

Background Art

[0001] Slice-and-view charged particle imaging is a necessary tool for sample reconstruction and inspection in biological research, semiconductor debugging, materials science, and many other applications. In slice-and-view charged particle imaging, successive surface layers of a sample are repeatedly imaged and then removed, and the resulting images are then combined to form a 3D reconstruction of the sample. For example, Figure 1 illustrates a sample process 100 for slice-and-view imaging with a charged particle microscope. Specifically, Figure 1 comprises three images depicting iterative steps of a slice-and-view imaging process.

[0002] Image 102 shows SEM imaging of the nth layer of sample 108. During SEM imaging, an electron beam 110 scans the surface 108 of the nth layer of sample 108, and a detector 112 detects emissions 114 generated by the sample being irradiated by the electron beam. Image 104 then depicts the process of removing and / or otherwise delaying the nth layer of sample 108. Although image 104 depicts removing the nth layer with a focused ion beam 116, other mechanisms such as an electron beam, a laser, a diamond blade, etc. can be used to remove the nth layer. As shown in image 106, SEM imaging can then be repeated for the n+1th layer of sample 108. The detected emissions 114 from images 102 and 106 are then used to generate images of the nth layer of sample 108 and the n+1th layer of sample 108, respectively. Since slice-and-view charged particle imaging acquires a series of cross-sectional images of a sample at various depths along the z-axis, the acquired images and / or the data resulting therefrom can be used to reconstruct a 3D representation of the sample.

[0003] However, a fundamental limitation of slice and view charged particle imaging is the increasing blurring of the 3D reconstruction along the z-axis of the sample (i.e., z-blurring), which occurs when the thickness of the removed sample slice is reduced below the electron interaction depth of the charged particle beam. The electron interaction depth corresponds to the vertical distance from the surface of the sample 108 to another part of the region of the sample where the imaging beam of electrons / charged particles of the sample interacts, such that molecules / elements / features within the region of the sample release emissions 114. For example, images 102 and 106 illustrate electron interaction regions 118 (i.e., the regions of the sample where electrons introduced by the electron beam 110 interact with the sample) extending below the thickness of the nth and n+1th sample layers, respectively. Since the detected emissions 114 contain information describing the electron interaction regions 118 from which they are generated, the images generated from such detected emissions contain information from a z-depth greater than the slice thickness. That is, although the images are intended to depict the corresponding slices of the sample, when the electron interaction depth exceeds the thickness of the removed sample slice, the images contain information from multiple slices. This phenomenon is the cause of the progressive blurring of the 3D reconstruction along the z-axis described above.

[0004] To address this issue, the current practice is to reduce the voltage of the charged particle beam so that there is a smaller electron interaction region 120. While this reduces the depth of the electron interaction region, the reduction in the voltage of the charged particle beam causes a corresponding reduction in the emissions 114, which in turn reduces the signal-to-noise ratio (SNR) of the signal obtained by the detector 112. In addition, the reduction in the voltage of the charged particle beam also makes it more difficult to shape the beam. This change in the beam shape results in a reduction in the xy resolution of the 3D reconstruction. Therefore, a solution is needed that allows for the removal of thinner sample slices (e.g., 2 to 5 nm or less) without reducing the xy resolution or SNR. Summary of the Invention

[0005] Methods and systems for high-resolution reconstruction of an imaged 3D sample using a slice-and-view process are disclosed, where the electron interaction depth of the imaging beam is greater than the slice thickness. Example methods include: obtaining first data associated with a first layer of a sample, the first data having been obtained by first irradiating the first layer of the sample with a charged particle beam; and subsequently obtaining second data associated with a second layer of the sample, the second data having been obtained by second irradiating the second layer of the sample with a charged particle beam, where the first layer of the sample is removed between the first irradiation and the second irradiation, and where the electron interaction depth of the charged particle beam is greater than each of the thickness of the first layer and the thickness of the second layer. The first data and the second data are then enhanced with a depth blur reduction algorithm (e.g., a 3D neural network-based algorithm, a 3D blind neural network-based algorithm, a trained 3D blind deconvolution algorithm, etc.), the depth blur reduction algorithm being configured to reduce depth blur caused by portions of the first data and the second data generated by electron interactions outside the first layer and the second layer, respectively, to produce enhanced first data and enhanced second data. The enhanced first data and the enhanced second data are then used to generate a high-resolution 3D reconstruction of the sample. In some embodiments, the depth blur reduction algorithm may be selected from a group of such algorithms independently configured for certain microscope conditions (e.g., imaging beam voltage, imaging beam type, spot size, scan speed, etc.), sample conditions (e.g., sample type, sample material, sample features, etc.), or a combination thereof.

[0006] A system for high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness, the system includes: a sample holder configured to hold a sample; an electron beam source configured to emit an electron beam toward the sample; an electron beam column configured to direct the electron beam onto the sample; a delay component configured to remove a layer having a known thickness from the surface of the sample; and one or more detectors configured to detect emissions generated by the electron beam irradiating the sample. The system further includes one or more processors and a memory storing instructions that, when executed on the one or more processors, cause the system to: obtain first data associated with a first layer of the sample and second data associated with a second layer of the sample by slice-and-view imaging, where the electron interaction depth of the imaging is greater than the thickness of the first layer and the thickness of the second layer; enhance the first data and the second data using a depth blur reduction algorithm; and subsequently generate a high-resolution 3D reconstruction of the sample using the enhanced data.

[0007] Other example methods for generating a high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness, the example method comprising training a depth blur reduction algorithm based at least in part on a first set of training data obtained by slice-and-view charged particle imaging of a training sample and a second set of training data corresponding to a labeled reconstruction of the training sample, where the electron interaction depth of the imaging beam is greater than the corresponding slice thickness. The first data and the second data are (respectively) associated with a first layer of the sample and a second layer of the sample imaged by slice-and-view, where the electron interaction depth of the charged particle beam is greater than each of the thickness of the first layer and the thickness of the second layer. Subsequently, the first data and the second data are enhanced with the depth blur reduction algorithm configured to reduce depth blur caused by portions of the first data and the second data respectively generated by electron interactions outside the first layer and the second layer to produce enhanced first data and enhanced second data. Subsequently, a high-resolution 3D reconstruction of the sample is generated using the enhanced first data and the enhanced second data. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The detailed description has been made with reference to the accompanying drawings. In the drawings, one or more digits at the leftmost of the reference numerals identify the drawing in which the reference numeral first appears. The same reference numerals in different drawings indicate like or identical items.

[0009] Figure 1 Illustrates a sample process for slice-and-view imaging with a charged particle microscope.

[0010] Figure 2 Depicts a sample process for using, training, optimizing, and / or retraining a depth blur reduction algorithm for generating a high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness.

[0011] Figure 3 Displays a set of images illustrating a process for generating a high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness.

[0012] Figure 4 Displays a set of schematics illustrating a process for generating a high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness.

[0013] Figure 5 and 6 Shows experimental results demonstrating slice-and-view images processed using the process of the present invention compared to slice-and-view images of prior art systems.

[0014] Throughout several views of the accompanying drawings, like reference numerals refer to corresponding parts. Generally, in the drawings, elements that may be included in a given instance are shown in solid lines, while elements that are optional for a given instance are shown in dashed lines. However, elements shown in solid lines are not required for all instances of the present disclosure, and elements shown in solid lines may be omitted from a particular instance without departing from the scope of the present disclosure. Detailed Description

[0015] Methods and systems are disclosed herein for high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness. Specifically, the methods and systems employ a depth blur reduction algorithm on data acquired by slice-and-view imaging to allow for the generation of a 3D reconstruction of the sample without z-direction blurring caused by portions of data resulting from electron interactions outside the corresponding sample layer. In some embodiments of the invention, the depth blur reduction algorithm is trained using a first set of training data acquired by slice-and-view charged particle imaging of a training sample and a second set of training data corresponding to a labeled reconstruction of the training sample, where the electron interaction depth of the imaging beam is greater than the corresponding slice thickness.

[0016] The methods and systems of the present disclosure allow for high-resolution 3D reconstruction of a sample using a slice-and-view process without having to reduce the imaging beam energy such that the electron interaction depth is less than the layer thickness. Additionally, by enabling such slice-and-view processes to use a higher energy beam, the SNR of the acquired data is reduced, and the beam can be consistently shaped to allow for high-resolution data acquisition.

[0017] Figure 2 FIG. 200 is an illustration of an environment for using, training, optimizing, and / or retraining a depth blur reduction algorithm for high-resolution reconstruction of an imaged 3D sample using a slice-and-view process, where the electron interaction depth of the imaging beam is greater than the slice thickness. Specifically, Figure 2 FIG. 202 shows an example environment that includes an example microscope system 204 for generating slice-and-view images of a sample 206, and one or more computing devices 208 for using, training, optimizing, and / or retraining a depth blur reduction algorithm for high-resolution reconstruction of an imaged 3D sample with reduced z-dimension blurring. It should be noted that the present disclosure is not limited to environments that include microscopes, and in some embodiments, environment 200 may include different types of systems configured to generate relevant images, or may not include a system for generating images at all.

[0018] The example microscope system 204 can be or include one or more different types of optical and / or charged particle microscopes, such as but not limited to a scanning electron microscope (SEM), a scanning transmission electron microscope (STEM), a transmission electron microscope (TEM), a charged particle microscope (CPM), a cryo-compatible microscope, a focused ion beam microscope (FIB), a dual-beam microscopy system, or a combination thereof. Figure 2 The example microscope system 204 is shown as a dual-beam microscopy system that includes an SEM column 210 and an FIB column 212.

[0019] Figure 2 The example microscope system 204 is depicted as including an SEM column 210 for imaging layers of a sample 206 during sectioning and view imaging. The SEM column 210 includes an electron source 214 (e.g., a thermionic source, a Schottky emission source, a field emission source, etc.), which emits an electron beam 216 along an electron emission axis 218 and toward the sample 206. The electron emission axis 218 is a central axis that extends along the length of the example microscope system 204 from the electron source 214 and through the center of the sample 206. Although Figure 2 The example microscope system 204 is depicted as including an electron source 204, but in other embodiments, the example microscope system 204 can include a charged particle source, such as an ion source, configured to emit a plurality of charged particles toward the sample 206.

[0020] An accelerator lens 220 accelerates / decelerates, focuses, and / or directs the electron beam 216 to an electron focusing column 222. The electron focusing column 222 focuses the electron beam 212 such that it impinges on at least a portion of the sample 206. Additionally, the focusing column 222 can correct and / or adjust aberrations (e.g., geometric aberrations, chromatic aberrations) of the electron beam 216. In some embodiments, the electron focusing column 222 can include one or more of an aperture, a deflector, a transfer lens, a scanning coil, a condenser lens, an objective lens, etc., which together focus electrons from the electron source 214 onto a small spot on the sample 206. Different positions of the sample 206 can be scanned by adjusting the direction of the electron beam via a deflector and / or a scanning coil. In this way, the electron beam 216 acts as an imaging beam that scans the surface layer of the sample (i.e., the surface of the layer proximate to the SEM column 204 and / or irradiated by the electron beam 216). This irradiation of the surface layer of the sample 206 causes the constituent electrons of the electron beam 216 to interact with the constituent elements / molecules / features of the sample, such that the constituent elements / molecules / features cause an emission 223 to be emitted from the sample 206. The particular emission released is based on the corresponding element / molecule / feature that causes it to be emitted, such that the emission can be analyzed to determine information about the corresponding element / molecule.

[0021] Although the electron beam 216 is incident on the surface layer of the sample 206, a portion of its constituent electrons penetrate the sample and interact with elements / molecules / features at depths different from the surface of the sample. The electron interaction depth of the electron beam 216 corresponds to the distance from the sample surface that encompasses 95% of the elements / molecules / features of the sample that interact with the electrons of the electron beam 216 during irradiation. In this manner, the emission 223 can be analyzed to determine information regarding the elements / molecules present in the sample 206 from the surface to the electron interaction depth.

[0022] Figure 2 Further shown is a detector system 224 for detecting emissions generated by the electron beam 216 incident on the sample 206. The detector system 224 can include one or more detectors positioned or otherwise configured to detect such emissions. In various embodiments, different detectors and / or different portions of a single detector can be configured to detect different types of emissions, or configured such that the parameters of the emissions detected by different detectors and / or different portions are different. The detector system 224 is further configured to generate data / data signals corresponding to the detected emissions and transmit the data / data signals to one or more computing devices 208.

[0023] Although Figure 2 The example microscope system 204 is depicted as including an FIB column 212 for removing layers of the sample 206 during sectioning and view imaging. In other embodiments, the example microscope system 204 can include other types of ablation components, such as lasers, mechanical blades (e.g., diamond blades), electron beams, etc. The FIB column 212 is shown as including a charged particle emitter 226 configured to emit a plurality of ions 228 along an ion emission axis 230.

[0024] The ion emission axis 230 is a central axis extending from the charged particle emitter 226 and passing through the center of the sample 206. The FIB column 212 further includes an ion focusing column 232 that includes one or more of an aperture, deflector, transfer lens, scan coil, condenser lens, objective lens, etc., which together focus the ions from the charged particle emitter 226 onto a small spot on the sample 206. In this manner, the elements in the ion focusing column 232 can ablate or otherwise remove one or more portions of the sample with the ions emitted by the charged particle emitter 226. For example, during sectioning and view imaging, the FIB column 212 can be configured to remove a surface layer of the sample 206 having a known thickness from the sample 206 between image acquisitions.

[0025] Figure 2The example microscope system 204 is further shown as additionally including a sample holder 234. The sample holder 234 is configured to hold a sample 206 and can translate, rotate, and / or tilt the sample 102 relative to the example microscope system 204.

[0026] The environment 200 is also shown as including one or more computing devices 208. Those skilled in the art will appreciate that Figure 2 the computing device 208 depicted in is merely illustrative and is not intended to limit the scope of the present disclosure. Computing systems and devices can include any combination of hardware or software that can perform the indicated functions, including computers, network devices, network appliances, PDAs, wireless telephones, controllers, oscilloscopes, amplifiers, etc. The computing device 208 can also be connected to other devices not shown or can actually operate as a stand-alone system.

[0027] It should also be noted that one or more computing devices 208 can be components of the example microscope system 204, can be stand-alone devices separate from the example microscope system 204 that communicate with the example microscope system 204 through a network communication interface, or a combination thereof. For example, the example microscope system 204 can include a first computing device 208 that is an integral part of the example microscope system 204 and that acts as a controller that drives the operation of the example charged particle microscope system 204 (e.g., adjusts the scan position on the sample by operating a scan coil, etc.). In such embodiments, the example microscope system 204 can also include a second computing device 208 that is a desktop computer separate from the example microscope system 204 and that can execute to process data received from the detector system 224 to generate an image of the sample 206 and / or perform other types of analysis or post-processing on the detector data. The computing device 208 can be further configured to receive user selections via a keyboard, mouse, touchpad, touch screen, etc.

[0028] The computing device 208 is configured to generate an image of the surface layer of the sample 206 within the example microscope system 204 based on data and / or data signals from the detector system 224. Specifically, because the data and / or data signals from the detector system 224 are based on emissions 223 emitted from the sample 206 during irradiation of the surface of the sample, the data and / or data signals can be analyzed to determine the composition (i.e., constituent elements / molecules / features) of the sample between the surface of the sample and the depth of interaction of the electrons of the electron beam 216, such that an image of the surface layer of the sample can be generated. In some embodiments, the image is a grayscale image that shows contrast indicative of the shape and / or material of the sample.

[0029] In addition, since the FIB column 212 is capable of removing layers of the sample 206 having a known thickness, the computing device 208 is able to determine the position of the sample 206, and each image in the series of images corresponds to the position. In this way, during the process of sectioning and view imaging, the computing device 208 generates a series of images of the layers of the sample at periodic depths. However, since the images contain information related to elements / molecules / features between the surface and the depth of electron interaction of the electron beam 216, when the thickness of the layer removed by the FIB column 212 is less than the depth of electron interaction of the electron beam 216, the images generated by the computing device 208 depict the elements / molecules / features of the sample from multiple layers of the sample. For example, in an embodiment where the FIB column 212 removes a 2 nm layer from the surface of the sample 206 and the depth of electron interaction of the electron beam 216 is 10 nm, each image generated by the computing device 208 will depict the elements / molecules / features of the sample from 5 different layers of the sample. Since the elements / molecules / features are considered to be present in multiple layers of the sample, but in fact they are only present in one layer, the 3D reconstruction of the sample 206 based on such images will have a blurred reconstruction.

[0030] According to the present invention, the computing device 208 is further configured to apply a depth blur reduction algorithm to one or more of the series of generated images to remove information from individual images corresponding to elements / molecules / features present in layers of the sample 206 that do not correspond to the individual images. In other words, the depth blur reduction algorithm applied by the computing device 208 reduces the image information generated from the data / sensor data of emissions that can be generated by electron interactions outside the surface layer (i.e., the layer of the sample between the surface of the sample and the thickness of the layer removed during sectioning and view imaging). In this way, the depth reduction algorithm produces an enhanced version of the sectioning and view images, with a reduced amount of information generated by electron interactions outside the layers of the sample depicted in the corresponding samples of the enhanced version.

[0031] In some embodiments, when applying the depth blur reduction algorithm, the computing device 208 may be configured to first access a plurality of depth blur reduction algorithms and select a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more microscope conditions (e.g., beam voltage, spot size of the beam, beam type, etc.), sample conditions (e.g., sample type, sample material, sample features, etc.), or a combination thereof. Each of the depth blur reduction algorithms in the plurality of depth blur reduction algorithms may be trained for different corresponding microscope conditions, sample conditions, or combinations thereof. In such embodiments, the computing device 208 may select the depth blur reduction algorithm trained for the corresponding microscope conditions and / or sample conditions associated with the plurality of generated images.

[0032] In various embodiments, a depth blur reduction algorithm may include determining portions of image information and / or data / data streams of multiple sample layers attributable to the same element / molecule / feature, and then determining the layer or set of layers of the multiple sample layers in which the element / molecule / feature actually exists. For example, the depth blur reduction algorithm may determine that the element / molecule / feature is in the middle layer of the multiple layers or in the layer in which the portion of the image information and / or data / data stream attributable to the element / molecule / feature is optimal (i.e., for pixels, the clearest, strongest, highest percentage of signal, etc.).

[0033] Alternatively or additionally, the depth blur reduction algorithm may include a machine learning algorithm configured to process three or more dimensions, such as data in a 3D neural network. For example, the depth blur reduction algorithm may include a trained 3D neural network configured to receive multiple images and / or data generated based on slice and view imaging of a sample, and remove portions of the multiple images and / or data attributable to elements / molecules / features outside the corresponding layer of the sample, where the electron penetration depth of the imaging beam is greater than the slice thickness. According to the present invention, such a neural network may be trained using a first set of training data obtained by slice and view charged particle imaging of a sample and a second set of training data corresponding to a labeled reconstruction of the sample used as a reference, where the electron interaction depth of the imaging beam is greater than the corresponding slice thickness. In some embodiments, the second set of training data is obtained using a low voltage slice and view process, where the electron interaction depth of the imaging beam is less than the corresponding slice thickness. Alternatively, the second set of training data may be obtained at least in part by applying a deconvolution algorithm to at least a portion of the first set of training data. In another embodiment, the second set of training data is a simulated reconstruction of the sample based on a map of the sample, a description of the sample, known properties of the sample, or a combination thereof.

[0034] Figure 2 Further included is a schematic diagram showing an example computing architecture 250 of the computing device 208. The example computing architecture 250 shows additional details of the hardware and software components that may be used to implement the techniques described in the present disclosure. Those skilled in the art will understand that the computing architecture 250 may be implemented in a single computing device 208 or in multiple computing devices. For example, the individual modules and / or data constructs depicted in the computing architecture 250 may be executed by different computing devices 208 and / or stored on the different computing devices. In this way, the different process steps of the inventive method disclosed herein may be carried out and / or executed by separate computing devices 208 and in various orders within the scope of the present disclosure. In other words, in some embodiments, the functions provided by the shown components may be combined in fewer components or distributed in additional components. Similarly, in some embodiments, the functions of some of the shown components may not be provided and / or other additional functions may be used.

[0035] In an example computing architecture 250, a computing device includes one or more processors 252 and a memory 254 communicatively coupled to the one or more processors 252. The example computing architecture 250 can include an image generation module 256, a data enhancement module 258, a reconstruction generation module 260, a training module 262, and a control module 264 stored in the memory 254. The example computing architecture 250 is further shown to include a depth blur reduction algorithm 266 stored on the memory 254. As used herein, the term “module” is intended to represent an example partitioning of executable instructions for purposes of discussion and is not intended to denote any type of requirement or required methodology, manner, or organization. Thus, while various “modules” are described, their functionality and / or similar functionality can be arranged differently (e.g., combined into a smaller number of modules, broken down into a large number of modules, etc.). Additionally, while specific functions and modules are described herein as being implemented by software and / or firmware executable on a processor, in other examples, any one or all of the modules can be implemented in whole or in part by hardware (e.g., a specialized processing unit, etc.) to perform the described functions. As discussed above in various embodiments, the modules described herein associated with the example computing architecture 250 can be executed in multiple computing devices 208.

[0036] The image generation module 256 can be executed by the processor 252 to generate an image of a surface layer of a sample 206 within the example microscope system 204 based on data and / or data signals from the detector system 224. Specifically, because the data and / or data signals from the detector system 224 are based on emissions 223 emitted from the sample 206 during irradiation of the surface of the sample, the data and / or data signals can be analyzed to determine the composition (i.e., constituent elements / molecules / features) of the sample between the surface of the sample and the depth of electron interaction of the electron beam 216, such that an image of the surface of the sample can be generated. In some embodiments, the image is a grayscale image that shows contrast indicative of the shape and / or material of the sample.

[0037] The data enhancement module 256 may be executed by the processor 252 to apply a depth deblurring reduction algorithm to one or more of a series of generated images and / or detector data / data signals to remove information corresponding to elements / molecules / features present in layers of the sample 206 that do not correspond to the individual images from their respective portions. In other words, the depth deblurring reduction algorithm applied by the computing device 208 reduces the information / data generated by the emitted data / sensor data that may be generated by electronic interactions external to the surface layer (i.e., the layer of the sample between the surface of the sample and the thickness of the layer removed during sectioning and view imaging). In this way, the depth reduction algorithm produces an enhanced version of the sectioned and view images, with a reduced amount of information generated by electronic interactions external to the layer of the sample depicted by the corresponding sample in the enhanced version.

[0038] The depth deblurring reduction algorithm 266 may include a module or set of modules that determines portions of image information and / or data / data streams attributable to multiple sample layers of the same element / molecule / feature and then determines the layers of the multiple sample layers in which the element / molecule / feature is actually present. For example, the depth deblurring reduction algorithm may determine that the element / molecule / feature is located in the middle-most layer of the multiple layers or in the layer in which the portion of the image information and / or data / data stream attributable to the element / molecule / feature is optimal (i.e., for pixels, the clearest, strongest, highest percentage of signal, etc.). In this way, the depth deblurring reduction algorithm then removes the portions of the image / data signal attributable to elements / molecules / features located external to the corresponding layer of the sample.

[0039] In various embodiments, the depth deblurring reduction algorithm may include a trained machine learning module (e.g., an artificial neural network (ANN), a convolutional neural network (CNN), a fully convolutional neural network (FCN), etc.) that is capable of identifying portions of an image and / or detector data / data signals corresponding to elements / molecules / features present in layers of the sample 206 that are different from the layers corresponding to the individual image / data signal. For example, the depth deblurring reduction algorithm may include a machine learning algorithm configured to process data in three or more dimensions, such as in a 3D neural network. In some embodiments, the depth deblurring reduction algorithm may include a trained 3D neural network that is trained to receive multiple images and / or data generated based on sectioning and view imaging of a sample and remove portions of the multiple images and / or data attributable to elements / molecules / features external to the corresponding layer of the sample, where the electron penetration depth of the imaging beam is greater than the section thickness.

[0040] In some embodiments, the data enhancement module 256 may be executed by the processor 252 to first access a plurality of depth blur reduction algorithms 264, and then select a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more microscope conditions (e.g., beam voltage, spot size of the beam, beam type, etc.), sample conditions (e.g., sample type, sample material, sample features, etc.), or a combination thereof. Each of the depth blur reduction algorithms in the plurality of depth blur reduction algorithms may be trained for different corresponding microscope conditions, sample conditions, or combinations thereof. In such embodiments, the data enhancement module 256 may select a depth blur reduction algorithm trained for the corresponding microscope conditions and / or sample conditions associated with the plurality of generated images.

[0041] The reconstruction generation module 260 may be executed by the processor 252 to generate a 3D reconstruction of the sample 206 based on the enhanced image and / or data that has been enhanced by the depth blur reduction algorithm 266. Since the FIB column 212 is capable of removing layers of the sample 206 having a known thickness, the reconstruction generation module 260 is capable of determining the position of the sample 206 corresponding to each image in a series of images, and using this information to construct a 3D model / rendering of the sample 206.

[0042] The computing architecture 250 may include a training module 262 that may be executed to train the depth blur reduction algorithm 266 and / or its component machine learning algorithms to identify key points in the image at prominent features of the image. The training module 170 facilitates training the depth blur reduction algorithm 266 and / or the component machine learning algorithms using a first set of training data obtained by slice and view charged particle imaging of the sample and a second set of training data corresponding to a labeled reconstruction of the sample, where the electron interaction depth of the imaging beam is greater than the corresponding slice thickness. In some embodiments, the second set of training data is obtained using a low voltage slice and view process, where the electron interaction depth of the imaging beam is less than the corresponding slice thickness. Alternatively, the second set of training data may be obtained at least in part by applying a deconvolution algorithm to at least a portion of the first set of training data. In another embodiment, the second set of training data is a simulated reconstruction of the sample based on a map of the sample, a description of the sample, known properties of the sample, or a combination thereof. The training module 262 may be configured to perform additional training with new training data and then transmit updates that improve the performance of the depth blur reduction algorithm 266 and / or its component machine learning algorithms.

[0043] The control module 264 may be executed by the processor 252 to cause the computing device 208 and / or the example microscope system 204 to take one or more actions. For example, the control module 264 may cause the example microscope system 204 to perform slice and view processing of the sample 206, where the electron interaction depth of the imaging beam is greater than the thickness of the layer removed by the delay component (e.g., the FIB column 212).

[0044] As discussed above, computing device 208 includes one or more processors 252 configured to execute instructions, applications, or programs stored in a memory 254 accessible by the one or more processors. In some instances, the one or more processors 252 may include hardware processors including, but not limited to, a hardware central processing unit (CPU), a graphics processing unit (GPU), and the like. Although in many cases techniques are described herein as being performed by one or more processors 252, in some cases the techniques may be implemented by one or more hardware logic components such as a field programmable gate array (FPGA), a complex programmable logic device (CPLD), an application specific integrated circuit (ASIC), a system on a chip (SoC), or a combination thereof.

[0045] The memory 254 accessible by the one or more processors 252 is an example of a computer-readable medium. Computer-readable media can include two types of computer-readable media, namely computer storage media and communication media. Computer storage media can include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD), or other optical storage devices, magnetic tape cartridges, magnetic tape, magnetic disk storage devices, or other magnetic storage devices, or any other non-transmission medium that can be used to store the desired information and is accessible by a computing device. Generally, computer storage media can include computer-executable instructions that, when executed by one or more processing units, cause the performance of the various functions and / or operations described herein. In contrast, communication media embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transmission mechanism. As defined herein, computer storage media does not include communication media.

[0046] Those skilled in the art will also understand that, for purposes of memory management and data integrity, items or portions thereof may be transferred between the memory 254 and other storage devices. Alternatively, in other embodiments, some or all of the software components may be executed in a memory on another device and communicate with the computing device 208. Some or all of the system components or data structures may also be stored (e.g., as instructions or structured data) on a non-transitory computer-accessible medium or portable article for reading by an appropriate drive, various examples of which were described above. In some embodiments, instructions stored on a computer-accessible medium separate from the computing device 208 may be transmitted to the computing device 208 via a transmission medium or a signal such as an electrical, electromagnetic, or digital signal communicated via a communication medium such as a wireless link. Various embodiments may also include receiving, sending, or storing instructions and / or data implemented on a computer-accessible medium as described above.

[0047] Figure 3 Depicts a visual flowchart 300 according to the present invention that includes a plurality of images that together depict an example process that can be performed by the environment 200 to produce a high-resolution reconstruction of a 3D sample using a slice and view process imaging, where the electron interaction depth of the imaging beam is greater than the slice thickness. Images 302 to 306 show an example slice and view imaging process where the electron penetration depth 316 of the imaging beam 318 is greater than the thickness 320 of the layer removed by the delay component 322. In image 304, the delay component 322 is shown as a FIB, however, the present disclosure is not limited to this embodiment, and those skilled in the art will understand that other known delay techniques may be used within the scope of the present disclosure. Emissions generated by the interaction of the electrons of the imaging beam with the elements / molecules / features of the sample within the electron penetration depth are detected by the detector 324. As Figure 3 shown, the imaging and delay processes of the sample may be repeated until the desired portion of the sample is imaged (e.g., the desired region, multiple layers, the desired depth of the sample, the entire sample, a percentage of the sample, one or more features of the sample, etc.).

[0048] Image 308 shows multiple images of a sample layer generated based on data from detector 324 obtained during the slicing and view imaging process of images 302 to 306. Each of the multiple images is reconstructed using data from detector 324 during the imaging step of the corresponding layer. A depth blur reduction algorithm symbolically depicted in image 310 is applied to the multiple generated images of image 308 and / or the data used to construct them, such that information generated from elements / molecules / features located outside the corresponding layer of the sample is removed. In this way, an enhanced image as shown in image 312 can be created. These enhanced images in 312 are enhanced versions of the slice and view images in 308, with a reduced amount of information generated by electron interactions outside the layer of the sample depicted by the corresponding sample.

[0049] Image 314 shows a 3D reconstruction of the sample based on the enhanced image and / or data. Thus, using the method of the present disclosure, a higher resolution 3D reconstruction of the sample with reduced z-dimension blur due to electron interactions occurring outside the individual slices of the slicing and view processing can be created.

[0050] Images 316 to 320 illustrate an optional process for training one or more neural network components of the depth blur reduction algorithm. In this process, training data obtained using slice and view examination of a training sample (shown in image 316) is obtained, where the imaging beam is set such that its electron interaction depth is greater than the layer thickness. This training data is input into the neural network training process shown in image 320 along with the labeled data shown in image 318. The depth blur reduction algorithm can be retrained periodically and / or multiple depth blur reduction algorithms can be trained such that each depth blur reduction algorithm is optimized for a specific microscope and / or sample setup. Those skilled in the art will understand that while image 320 depicts a U-net training process, in various other embodiments, other training processes can be used to train the depth blur reduction algorithm.

[0051] Figure 4 is a flow chart of an illustrative process depicted as a collection of blocks in a logic flow chart, the blocks representing a sequence of operations that can be implemented in hardware, software, or a combination thereof. In the context of software, the blocks represent computer-executable instructions stored on one or more computer-readable storage media that, when executed by one or more processors, perform the recited operations. In general, computer-executable instructions include routines, programs, objects, components, data structures, etc. that perform a particular function or implement a particular abstract data type. The order in which the operations are described is not intended to be construed as a limitation, and any number of the described blocks can be executed in any order and / or in parallel combination to implement the process.

[0052] Specifically, Figure 4Flowchart of illustrative process 200 for generating high-resolution reconstructions of 3D samples using slice-and-view processes, where the electron interaction depth of the imaging beam is greater than the slice thickness. Process 400 may be implemented in environment 300 and / or by one or more computing devices 208 and / or by computing architecture 250 and / or in other environments and computing devices.

[0053] At 402, training data is optionally obtained, where the training data is generated using slice-and-view processes of a sample, where the electron depth of the imaging beam is greater than the slice thickness. The training data may be obtained by processing the sample via a microscope system, such as the dual-beam microscope system described in environment 200. Alternatively, the data may be obtained by a computing device via a wired or wireless connection (e.g., WAN, LAN, cable connection, etc.) or from another memory device storing the training data (e.g., another computer memory, hard drive, CD-ROM, portable memory device, etc.).

[0054] At 404, labeled training data of the sample is optionally obtained. The labeled training data corresponds to slice-and-view images of the sample with reduced depth blurring compared to the training data obtained in step 402. In some embodiments, the labeled training data is obtained using a low-voltage slice-and-view process, where the electron interaction depth of the imaging beam is less than the corresponding slice thickness. In another embodiment, the labeled training data may be obtained at least in part by applying a deconvolution algorithm, a noise reduction algorithm, and / or another type of data cleaning algorithm to at least a portion of the training data obtained in step 402. Alternatively or additionally, at least a portion of the labeled training data may be a simulated reconstruction of the sample based on a map of the sample, a description of the sample, known properties of the sample, or a combination thereof.

[0055] At 406, a depth blurring reduction algorithm is optionally trained at least in part based on the training data and the labeled data. For example, the depth blurring reduction algorithm may include a neural network component (e.g., a 3D neural network), which is trained by inputting the training data and the labeled data into a 3D U-net. In some embodiments, the depth blurring reduction algorithm may be retrained periodically to improve performance or to adapt to specific microscope and / or sample settings. By repeating such training, the performance of the depth blurring reduction algorithm can be improved. Alternatively or additionally, multiple depth blurring reduction algorithms may be trained such that each depth blurring reduction algorithm is optimized for a specific microscope and / or sample setting. In such embodiments, the training data in 402 is obtained via the desired microscope and / or sample setting, and the resulting depth blurring reduction algorithm is optimized for that setting.

[0056] At 408, image data of a sample is obtained, where the image data is generated using a slice and view process, and where the electron depth of the imaging beam is greater than the slice thickness. In some embodiments, the sample is processed by a microscope system, such as the dual beam microscope system described in environment 200, to obtain the image data of the sample. However, in other embodiments, obtaining the data may include obtaining the image data via a wired or wireless connection (e.g., WAN, LAN, cable connection, etc.) or from another memory device storing training data (e.g., another computer memory, hard drive, CD-ROM, portable memory device, etc.).

[0057] At 410, optionally, a plurality of images are generated from the image data, where each image corresponds to a layer of the sample imaged using slice and view imaging. For example, a plurality of grayscale images are shown that indicate the contrast of the shape and / or material of the individual layer of the sample corresponding to each image. Since the image data of the sample is obtained using a slice and view process where the electron interaction depth is greater than the slice thickness, the data obtained during the imaging of each individual slice contains information related to molecules / elements / features located outside the particular sample layer being imaged. Thus, each of the plurality of images also contains information related to molecules / elements / features that do not actually exist in the sample layer depicted. This phenomenon causes depth blurring of the images.

[0058] At 412, the image data is enhanced using a depth blur reduction algorithm. For example, the depth blur reduction algorithm can be applied to the image data and / or the plurality of images to remove from the individual images information corresponding to elements / molecules / features that do not actually exist in the layer of the sample corresponding to the individual image / image data. In other words, the depth blur reduction algorithm reduces the image information generated by the data / sensor data of emissions that can be caused by electron interactions outside the surface layer (i.e., the layer of the sample between the surface of the sample and the thickness of the layer removed between the imaging during slice and view imaging). In this way, the depth reduction algorithm produces an enhanced version of the slice and view image, with a reduced amount of information generated by electron interactions outside the layer of the sample depicted by the corresponding sample of the enhanced version. In some embodiments, the depth reduction algorithm can be selected from a variety of depth blur reduction algorithms based on one or more microscope conditions (e.g., beam voltage, spot size of the beam, beam type, etc.), sample conditions (e.g., sample type, sample material, features of the sample), or a combination thereof. Each of the depth blur reduction algorithms among the plurality of depth blur reduction algorithms can be trained for different corresponding microscope conditions, sample conditions, or a combination thereof.

[0059] At 414, multiple enhanced images are optionally generated from the enhanced image data. Because the amount of information in the enhanced image data associated with each image related to molecules / elements / features that do not exist in the sample layer depicted in the image is reduced, the images will have higher resolution, higher clarity, and / or reduced blurriness.

[0060] At 416, an enhanced 3D reconstruction of the sample is generated based on the enhanced image data and / or the enhanced images. Because the amount of information in the enhanced image data associated with each image regarding molecules / elements / features that do not exist in the sample layer depicted in the image is reduced, the amount of depth blurriness in the 3D reconstruction will be reduced, which is a problem in prior art slicing and view imaging processes.

[0061] Figure 5 and Figure 6 Slices and view images processed with the inventive process of the present disclosure are shown compared to slices and view images of prior art systems. For example, images 502 and 602 show images of layers of a sample obtained via slicing and view processing, where the electron penetration depth of the imaging beam is greater than the slice thickness. Images 504 and 604 respectively show the results of applying an image enhancement filter to the raw data of images 502 and 602. Images 506 and 606 respectively show the results of applying the depth reduction algorithm of the present disclosure to the raw data of images 502 and 602. Images 508 and 608 are the results of applying a deconvolution algorithm to post-process the raw data of images 502 and 602. It can be seen that compared to prior art systems, the methods and systems of the present disclosure improve the resolution of the sample while also increasing the signal-to-noise ratio of the images.

[0062] Examples of the inventive subject matter according to the present disclosure are described in the paragraphs listed below.

[0063] A1. A method for improving resolution reconstruction through sliced and viewed charged particle imaging, the method comprising: obtaining first data related to a first layer of a sample, the first data having been obtained by first irradiating the first layer of the sample with a charged particle beam; obtaining second data related to a second layer of the sample, the second data having been obtained by second irradiating the second layer of the sample with a charged particle beam, wherein: the first layer of the sample is removed between the first irradiation and the second irradiation; and the electron interaction depth of the charged particle beam is greater than the thickness of the first layer and the thickness of the second layer; enhancing the first data with a depth blur reduction algorithm to create enhanced first data, wherein the depth blur reduction algorithm reduces the depth blur caused by a first portion of the first data generated by electron interactions outside the first layer; enhancing the second data with a depth blur reduction algorithm to create enhanced second data, wherein the depth blur reduction algorithm reduces the depth blur caused by a second portion of the second data generated by electron interactions outside the second layer; and constructing a high-resolution 3D reconstruction of the sample using the enhanced first data and the enhanced second data.

[0064] A2. The method according to technical solution 1, further comprising the steps of: generating a first image of the first layer of the sample; and generating a second image of the second layer of the sample.

[0065] A3. The method according to any one of technical solutions A1 to A2, wherein enhancing the first data comprises generating an enhanced first image with reduced information content based on electron interactions outside the first layer.

[0066] A3.1. The method according to paragraph A3, wherein enhancing the second data comprises generating an enhanced second image with reduced information content based on electron interactions outside the second layer.

[0067] A4. The method according to any one of paragraphs A1 to A3.1, wherein reducing the depth blur corresponds to: removing a first portion of the first data generated by electron interactions outside the first layer from the first data; and removing a second portion of the second data generated by electron interactions outside the second layer from the second data.

[0068] A5. The method according to any one of paragraphs A1 to A4, wherein the depth blur corresponds to the presence of image data generated by an electron interaction depth of the charged particle beam greater than the thickness of the first layer and / or the second layer.

[0069] A6. The method according to any one of paragraphs A1 to A5, wherein the charged particle beam is an electron beam.

[0070] A6.1. The method according to paragraph A6, wherein the electron beam is a single energy electron beam.

[0071] A7. The method according to any one of paragraphs A1 to A6.1, wherein the first layer of the sample is removed by one or more of the following: focused ion beam; laser; electron beam; and diamond blade.

[0072] A8. The method according to any one of paragraphs A1 to A7, wherein the depth blur reduction algorithm is a neural network.

[0073] A8.1. The method according to paragraph A8, wherein the depth blur reduction algorithm is a 3D neural network.

[0074] A8.2. The method according to any one of paragraphs A8 to A8.2, wherein the neural network is trained using: a first set of training data obtained by slice and view charged particle imaging of the sample, wherein the electron interaction depth of the imaging beam is greater than the corresponding slice thickness; and a second set of training data corresponding to the labeled reconstruction of the sample.

[0075] A8.2.1. The method according to paragraph A8.2, wherein the second set of training data is obtained using low voltage slice and view charged particle imaging processing, wherein the electron interaction depth of the imaging beam is less than the corresponding slice thickness.

[0076] A8.2.2. The method according to any one of paragraphs A8.2 to A8.2.1, wherein the second set of training data is obtained at least in part by applying a deconvolution algorithm to the first set of data.

[0077] A8.2.3. The method according to any one of paragraphs A8.2 to A8.2.2, wherein the second set of training data is obtained by simulation based on mapping of the sample, description of the sample, known properties of the sample, or a combination thereof.

[0078] A8.3. The method according to any one of paragraphs A8 to A8.2.3, wherein the method further comprises training a neural network with a first set of training data obtained by slice and view charged particle imaging of the sample and a second set of training data corresponding to the labeled reconstruction of the sample, wherein the electron interaction depth of the imaging beam is greater than the corresponding slice thickness.

[0079] A.8.4. The method according to any one of paragraphs A8 to A8.3, wherein the depth blur reduction algorithm is a training-based 3D blind deconvolution algorithm.

[0080] A.8.5. The method according to any one of paragraphs A8 to A8.4, wherein the depth blur reduction algorithm is an algorithm based on a 3D blind neural network.

[0081] A9. The method according to any one of paragraphs A1 to A8.5, wherein the method further comprises accessing a plurality of depth blur reduction algorithms and selecting a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more microscope conditions.

[0082] A9.1. The method according to paragraph A9, wherein the microscope conditions include the charged particle beam voltage, the spot size of the charged particle beam, and the type of the charged particle beam.

[0083] A9.2. The method according to any one of paragraphs A9 to A9.1, wherein individual depth blur reduction algorithms among the plurality of depth blur reduction algorithms are trained for different corresponding microscope conditions.

[0084] A10. The method according to any one of paragraphs A1 to A9.2, wherein the method further comprises accessing a plurality of depth blur reduction algorithms and selecting a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more sample conditions.

[0085] A10.1. The method according to paragraph A10, wherein the one or more sample conditions include the type of the sample, the material of the sample, the features of the sample, or a combination thereof.

[0086] A10.2. The method according to any one of paragraphs A10 to A10.1, wherein individual depth blur reduction algorithms among the plurality of depth blur reduction algorithms are trained for different corresponding sample conditions.

[0087] B1. A charged particle microscope system, the system comprising: a sample holder configured to hold a sample; an electron beam source configured to emit an electron beam towards the sample; an electron beam column configured to direct the electron beam onto the sample; a delay component configured to remove a layer having a known thickness from the surface of the sample; one or more detectors configured to detect emissions generated by the electron beam irradiating the sample; one or more processors; and a memory storing instructions which, when executed on the one or more processors, cause the charged particle microscope system to perform the method according to any one of paragraphs A1 - A10.2.

[0088] C1. A method for improving resolution reconstruction through sliced-view charged particle imaging, the method comprising: training a deep blur reduction algorithm based at least in part on: a first set of training data obtained by sliced and view charged particle imaging of a training sample, wherein the electron interaction depth of the imaging beam is greater than the corresponding slice thickness; and a second set of training data corresponding to a labeled reconstruction of the training sample; obtaining first data related to a first layer of a sample, the first data having been obtained by first irradiating the first layer of the sample with a charged particle beam; obtaining second data related to a second layer of the sample, the second data having been obtained by second irradiating the second layer of the sample with a charged particle beam, wherein: the first layer of the sample is removed between the first irradiation and the second irradiation; and the electron interaction depth of the charged particle beam is greater than the thickness of the first layer and the thickness of the second layer; enhancing the first data with the deep blur reduction algorithm to create enhanced first data, wherein the deep blur reduction algorithm reduces the depth blur caused by a first portion of the first data generated by electron interactions outside the first layer; enhancing the second data with the deep blur reduction algorithm to create enhanced second data, wherein the deep blur reduction algorithm reduces the depth blur caused by a second portion of the second data generated by electron interactions outside the second layer; and constructing a high-resolution 3D reconstruction of the sample using the enhanced first data and the enhanced second data.

[0089] C2. The method according to paragraph C1, further comprising the steps of: generating a first image of the first layer of the sample; and generating a second image of the second layer of the sample.

[0090] C2.1. The method according to paragraph C2, wherein enhancing the first data corresponds to generating an enhanced first image with reduced information content based on electron interactions outside the first layer.

[0091] C2.2. The method according to paragraphs C2 to C2.1, wherein enhancing the second data corresponds to generating an enhanced second image with reduced information content based on electron interactions outside the second layer.

[0092] C3. The method according to any one of paragraphs C1 to C2.2, wherein reducing the depth blur corresponds to: removing a first portion of the first data generated by electron interactions outside the first layer from the first data; and removing a second portion of the second data generated by electron interactions outside the second layer from the second data.

[0093] C4. The method according to any one of paragraphs C1 to C3, wherein the depth blur corresponds to the presence of image data generated by an electron interaction depth of the charged particle beam greater than the thickness of the first layer and / or the second layer.

[0094] C5. The method according to any one of paragraphs C1 to C4, wherein the charged particle beam is an electron beam.

[0095] C5.1. The method according to aspect C5, wherein the electron beam is a single energy electron beam.

[0096] C6. The method according to any one of paragraphs C1 to C5.1, wherein the first layer of the sample is removed by one or more of the following: a focused ion beam; a laser; an electron beam; and a diamond blade.

[0097] C7. The method according to any one of paragraphs C1 to C6, wherein the depth blur reduction algorithm is a neural network.

[0098] C7.1. The method according to aspect C7, wherein the depth blur reduction algorithm is a 3D neural network.

[0099] C8. The method according to any one of paragraphs C1 to C7.1, wherein a second set of training data is obtained using low voltage slicing and view charged particle imaging processing, wherein the electron interaction depth of the imaging beam is less than the corresponding slice thickness.

[0100] C9. The method according to any one of paragraphs C1 to C8, wherein the second set of training data is obtained at least in part by applying a deconvolution algorithm to the first set of data.

[0101] C10. The method according to any one of paragraphs C1 to C9, wherein the second set of training data is obtained by simulation based on mapping of the sample, description of the sample, known properties of the sample, or a combination thereof.

[0102] C11. The method according to any one of paragraphs C1 to C10, wherein the method further comprises accessing a plurality of depth blur reduction algorithms and selecting a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more microscope conditions.

[0103] C11.1. The method according to aspect C11, wherein the microscope conditions include charged particle beam voltage, spot size of the charged particle beam, type of the charged particle beam.

[0104] C11.2. The method according to aspect C11, wherein individual depth blur reduction algorithms among the plurality of depth blur reduction algorithms are trained for different corresponding microscope conditions.

[0105] C12. The method according to any one of paragraphs C1 to C11.2, wherein the method further comprises accessing a plurality of depth blur reduction algorithms and selecting a depth blur reduction algorithm from the plurality of depth blur reduction algorithms based on one or more sample conditions.

[0106] C12.1. The method according to aspect C12, wherein one or more sample conditions include the type of the sample, the material of the sample, the characteristics of the sample, or a combination thereof.

[0107] C12.1. The method according to aspect C12, wherein an individual depth blur reduction algorithm among a plurality of depth blur reduction algorithms is trained for different corresponding sample conditions.

[0108] D1. Use the system of B1 to perform the method according to any one of paragraphs A1 to A10.2 or C1 to C12.1.

[0109] The systems, devices, and methods described herein should not be construed in any way as restrictive. In fact, this disclosure is directed to all novel and non - obvious features and aspects of the various disclosed embodiments, whether individually or in various combinations and sub - combinations formed with each other. The disclosed systems, methods, and devices are not limited to any specific aspect or feature or combination thereof, and the disclosed systems, methods, and devices do not require the presence of any one or more specific advantages or the solution of any one or more specific problems. Any theory of operation is for ease of explanation, but the disclosed systems, methods, and devices are not limited to such theory of operation.

[0110] Although the operations of some of the disclosed methods are described in a particular sequential order for ease of presentation, it should be understood that this description covers rearrangements unless the specific language set forth below requires a particular ordering. For example, in some cases, the operations described in sequence can be rearranged or performed concurrently. In addition, for simplicity, the figures may not show the various ways in which the disclosed systems, methods, and devices can be used in conjunction with other systems, methods, and devices. Additionally, the description sometimes uses terms such as "determine", "identify", "generate", and "provide" to describe the disclosed methods. These terms are high - level abstractions of the actual operations performed. The actual operations corresponding to these terms will vary depending on the particular implementation and are readily discernible by those skilled in the art.

Claims

1. A method for generating an improved 3D reconstruction of a sample through sliced and viewed charged particle imaging, the method comprising: acquiring first data related to a first layer of the sample, the first data having been acquired by first irradiating the first layer of the sample with a charged particle beam; acquiring second data related to a second layer of the sample, the second data having been acquired by second irradiating the second layer of the sample with the charged particle beam, wherein: removing the first layer of the sample between the first irradiation and the second irradiation; and the electron interaction depth of the charged particle beam is greater than the thickness of the first layer and the thickness of the second layer; enhancing the first data with a trained 3D blind deconvolution algorithm to create enhanced first data, wherein the trained 3D blind deconvolution algorithm reduces depth blur caused by a first portion of the first data generated by electron interactions outside the first layer; enhancing the second data with the trained 3D blind deconvolution algorithm to create enhanced second data, wherein the trained 3D blind deconvolution algorithm reduces depth blur caused by a second portion of the second data generated by electron interactions outside the second layer; and using the enhanced first data and the enhanced second data to construct a high-resolution 3D reconstruction of the sample.

2. The method according to claim 1, further comprising the steps of: generating a first image of the first layer of the sample; and generating a second image of the second layer of the sample.

3. The method according to claim 1, wherein enhancing the first data comprises generating an enhanced first image with reduced information content based on electron interactions outside the first layer, and enhancing the second data comprises generating an enhanced second image with reduced information content based on electron interactions outside the second layer.

4. The method according to claim 1, wherein reducing the depth blur corresponds to: removing the first portion of the first data generated by electron interactions outside the first layer from the first data; and removing the second portion of the second data generated by electron interactions outside the second layer from the second data.

5. The method according to claim 1, wherein the depth blur corresponds to the presence of image data generated by the electron interactions occurring outside the sample layer corresponding to the image.

6. The method according to claim 1, wherein the charged particle beam is a single energy electron beam.

7. The method according to claim 1, wherein the first layer of the sample is removed with one or more of the following: a focused ion beam; a laser; an electron beam; and a diamond blade.

8. The method according to any one of claims 1-6, wherein the trained 3D blind deconvolution algorithm comprises a 3D neural network.

9. The method according to claim 8, wherein the 3D neural network is trained using the following: A first set of training data obtained by slicing a training sample and performing view charged particle imaging, wherein the electron interaction depth of the imaging beam is greater than the corresponding slice thickness; and A second set of training data corresponding to the labeled reconstruction of the training sample.

10. The method according to claim 9, wherein the second set of training data is obtained using low voltage slicing and view charged particle imaging processing of the training sample, wherein the electron interaction depth of the imaging beam is less than the corresponding slice thickness.

11. The method according to claim 9, wherein the second set of training data is obtained at least in part by applying a deconvolution algorithm to the first set of training data.

12. The method according to claim 9, wherein the second set of training data is obtained by simulation based on a map of the training sample, a description of the training sample, known properties of the training sample, or a combination thereof.

13. The method according to claim 1, wherein the method further comprises accessing a plurality of trained 3D blind deconvolution algorithms and selecting a depth blur reduction algorithm from the plurality of trained 3D blind deconvolution algorithms based on one or more microscope conditions, wherein individual trained 3D blind deconvolution algorithms in the plurality of trained 3D blind deconvolution algorithms are trained for different corresponding microscope conditions.

14. The method according to claim 1, wherein the method further comprises accessing a plurality of trained 3D blind deconvolution algorithms and selecting the trained 3D blind deconvolution algorithm from the plurality of trained 3D blind deconvolution algorithms based on one or more sample conditions, wherein individual trained 3D blind deconvolution algorithms in the plurality of trained 3D blind deconvolution algorithms are trained for different corresponding sample conditions.

15. A charged particle microscope system, the system comprising: A sample holder configured to hold a sample; An electron beam source configured to emit an electron beam towards the sample; An electron beam column configured to direct the electron beam onto the sample; A delay component configured to remove a layer from the surface of the sample, the layer having a known thickness; One or more detectors configured to detect emissions generated by irradiating the sample with the electron beam; One or more processors; and A memory storing instructions that, when executed on the one or more processors, cause the charged particle microscope system to perform the following operations: Obtain first data related to a first layer of the sample, the first data having been obtained by first irradiating the first layer of the sample with the electron beam; Obtain second data related to a second layer of the sample, the second data having been obtained by second irradiating the second layer of the sample with the electron beam, wherein: The first layer of the sample is removed by the delay component between the first irradiation and the second irradiation; and The electron interaction depth of the electron beam is greater than the thickness of the first layer and the thickness of the second layer; Enhance the first data using a training-based 3D blind deconvolution algorithm to create enhanced first data, wherein the training-based 3D blind deconvolution algorithm reduces depth blur caused by a first portion of the first data generated by electron interactions outside the first layer; Enhance the second data using the training-based 3D blind deconvolution algorithm to create enhanced second data, wherein the training-based 3D blind deconvolution algorithm reduces depth blur caused by a second portion of the second data generated by electron interactions outside the second layer; and Construct a high-resolution 3D reconstruction of the sample using the enhanced first data and the enhanced second data.

16. The system according to claim 15, wherein reducing the depth blur corresponds to: Removing the first portion of the first data generated by electron interactions outside the first layer from the first data; and Removing the second portion of the second data generated by electron interactions outside the second layer from the second data.

17. The system according to any one of claims 15-16, wherein the training-based 3D blind deconvolution algorithm comprises a 3D neural network.

18. The system according to claim 17, wherein the 3D neural network is trained using: A first set of training data obtained by slicing and view charged particle imaging of a training sample, wherein the depth of electron interaction of the imaging beam is greater than the corresponding slice thickness; and A second set of training data corresponding to a labeled reconstruction of the training sample.

19. The system according to claim 18, wherein the second set of training data is obtained using low-voltage slicing and view charged particle imaging processing of the training sample, wherein the depth of electron interaction of the imaging beam is less than the corresponding slice thickness.

20. The system according to claim 18, wherein the second set of training data is obtained at least in part by applying a deconvolution algorithm to the first set of training data.

Images