Classification of substrate regions

Abstract

To provide a method for classifying regions of a substrate prior to performing an analytic procedure involving analyte particles supported on the substrate.SOLUTION: Regions of a substrate are sparsely scanned using low-energy electron point projection imaging. The regions are classified as suitable for the analytic procedure (or not) based on defects visible in the respective images. Given one suitable region, neighboring regions are scanned to increase the yield of suitable regions. Based on a distribution of suitable regions, a deposition pattern is planned, and analyte is deposited according to the plan. Following the deposition, the suitable regions are scanned again to identify or count visible analyte molecules. Based on the number of analyte molecules found, regions are earmarked for analyte characterization, e.g. by low-energy electron holography and reconstruction. A trained machine learning classifier provides consistent, accurate image classification across a range of defect types.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] Graphene has emerged as a suitable substrate capable of supporting proteins for low-energy electron holography (LEEH) and other similar transmission-mode analytical procedures. However, surface contaminants and structural defects can impair imaging performance and reduce yield. Therefore, improved techniques for improving yield in protein structural studies and similar applications remain needed. Summary of the Invention

[0002] Briefly, examples of the disclosed technology use low-energy electron point projection (LEEPP) imaging to identify regions of a substrate as suitable for analyte deposition and characterization. In one aspect, sparse scanning of regions can balance substrate area coverage and scan time, and secondary scans near suitable regions can efficiently increase the number of suitable regions available for analyte study. In another aspect, analyte deposition patterns can be established based on the distribution of suitable regions. In a third aspect, pre-analyte deposition classification can be used to identify regions suitable for analyte deposition, and post-analyte deposition classification can be used to identify regions with suitable amounts of analyte for further study. In a fourth aspect, images of scanned regions can be classified by trained machine learning tools.

[0003] The foregoing and other objects, features, and advantages of the present invention will become more apparent from the following detailed description which proceeds with reference to the accompanying drawings. [Brief explanation of the drawings]

[0004] [Figure 1] FIG. 1 illustrates an exemplary environment in which the disclosed technology can be implemented. [Figure 2]1 is a flowchart of an exemplary analysis workflow according to the disclosed technology. [Figure 3] 1 illustrates an example of LEEPP imaging in accordance with the disclosed technology. [Figure 4] 1 is a set of images illustrating examples of LEEPP images of different classes of graphene substrates. [Figure 5] 1 is a pair of schematic drawings illustrating an exemplary application of the disclosed technology. [Figure 6] 3 is a set of exemplary images according to the workflow of FIG. 2. [Figure 7] 1A-1C are views of a substrate illustrating exemplary regions and exemplary deposition patterns according to the disclosed technology. [Figure 8] 1 is a flowchart of a first exemplary method in accordance with the disclosed technology. [Figure 9] 10 is a flowchart of a second exemplary method in accordance with the disclosed technology. [Figure 10] 10 is a flowchart of a third exemplary method in accordance with the disclosed technology. [Figure 11A] 11A-11C are data flow diagrams of the respective phases of the method of FIG. 10. [Figure 11B] 11A-11C are data flow diagrams of the respective phases of the method of FIG. 10. [Figure 12] 11 is a flowchart of an exemplary extension of the method of FIG. 10 in accordance with the disclosed technology. [Figure 13] 1 illustrates a generalized example of a suitable computing environment in which embodiments, techniques, and technologies related to substrate characterization or analysis workflows according to the disclosed technology may be implemented. DETAILED DESCRIPTION OF THE INVENTION

[0005] Introduction Graphene has emerged as an attractive conductive substrate material, particularly for protein structure studies. Graphene monolayers and bilayers can be suspended over an aperture in a support member and have electron transparency exceeding 70% or 30%, respectively, at the typical low electron energies (50-600 eV) used in point projection imaging. This means that a high percentage of electrons can pass through the thin graphene layer without interaction. At the same time, 1-2 × 10 6 The conductivity of S / m is sufficient to prevent charge buildup and to provide a planar ground plane for electric field formation between the electron emitter and the substrate.

[0006] Analyte particles, such as proteins, can be deposited on graphene and studied by holography and other techniques. Unlike cryogenic transverse electron microscopy (cryoTEM), in which the analyte particles are immobilized, studies with the disclosed technique can be performed without analyte immobilization, allowing dynamic particle behavior to be studied.

[0007] However, analyte studies require that the graphene be clean, i.e., free of defects. Defects can interfere with the measurement of nearby analyte particles. Also, some defects pose a risk of arcing under high magnification conditions with high electric fields. Arc damage can result in significant instrument downtime of several hours or even days, in addition to wasting the substrate and deposited analytes.

[0008] The manufacturing process often results in graphene that is not sufficiently clean for analyte studies. For example, graphene foils can be produced by deposition onto a metal substrate and then transferred to a target substrate in a four-step process: (1) applying polymethyl methacrylate (PMMA) onto the graphene, (2) etching away the original metal substrate, (3) transferring to the target substrate, and (4) etching away the PMMA backing. This transfer process can leave PMMA or solvent residues or other contaminants on the transferred graphene. Repeated handling of graphene, even with techniques such as ultra-high vacuum (UHV) briefcases, presents further opportunities for contamination.

[0009] Contaminants are called "extrinsic defects" on graphene. Graphene can also develop "intrinsic defects," which are imperfections in the graphene crystal structure that can manifest as breaks, cracks, strand defects, bubbles, crystal dislocations, or other structural defects. Substrates of other 2D materials are feasible but face similar challenges.

[0010] Some examples of the disclosed techniques involve characterizing substrate regions for cleanliness. Regions that are free of defects or have defects below a severity threshold can be assigned as suitable for analyte deposition. After deposition, these regions can be imaged again. Any new particles observed can be considered analyte particles with a high degree of confidence.

[0011] In particular, the entire analyte characterization workflow described herein, including substrate characterization, analyte deposition, analyte mapping, and analyte characterization, can be performed in-situ in a single process chamber, substrate handling can be minimized, and an ultra-high vacuum of less than about 1 micropascal (7.5 nanotorr) can be maintained throughout.

[0012] While cleaning techniques exist, they can have limited success. In some cases, only 5-10% of a graphene substrate may be clean enough for analyte characterization. However, if analyte particles can be deposited in locally clean areas of the substrate, the need to clean the entire substrate can be avoided. Therefore, some disclosed techniques involve scanning areas of the substrate to identify locally clean areas.

[0013] Low-energy electron point projection (LEEPP) is an attractive technique for imaging graphene substrates because it can provide good contrast between the transparency of graphene and the visibility of features of both intrinsic and extrinsic defects. Furthermore, the image quality and magnification can be sufficient to characterize defects. An alternative technique, atomic force microscopy, is orders of magnitude slower, taking 10–20 minutes, including positioning, to scan a tile that can be imaged in about 5 seconds with LEEPP.

[0014] Deposition equipment can deposit over fairly wide patterns, often with lateral extents of 50-500 μm, often 100-200 μm, due in part to beam drift between the analyte source and the substrate. Therefore, it is desirable for the scanned area to cover the entire lateral extent of the substrate. As further described herein, there is a trade-off between high magnification (for better defect visibility) and a wide field of view (to efficiently cover the substrate with fewer images, thus reducing the duration). While areas with a lateral extent of a few μm (e.g., 3-10 μm or about 4-6 μm) are practical for defect visibility, scanning an entire substrate with a lateral extent of several hundred μm can take several hours. In various scenarios, a comprehensive scan of such a substrate can take 4-72 hours, or 12-24 hours.

[0015] Some examples of the disclosed technology employ a sparse scan approach to reduce scan time while still covering the lateral extent of the substrate. The substrate area can be divided into tiles. The sparse sampling of tiles can vary from 1 / 4 to 1 / 400 of the available tiles, reducing the total time to scan the substrate by a factor of 4 to 1 / 400. At approximately 1-10 seconds per tile, the entire sparse scan can be completed in less than an hour, or even just a few minutes. In various scenarios, a sparse scan can take 40 seconds to 6 hours, 2 minutes to 2 hours, or 10-40 minutes.

[0016] In many cases, defects are common and the percentage of area found to be defect-free can range from 1-20%, and in some cases can be as low as 5%. Furthermore, the density of defect-free areas can vary considerably across a substrate.

[0017] To increase the yield of analyte characterization data, it may be desirable to identify cleaner areas than can be observed using sparse scans. Defects may be spatially correlated, meaning that a clean tile is more likely to have additional clean neighboring tiles than a tile containing a defect. Thus, some examples of the disclosed technology employ the approach of performing secondary scans around an identified clean tile to increase the total number of clean tiles available for subsequent analyte studies.

[0018] Upon completion of the sparse scan and secondary scan, several tiles may be identified as suitable for the analytical procedure (e.g., analyte deposition and characterization, such as proteins). Some examples of the disclosed technology employ an approach to planning the deposition pattern to optimize the utilization of suitable tiles. As further described herein, various factors can be considered to optimize the deposition pattern. Because the deposition pattern may be non-uniform (often bell-shaped), the deposition pattern can be ordered to optimize suitable tiles within the preferred zones of the deposition pattern. In particular, near the edges of the deposition pattern, tiles may receive too few (or even no) analyte particles, which may be suboptimal. Conversely, near the center of the deposition pattern, tiles may receive too many analyte particles, and the signals of such particles may interfere with each other during subsequent characterization. The parameters of the deposition beam and the relative positioning of the source and substrate can be configured to obtain a suitable deposition pattern.

[0019] Analyte deposition can then be performed according to the plan, resulting in some suitable tiles with analyte particles ready for characterization, however it may be useful to map these tiles to avoid wasting characterization time on tiles with no analyte particles or too many analyte particles.

[0020] Some examples of the disclosed technology take the approach of repeating an imaging scan (e.g., LEEPP) over only suitable tiles to identify a subset of suitable tiles that actually have one or more visible analyte particles suitable for characterization.

[0021] Analyte particles within this subset of tiles can then be characterized. Low-energy electron holography can be used using the same or similar equipment as for LEEPP imaging, but configured for higher magnification. The holograms can be processed by a computer to reconstruct a structural description of the analyte particles.

[0022] In another aspect, consistent classification of LEEPP images can be challenging. When performed by humans, two people may disagree, or even one person may change the threshold for identifying defects over time. Machine vision tools, for example, using edge detection or shape detection techniques, can be effective, but may be based on heuristics and may fail when encountering images outside of a limited range. LEEPP imaging of 2D materials is also prone to variations in lighting across a single image or between images, which can pose additional challenges to machine vision heuristics.

[0023] Accordingly, some examples of the disclosed technology employ a trained machine learning tool to classify LEEPP images of substrate regions. The trained ML tool can be implemented as a convolutional neural network (CNN) trained on a labeled dataset with LEEPP images of similar substrates that have been classified by human experts. This approach can provide consistent classifications, can learn to mimic the classifications of human experts, and can provide more accurate results (in terms of agreement with human experts) than machine vision heuristics. Furthermore, as technology and analyte research requirements evolve, the ML workflow can be easily adapted to new substrates, image sets, or target classifications.

[0024] In tests, the trained ML classifier described herein demonstrated an accuracy of over 96%, with discrepancies occurring primarily in images for which the human experts had low confidence in their classification. For images for which the human experts had high confidence in their classification, the accuracy of the ML classifier was well over 99%.

[0025] The above description is illustrative and does not attempt to incorporate all innovative features disclosed herein. Further details are described below in the context of the drawings. Furthermore, any example of the disclosed technology can employ, omit, or modify any of the features described above, in any combination. In particular, the disclosed technology can be applied to other substrate materials or other imaging modalities.

[0026] term The usage and meaning of all terms cited in this section apply throughout this disclosure unless expressly indicated otherwise or contrary to context. The following terms expand to the related word forms:

[0027] An "analyte" is a material species used as a sample for an analytical procedure. In some disclosed examples, the analytical procedure can be holographic imaging of the analyte using an electron point projection microscope. Non-limiting examples of analytes can include macromolecules such as proteins, lipids, nanotubes, nucleic acids, polymers, capsids, biomolecule-ligand complexes, protein-protein complexes, RNA, viruses, virus-like assemblies, molecular component bioinspired materials, or inorganic nanoparticles. Other analytical procedures can include electron backscattering analysis, electron microscopy, etching, imaging, mass spectrometry, materials analysis, metrology, nanoprobing, spectroscopy, sample preparation, or surface treatment. An instrument used to perform an analytical procedure is called an "analytical instrument." Analyte particles are often molecules, although this is not a requirement; in some examples, analyte particles can be clusters or fragments of molecules.

[0028] An "aperture" or "pore" is a hole in a solid object that provides a distinct, linear path through the object from one exterior surface to the opposite exterior surface. Materials with multiple apertures are referred to as "porous." In some examples of the disclosed technology, the substrate can be supported on a porous support, such as a holey silicon nitride ("holey SiN") membrane, with aperture diameters typically in the range of 400-1000 nm.

[0029] A "beam" is a directed stream of particles or energy. Common beams of interest in this disclosure are electron beams, ion beams, or light beams, but the term is not limited to these. Ion beams and electron beams are streams of ions or electrons, respectively. Each particle in such beams has an electric charge and is a "charged particle." The "energy" of a charged particle beam is the average kinetic energy of the individual charged particles in the beam. A light beam is a stream of electromagnetic energy, which can be thought of as a wave or (particle-like) photons. A beam may have a finite extent transverse to the main longitudinal direction of the stream. The line connecting the centers of gravity of two or more cross sections of the beam is the "axis" of the beam, i.e., the "beam axis." An ion beam that deposits material on a substrate is a "deposition beam," and its axis is called the "deposition axis."

[0030] "Charged particle optics" refers to devices that can change the propagation characteristics of a charged particle beam. Exemplary propagation characteristics can include direction of propagation, focusing (e.g., converging or diverging), transverse shape, energy, or dispersion. Such devices can apply electric or magnetic fields to the beam and are collectively referred to as "electromagnetic devices." "Electromagnetic deflection elements" are electromagnetic devices configured to control the propagation direction of a charged particle beam.

[0031] "Classify" and "classification" refer to the act of assigning an item to one or more options from a finite, predetermined set of options. In some examples of interest herein, the classified items may be imaged regions of a sample, and each of the options may indicate a type or severity of a defect within the region. In various examples, classification may be performed by trained machine learning software, by other automated software, or by a user (e.g., using interactive software). The software program that performs the classification is called a "classifier." The classes assigned to training data are called "labels."

[0032] "Cleanliness" is a parameter that characterizes the severity of defects on a substrate or region thereof. Cleanliness can have a numeric value that varies proportionally or inversely with the amount of defects, or it can have a logical value (e.g., clean or not clean).

[0033] A "contaminant" is an undesired material species on or in a sample. In some disclosed examples, contaminants on a graphene sample can include organic residues from a polymethyl methacrylate (PMMA) transfer procedure. Contaminants can adversely affect downstream analytical processes, for example, by reducing resolution or otherwise degrading images performed using the sample. Contaminants can also cause equipment damage through arcing.

[0034] A "controller" is an electronic device coupled to one or more actuators for effecting a change in a physical parameter or coupled to one or more sensors for monitoring a physical parameter. Some controllers may include a microprocessor that can be programmed to execute machine-readable instructions. The description of a computing device herein is generally applicable to such controllers. Such controllers may include additional electronic circuitry, such as an analog-to-digital converter (ADC), a digital-to-analog converter (DAC), switches, comparators, filters, or amplifiers. Other controllers may include analog circuitry without a microprocessor.

[0035] A "criterion" is a condition or basis for making a category determination.

[0036] In the context of a substrate region, a "defect" refers to any characteristic of the substrate region that adversely affects the utility of supporting an analyte, imaging an analyte, or performing other analyses on the analyte. A defect is "visible" if it is recognizable on a LEEPP image of the substrate region. In particular, defects over openings in a perforated SiN support may be visible, but similar defects away from any openings may be invisible because the underlying Si support is opaque to the low-energy electron beam. "Intrinsic" defects are parts of the substrate (graphene) structure, such as bubbles, ruptures, cracks, dislocations, ridges, or voids. "Extrinsic" defects are on the surface of the substrate and can be contaminants in the form of particles, adsorbates, droplets, or films. Previously deposited analyte molecules can also be extrinsic defects for another future analyte deposition.

[0037] "Deposition" refers to the act of depositing a material (e.g., a protein or other analyte) onto a substrate. The spatial distribution of the deposited material is called the "deposition pattern."

[0038] A "detector" is a device for measuring a received signal, which may be in the form of light, an electrical signal, or an electron beam. Detectors can incorporate photodiodes, complementary metal oxide semiconductor (CMOS) elements, charge coupled devices (CCDs), microchannel plates, photomultiplier tubes, or similar devices, either alone or as an array. Some detectors can be pixelated and form an image from a spatially distributed signal. However, a pixelated detector is not a requirement for forming an image. In other examples, spatial variations in an image can be obtained, for example, by scanning a narrow beam or point beam across a sample (as in an SEM) or by moving a small detector across a spatially distributed signal. Detectors can be part of analytical instruments such as a spectrometer or an electron point projection microscope.

[0039] An "electron microscope" is a type of analytical instrument that irradiates a sample with an electron beam and uses the resulting particles or electromagnetic radiation to form a spatially resolved image. An "electron point projection microscope" uses a diverging electron beam to image a sample (e.g., an analyte) by detecting transmission through or around the sample, or diffraction or scattering from the sample's edges. A scanning electron microscope (SEM) images a sample surface based on reflected, secondary, or backscattered particles or radiation from one or more surfaces of the sample. An electron microscope can include an electron source (sometimes called an "electron emitter") and an imaging detector.

[0040] A "hologram" is an image that incorporates the amplitude and phase information of interfering waves. The waves can be electromagnetic waves (e.g., light waves) or de Broglie waves of interfering electrons (e.g., in an electron point projection microscope). Various techniques exist for reconstructing the spatial description of the imaged object. As an example, a hologram can be Fourier transformed into momentum space, propagated in momentum space, and then transformed back into physical space to recover the spatial description of the imaged object.

[0041] An "image" is a two-dimensional representation of parameter values ​​over an area of ​​interest of a sample. In examples, the area of ​​interest may be a tile of a substrate, a portion of a substrate over an opening in a support, or a smaller area enclosing one or a few analyte molecules. The parameter may be the amplitude of electrons reaching a detector. An "imager" is a device for generating an image and may include an illumination source and a detector.

[0042] "Low-energy electron point projection" ("LEEPP") is a technique for imaging samples using an electron beam with energies in the range of 50-1000 eV. While voltage and magnification can be varied independently as described herein, lower-magnification imaging (wider field of view) can be performed using higher voltages. Thus, a maximum of approximately 600 eV is suitable for tiles with a lateral range of 1-10 μm, and LEEPP imaging of entire 500 μm substrates can be performed using energies up to approximately 1000 eV. At higher energies, defects may become transparent, reducing the contrast between clean graphene and defects. At higher magnifications, interference effects in the image may become more pronounced, and the image can be considered a hologram. In this regime, LEEPP can be referred to as "low-energy electron holography" ("LEEH").

[0043] "Machine learning" (or "ML") refers to techniques for improving the performance of a software tool through experience (called "training"), and the tool is called a "machine learning tool." The adjective "trained" modifier indicates that the ML tool has been trained to achieve performance at least equal to a predetermined threshold. Training can be performed in a "training phase," in which training data (for which the desired output is known) is applied to the tool's inputs, and deviations between the tool output and the desired output are used to adjust parameter values ​​within the ML tool, for example, by backpropagation. After the ML tool has been trained, it can be fed new data and its output can be used as needed. This phase is sometimes called the "inference phase." A neural network is an example of a software tool that can be trained by machine learning.

[0044] A "major surface" of a substrate or sample is a surface of the substrate or sample that is not substantially larger in area than any other surface of the substrate or sample. For purposes of discussion, a substrate or sample is considered to have top and bottom major surfaces, with the bottom surface being fixed to a support. In some instances, a low-energy electron beam can be directed at the top surface of the substrate, and analyte deposition can occur on the bottom surface, e.g., through an opening in the support.

[0045] A "molecule" is the smallest unit of a material that has the same composition and chemical properties as the material. While many molecules contain two or more atoms (CO, O), this is not a requirement; a single atom (e.g., He) can also be considered a molecule. Some molecules described herein may be in an ionized state, for example, in a deposition beam.

[0046] A first and second region are "adjacent" if they are adjacent to one another, including along a diagonal. A gap between two regions does not exclude them from being adjacent, unless another region is present within the gap.

[0047] A "neural network" is an artificial network of "units" (or "cells") with linkages modeled on the behavior of biological neurons and that may be implemented by a software program on a computer. A neural network is an example of a machine learning tool. Some neural networks described herein may be "convolutional neural networks" (or "CNNs") with connections between cells that are independent of the spacing between cells. A CNN is a multi-layer neural network incorporating at least one convolutional layer, the connectivity and parameters of which are determined by applying a convolution operation uniformly across the cells of the previous layer to obtain the current layer.

[0048] A "notification" is a message that may or may not require a substantive response. A notification may be presented as a computer communication (e.g., over a bus or network or written to a shared memory location), as a visual display on a graphical user interface (GUI) or annunciator, or by audio or tactile means. In contrast, a "request" is a message that expects a substantive response (e.g., beyond a simple acknowledgment).

[0049] A "parameter" is an attribute or characteristic of a physical system that can be measured or observed. The result of such a measurement or observation is the "value" of the parameter. A parameter can be an attribute or characteristic of a physical system that can be controlled. In this context, a setting applied to a parameter is its value. A parameter can have one or more "values." While parameters often have numeric values, this is not a requirement; some parameter values ​​can be Boolean values ​​(e.g., whether the parameter is above or below a threshold, or on or off), categorical variables (e.g., a defect type or severity classification), strings (e.g., describing the parameter), or data structures (e.g., a LEEPP image).

[0050] A "protein" is a molecule comprising a chain of amino acids linked by peptide bonds. In the case of a protein analyte, the analyte particle is the protein molecule.

[0051] A "region" is a contiguous, generally two-dimensional extent of a substrate and can include both the top and bottom surfaces, the volume enclosed therein, and extrinsic defects on either surface. In various examples, a region can be a portion of a substrate suspended over a single aperture, or a tile covering multiple apertures. Descriptions of regions herein encompass both apertures and larger tiles, as well as areas smaller than a single aperture. A "suitable" region is one that meets defect severity criteria (e.g., few or no defects) and therefore can be used for subsequent analysis operations, including analyte deposition.

[0052] A "region of interest" (ROI) is a portion of a substrate or sample that contains a feature or structure (e.g., an analyte molecule, a substrate defect, or a defect-free substrate region) that is the subject of imaging or other analysis. The term ROI does not refer to any human interest.

[0053] "Rotation" refers to a change in the angular orientation of an object relative to another object or a fixed coordinate system. In the former case, rotation can be achieved by moving either one or both of the objects. In some examples, the deposition axis of the analyte ion beam can be rotated relative to the substrate. Such rotation can have two degrees of freedom: tilt with respect to the surface normal of the substrate, and azimuthal rotation about the surface normal.

[0054] A "sample" is a physical object on which an analytical, fabrication, or preparatory procedure is performed. A "substrate" is another physical object, distinct from a sample, on which the sample is mounted for such a procedure. An analytical, fabrication, or preparatory procedure can also be performed on a substrate. In a series of operations, the sample on which one operation is performed can be the substrate for another operation. Illustratively, the disclosed techniques can be used to characterize a graphene sample mounted on a holey silicon nitride substrate. The disclosed techniques can also be used to prepare analyte samples on characterized regions of the graphene substrate. Holographic analysis can then be performed on the analytes deposited on the graphene, with the graphene serving as the substrate. In an example, the sample and substrate can be sequentially secured to a support (e.g., holey SiN), a mount (e.g., a clamping fixture), and a stage. Some graphene substrates described herein can have lateral extents of about 200-1000 μm, or about 400-500 μm.

[0055] "Severity" is a measure of the extent to which one or more defects affect, for example, the usability of substrate area for analytical procedures. In various examples, severity can be expressed as a binary variable, a categorical variable, or a number.

[0056] "Software" refers to computer-executable programs, instructions, or associated data structures. Software may be in an active state or a quiescent state. In an active state, software may be loaded into memory or undergoing execution by one or more processors. In a quiescent state, software may be stored on a computer-readable medium, awaiting transmission or execution.

[0057] A "source" is a device configured to deliver material or energy. Exemplary sources described herein include an electron source, an analyte source, or a light source. Illustratively, imaging can be performed using an electron beam generated by an electron source. An exemplary analyte source can deliver a beam of ionized proteins onto a clean graphene sheet for subsequent imaging. A light source can generate light at one or more wavelengths.

[0058] "Sparse" refers to a sample whose concentration is at most 25% of the concentration of the population or space of which it is a part, and whose members form at least two disjoint groups. In some examples, a sparse set of images has a lateral separation distance (e.g., gap) between images that is equal to or greater than the lateral extent of each image.

[0059] A "stage" is a device to which a substrate or sample can be attached and which has actuators for adjusting the position or orientation of the sample. A "multi-axis stage" has multiple degrees of freedom, providing spatial adjustment of the sample in various combinations of translation or rotation.

[0060] "Steering" refers to the act of changing the direction of a beam relative to a physical object or a fixed frame of reference. Steering of charged particle beams can be achieved using electromagnetic deflection elements.

[0061] "Training" refers to the process of determining values ​​(coefficients) to be applied within an ML tool, for example, in the neurons of a neural network, to enable the ML tool to perform its desired function. In the context of a classifier, training can be performed using a training dataset containing images whose desired classifications are already known. Comparing the actual output of the trained ML classifier with the desired classification can provide a loss function, which can be propagated backward through the ML classifier, applying gradient descent or another established technique to adjust the coefficients at each layer. As training progresses, the classifier output can converge to the desired classification and the magnitude of the loss function can be reduced. If the loss function reduces below a predetermined threshold, the classifier can be validated and considered "trained." A trained classifier can be deployed to classify new data (e.g., LEEPP images) whose correct classification is not yet known.

[0062] The term "translation" refers to a change in the position of a physical object relative to another object or relative to a fixed coordinate system. In the former case, translation can be achieved by moving either one or both of the objects. A translation from coordinates (x1, y1, z1) to (x2, y2, z2) can be performed in a straight line (i.e., as a "linear translation"), but need not be unless explicitly stated. Translation can be performed in a circular, wavy, or other path. Similarly, translation does not require or exclude a change in orientation. However, a change in orientation without a change in position is not considered a translation.

[0063] Terms such as "top," "bottom," "up," "down," "upper," "lower," "horizontal," and "vertical" are used for convenience with respect to the common configuration in which the exposed major surface of a horizontal sample has an outward normal pointing upward, e.g., the sample is accessible to a process tool from above. The term "lateral" refers to a direction within or parallel to the major surface of the specimen. "longitudinal" refers to a direction along the beam axis. Those skilled in the art will understand from this disclosure that the actual orientation selection can be varied without departing from the scope of the disclosed technology.

[0064] A "two-dimensional material" (or "2D material") is a thin sheet of material with a thickness of 1 to 10 atomic layers. Some 2D materials are highly transparent to electrons, light, or other bands of electromagnetic radiation. Some 2D materials have a single atomic layer and are called "monolayers" or "monolayers." Some 2D materials have two atomic layers and are called "bilayers" or "bilayers." An example of the disclosed technology uses monolayer graphene as the substrate. Bilayer graphene or other 2D materials can also be used. Exemplary alternatives to graphene include hexagonal boron nitride, molybdenum disulfide, other transition metal dichalcogenides, phosphorene, silicene, platinum, nickel, melamine, 2DPA-1, or similar sheet polymers.

[0065] "Validation" is the process or act of verifying that a software module meets its specifications. In the context of ML classifiers, the specifications may require that the agreement between the classifier output and a human expert meets or exceeds a predetermined threshold. The threshold may be in the range of 80-100%, e.g., 90%, 95%, 98%, 99%, 99.5%, 99.8%, or 99.9%.

[0066] Example Environment and Analysis Workflow Figure 1 is a diagram 100 of an exemplary environment in which the disclosed techniques can be implemented. Shown in Figure 1 are an exemplary defective substrate, mounting device, analytical instrument, and controller. Figure 1 will be discussed in conjunction with Figure 2, which is an example of an analytical workflow for an analyte species.

[0067] The substrate 110 can be fixed to a support 130, which can be held on a stage 135 by a mounting bracket 133. The stage 135 can be operated by a controller 170 to provide rotation or translation of the substrate 110 with one or more degrees of freedom. The substrate 110 can incorporate graphene or another 2D material. The support 130 can incorporate silicon (e.g., in the form of holey silicon), silicon nitride (SiN), or another porous material. As illustrated, the openings 132 are regularly distributed on the support 130, although this is not a requirement.

[0068] The substrate 110 can have a classification of defects. Exemplary defects can include extrinsic defects, such as contaminant particles 112, 114 on the top or bottom surface of the substrate 110, respectively, or intrinsic defects 116 shown as breaks. The illustrated defect locations are exemplary. Other types of intrinsic defects can be located on the surface of the substrate 110, and extrinsic defects can also penetrate through openings in the substrate 110. Defects are further described herein.

[0069] FIG. 1 also shows a LEEPP imaging subsystem, including an electron emitter 120 configured to generate a low-energy electron beam 122 and an electron detector 125, which may be a pixelated detector. A controller 170 can control the emitter 120 and read out the detector 125. The LEEPP imaging subsystem can be used to characterize regions of the substrate 110 according to defect severity, for example, in block 210 of FIG. 2. The low-energy electrons 122 can penetrate the substrate 110 but make defects 112, 114, and 116 visible. Representative LEEPP images showing the presence of such defects are described further herein. While FIG. 1 shows the beam 122 incident on the top surface 117 of the substrate 110, this is not a requirement; in other examples, LEEPP imaging can be performed from the opposite side of the substrate 110.

[0070] For clarity of illustration, the beam 122 is shown illuminating a substrate area corresponding to a single aperture 132 in the support 130, although this is not a requirement. In the example, a suitable LEEPP imaging geometry may be constrained by several factors. In one aspect, a larger beam width may be subject to variations in beam intensity across the cross section, resulting in a corresponding degradation in image quality. The actual lateral size of the detector 125 may be another constraint. For a given actual detector size and beam width, there may be a trade-off between magnification and imaging area. Thus, moving the substrate 110 closer to the source 120 while keeping the detector 125 fixed increases the magnification but reduces the field of view (e.g., the imaged area of ​​the substrate 110), thereby requiring more time to scan a given amount of substrate surface area and reducing scanning efficiency. Conversely, moving the substrate 110 closer to the detector 125 while keeping the emitter 120 fixed increases the field of view but reduces the magnification. At low magnification, significant defects may not be discernible or may be prone to incorrect classification.

[0071] Substrate 110 may desirably be used as a support for analyte particles 147. To this end, analyte deposition source 140 is also shown. Illustratively, source 140 may incorporate an analyte source 142, a conditioner 144, and a delivery subsystem 146, some or all of which may be controlled by controller 170. In some examples, source 142 may be an electrospray ionizer or a matrix-assisted laser desorption / ionization (MALDI) source, which may be configured to introduce analyte ions into conditioner 144. Conditioner 144 may be a mass spectrometer and may function as a mass filter, transferring only particles within a narrow window centered around a desired m / Z ratio to subsystem 146. For ionized analytes, delivery subsystem 146 may be implemented using charged particle optics. In an example, optics 146 may deliver a collimated beam 148 having a beam width of 100-500 μm through a series of electromagnetic lenses.

[0072] In an example, an electric field can be applied to the analyte stream 148 to apply braking to the analyte ions as they approach the substrate 110, providing a soft landing with particle kinetic energies of about 5-20 eV, or less than about 50 eV, at a surface potential of about 0.2-5 V, typically about 1 V. The soft landing can prevent fragmentation of the analyte molecules 147 upon collision with the substrate 110 and can also prevent damage to the substrate 110 from this collision. While FIG. 1 shows the beam 148 incident on the bottom surface 119 of the substrate 110, this is not a requirement, and in other examples, analyte deposition can be performed from the opposite side of the substrate 110.

[0073] FIG. 1 also illustrates certain attributes of the analyte deposition beam 148, which is shown as having a deposition axis 145. In particular, the beam 148 can be translated relative to the substrate 110, as indicated by arrow 141. The beam 148 can also be rotated relative to the surface normal 115. Arrow 143 represents two types of rotation: tilt, which changes the angle between axis 145 and the normal 115, and azimuthal rotation about the normal 115. Charged particle optics 146 can also be used to configure the width, shape (e.g., circular or elliptical), and focusing characteristics (e.g., converging, collimated, or diverging) of the deposition beam 148. Through control of these beam attributes, for example, by controller 170, the shape, orientation, and position of the analyte deposition pattern can be controlled. In particular, as shown in block 220 of FIG. 2, analyte deposition can be planned based on defect characterization using LEEPP imaging in block 210. Following block 220, in block 230, an analyte 147 may be deposited onto the substrate 110, as described above.

[0074] Typical ranges for beam parameters are analyte beam currents of 1-30 pA, deposition times up to about 10 minutes, and ionic charge states (Z) in the range of 5-30 for proteins and similar analyte macromolecules, or total deposited charge in the range of 0.1-10 nC (often 0.5-2.0 nC) when Z is in the range of 10-20 or about 15.

[0075] Following deposition of the analyte 147, in block 240, the LEEPP imaging subsystem can again be used to identify or map, for example, areas of the substrate 110 that were previously defect-free and now contain one or more deposited analyte particles 147. In some analyte studies, it may be desirable to have exactly one particle 147 within a given aperture 132 to avoid interference between subsequent measurements on two nearby analyte particles. However, in other studies, having two or more (but not too many) analyte particles 147 within the aperture 132 allows for the collection of more analyte characterization data within a given experiment duration or over a given substrate area. Time may be particularly important in some studies due to natural degradation of analyte particles or the actual duration that temperature, pressure, or other process conditions can be maintained on the substrate 110 and sample 147. Each analyte study may have a respective target range of analyte particles desired to be deposited within an area. Illustrative examples of target ranges can be 3 to 20 (e.g., multiple analyte particles for faster studies), 1 to 10, or 1 to 1 (e.g., a single analyte particle for minimal interference between particles 147).

[0076] Once a suitable analyte-containing region has been identified, further analysis of the analyte 147 can be performed in block 250. In some examples, the LEEPP imaging subsystem (or a separate but similar subsystem) can be operated at high magnification to acquire holograms of the analyte 147, for example in LEEH mode. These holograms can be used to reconstruct a spatial description of the analyte 147. However, the disclosed techniques and environments are not limited to LEEH and can apply a wide range of imaging, spectroscopy, or other analytical techniques.

[0077] Thus, the illustrated environment may also include one or more additional subsystems. In some examples, subsystem 150 may represent an alternative analyzer. In other examples, subsystem 150 may represent a cleaning tool for reducing extrinsic defects on substrate 110. In-situ cleaning may be performed in optional block 215 prior to surface characterization to improve yield of areas suitable for analyte studies. Cleaning may also be performed after analyte studies are completed to remove analyte particles 147 as well as extrinsic defects, for example, to render substrate 110 reusable for further analyte studies.

[0078] In some examples, a single controller may control and monitor all of the subsystems illustrated in FIG. 1, while in other examples, multiple controllers may be used for the various subsystems in any combination.

[0079] The disclosed technology facilitates performing an analytical procedure for an analyte on a substrate. In an example, referring to Figure 2, the analytical procedure can include block 235 or any one or more of blocks 230, 240, and 250. As indicated by arrow 255, the workflow of Figure 2 can be repeated using a single substrate or using different substrates.

[0080] Exemplary LEEPP imaging FIG. 3 is a diagram 300 illustrating an exemplary LEEPP imaging system. An electron source 320 can be configured to emit a low-energy electron beam 322 toward an electron detector 325. Along the way, the beam 322 can encounter a support 330 to which a substrate (similar to 110) can be attached. Conceptually, the thin substrate can be transparent to the electrons 322 and is not shown separately in FIG. 3. As shown, the support 330 can be porous and have openings 332 through which the electrons 322 can propagate, while the support material itself is opaque to the electrons 322. The substrate, affixed to the support 330, is located in the object plane of the illustrated imaging system 300.

[0081] When possible, electrons 322 propagate beyond support 330 along projection path 360 toward detector 325. Detector 325 is at an image plane. As shown, the image formed at detector 325 may be a projection of the structure at the object plane. Figure 3 illustrates the projection of several apertures 332 within the field of view of beam 322. Similarly, an image 368 of defect 318 may also be projected onto detector 325.

[0082] Measured from the tip of the electron source, the substrate on support 330 is at an object distance Z0 372, and the detector 325 is at an image distance Z1 325. According to geometric optics, the magnification M of this configuration is M = Z1 / Z0. In examples, LEEPP can be performed at an object distance Z1 372 of about 10-50 μm and an image distance Z1 of about 50-300 mm or about 100-150 mm. In contrast, LEEH can be performed at an object distance Z1 372 of about 1 μm. The transition between geometric and holographic imaging can occur around about 5 μm for typical Z1 distances.

[0083] LEEPP Image Class Example Figure 4 is a set of LEEPP images of a graphene substrate illustrating examples of various classes of defects. For reference, image 402 shows a defect-free area on a single opening in a perforated silicon nitride substrate. Images 410, 415, 420, 430, 440, and 450 show various exemplary defect types.

[0084] Image 410 shows a single large contaminant object 412 that interferes with electron transport. The contaminant 412 could be, for example, a large catalyst particle or flake of material used in the transfer process to transfer the graphene substrate from a catalyst or other growth medium onto a holey silicon nitride support.

[0085] For example, under high magnification conditions for LEEH imaging, the object distance ZO (372) is very small and the electric field can be high even for the low electron energies used. Under these conditions, large contaminants 412 pose a high risk of arcing the electron emitter (320). Arcing can damage the electron emitter beyond repair, requiring hours or even days to replace the electron emitter. Therefore, the imaging region 410 may be unsuitable for analyte studies.

[0086] Image 415 shows several smaller defects 417, which may be minute catalyst particles on one or both surfaces of the substrate. These are also opaque to LEEPP electrons. Area image 415, although smaller than image 410, does pose some arcing risk. Particles 417 are likely to interfere with downstream holographic imaging of any analyte particles, and area 415 may also be unsuitable for analyte studies.

[0087] Image 430 shows haze on the substrate surface: objects of significant size are not easily discernible and the risk of arcing is relatively low, but the haze significantly reduces electron transmission, likely resulting in poor signal-to-noise ratios for any analyte holograms.

[0088] Images 410, 415, and 430 show extrinsic defects, while images 420, 440, and 450 show examples of intrinsic defects. Image 420 shows graphene strands 422 that may present an arcing risk. Image 440 shows bright edges 442 indicative of missing graphene and distortions at the hole edges, which may also pose some relatively low risk of electric field concentration and arcing. However, viewing conditions for analyte studies in other parts of the imaged area are good. Finally, image 450 shows a small number of defects, sometimes called "bubbles," which represent small localized accumulations of positive charge. These do not significantly impede electron transport or pose a significant arcing risk. In some studies, the severity of the defects decreases continuously from image 410 to image 450.

[0089] Different analyte studies may have different sensitivities to defects. In some LEEH studies, all images 410-450 may be deemed unsuitable for the analyte study, with only images such as 402 being suitable for analyte imaging. In other LEEH studies, image 450 may be deemed suitable (in addition to image 402), with images 410-440 being unsuitable. In a further example, images {402, 440, 450} may be deemed suitable, with the remaining defect examples determined to be unsuitable.

[0090] Exemplary Analyte Mapping Using LEEPP Imaging FIG. 5 shows a pair of schematic drawings 501 and 502 illustrating LEEPP images acquired before and after analyte deposition, respectively. Drawing 501 shows a substrate 510 with a defect 530, which may be an intrinsic defect (e.g., a break in the graphene substrate) or an extrinsic defect (e.g., a surface contaminant). Also shown is an analyte particle 540 that may have been previously deposited and characterized. Thus, drawing 501 may represent a LEEPP image taken in block 210 in the second iteration of the workflow of FIG. 2. In drawing 501, a dashed outline 551 indicates a clean area on substrate 510.

[0091] Drawing 502 shows substrate 510 at a later point in time after further analyte deposition, for example, in block 240 of the third iteration of the workflow of Figure 2. Defects 530 and previous analytes 540 remain in place, but analyte particles 552 have been deposited and are visible in LEEPP image 502. Thus, a comparison of image 502 with image 501 can show that newly deposited analyte molecules 552 are present in previously clean areas 551. Analyte molecules 552 can be further characterized, for example, by LEEH imaging in block 250 of the third workflow iteration.

[0092] Exemplary Image Sequence Figure 6 is an example set of images according to the workflow of Figure 2. First, in block 210, a LEEPP scan of the substrate can identify, in image 610, defect-free areas of the substrate.

[0093] In the first example, in block 240, the same region can generate a second LEEPP image 640. In image 640, 10-20 protein molecules 641 are recognizable. LEEH can then be used to acquire a higher magnification image 650 in block 250. The hologram 651 is well separated from images of other protein molecules and can be used for high-quality spatial reconstruction of the protein molecules 641.

[0094] In a second example, an image similar to 645 may be acquired in block 240. Image 645 shows approximately 50 protein molecules, some of which are naturally clustered into chains. A corresponding LEEH image 655 may be acquired in block 250. Image 655 has a lower magnification than image 650, which may be suitable for certain studies. However, at higher magnifications, interference between holograms of neighboring protein molecules increases, making the hologram unsuitable for protein characterization.

[0095] Exemplary Regions and Deposition Patterns FIG. 7 is a diagram 700 of a substrate illustrating exemplary regions and exemplary deposition patterns according to the disclosed techniques.

[0096] The available area of ​​the substrate 702 can be subdivided into tiles, such as 711-714 or 780. Initially, a sparse set of tiles 711-714, 721-724, 731-734, and 741-744 can be imaged by LEEPP, and other tiles 780 can be omitted. As described herein, the tile size and LEEPP magnification can be selected together to optimize scanning efficiency. Some imaged tiles can be found to have defects of an acceptable severity level (including cases where there are no visible defects), while other imaged tiles can be found to have defects of an unacceptable severity level.

[0097] In the description, for purposes of illustration, it is assumed that tiles 732, 743 are found to have acceptable defect severity and are therefore suitable for subsequent analytical procedures, e.g., analyte characterization. A secondary LEEPP imaging scan can then be performed on tiles 751-758 adjacent to tile 732. Similarly, a secondary LEEPP imaging scan can be performed across tiles 761-768 adjacent to suitable tile 743. Because both intrinsic and extrinsic substrate defects can often be spatially correlated, the vicinity of suitable (e.g., clean) tiles 732, 743 can have a higher probability of additional acceptable tiles compared to the vicinity of unsuitable (e.g., defect-containing) tiles.

[0098] Figure 7 also shows several circles and ellipses, or portions thereof, representing analyte deposition patterns. Circles 774-776 represent one such pattern. Analyte deposition density from a deposition beam similar to 148 in Figure 1 can have a bell-shaped, Gaussian, or parabolic shape that decreases with distance from axis (145). Also, as described herein, a given study can have a target range for how many visible analyte particles are desired to be deposited in an area (tile). Illustratively, circles 774 and 775 can be loci where the expected analyte deposition density times the tile area equals the upper and lower limits of the target range, respectively. Thus, tiles within the annular region bounded by circles 774 and 775 have an expected number of analyte particles within the target range. Tiles inside circle 774 may have an expected number of analyte particles greater than the desired number, which can result in interference between high-magnification holograms of nearby analyte particles. Tiles outside of circle 775 may receive fewer analyte particles than desired. The analyte particles in such tiles can nevertheless be characterized, but more analyte particles may better utilize the available substrate area. Another dashed line 776 then indicates the extent of the deposition pattern. In some examples, circle 776 may correspond to a deposition density of 0.5 analyte particles per tile; in other examples, the delivery subsystem (146) may include a clipping aperture, and circle 776 may represent the sharp edge of the current deposition pattern.

[0099] Although the contours of deposition patterns 774-776 are shown as concentric circles, this is not a requirement. By controlling the beam shape and tilt of the deposition axis, other deposition patterns can be achieved. Patterns 771-773 have roughly circular contours that are not concentric, while patterns 778-779 have elliptical contours that are concentric. Notably, the entire interior of ellipse 778 is within the target range of deposited analyte. There is no inner contour similar to 774 that corresponds to the upper limit of the desired deposition density.

[0100] Additionally, as illustrated by patterns 771-773, 774-776, 778-779, deposition patterns can be suitably positioned across, or in some cases away from, substrate 702 to optimize utilization of substrate 702, according to various criteria described herein, over one or more iterations of a workflow similar to that of FIG. 2.

[0101] Each illustrated deposition pattern can reflect the current deposition shape of the analyte beam (148) as well as the position drift between the source (140) and the substrate (110) over time.

[0102] First Exemplary Method—Identifying a Suitable Region 8 is a flowchart 800 of a first exemplary method for identifying regions suitable for a subsequent analytical procedure. In this method, an imaging scan is performed in two phases to efficiently identify regions suitable for the analytical procedure across the entire substrate area. This method is described with reference to FIG. 7. This method may be used to implement block 210 of FIG. 2.

[0103] Process block 810 marks the beginning of a first iterative loop over the first regions {Rj}. In block 810, an initial first region (e.g., tile 711) may be selected in the first iteration, or a next first region (e.g., tiles 712, 713, ... 743, 744) may be selected in a subsequent iteration. In block 812, a beam (e.g., a low-energy electron beam similar to 122 in FIG. 1) may be directed to the current region Rj, and in block 814, a first image Ij of region Rj may be acquired.

[0104] At decision block 816, based on image Ij, a determination can be made whether region Rj satisfies criterion C1 for performing subsequent analysis procedures. Criterion C1 can be based on visible defects identified in image Ij. Criterion C1 can be based on the type or number of defects. Illustratively, a single large contaminant or substrate strand may be sufficient for a region to fail criterion C1, while no more than a threshold number of small contaminants may be acceptable, and any number of small intrinsic defects (e.g., bubbles) may be acceptable.

[0105] If region Rj is unacceptable, the method can proceed via the N branch directly to decision block 818, where a check is made to see if additional regions {Rj} remain to be scanned. If all regions {Rj} have been scanned, the method can proceed via the Y branch to end block 890, where the method can end. Otherwise, the method can return via the N branch from block 818 to block 810 to repeat the first loop for the next region Rj.

[0106] However, if region Rj is acceptable, the method may follow the Y branch to block 820, which is the beginning of a second iterative loop over second regions {Sk}. In block 820, the first second region Sk (e.g., tiles 751, 761) may be selected in the first iteration, or the next second region (e.g., tiles 752...758 or tiles 762...768) may be selected in a subsequent iteration. In block 822, a beam (e.g., 122) may be directed to the current region Sk, and in block 824, a second image Hk of region Sk may be acquired.

[0107] At decision block 826, a determination can be made based on image Hk whether region Sk satisfies criterion C2 for performing subsequent analysis procedures. Criterion C2 can often be the same as criterion C1, but this is not a requirement.

[0108] If region Sk is unacceptable, the method can proceed directly via the N branch to decision block 828, where a check is made to see if additional regions {Sk} remain to be scanned for the current region Rj. If all regions {Sk} (e.g., tiles 751-758) have been scanned, the method can proceed via the Y branch to end block 832. Otherwise, the method can return via the N branch from block 828 to block 820 to repeat the second loop for the next region Sk.

[0109] If region Sk is acceptable, the method can follow the Y branch from block 826 to block 830, where Sk can be added to the set of preferred second regions {S+}. The method then continues to decision block 828, described above.

[0110] In block 830, a notification of suitable regions can be output. This can include the current region Rj and all regions {S+} that meet the criteria C2. In some cases, there may be no regions {S+} that meet the criteria C2. The notification can inform a controller or other recipient that the noted region is suitable for subsequent analysis procedures.

[0111] Numerous extensions or variations of the first method can be implemented within the scope of the disclosed technology. The substrate can be graphene, e.g., a single layer. The beam can be an electron beam, such as low-energy electron beam 122. In some examples, directing the beam in block 812 or block 822 can be implemented using one or more electromagnetic deflection elements. In other examples, the substrate can be mounted on a stage (135), the beam can be generated by a source (120), and directing the beam in block 812 or block 822 can be implemented by relative translation between the beam source and the stage.

[0112] In some examples, criterion C1 may be that the severity of defects visible in each first image Ij does not exceed a first predetermined threshold. Similarly, criterion C2 may be that the severity of defects visible in each second image Hk does not exceed a first predetermined threshold. Various techniques may be used to make the determination at block 816 or 826. In some examples, each image Ij or Hk may be displayed to a user, and an input may be received from the user indicating whether a region Rj or Sk satisfies criterion C1 or C2, respectively. In other examples, a signal may be calculated from image Ij or Sk indicating the severity of defects in that image. The calculated signal may be compared to a predetermined threshold to determine whether criterion C1 or C2 is met. Illustratively, the signal may be calculated using machine vision heuristics based on pixel intensity changes, edge detection, or shape detection. In a further example, each image Ij or Hk may be input to a trained machine learning classifier, and a corresponding output may be received. The output may indicate whether criterion C1 or C2 is met.

[0113] In a further example, the method can be extended to performing a subsequent analysis procedure on the first region Rj (meeting criterion C1). The subsequent analysis procedure can include deposition of analyte species (in the form of analyte particles or molecules) onto the substrate over a time interval by a deposition beam having a beam axis. The beam axis and time interval can be selected so that the preferred first region Rj falls within a predetermined zone of the deposition pattern (a zone similar to the inside of the annulus or ellipse 778 between 771 and 772 in FIG. 7). That is, the deposition pattern can be planned (similar to block 220) so that the region Rj is expected to meet a target range of analyte particles. With the beam axis and time interval set, analyte deposition can be performed (230).

[0114] In an additional example, following analyte deposition 230, the beam 122 can be directed once again to region Rj, and a third image can be acquired. If the third image indicates that region Rj still meets criterion C1, region Rj can be reused in a subsequent iteration of the workflow of FIG. 2. If the third image indicates that region Rj has at least one analyte particle, but not beyond a predetermined limit, those analyte particles can be characterized. In some examples, the third image can be a high-magnification hologram (LEEH), and holographic reconstruction can be performed to derive a fourth image from the third image, and the reconstructed fourth image can be used to characterize the analyte.

[0115] In various examples, the criteria C1 and C2 can be applied per tile or aperture. In some examples with multiple apertures per tile image, the defect severity criteria can be applied separately to each aperture in the tile image, and a decision regarding tile suitability can be made based on an evaluation of the criteria for the various apertures. In some examples, a tile can be considered suitable if all apertures in the tile image meet the defect severity criteria, and in other examples, a tile can be considered suitable if at least a predetermined number of apertures in the tile meet the defect severity criteria.

[0116] In another example, the set of regions {Sk} scanned in the second loop (blocks 820-828) may also be sparse: the second loop may be run multiple times on a single region Rj using different regions {Sk} to progressively identify additional suitable regions.

[0117] In a further example, the controller (170) can be configured to perform the first method or any of its variations or extensions, and the disclosed techniques can be embodied in an apparatus including a controller, a beam source, and a detector. The beam source can be controlled by the controller in blocks 812 and 822 and configured to direct the beam to a selected portion of the substrate. The detector can be configured to image the selected portion of the substrate illuminated by the beam. The detector can incorporate a pixelated electronic detector.

[0118] Second Exemplary Method—Analyte Deposition 9 is a flowchart 900 of a second exemplary method for analyte deposition, in which a deposition pattern is arranged based on the severity of defects in various regions of a substrate, and an analyte is deposited according to the deposition pattern. This method can be used to implement blocks 220 and 230 of FIG. 2.

[0119] Substrate regions that meet defect severity criteria may be identified in process block 910. In various examples, as discussed herein, the defect severity criteria may require that a region be free of all visible defects or may tolerate a particular type or number of visible defects.

[0120] As described herein, a deposition pattern can have preferred zones where the expected number of deposited analyte particles is within a predetermined target range, while areas outside the preferred zones are expected to have too many or too few deposited particles. In block 920, the location of the preferred zones of the deposition pattern can be determined such that at least one of the regions identified in block 910 is within the preferred zone. A region can be considered to be within the preferred zone if the center of gravity of the region is within the preferred zone.

[0121] In block 930, the source and substrate combination can be configured based on the geometry determined in block 920. In various examples, this can include configuring only the source, only the substrate, or both the source and the substrate. For example, the position or orientation of the source or the substrate can be adjusted. Additionally, source parameters can be configured, including the duration, intensity, size, or shape of the deposition beam.

[0122] The supply source can then be activated in block 940 to deposit the analyte onto the substrate according to the deposition pattern determined in block 920 .

[0123] Numerous extensions or variations of the first method can be implemented within the scope of the disclosed technology. The analyte can include a protein. Identifying can include performing a sparse scan of a substrate area similar to 711-714, 721-724, 731-734, and 741-744 in FIG. 7. Illustratively, 4 μm×4 μm tiles can be imaged on a 25 μm×25 μm grid on a 500 μm×500 μm substrate, thereby imaging 400 tiles out of 15,625 tiles.

[0124] The configuration in block 930 can include adjusting the distance (149) between the source and the substrate, adjusting the analyte release rate of the source (140, 142), adjusting the duration of analyte deposition, or adjusting the angle (143) between the deposition axis (145) and the surface normal (115) of the substrate. The target range defining the preferred zone can be between a first threshold and a second threshold. Exemplary target ranges can include 1 to 1, 1 to 2, 1 to 10, 2 to 5, 2 to 10, 3 to 10, 3 to 20, 3 to 30, 5 to 10, 5 to 20, 5 to 30, 10 to 20, 10 to 30, 10 to 50, or 20 to 50. The deposition axis (145) can target the portion of the substrate (110) having the highest density of preferred regions or the centroid of the preferred regions.

[0125] In an example, the determination in block 920 may optimize the number of identified regions that fall within the preferred zone (e.g., satisfy defect severity criteria), e.g., maximize this number. To illustrate, a pattern configuration with 12 identified regions within the preferred zone may be selected in preference to another pattern configuration with only two identified regions within the preferred zone. Block 920 may consider various factors. First, several identified regions within the preferred zone may be used. Having more identified regions within the preferred zone may result in more deposited analyte particles suitable for characterization downstream, e.g., in block 250. Second, the analyte supply may be limited, and a deposition pattern that consumes less analyte supply may be preferable to another deposition pattern that consumes more analyte. Thus, a second factor that may be considered is the total amount of analyte consumed in the supply source (142) during analyte deposition.

[0126] Analyte deposition may then leave some identified regions without analyte particles, for example, if such regions are outside the deposition pattern, and these regions may be available for reuse, for example, in another iteration of the workflow of FIG. 2. Thus, a third factor that may be considered in block 920 is the expected second number of identified regions that will still meet the defect severity criteria after analyte deposition. To illustrate with reference to FIG. 7, consider a scenario with regions 714, 721, 723, and 732 identified as defect-free. Patterns 774-776 then include two regions 714 and 723 within the preferred zone, while regions 721 and 732 are outside pattern perimeter 776 and therefore may be available for use in a subsequent workflow iteration. Alternative patterns 771-773 also include two regions 723 and 732 within the preferred zone, but only one identified region 714 is outside perimeter 773. Because region 721 is within perimeter 773, it may receive a suboptimal number of analyte particles in the current workflow iteration, but may also not be suitable for use in subsequent workflow iterations. Thus, although patterns 774-776 and 771-773 both provide the same number (2) of identified regions within the preferred zone, the former may be preferred due to the higher availability (2 vs. 1) of regions that are likely to be available for later use.

[0127] Optimization can be performed in various ways. Some parameters can be held fixed, while others can be varied. In some examples, for example, when there are few variable parameters, optimization can be performed by scanning parameter values ​​across multiple configurations and, for each configuration, predicting results for various factors, such as the count of deposited analyte particles in an identified region. One of the scanned configurations can then be selected as the optimal for the factor under consideration. Parameter values ​​can be monotonically scanned using a binary search, random selection, or a combination of these techniques. In other examples, a cost function can be defined based on the factor under consideration, and parameters can be varied to optimize the cost function. Gradient descent or steepest descent techniques can be used.

[0128] In a further example, the controller (170) can be configured to perform the second method or any of its variations or extensions, and the disclosed techniques can be embodied in a system including the controller and a source (140). The deposition source can include an electrospray ionizer (142), a mass spectrometer (144), and charged particle optics (146). The ionizer can be coupled to deliver the first ionized analyte molecules to an input of the mass spectrometer. The charged particle optics can be coupled to receive a portion of the first ionized analyte molecules, e.g., filtered by mass-to-charge ratio, and can be configured to deposit the second analyte molecules onto a substrate.

[0129] Third Exemplary Method - LEEPP Image Classification Figure 10 is a flowchart 1000 of a third exemplary method employing a trained machine learning (ML) tool to classify LEEPP substrate images. Figures 11A and 11B provide data flow diagrams 1101-1102 of respective phases of the method of Figure 10 and are used to facilitate discussion of Figure 10. Portions of this method can be used to implement blocks 210, 240 of Figure 2, 816, 826 of Figure 8, or block 910 of Figure 9.

[0130] 10, blocks 1010-1035 constitute the training phase of ML classifier 1130A, the purpose of which is to validate classifier 1130A for classification of LEEPP substrate images, e.g., with respect to defect severity. Classifier 1130A in FIG. 11A and classifier 1130B in FIG. 11B can represent the same classifier before and after validation, respectively.

[0131] At process block 1010, the labeled dataset 1112 can be partitioned into a training dataset and a test dataset. Each record 1122 in the dataset 1112 can include a LEEPP image 1124 and one or more classifications 1126 assigned by one or more users (e.g., human experts). The user-assigned classifications 1126 can indicate visible defects in the image 1124 selected from a plurality of predetermined classifications. In some examples, each image 1124 can span a single aperture (132), although this is not a requirement; in other examples, the classifier 1130A can be trained on tile images, each encompassing multiple apertures.

[0132] At block 1020, an ML tool 1130A may be trained using a training dataset. Illustratively, sequential images 1124 may be provided as input 1132 to a trainee tool 1130A, and a corresponding output classification 1134 may be received. The output 1134 may be compared to a corresponding user-assigned classification 1126 by a software module 1140 to obtain a loss function 1142, which may be fed back to the ML tool 1130A to update coefficients within the tool 1130A, for example, by backpropagation.

[0133] Then, in block 1030, a test dataset can be applied to tool 1130A to generate test classifications. Up to this point, training has been applied to tool 1130A, but tool 1130A has not been validated. Illustratively, test images 1124 can be provided as input 1132 to trained tool 1130A, which can receive corresponding test classifications 1134. In block 1135, software module 1140 (which can be the same or different from the software module used in block 1020) can compare the test classifications with user-assigned classifications 1126, and the resulting comparison 1144 can be used to validate trained ML tool 1130A.

[0134] Blocks 1050 and 1060 constitute the inference phase, in which the trained ML classifier 1130B can be used to characterize the LEEPP image of the substrate region. FIG. 11B illustrates the corresponding data flow. As indicated by block 1050, the common processing of blocks 1052, 1054, and 1056 can be applied to multiple images 1152 of each substrate region. At block 1052, the current image 1152 can be input to the trained ML classifier 1130B, and at block 1154, one or more corresponding image classifications can be received. Then, at block 1056, a determination can be made based on the one or more classifications as to whether each substrate region of the current image meets criteria for subsequent analysis procedures. For illustrative purposes, the description will consider the case where at least one of the imaged substrate regions meets the criteria. Then, at block 1060, a notification can be issued indicating that the given substrate region is suitable for subsequent analysis procedures.

[0135] Many extensions or variations of the first method can be implemented within the scope of the disclosed technology. The subsequent analysis procedure can include the group of operations 235 in Figure 2, or any one or more of operations 230, 240, and 250. The ML classifier can be implemented as a convolutional neural network.

[0136] In one example, the ML classifiers 1130A, 1130B were implemented as CNNs with 23 convolutional layers and 3 fully connected ("FC") layers, operating on 128x128 pixel images. A labeled dataset 1112 with approximately 7000 records 1122 was used for training. Other CNNs with 10-50 convolutional layers and 2-10 fully connected layers can also be used, operating on input image sizes from 32x32 to 1024x1024 pixels.

[0137] Corresponding to the labels 1126 applied to the training dataset, the classifications 1154 can be organized in a number of ways. In some examples, the classifications 1154 can be binary, indicating that a given image is either suitable or unsuitable for the analysis procedure. As described herein, various criteria can be used for suitability, such as the image having no visible defects, the image having at most defects of a particular benign class, the image having at most a limited number of defects of another class, or a combination thereof. In other examples, more granular classifications can be used, such as the image having no visible defects, the presence of at least one intrinsic defect, or the presence of at least one extrinsic defect. Additional or alternative classifications can be used to indicate the presence of at least one analyte molecule or whether the number of analyte molecules is below, within, or above a target range. Alternatively, the classification can indicate the most severe type of defect present, for example, among the classes of defects described in the context of FIG. 4 . Additionally, a single image may receive multiple classifications, for example, separate classifications indicating the presence of both intrinsic and extrinsic defects, or a binary presence / absence classification for each of multiple defect types.

[0138] In some examples, LEEPP images can be acquired in tiles spanning multiple apertures (similar to 132), while classification can be performed on smaller regions containing a single aperture. Thus, prior to block 1052, the acquired image can be divided into individual images to be processed by block 1050. Machine vision techniques can be used to identify the circular regions of the individual apertures from the surrounding low intensity regions corresponding to the opaque support, as seen in Figures 3 and 4.

[0139] FIG. 12 is a flowchart 1200 of another exemplary extension of the method of FIG. 10 . In this extension, a classifier is used to map analyte deposition, for example, in block 240 of the workflow of FIG. 2 . In block 1270 (similar to block 230), analyte deposition can be performed as all or part of a subsequent analysis procedure over at least the given region determined to be suitable for the analysis procedure in block 1056. Thereafter, in block 1280, a new LEEPP image of the given region can be classified by the trained ML classifier 1130B, and in block 1282, it can be determined that the given region contains several analyte molecules within a target range. In some examples, the classification received in block 1280 can directly indicate that the number of analyte molecules is within the target range; in other examples, the classification in block 1280 can indicate the presence of analyte molecules, which can be counted in block 1282 using, for example, machine vision techniques for shape detection. A notification can then be output indicating that the given region is suitable for characterization of one or more analyte molecules at block 1290. As described herein, this characterization can be performed by LEEH, although this is not a requirement and other analyses can also be performed.

[0140] Additional Examples Additional examples of the disclosed technology are described below.

[0141] Example 1 is a computer-implemented method including: sequentially directing a beam to a plurality of first regions of a substrate; acquiring a respective first image for each of the first regions; determining, based on the respective first images, whether the first regions satisfy a first criterion for a subsequent analysis procedure, where at least one given one of the first regions satisfies a first criterion; sequentially directing, for a given first region, a plurality of second regions adjacent to the given first region; acquiring a respective second image for each of the second regions; determining, based on the respective second images, whether the second region satisfies a second criterion for a subsequent analysis procedure; and outputting a notification that the given first region and all of the second regions that satisfy the first criterion are suitable for the subsequent analysis procedure.

[0142] Example 2 includes the subject matter of example 1, and further specifies that one or more of the directing actions include steering the beam with one or more electromagnetic deflection elements.

[0143] Example 3 includes the subject matter of any of Examples 1 and 2, and further specifies that the beam is generated by a beam source, the substrate is mounted on a stage, and one or more of the directing acts includes performing a relative translation of the beam source and the stage.

[0144] Example 4 includes the subject matter of any of Examples 1-3, and further specifies that the first criterion is that the defects visible in each first image have a severity that does not exceed a first predetermined threshold.

[0145] Example 5 includes the subject matter of any of Examples 1-4, further specifying that the second criterion is that the visible defects in each second image have a severity that does not exceed a second predetermined threshold.

[0146] Example 6 includes the subject matter of any of Examples 1-5, and further specifies that determining whether the first region satisfies the first criterion includes: (a) displaying each first image to a user and receiving input from the user indicating whether the first region satisfies the first criterion; (b) calculating a signal indicative of the severity of the defect in each first image and comparing the calculated signal to a predetermined threshold, where a result of the comparing indicates whether the first region satisfies the first criterion; or (c) providing each first image as an input to a trained classifier and receiving a corresponding output from the trained classifier indicating whether the first region satisfies the first criterion.

[0147] Example 7 includes the subject matter of any of Examples 1-6, further including performing a subsequent analytical procedure on the first region.

[0148] Example 8 includes the subject matter of Example 7, and further specifies that the subsequent analysis procedure includes deposition of an analyte species over a time interval by a deposition beam having a beam axis, and performing further includes selecting the beam axis and the time interval such that the given first region is within a predetermined zone of the deposition pattern, and activating the deposition beam to deposit the analyte species on the substrate.

[0149] Example 9 includes the subject matter of any of Examples 7 and 8, and further specifies that the subsequent analytical procedure includes deposition of an analyte species, and the method includes, subsequent to performing: directing a beam to a given first region; acquiring a third image of the first region; and determining whether the third image satisfies a first criterion; determining whether the third image indicates the presence of a number of analyte particles in a range from 1 to a predetermined limit; or outputting a fourth image derived from the third image, used to characterize the analyte species.

[0150] Example 10 includes the subject matter of any of Examples 1-9, and further specifies that the substrate is graphene.

[0151] Example 11 is one or more computer-readable media that store instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform any of the methods of Examples 1 to 10.

[0152] Example 12 is an apparatus including a controller, the controller having a memory coupled to the controller and configured to perform the method of any of Examples 1 to 10; a beam source configured to direct a beam to a selected portion of a substrate; and a detector configured to image the selected portion illuminated by the beam.

[0153] Example 13 includes the subject matter of Example 12, and further specifies that the beam source is an electron source and the beam is an unfocused electron beam having an energy between 50 eV and 600 eV.

[0154] Example 14 includes the subject matter of any of Examples 12 and 13, and further specifies that the detector is a pixelated electronic detector.

[0155] Example 15 is a computer-implemented method for controlling analyte deposition on a substrate by a source having a deposition axis, the method including: identifying a plurality of regions of the substrate that meet defect severity criteria; determining a preferred zone arrangement such that at least one of the identified regions is within a preferred zone of a deposition pattern on the substrate; configuring a combination of the source and the substrate according to the determined arrangement; and activating the source to deposit the analyte on the substrate according to the deposition pattern.

[0156] Example 16 includes the subject matter of Example 15, and further specifies that the analyte includes a protein.

[0157] Example 17 includes the subject matter of any of examples 15 and 16, and further specifies that identifying includes performing a sparse scan on the substrate.

[0158] Example 18 includes the subject matter of any of Examples 15-17, and further specifies that configuring includes relative translation between the source and the stage supporting the substrate, or relative rotation between the deposition axis and the stage supporting the substrate.

[0159] Example 19 includes the subject matter of any of Examples 15-18, and further specifies that configuring includes controlling the deposition pattern by adjusting one or more parameters including a distance between the source and the substrate, an analyte release rate of the source, a duration of analyte deposition, or an angle between the deposition axis and a surface normal of the substrate.

[0160] Example 20 includes the subject matter of any of Examples 15-19, further specifying that regions within the preferred zone have a first probability of receiving a number of analyte molecules within a predetermined range during analyte deposition, the first probability being greater than or equal to a first threshold and less than or equal to a second threshold, and regions outside the preferred zone have respective probabilities of receiving a respective number of analyte molecules within a predetermined range during analyte deposition, the respective probabilities being less than or equal to the first threshold or greater than or equal to the second threshold.

[0161] Example 21 includes the subject matter of any of Examples 15-20, and further specifies that the determining optimizes the number of identified regions that are within the preferred zone.

[0162] Example 22 includes the subject matter of any of Examples 15-21, further specifying that the factors used in determining include a first number of identified regions within the preferred zone, an expected value of a second number of identified regions that meet the defect severity criteria after analyte deposition, or a total amount of analyte consumed at the source during analyte deposition.

[0163] Example 23 is a system including a controller, the controller having a memory coupled to the controller and configured to perform the method of any of Examples 15 to 22, and a source.

[0164] Example 24 includes the subject matter of Example 23, further specifying that the source includes a mass spectrometer; an ionization source coupled to deliver analytes in charged particles to an input of the mass spectrometer; and charged particle optics coupled to receive ionized analyte molecules having a selected mass-to-charge ratio from an output of the mass spectrometer and configured to deposit second analyte molecules on the substrate.

[0165] Example 25 includes, during a training phase, partitioning a labeled dataset including a plurality of records, each record including a first Low Energy Electron Point Projection (LEEPP) image of a respective substrate region and one or more respective user-assigned classifications, the user-assigned classifications being selected from a plurality of predetermined classifications, the one or more respective user-assigned classifications indicating defects visible in the respective first LEEPP image, into a training dataset and a test dataset; training a machine learning tool using the training dataset; applying the test dataset to the trained machine learning tool to generate test classifications; and labeling the test classifications with the respective user-assigned classifications. and validating the trained machine learning tool by comparing the second LEEPP image with a user-assigned classification; and, during characterization of the given substrate, for each of a plurality of second LEEPP images of respective regions of the given substrate, providing the respective second LEEPP image to the trained machine learning tool; receiving one or more of the predetermined classifications from the trained machine learning tool; determining whether each region of the given substrate, where at least a given one of the regions of the given substrate satisfies a criterion for a subsequent analysis procedure, based on the received classification; and outputting a notification that the given region is suitable for the subsequent analysis procedure.

[0166] Example 26 includes the subject matter of Example 25, and further specifies that the machine learning tool is a convolutional neural network (CNN), a vision transformer neural network (ViT), or other neural network.

[0167] Example 27 includes the subject matter of any of Examples 25 and 26, and further specifies that the predetermined classification is binary, with the two classifications indicating that the corresponding substrate region is suitable or unsuitable for subsequent analytical procedures, respectively.

[0168] Example 28 includes the subject matter of any of Examples 25-27, further specifying that the predetermined classifications include at least a first classification indicating an absence of defects in the corresponding substrate region, a second classification indicating the presence of at least one intrinsic defect, and a third classification indicating the presence of at least one extrinsic defect.

[0169] Example 29 includes the subject matter of any of Examples 25-28, further specifying that the predetermined classifications include at least a first classification indicating an absence of defects in the corresponding substrate region, and a plurality of additional classifications indicating the presence of respective classes of intrinsic defects or the presence of respective classes of extrinsic defects.

[0170] Example 30 includes the subject matter of any of Examples 25-29, further specifying that the subsequent analysis procedure includes depositing analyte molecules, and the predetermined classifications include at least a first classification indicative of an absence of defects in the corresponding substrate region, a second classification indicative of the presence of at least one defect, and a third classification indicative of the presence of at least one of the analyte molecules.

[0171] Example 31 includes the subject matter of any of Examples 25-30, further including, prior to providing, dividing the acquired LEEPP images of the given substrate, each spanning a plurality of respective regions, into second LEEPP images, each spanning one region.

[0172] Example 32 includes the subject matter of any of Examples 25-31, and further specifies that the subsequent analysis procedure includes deposition of analyte molecules, and the method further includes, following deposition, during the analyte detection phase, classifying a new region image of the given LEEPP with the trained machine learning tool to obtain a new classification; determining, based on the new classification, that the given region contains a number of analyte molecules within a target range; and outputting a notification that the given region is suitable for characterization of one or more analyte molecules.

[0173] Example 33 is one or more computer-readable media that store instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform any of the methods of Examples 25 to 32.

[0174] Generalized Computer Environment 13 illustrates a generalized example of a suitable computing system 1300 on which the described examples, techniques, and technologies for substrate classification and analysis procedures can be implemented. The computing system 1300 is not intended to suggest any limitations on the scope of use or functionality of the present disclosure, as the innovations can be implemented in a variety of general-purpose or special-purpose computing systems. The computing system 1300 can control or monitor electronic point projection imagers, stages, analyte sources, analyzers, or ancillary equipment, or can acquire, process, output, or store measurement, classification, and characterization data.

[0175] Referring to FIG. 13, a computing environment 1310 includes one or more processing units 1322 and memory 1324. In FIG. 13, this basic configuration 1320 is included within the dashed line. The processing unit 1322 can execute computer-executable instructions, such as for control or data acquisition, as described herein. The processing unit 1322 may be a general-purpose central processing unit (CPU), a processor in an application-specific integrated circuit (ASIC), or any other type of processor. In a multiprocessing system, multiple processing units execute computer-executable instructions to increase processing power. The computing environment 1310 may also include a graphics processing unit or co-processing unit 1330. The tangible memory 1324 may be volatile memory (e.g., registers, cache, or RAM), non-volatile memory (e.g., ROM, EEPROM, or flash memory), or some combination thereof, accessible by the processing unit 1322, the processing unit 1330. Memory 1324 stores software 1380, in the form of computer-executable instructions suitable for execution by processing unit 1322, processing unit 1330, that implements one or more of the innovations described herein. For example, software 1380 can include software 1386 for controlling an imager, software 1381 for stage control, software 1382 for scanning logic, software 1383 for performing classification, software 1385 for deposition pattern planning, software 1384 for controlling an analyte source, or other software 1387 (including a user interface, a host interface, or hologram reconstruction). The inset shown for software 1380 in storage 1340 is equally applicable to software 1380 elsewhere in FIG. 13 . Memory 1324 can also store control parameters, calibration data, measurement data, spectroscopic libraries, training data, or other database data.The memory 1324 may also store configuration or operational data.

[0176] Computing system 1310 may have additional functionality, such as one or more of storage devices 1340, input devices 1350, output devices 1360, or communication ports 1370. An interconnection mechanism (not shown), such as a bus, controller, or network, interconnects the components of computing environment 1310. Typically, operating system software (not shown) provides an operating environment for other software executing in computing environment 1310 and coordinates the activities of the components of computing environment 1310.

[0177] Tangible storage 1340 may be removable or non-removable and includes magnetic disks, magnetic tapes or cassettes, CD-ROMs, DVDs, or any other medium that can be used to store information in a non-transitory manner and that is accessible within computing environment 1310. Storage 1340 stores instructions (including instructions and / or data) for software 1380 that implements one or more of the innovations described herein. Storage 1340 may also store image data (including LEEPP or LEEH images), measurement data, workflow programs, reference data, calibration data, configuration data, sample data, or other databases or other data structures described herein.

[0178] The input devices 1350 may be mechanical, touch-sensitive, or proximity-sensitive input devices such as a keyboard, mouse, pen, touchscreen, or trackball, a voice input device, a scanning device, or another device that provides input to the computing environment 1310. The output devices 1360 may be a display, a printer, speakers, an optical disc burner, or another device that provides output from the computing environment 1310. The input or output may also be communicated to or from remote devices via a communication port 1370 over a network connection.

[0179] Communications port 1370 enables communication over a communications medium to another computing entity. The communications medium conveys information such as computer-executable instructions, audio or video input or output, or other data in a modulated data signal. A modulated data signal is a signal that has one or more characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, the communications medium may use an electrical, optical, RF, acoustic, or other carrier wave.

[0180] A data acquisition system can be integrated into the computing environment 1310 either as an input device 1350 or coupled to a communication port 1370 and can include analog-to-digital converters or connections to the instrumentation bus. An instrumentation control system can be integrated into the computing environment 1310 either as an output device 1360 or coupled to a communication port 1370 and can include digital-to-analog converters, switches, or connections to the instrumentation bus.

[0181] In some embodiments, computer system 1300 may also include a computing cloud 1390 in which instructions for performing all or a portion of the disclosed techniques are implemented. Any combination of memory 1324, storage 1340, and computing cloud 1390 may be used to store software instructions and data for the disclosed techniques.

[0182] The innovation may be described in the general context of computer-executable instructions, such as instructions contained in program modules, being executed in a computing system on a target real or virtual processor. Generally, program modules or components include routines, programs, libraries, objects, classes, components, data structures, etc. that perform particular tasks or implement particular data types. The functionality of the program modules may be combined or split between program modules as desired in various embodiments. The computer-executable instructions of the program modules may be executed in a local computing system or in a distributed computing system.

[0183] The terms "computing system," "computing environment," and "computing device" are used interchangeably herein. Unless the context clearly dictates otherwise, no term is intended to limit the type of computing system, computing environment, or computing device. In general, a computing system, computing environment, or computing device may be local or distributed and may include any combination of dedicated hardware and / or general-purpose hardware and / or virtualized hardware, along with software that implements the functionality described herein.

[0184] General Considerations As used in this application and the claims, the singular forms "a," "an," and "the" include the plural forms unless the content clearly dictates otherwise. Additionally, the term "includes" means "comprises." Furthermore, the term "coupled" does not exclude the presence of intermediate elements between the coupled items. Furthermore, as used herein, the terms "or" and "and / or" refer to any one item or combination of items in a phrase. Terms connected by "or" or "and / or" are not necessarily mutually exclusive.

[0185] The systems, methods, and apparatus described herein should not be construed as limiting in any way. Instead, the present disclosure is directed to all novel and non-obvious features and aspects of the various disclosed embodiments, alone and in various combinations and subcombinations with one another. The disclosed systems, methods, and apparatus are not limited to any specific aspects or features or combinations thereof, and the disclosed systems, methods, and apparatus do not require that any one or more specific advantages be present or problems be solved. Techniques from any example can be combined with techniques described in any one or more of the other examples. Any theory of operation is for ease of explanation, but the disclosed systems, methods, and apparatus are not limited to such theories of operation.

[0186] Although some operations of the disclosed methods are described in a particular sequential order for convenient presentation, it should be understood that the format of the specification encompasses reordering unless a specific ordering is required by specific language set forth below. For example, operations described sequentially may, in some cases, be reordered or performed simultaneously. Moreover, for the sake of simplicity, the accompanying drawings may not show the various ways in which the disclosed systems, methods, and apparatuses may be used in conjunction with other systems, methods, and apparatuses. Furthermore, the description may use terms such as “acquire,” “activate,” “analyze,” “apply,” “calculate,” “characterize,” “compare,” “configure,” “feed,” “deposit,” “determine,” “instruct,” “display,” “establish,” “generate,” “identify,” “show,” “obtain,” “output,” “segment,” “plan,” “produce,” “provide,” “receive,” “repeat,” “rotate,” “scan,” “select,” “steer,” “train,” “convert,” “transmit,” “use,” or “verify” to describe the disclosed technology. These terms are high-level abstractions of the actual operations that take place. The actual operations corresponding to these terms will vary depending on the particular implementation and will be readily discernible to one of ordinary skill in the art having this disclosure in hand.

[0187] In some examples, values, procedures, or devices may be referred to as "minimum," "best," "maximum," "optimum," "extreme," etc. It will be understood that such descriptions are intended to indicate that a choice from few or many alternatives is possible, and that such a choice is not necessarily preferred over lower, better, lesser, or other options not otherwise considered.

[0188] Any theories of operation, scientific principles, or other theoretical explanations presented herein with reference to the devices or methods of the present disclosure are offered for the purpose of better understanding and are not intended to limit the scope, and the devices and methods in the appended claims are not limited to devices and methods that function in a manner described by such theories of operation.

[0189] Any of the disclosed methods may be controlled by or implemented as computer-executable instructions or computer program products stored on one or more computer-readable storage media, such as tangible, non-transitory computer-readable storage media, and executed on a computing device (e.g., any available computing device, including a tablet, smartphone, or other mobile device containing computing hardware). A tangible computer-readable storage medium is an available, tangible medium that can be accessed within a computing environment (e.g., one or more optical media disks, such as a DVD or CD; a volatile memory component (e.g., DRAM or SRAM); or a non-volatile memory component (e.g., flash memory or a hard drive)). As an example, referring to FIG. 13 , a computer-readable storage medium includes memory 1324 and storage device 1340. The terms computer-readable medium or computer-readable storage medium do not include signals and carrier waves. Furthermore, the terms computer-readable medium or computer-readable storage medium do not include communication ports (e.g., 1370).

[0190] Any computer-executable instructions for carrying out the disclosed techniques, as well as any data created and used during the practice of the disclosed embodiments, can be stored on one or more computer-readable storage media. The computer-executable instructions can be, for example, a dedicated software application or part of a software application accessed or downloaded via a web browser or other software application (such as a remote computing application). Such software can be executed, for example, on a single local computer (e.g., any suitable commercially available computer) or in a networked environment using one or more networked computers (e.g., via the Internet, a wide area network, a local area network, a client-server network, a cloud computing network, or other such network).

[0191] For clarity, only certain selected aspects of the software-based implementation are described. Other details well known in the art are omitted. For example, it should be understood that the disclosed technology is not limited to a particular computer language or program. For example, the disclosed technology may be implemented by software written in Adobe Flash, C, C++, C#, Curl, Dart, Fortran, Java, JavaScript, Julia, Lisp, Matlab, Octave, Perl, Python, Qt, R, Ruby, Rust, SAS, SPSS, SQL, WebAssembly, any derivatives thereof, or any other suitable programming language, or in some embodiments, a markup language such as HTML or XML, or any combination of suitable languages, libraries, and packages. Similarly, the disclosed technology is not limited to any particular computer or type of hardware. Specific details of suitable computers and hardware are well known and need not be described in detail in this disclosure.

[0192] Additionally, any of the software-based embodiments (e.g., including computer-executable instructions for causing a computer to perform any of the disclosed methods) may be uploaded, downloaded, sideloaded, or remotely accessed through suitable communication means, including, for example, the Internet, the World Wide Web, an intranet, a software application, cable (including fiber optic cable), magnetic communication, electromagnetic communication (including RF communication, microwave communication, infrared communication, and optical communication), electronic communication, or other such communication means.

[0193] In view of the many possible embodiments to which the principles of the disclosed subject matter may be applied, it should be recognized that the illustrated embodiments are merely preferred examples of the disclosed subject matter and should not be construed as limiting the scope of the claims. Rather, the scope of the claimed subject matter is defined by the following claims. We therefore claim all that comes within the scope and spirit of those claims.

Claims

1. 1. A computer-implemented method comprising: sequentially directing a beam onto a plurality of first regions of a substrate; For each of the first regions: acquiring a respective first image; determining, based on the respective first images, whether the first regions satisfy first criteria for a subsequent analysis procedure; at least a given one of the first regions satisfies the first criterion; For said given first region: sequentially directing the beam to a plurality of second regions adjacent to the given first region; For each of the second regions: acquiring a respective second image; determining whether the second region satisfies second criteria for the subsequent analysis procedure based on the respective second images; and outputting a notification that the given first region and all of the second regions that meet the first criteria are suitable for the subsequent analysis procedure.

2. The computer-implemented method of claim 1 , wherein the first criterion is that defects visible in the respective first images have a severity that does not exceed a first predetermined threshold.

3. Determining whether the first region satisfies the first criterion includes: (a) displaying each of the first images to a user and receiving input from the user indicating whether the first region meets the first criteria; (b) calculating a signal indicative of the severity of the defect in each of the first images and comparing the calculated signal with a predetermined threshold, the result of the comparing indicating whether the first region satisfies the first criterion; or 2. The computer-implemented method of claim 1, comprising: (c) providing each of the first images as input to a trained classifier and receiving a corresponding output from the trained classifier indicating whether the first region satisfies the first criterion.

4. The computer-implemented method of claim 1 , further comprising performing the subsequent analysis procedure on the given first region.

5. the subsequent analysis procedure includes deposition of analyte species over a time interval with a deposition beam having a beam axis, said performing comprising: selecting the beam axis and the time interval such that the given first region is within a predetermined zone of a deposition pattern; The computer-implemented method of claim 4 , further comprising activating the deposition beam to deposit the analyte species on the substrate.

6. the subsequent analytical procedure includes deposition of an analyte species, and the method, subsequent to said performing, comprises: directing the beam to the given first region; acquiring a third image of the given first region; determining whether the third image satisfies the first criterion; determining whether the third image indicates the presence of a number of analyte particles in the range of 1 to a predetermined limit; or 5. The computer-implemented method of claim 4, further comprising: outputting a fourth image derived from the third image used to characterize the analyte species.

7. The computer-implemented method of claim 1 , wherein the substrate is graphene.

8. 1. An apparatus comprising: a controller coupled to the memory and configured to perform the method of claim 1; a beam source configured to direct the beam onto a selected portion of the substrate; a detector configured to image the selected portion illuminated by the beam.

9. 9. The apparatus of claim 8, wherein the beam source is an electron source and the beam is an unfocused electron beam having an energy between 50 eV and 600 eV.

10. 1. A computer-implemented method for controlling analyte deposition on a substrate by a source having a deposition axis, comprising: identifying a plurality of regions of the substrate that meet defect severity criteria; determining a location of a preferred zone of a deposition pattern on the substrate such that at least one of the identified regions is within the preferred zone; configuring the source and substrate combination according to the determined placement; and activating the source to deposit an analyte onto the substrate according to the deposition pattern.

11. The computer-implemented method of claim 10 , wherein the analyte comprises a protein.

12. The computer-implemented method of claim 10 , wherein the identifying comprises performing a sparse scan on the substrate.

13. The configuring includes controlling the deposition pattern by adjusting one or more parameters, the one or more parameters being: the distance between the source and the substrate; the analyte release rate of the source; the duration of said analyte deposition; or The computer-implemented method of claim 10 , including an angle between the deposition axis and a surface normal of the substrate.

14. The computer-implemented method of claim 10 , wherein said determining optimizes the number of said identified regions that are within said preferred zone.

15. The factors used in the determination are: a first number of the identified regions that are within the preferred zone; an estimate of a second number of the identified regions that meet the defect severity criteria after analyte deposition; or The computer-implemented method of claim 10 , including a total amount of analyte consumed in the source during the analyte deposition.

16. 1. A system comprising: a controller coupled to the memory and configured to perform the method of claim 10; the source.

17. 1. A method comprising: During the training phase, Partitioning a labeled data set including a plurality of records, each record including a first Low Energy Electron Point Projection (LEPP) image of a respective substrate area and one or more respective user-assigned classifications, the user-assigned classifications selected from a plurality of predetermined classifications, the one or more respective user-assigned classifications indicative of defects visible in the respective first LEEPP image, into a training data set and a test data set; training a machine learning tool using the training dataset; applying the test dataset to the trained machine learning tool to generate a test classification; comparing the test classifications with the respective user-assigned classifications to validate the trained machine learning tool; During the characterization of a given substrate, For each of a plurality of second LEEPP images of a respective region of the given substrate: providing the trained machine learning tool with each second LEEPP image; receiving one or more of the predetermined classifications from the trained machine learning tool; determining whether the respective region of the given substrate meets criteria for a subsequent analysis procedure based on the received classification; at least a given one of said given substrates satisfies said criteria; and outputting a notification that the given region is suitable for the subsequent analysis procedure.

18. The method of claim 17 , wherein the machine learning tool is a neural network.

19. 18. The method of claim 17, wherein the predetermined classification is binary, each of the two classifications indicating whether the corresponding substrate region is suitable or unsuitable for the subsequent analytical procedure.

20. the subsequent analytical procedure comprises deposition of analyte molecules, the method comprising: Following said deposition, during an analyte detection phase: classifying new LEEPP images of the given area with the trained machine learning tool to obtain new classifications; determining that the given region contains a number of the analyte molecules within a target range based on the new classification; 20. The method of claim 17, further comprising: outputting a notification that the given region is suitable for characterizing one or more analyte molecules.

Images