Real-time, rapid, and non-invasive " atomic lock-on" in the scanning transmission electron microscope with picometer precision

Abstract

A procedure for using a scanning transmission electron microscope to determine atomic position is disclosed. A diagnostic scan pattern is executed and areas of high intensity in the resulting detector output are identified. These areas may represent the position of atoms in the sample. This output is then compared to the known or measured lattice structure of the sample, and the coordinates of the identified atoms are adjusted based on the comparison. This procedure avoids exposing a target region of interest, compensates for drift, and allows for precision to be improved compared to the case where the procedure is not used, reaching values that may be better than 10 picometers.

Description

[0001] Real-time , rapid, and non-invasive "atomic lock-on" in the scanning transmission electron microscope with picometer precision

[0002] This application claims priority of U . S . Provisional Patent Application Serial No . 63 / 601 , 529 , filed November 21 , 2023 , the disclosure of which is incorporated herein by reference in its entirety .

[0003] Field

[0004] This disclosure describes a method of positioning an electron beam with improved precision and accuracy.

[0005] Background

[0006] State-of-the-art aberration corrected scanning transmission electron microscopes provide unprecedented spatial resolution by scanning a narrow electron beam across the sample . The accuracy in placing the electron beam in real time is key for many applications , foremost imaging of atomic nanostructures , obtaining spectroscopic information or using the electron probe for guided material manipulation on the atomic scale . For all such applications and for many unrealized applications , both speed and atomic precision, or most ideally, a precision that is well below the inter-atomic bond length ( distance between two atoms ) , are crucial in placing the focused electron probe at the correct position with respect to the atomic lattice . However, this degree of control is currently absent in electron microscopy . In terms of precision, dri ft-corrected imaging techniques can measure physical properties , such as strain, down to the few picometer level . However, the precision of beam positioning relative to a target site in these techniques is not in the few- picometer level , because scan distortions and specimen dri ft prevent accurately positioning the beam . Instead, in these techniques , positioning the beam is usually done relative to a feature of interest in an image acquired immediately beforehand . The precision of beam placement relative to this image is strongly af fected by scan distortions , and with accumulated time, specimen dri ft as well . One of the most visually obvious and well-known scan distortions is the so-called flyback distortion, where the beam undergoes a rapid acceleration change after reaching the end of each horizontal scan line , which manifests as a strong distortion typically on the left-hand side of the image (where the scan begins ) . Even when using a scan that has a uniform acceleration (Archimedean spiral ) , other spatiotemporal distortions are present , which are not well understood . Regardless of scan choice , distortions will be present that prevent precise beam positioning . Alternatively, one may perform a small ( few angstrom to few nanometer ) " subscan" relative to an initially acquired image to manually correct for the distortions and dri ft to ensure proper positioning of the beam . However, this imparts a very large amount of electron dose to a small region, which usually results in an unintended or uncontrolled damage or change to the structure .

[0007] Most of the existing dri ft correction approaches use postprocessing of data . For this , special techniques for data collection are often required, which utili ze higher doses to the sample . This high dose limits their applicability to samples that are dose-insensitive . A well-known approach in the microscope community is RevSTEM, where a series of images is required to remove drift and scan distortions from already collected data . For real-time dri ft correction that is necessary while conducting an experimental procedure, no approach exists that can reach atomic precision . There is one software package from the company Protochips , called 'AXON' , which allows dri ft correction during an experiment . In this approach, images are taken and changes between images are computed using a standard cross correlation algorithm, and an appropriate beam shi ft is applied in time to compensate for the measured specimen drift . This approach is not applicable to procedures that require low dosages to minimi ze unwanted beam damage , fast execution to have small time-accumulated drift , or high spatial precision . Other common algorithms used to correct for dri ft are SIFT , RANSAC or ORB .

[0008] While modern microscope sample stages benefit from greatly improved stability, lateral drift rates on the order of nanometers per minute are very common, making experiments that require high spatial precision very complicated or often times rendering them impossible .

[0009] Therefore , there is a need for a procedure that overcomes these limitations and provides real-time , picometer level accuracy for electron beam placement that is fast with minimal dose .

[0010] Summary A procedure for using a scanning transmission electron microscope to determine atomic positions relative to the electron beam in a fast and minimally invasive manner is disclosed . A diagnostic scan pattern is executed and areas of high intensity in the resulting detector output are identi fied . These areas may represent the position of atoms in the sample or other periodic or even non-periodic structures , provided a reference is available . This output is then compared to the known or measured lattice structure of the sample , and the coordinates of the identified atoms are adj usted based on the comparison . This procedure avoids exposing a target region of interest , compensates for dri ft, and allows for precision to be improved compared to the case where the procedure is not used, reaching values that may be better than 10 picometers .

[0011] According to one embodiment , a method for using a scanning transmission electron microscope to determine atomic positions of a sample relative to an electron beam is disclosed . The method comprises performing a scan to obtain data, wherein the scan is performed using a diagnostic scan pattern; selecting points from the data based on intensity; obtaining an artificial lattice of the sample ; and mapping the points onto the arti ficial lattice by performing a mathematical optimi zation . In some embodiments , the artificial lattice is generated using known lattice constants of the sample . In some embodiments , the arti ficial lattice is generated by supplying the data to a neural network . In some embodiments , the mathematical optimi zation comprises a least squares fit or another similar algorithm . In some embodiments , the diagnostic scan pattern comprises one or more loops surrounding a target region . In certain embodiments , the diagnostic scan pattern comprises a plurality of loops . In certain embodiments , a group of closely spaced points is defined as a cluster and a center of mass is determined for the cluster and the group of closely spaced points is replaced by one point disposed at the center of mass . In certain embodiments , electrons are not irradiated into the target region during the scan . In certain embodiments , a task is performed in the target region after the mapping . In some embodiments , the diagnostic scan pattern surrounds at least a portion of a target region . In certain embodiments , the diagnostic scan pattern comprises a portion of a loop . In certain embodiments , electrons are not irradiated into the target region during the scan . In certain embodiments , a task is performed in the target region after the mapping .

[0012] According to another embodiment , a system for determining atomic positions of a sample is disclosed . The system comprises a scanning transmission electron microscope ( STEM) to create an electron beam; a detector; and a processing device , wherein the processing device is configured to : instruct the STEM to perform a scan using a diagnostic scan pattern, wherein the sample is disposed between the detector and the STEM, and the detector captures data as the electron beam passes through the sample ; select points from the data based on intensity; and map the points onto an artificial lattice by performing a mathematical optimi zation . In some embodiments , the detector comprises an annular dark field (ADE) detector, an annular bright field (ABF) detector, a bright field (BF) detector, a secondary electron ( SE ) detector, or a charge coupled device ( CCD) , complementary metal- oxide semiconductor (CMOS ) , or direct electron ( DE ) detector . In some embodiments , the artificial lattice is generated using known lattice constants of the sample. In some embodiments, the artificial lattice is generated by supplying the data to a neural network. In some embodiments, the diagnostic scan pattern comprises one or more loops surrounding a target region. In some embodiments, the diagnostic scan pattern does not include the target region.

[0013] Brief Description of the Drawings

[0014] For a better understanding of the present disclosure, reference is made to the accompanying drawings, in which like elements are referenced with like numerals, and in which:

[0015] FIG. 1 shows a flowchart of a procedure to identify atomic positioning according to one embodiment;

[0016] FIG. 2 shows the diagnostic scan pattern superimposed on an annular dark field (ADF) image of one exemplary sample;

[0017] FIG. 3 shows the resulting annular dark field (ADF) data for the diagnostic scan pattern in FIG. 2;

[0018] FIG. 4 shows a graph of the intensity as a function of angle for the ADF data shown in FIG. 3;

[0019] FIG. 5 shows the peaks found in FIG. 4 superimposed on the sample ;

[0020] FIG. 6 shows the actual lattice points as compared to the points generated by this procedure;

[0021] FIG. 7 shows the effects of drift and the correction performed by this procedure;

[0022] FIG. 8 illustrates the accuracy of the procedure of FIG. 1;

[0023] FIG. 9 shows a flowchart of a procedure to identify atomic positioning according to a second embodiment; FIG . 10A shows the sample ;

[0024] FIG . 10B shows the diagnostic scan pattern according to one embodiment ;

[0025] FIG . I OC shows the intensity peaks derived from FIG . 10B;

[0026] FIG . 10D shows the atoms placed in the diagnostic scan pattern;

[0027] FIGs . 11A-11B show the ef fect of the atomic lock-on process ;

[0028] FIG . 12A illustrates the experimentally obtained offset between the actual atom position and the experimentally obtained atom positions ; and

[0029] FIG . 12B illustrates a simulation of 1000 random targets .

[0030] Detailed Description

[0031] A procedure is disclosed that allows for ultra-precise "atomic lock-on" in real-time in a scanning transmission electron microscope ( STEM) . "Atomic lock-on" refers to the process of obtaining the atomic lattice information from a special electron beam scan pattern, and refining that information based on knowledge of the actual lattice structure . This process may be used for targeting speci fic atoms or bonds . The procedure is generally applicable in any scanning transmission electron microscope .

[0032] A scanning transmission electron microscope includes an electron source and a plurality of electromagnetic lenses , such as condenser lenses that serve to focus the electron beam on a sample , which may be disposed on a stage . These condenser lenses may be electromagnets . Prior to the stage may be an obj ective lens to focus the image . Additionally, after the stage is a detector that is used to form the image . The image can be obtained by an annular dark field (ADF) detector, but may also be obtained by any other detector such as an annular bright field (ABF) detector, a bright field (BF) detector, a secondary electron ( SE ) detector, or a charge coupled device (CCD) , complementary metal-oxide semiconductor ( CMOS ) , or direct electron ( DE ) detector that may capture di f fraction or energy loss images . Electrons that pass through and interact with the sample , referred to as transmitted electrons , are incident on the detector, and are used to create the image . The scanning transmission electron microscope also includes an electronic device ( referred to as a controller ) to control the beam as a function of time and space . The electron beam may be moved by the controller according to a predetermined scanning pattern so as to image the sample .

[0033] FIG . 1 shows the procedure according to a first embodiment . The procedure may be performed using any processing device , such as a personal computer, laptop computer or other suitable device . The processing device may be in communication with the STEM and may include the instructions to execute the procedures detailed herein . Note in some embodiments , the controller and the processing device may be the same component . The procedures described herein may be implemented using a software program, disposed on a computer readable non-transitory media, such as a memory or hard drive . Thus , in the disclosure , all references to the procedures are describing speci fic functions performed by the software program executed by the processing device .

[0034] As noted above, the processing device is in communication with the controller of the scanning transmission electron microscope ( STEM) so as to supply commands regarding position to the STEM and receive image data from the STEM . The processing device may also receive operator input .

[0035] First, the procedure is started, as shown in Box 1 . Next, as shown in Box 2 , a diagnostic scan pattern is established . This may be provided by an operator . In this disclosure, a "diagnostic scan pattern" refers to a scan pattern that surrounds all or at least a portion of a target region, which may be, for example, a few nanometers in both directions . For example , the diagnostic scan pattern may be one or more loops , where a loop is defined as a closed pattern having any suitable shape , such as circular, oval , elliptical , triangular, rectangular or any other shape , that surrounds the target region . In other embodiments , the diagnostic scan pattern is a portion of a loop, which surrounds only a portion of the target region, such as a semi-annular ring . In many embodiments , the diagnostic scan pattern is defined such that the target region is not irradiated by the electron beam. Rather, data from the diagnostic scan pattern is used to determine the position of atoms and atomic bonds within the target region . In other embodiments , the diagnostic scan pattern may include the target region, however, the dose of electrons in the target region may be intentionally kept to a minimum so as not to af fect the target region . In the present description, a circular scan pattern is assumed to illustrate the procedure . However, as noted above , a di f ferent shaped diagnostic scan pattern may also be used . Next , parameters associated with that diagnostic scan pattern are provided to the controller . These parameters may include the type of pattern, the inner and outer radii of the diagnostic scan pattern (which may be in the range of 0 . 1 nanometer to few nanometer ) , and the thickness of the diagnostic scan pattern which may be a number ( for example 1 pixel or few pixel whereas the pixel si ze in picometer depends on the outer radius and the pixel density of the scan pattern) .

[0036] FIG . 2 shows a representative diagnostic scan pattern 25 overlayed on a microscope image of the sample 20 . The sample 20 may be CrSBr or any material in which atomic contrast may be observed . The sample 20 , which in this example is CrSBr, has a crystalline structure that includes S+Br columns 21 that comprise alternating rows of Br and S , as well as Cr columns 22 , that are disposed between the S+Br columns 21 . Note that the S+Br columns 21 appear brighter than the Cr columns 22 . The sample may be tilted to show defined atomic columns in a certain proj ection . The sample 20 also exhibits an in-plane orientation that may be defined by the angle theta with respect to the fast scan axis of the electron probe . Here , in this example , the CrSBr is oriented with its c axis parallel to the electron beam for the image to be observed in this manner . This is also referred to as the

[0001] direction .

[0037] Next , as shown in Box 3 , the scan is performed and a detector signal is measured to collect the electrons and form the scan or image . In this case, annular dark field (ADF) data is collected but any detector may be used . FIG . 3 shows the ADF data 30 for this circular diagnostic scan pattern . Note that there are areas 31 that have a greater intensity than other areas .

[0038] In certain embodiments , as shown in Box 4 , the intensity of the ADF data 30 in two dimensions is mapped as a function of angle . In other words, a mathematical transformation is performed to map the two-dimensional intensity profile into a one-dimensional graph. In one embodiment, the center of the diagnostic scan pattern 25 (also referred to as the target region) is set as the center and the intensity associated with each angle is plotted. A representative graph is shown in FIG. 4. The original ADF data 30 is plotted as a function of angle. In some embodiments, a smoothing algorithm is applied to generate smoothed data 34. Next, a threshold level 32 is set. This threshold level may be predetermined or generated based on the data in the graph. Local maxima that are above the threshold level 32 are identified as peaks 33, as shown in Box 5. In some embodiments, the threshold level 32 is adjusted to achieve a predefined number of peaks, as shown in Box 6. Note that this step may be optional, as the threshold may also be used to distinguish between specific atomic column intensities.

[0039] Next, as shown in Box 7, these peaks 33 are then remapped back onto the two-dimensional image space, by mapping the angles of the peaks 33 back onto the two-dimensional space. This is illustrated in FIG. 5 where for illustration purposes a full image is presented with the detected positions of atoms shown as white dots. The distance from the center is determined based on the parameters of the diagnostic scan pattern. For example, for a circular scan pattern, all peaks are the same distance from the center. If the diagnostic scan pattern is elliptical, a transformation is performed to determine the distance from the center. Thus, each peak is given a set of coordinates. Next, as part of Box 8, the actual lattice structure is supplied to the processing device. This may be performed by an operator. In another embodiment, the image of the structure (see FIG. 2) may be provided as the input to a neural network that calculates the lattice structure as well as the positions of the atoms in that image. The procedure then maps the peaks 33 to the lattice structure using a mathematical optimization. In some embodiments, this may be a least sguares fit or another approach. The results of this operation may be seen in FIG. 6. The peaks 33 are mapped onto the actual lattice, which are indicated as the S+Br columns 60. Thus, the coordinates of the peaks 33 as determined above are translated so as to coincide with the known lattice structure, as shown in Box 9. In this way, the actual locations of the atomic lattice structure of the sample 20 can be determined .

[0040] FIG. 7 shows this result. Point 70 shows the actual position of the atomic column. Point 71 shows the location of the atomic column as originally measured by the STEM, that may be changed due to drift or scan-distortion. Point 72 shows the position of the atomic column after the correction that is applied, using the technique described above. The process of aligning the measured peaks 33 with the actual locations of the atomic columns may be referred to as "atomic lock-on".

[0041] Once the peak 33 has been correctly positioned, the operator may perform some experiment or measurement on a selected atom or atomic bond, as shown in Box 10. The operator may then determine whether experiments are to be performed at other positions, as shown in Box 11. If so, the processing device instructs the STEM to move to the new position (see Box 12) and the procedure is repeated. Otherwise, the procedure is finished (see Box 13) .

[0042] FIG. 8 shows the accuracy achieved by the approach. One thousand experiments were performed and a statistically simulated accuracy of 24.5 ± 8.9 picometers was obtained, well below typical interatomic bond distances. The precision that may be attained depends on the specific shape of the diagnostic scan pattern, the material under investigation and the signal to noise ratio of the ADF data. In some embodiments, a preferred diagnostic scan pattern may be an annulus with a given radius and thickness due to its radial symmetry and the sharp overlap with the Gaussian intensity distribution of an atomic column, allowing for high accuracy.

[0043] FIG. 9 shows a variation of this procedure. Some of the steps in this procedure are similar to those described with respect to FIG. 1. As shown in Boxes 92-93, a scan is initiated. As described above, parameters associated with the scan, such as pattern type, radius, thickness are input to the processing device. Additionally, in this embodiment, the diagnostic scan pattern may include a plurality of concentric or nearly concentric loops. As noted above, the loops may be any suitable shape. FIG. 10B shows a diagnostic scan pattern that is circular and includes 3 roughly concentric circles. Thus, in some embodiments, the number of loops and the spacing between these loops may also be parameters associated with the scan.

[0044] FIG. 10A shows the sample 20, as detected by the STEM. As described above, the S+Br columns 21 are brighter than the Cr columns 22. FIG. 10B shows a representative diagnostic scan pattern that may be used . In this pattern, there are three roughly concentric loops . However, there may be more or fewer loops . For example , there may be between 1 and 10 loops , where a higher number of loops increases electron dosage when similar detector noise is targeted . Additionally, the spacing between adj acent loops is also configurable , and may be between 0 and 500 picometers or more . Note that FIG . 10B represents the two-dimensional intensity profile .

[0045] Next , as shown in Box 94 , the information about the atomic positions of the sample 20 are provided . This may be done by manually entering the lattice parameters . Alternatively, as explained above, the image shown in FIG . 10A may be provided to a neural network that determines the positions and lattice structure of the atoms in the sample and furthermore provides a mean atomic column intensity, which may be used for thresholding in a later step . Next, as shown in Box 95, an arti ficial lattice is created that may be used as the reference lattice . I f the lattice parameters are provided, the arti ficial lattice is generated by the software program . I f a neural network is used, the program may select the closest atoms to the target, which may be an atom to create the artificial lattice .

[0046] As shown in Box 96 , an intensity threshold is selected and the threshold is applied to the profile shown in FIG . 10B . The threshold may be provided by the operator or may be supplied indirectly by the neural network . As shown in FIG . 10C, only those points 121 with an intensity greater than the threshold are selected . Note that there are several clusters of high intensity clusters , such as cluster 120 . These clusters may be the result of multiple loops passing through the same atom . The clusters may exist along the radial direction, along the circumferential direction or both . For example , cluster 120 includes points in both the radial direction and the circumferential direction, while cluster 122 only includes points in the circumferential direction . Thus , in certain embodiments , it is assumed that points 121 that are within a certain distance of one another are actually associated with the same atom . Thus , the software program may then compute the center of mass for this cluster and replace the cluster with a single atom 125 located at the center of mass . The result of this operation is shown in FIG . 10D . Thus , the atoms for which the loops pass over are provided with a set of coordinates , which may be polar coordinates or cartesian coordinates .

[0047] In some embodiments , it may be preferable to obtain at least some minimum number of atoms 125 to facilitate the subsequent optimi zation of the placement of the atom 125 on the arti ficial lattice . Thus , if less than this minimum number of atoms 125 are created, the software program may reduce the threshold and repeat the process shown in Box 96.

[0048] Next , in Box 98 , the actual lattice structure is obtained by using a mathematical optimi zation, using the atoms 125 in FIG . 10D and the arti ficial lattice . The mathematical optimi zation may be a least squares fit , or another similar algorithm . This results in a modi fication of the coordinates of each atom 125 to align with the arti ficial lattice . This process is referred to as "atomic lock-on" ( see Box 99 ) . FIG . 11A shows the results associated with Boxes 96- 99. The detected atomic positions 110 obtained in Box 96 are shown . Points 111 ( shown as dark dots ) show the artificial or theoretical lattice . Note that, in the leftmost graph, the detected atomic positions 110 ( shown as hollow dots ) have not been correlated with the arti ficial or theoretical lattice . Therefore , there is an of fset between these points . The middle graph shows the placement of the detected atomic positions 110 after the mathematical optimi zation has been performed . The rightmost graph shows the points in the middle graph superimposed on the ADF data . The positions after the mathematical optimi zation are in excellent agreement with the ground truth of the actual atomic positions of S+Br atomic columns as obtained from a neural network followed by a Gaussian refinement fitting .

[0049] FIG . 11B shows the benefits of this procedure . The hollow dots 115 represent the ideal or actual positions of the four nearest S+Br atomic columns . The cross-hatched dots 116 represent the experimentally obtained positions of these atomic columns . Note that without Boxes 98- 99, there is a significant di f ference between the experimentally obtained positions and the actual positions ( see bottom 5 graphs ) . In contrast, using "atomic lock- on" , the experimentally obtained positions correlate more closely with the actual positions ( see top 5 graphs ) .

[0050] Once "atomic lock-on" has been achieved, a task can then be performed on a bond or atom located in the target region, as shown in Box 100 . This task may be a spectroscopic or phase-sensitive measurement to locally probe electronic or magnetic properties or a controlled beam dwell pattern to manipulate the atomic structure of the targeted material. Repeated "atomic lock-on" can also be used to remain locked onto the lattice and to track and compensate sample drift in time. A user defined offset (Ax, Ay) can be used as input to perform targeting with respect to the atomic position, for example to target a nearby atom or an atomic bond. A decision can then be made as to whether to move to another position that may be an atom or atomic bond (see Box 102) .

[0051] FIGs. 12A-12B show the impact of the "atomic lock-on" process. In FIG. 12A, the hollow dots represent the offset in two dimensions of the experimentally obtained atom position as compared to the actual atom position without "atomic lock-on". The circular distribution of positions shows the smallest atomic bond distance of CrSBr. Note that the atom position of the experimentally obtained atoms is scattered over a distance that is a significant fraction of the bond distance. In contrast, the dark dots show the atom positions that were determined using the procedure shown in FIG. 9, where "atomic lock-on" is achieved. Note that the precision is about 16.5 picometers. In FIG. 12B, the accuracy achieved by this approach is shown. One thousand simulations were performed and a statistically simulated accuracy of 5.9 ± 2.9 picometers was obtained, well below typical interatomic bond distances.

[0052] While many of the figures show the sample 20 as being CrSBr, other materials may also be used. For example, tests using M0S2 and graphene have also been performed with similar results. Generally, this procedure is effective for any crystalline material, including mono-, di- and multi-atomic materials, as well as single layer materials and thicker materials. The procedure described herein has many advantages . The approach is ( 1 ) non-invasive , ( 2 ) fast, ( 3 ) integrable , and ( 4 ) highly precise and as such ( 5 ) allows automated beam experiments . First, the significantly reduced diagnostic scan pattern results in avoiding sample damage, particularly in the area of interest ( the target region) that is not exposed or minimally exposed during execution of the procedure . Second, the entire approach that includes the scan and mathematical optimi zation is extremely fast and executed in less than 150 milliseconds on a typical mid-tier CPU or GPU . Third, the procedure can be integrated in any commonly used experimental workflows , allowing one to automate beam experiments by removing the human operator, increasing repeatability, and reali zing new types of experiments that require atomic precision . Fourth, the approach is extremely precise , reaching a demonstrated precision below 20 picometers in initial test experiments , well below typical interatomic bond distances .

[0053] The present disclosure is not to be limited in scope by the specific embodiments described herein . Indeed, other various embodiments of and modi fications to the present disclosure , in addition to those described herein, will be apparent to those of ordinary skill in the art from the foregoing description and accompanying drawings . Thus , such other embodiments and modifications are intended to fall within the scope of the present disclosure . Further, although the present disclosure has been described herein in the context of a particular implementation in a particular environment for a particular purpose , those of ordinary skill in the art will recognize that its usefulness is not limited thereto and that the present disclosure may be beneficially implemented in any number of environments for any number of purposes . Accordingly, the claims set forth below should be construed in view of the full breadth and spirit of the present disclosure as described herein .

Claims

What is claimed is :1 . A method for using a scanning transmission electron microscope to determine atomic positions of a sample relative to an electron beam, comprising : performing a scan to obtain data, wherein the scan is performed using a diagnostic scan pattern; selecting points from the data based on intensity; obtaining an artificial lattice of the sample ; and mapping the points onto the arti ficial lattice by performing a mathematical optimi zation .2 . The method of claim 1 , wherein the artificial lattice is generated using known lattice constants of the sample .3 . The method of claim 1 , wherein the artificial lattice is generated by supplying the data to a neural network .4 . The method of claim 1 , wherein the mathematical optimization comprises a least squares fit or another similar algorithm.5 . The method of claim 1 , wherein the diagnostic scan pattern comprises one or more loops surrounding a target region .6 . The method of claim 5, wherein the diagnostic scan pattern comprises a plurality of loops .7 . The method of claim 6 , wherein a group of closely spaced points is defined as a cluster and a center of mass is determined for the cluster and the group of closely spaced points is replaced by one point disposed at the center of mass .8 . The method of claim 5 , wherein electrons are not irradiated into the target region during the scan .

9. The method of claim 5, wherein a task is performed in the target region after the mapping.

10. The method of claim 1, wherein the diagnostic scan pattern surrounds at least a portion of a target region.

11. The method of claim 10, wherein the diagnostic scan pattern comprises a portion of a loop.

12. The method of claim 10, wherein electrons are not irradiated into the target region during the scan.

13. The method of claim 10, wherein a task is performed in the target region after the mapping.

14. A system for determining atomic positions of a sample, comprising : a scanning transmission electron microscope (STEM) to create an electron beam; a detector; and a processing device, wherein the processing unit is configured to: instruct the STEM to perform a scan using a diagnostic scan pattern, wherein the sample is disposed between the detector and the STEM, and the detector captures data as the electron beam passes through the s amp 1 e ; select points from the data based on intensity; and map the points onto an artificial lattice by performing a mathematical optimization.

15. The system of claim 14, where the detector comprises an annular dark field (ADF) detector, an annular bright field (ABF) detector, a bright field (BE) detector, a secondary electron (SE) detector, or a charge coupled device (CCD) ,complementary metal-oxide semiconductor (CMOS) , or direct electron (DE) detector.

16. The system of claim 14, wherein the artificial lattice is generated using known lattice constants of the sample.

17. The system of claim 14, wherein the artificial lattice is generated by supplying the data to a neural network.

18. The system of claim 14, wherein the diagnostic scan pattern comprises one or more loops surrounding a target region .

19. The system of claim 18, wherein the diagnostic scan pattern does not include the target region.

Images