Pith. sign in

REVIEW 4 major objections 6 minor 63 references

Autonomous Fabrication of Tailored Defect Structures in 2D Materials using Machine Learning-enabled Scanning Transmission Electron Microscopy

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A closed-loop electron microscope can fabricate designed atomic-scale defect structures on its own: it decodes images, selects sulfur atoms, and aims the beam to grow MoS-nanowire edges in MoS2.

desk verdict A credible proof-of-concept for closed-loop autonomous STEM fabrication, but the 'reliability' claim needs quantitative classifier and yield metrics. read the letter →

arxiv 2507.01194 v1 pith:D243BCVM submitted 2025-07-01 cond-mat.mtrl-sci cond-mat.mes-hall

Zijie Wu , Kevin M. Roccapriore , Ayana Ghosh , Kai Xiao , Raymond R. Unocic , Stephen Jesse , Rama Vasudevan , Matthew G. Boebinger This is my paper · ORCID
classification cond-mat.mtrl-scicond-mat.mes-hall
keywords autonomousfabricationscanningtransmissionelectronmicroscopymachinelearningMoS2SnanowirebeammanipulationdefectengineeringHAADFimaging
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper establishes that a closed-loop scanning transmission electron microscope can fabricate designed atomic-scale defect structures without human intervention: machine-learning models decode each high-angle annular dark-field (HAADF) image into atom identities, an automated planner chooses sulfur sites to remove, and programmable beam control exposes exactly those sites. Using a monolayer of $\mathrm{MoS_2}$ as the testbed, the authors report reliable production of $\mathrm{MoS}$-nanowire ($\mathrm{MoS}$-NW) edge structures on pre-defined nanopore edges, free-standing nanowires between two pores, and nanowires grown directionally across the field of view. The reason this matters is practical: if the loop works generally, atomically precise defect structures with tailored electronic and mechanical properties can be manufactured by specifying a target morphology and letting the microscope execute it, rather than by manual operation and favorable irradiation statistics. The authors show image sequences as evidence and note that the machine-learning models were trained on only a few dozen hand-labeled examples.

What carries the argument

The load-bearing machinery is the closed feedback cycle of image decoding, site selection, and beam delivery. Atom positions come from an ELIT ensemble-learning iterative-training model, pretrained on multislice simulations of $\mathrm{MoS_2}$ and augmented with experimental images. Nanowire regions are segmented by a lightweight U-net, a convolutional network architecture for image segmentation. The remaining atomic sites are classified by a random forest that takes a hand-crafted vector of the intensity differences and distances to each atom's five nearest neighbors, with class probabilities averaged over trees as an uncertainty estimate. The fabrication side is an FPGA-controlled scan routine that can draw scan paths such as an Archimedean spiral of 2--5~\AA{} size, giving a radially symmetric dose profile with controllable dwell time. The argument rests on these components working together: drift correction between HAADF snapshots keeps the field of view aligned, so every fabrication step is planned from a freshly decoded image.

What would settle it

Run the closed loop on a fresh pristine $\mathrm{MoS_2}$ monolayer at a pixel size and defect morphology not seen in training, compare every targeted site against a high-quality post-fabrication image, and count how many beam exposures remove the intended sulfur atom; if mislabeled molybdenum sites or missed nanowire regions divert the beam often enough to prevent the intended $\mathrm{MoS}$-NW structure from appearing, the claim of reliable autonomous fabrication is falsified.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that a small-data machine-learning stack, coupled to a flexible beam scanner, is sufficient for autonomous feedback-controlled fabrication of complex defect structures. The workflow cycles through four stages: acquire a HAADF image; decode it by locating every atomic column with an ELIT ensemble of convolutional networks, segmenting $\mathrm{MoS}$-NW regions with a lightweight U-net, and classifying each remaining atom as $\mathrm{Mo}$, $\mathrm{S}$, or sulfur-vacancy-line with a random forest; choose which sulfur sites to expose according to a pre-defined growth strategy; and deliver a localized 60~kV electron beam dose to each site, using either the microscope's native scan or an FPGA-controlled Archimedean spiral for radially symmetric exposure. The authors report that targeting sulfur sites is controllable while targeting molybdenum sites is not, consistent with sulfur's lower displacement threshold, and that this distinction is what makes the three demonstrated outcomes possible: nanopore-edge $\mathrm{MoS}$-NWs, free-standing $\mathrm{MoS}$-NWs spanning two nanopores, and directional growth of a single nanowire across the full field of view.

Load-bearing premise

The demonstration rests on the machine-learning classifiers continuing to label sulfur, molybdenum, and nanowire sites correctly as the lattice distorts and the pixel size changes, even though they were trained on only 55 hand-labeled atom neighborhoods and roughly 30 images and the paper reports no measured accuracy for them.

Editorial extensions

If this is right

  • Sulfur-site targeting, rather than molybdenum-site irradiation, is the controllable route for electron-beam defect fabrication in monolayer $\mathrm{MoS_2}$ at 60 kV.
  • The workflow reliably produces $\mathrm{MoS}$-NW structures on pre-defined nanopore edges by repeatedly removing sulfur atoms along the pore edge.
  • Free-standing $\mathrm{MoS}$-NWs can be grown between two seeded nanopores without straining the surrounding lattice, by targeting atoms along the edge between the pores.
  • Directional growth of a single $\mathrm{MoS}$-NW across the full field of view is achieved by clustering detected nanowire sites, fitting the largest cluster to a line, and removing nearby sulfur atoms at both ends of the line.
  • The authors claim the framework is material-agnostic, so the same closed loop could be applied to other 2D materials and other defect or heterostructure motifs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • An extension the authors leave implicit is to use the random forest's per-class probabilities as a gating signal: when classification confidence drops, the system could re-image or expand the labeled support set instead of continuing to target, which would reduce misdirected beam exposure.
  • Because the paper's directional-growth strategy treats each frame independently and clusters nanowire sites afresh, adding temporal tracking of individual atoms and defect clusters across frames would likely stabilize target selection in the early iterations, where the authors note the algorithm can fluctuate between small clusters.
  • A natural transfer test would be to run the same loop on another 2D material with a known electron-beam removal site; the small training-set sizes suggest the bottleneck is not model capacity but whether the hand-crafted neighbor-distance and intensity features separate the relevant site classes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript reports an autonomous scanning transmission electron microscopy (STEM) workflow for fabricating tailored MoS-nanowire (MoS-NW) edge structures in monolayer MoS2. The system iteratively acquires HAADF images, decodes them using a machine learning pipeline (ELIT for atom localization, a U-net for MoS-NW segmentation, and a random forest for Mo/S/SVL classification), selects sulfur sites for removal according to predefined fabrication strategies, and exposes those sites with an FPGA-controlled beam. The authors demonstrate three fabrication outcomes: MoS-NW growth along selected nanopore edges, freestanding MoS-NWs between two nanopores, and directional growth of long MoS-NWs across the field of view. The code and training data are made available in an open-source repository.

Significance. If the central claims hold, the work is a meaningful step toward closed-loop atomic-scale manufacturing: it integrates real-time ML-based image decoding with programmable beam control and demonstrates the formation of complex 1D-2D heterostructures. The open-source release of code and data is a strength that supports reproducibility. However, the demonstrations are essentially qualitative. The authors assert 'reliable' and 'fully autonomous' operation, but no quantitative ML validation metrics, yield statistics, or reproducibility data are provided. Because the ML classifications directly determine every beam-targeting decision, the lack of validation of these classifiers is a load-bearing gap. The framework's material-agnostic ambition is plausible but remains a promise rather than a demonstrated transfer.

major comments (4)
  1. [Machine learning models and training; Conclusions] No quantitative evaluation of the random forest (trained on 55 hand-labeled instances) or the U-net (trained on ~30 images) is reported. There are no confusion matrices, precision/recall values, or held-out validation results across fabrication stages, defect morphologies, or pixel sizes. The Conclusion's statement that the ML framework 'consistently achieved reliable atom classification at various pixel sizes' is therefore unsupported. This is load-bearing because every beam-targeting decision in Figure 2a is made directly from these classifications; a misclassified S site as Mo, or a wrongly segmented NW region, would produce incorrect targeting. The authors should report classification metrics on held-out images representative of pristine, seeded-nanopore, and NW-growth stages at the pixel sizes used, and ideally quantify how classification errors propagate into targeting decisions.
  2. [Machine learning models and training] The random forest uses a vectorization based on the five nearest atoms identified by ELIT, and the Methods section explicitly notes that its performance depends on ELIT accurately locating atom sites. Yet no localization error statistics for ELIT on experimental images are provided, particularly in distorted regions near nanopores and MoS-NWs where atom positions are most challenging. If ELIT mislocates atoms, the random forest input is corrupted and the entire decoding chain degrades. The authors should quantify ELIT detection precision/recall and localization error on representative experimental images, or otherwise demonstrate that the downstream classification is robust to realistic localization noise.
  3. [Results, Section II; Figure 5] The central claim of 'reliably' forming tailored MoS-NW structures is supported only by a small number of image sequences, with no statistics on the number of attempts, success rates, or variability across independent runs. Terms such as 'reliably formed' and 'we found this strategy to work relatively well' appear without quantitative support. For a paper whose contribution is an autonomous fabrication method, the authors should report the number of experiments performed for each strategy (targeted growth, freestanding NW, directional growth), the success/failure criteria, and quantitative descriptors of the resulting structures (e.g., NW length, continuity, alignment, or deviation from the intended geometry). This is needed to substantiate the reliability claim and to allow readers to judge the method's practical utility.
  4. [Abstract; Results; Conclusions] The paper repeatedly describes the approach as 'fully autonomous,' but the workflow depends on human-predefined fabrication strategies (e.g., the triangular mask geometry, the DBSCAN cluster selection, the ~0.2-0.3 nm distance threshold for targeting S sites) and on the human operator choosing which strategy to apply. The autonomy lies in the execution of a pre-specified recipe, not in the discovery or selection of the strategy. This should be stated precisely and the title/abstract claim qualified accordingly, because the scope of the claim is part of the paper's contribution and current wording risks overstating the advance relative to prior automated feedback systems.
minor comments (6)
  1. [Conclusions vs. Methods] The number of training images is inconsistent: the Conclusions state '~20 images with labeled MoS-NW regions and ~40 labeled atoms,' while the Methods state '~30 labeled MoS2 HAADF images' and a support set of '55 hand-labeled instances' (21 Mo, 24 S, 10 SVL). Please reconcile these numbers.
  2. [Figure 5 caption] The phrase 'In all three subfigures, (i-v) represent raw ADF acquisition and decoded images' is unclear; consider rewriting as 'In each row, panels (i)-(v) show raw ADF images and decoded images with beam targets overlaid, from the beginning to the end of the iterative workflow.'
  3. [Results, Section I] The statement 'while S and Mo can potentially be classified by e.g. a simple pixel intensity threshold' is made without a citation or a quantitative comparison; either add a reference or provide a brief justification.
  4. [Throughout] The abbreviation 'MoS-NW' is inconsistent with the abstract's 'MoS-nanowire' and appears as 'MoS-NW' elsewhere; standardize the usage.
  5. [Machine learning models and training] The repository link is given as a URL only; for archival purposes, please provide a DOI or versioned release, since URLs can change.
  6. [References] Reference [38] is a conference abstract (Microscopy and Microanalysis) that may not provide sufficient methodological detail; consider citing a full peer-reviewed article if one exists.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the experimental feedback loop is self-contained; the ML classifiers are trained on independent labeled sets and the fabrication claim is anchored to physical HAADF images, not to the fitted models.

full rationale

The paper's derivation chain is an experimental control loop: acquire HAADF image, decode atom positions and species with ML models, designate sulfur sites for beam exposure, blast, and re-image. No equation in the paper defines an output quantity in terms of the input quantity it is said to predict. The random forest is trained on 55 hand-labeled neighborhoods (21 Mo, 24 S, 10 SVL) and the U-net on about 30 labeled images, then applied at inference time to fresh frames; the resulting target lists drive beam positions. The final evidence is the raw ADF image series in Figure 5 showing actual MoS-NW edge structures, so the success criterion is not the classifier's own training labels. Prior work from the same group is cited for the ELIT atom-localization model and for previously found electron-beam fabrication parameters; these are externally published models and parameters, not a self-citation chain that forces the present conclusion. The paper itself flags the main weakness: it asserts 'reliable atom classification at various pixel sizes' while reporting no confusion matrices or held-out metrics, and it admits that no robust temporal tracking algorithm has been implemented. Those are validation gaps, not circular reductions. There is therefore no step where a 'prediction' is equivalent to a fitted input by construction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper does not introduce new physical entities. It relies on several experimental parameters chosen by hand (dwell time, thresholds, mask shapes) and on domain assumptions about Z-contrast imaging and sulfur-selective sputtering. The central claim depends on these choices and on the generalization of the ML classifiers trained on small data sets.

free parameters (5)
  • electron beam dwell time per sulfur site = 1-2 s
    Chosen by experimenters to reliably eject sulfur atoms; not derived from data.
  • distance threshold for targeting sulfur sites near nanowire segments = ~0.2-0.3 nm
    The paper states 'we currently found a good threshold to be ~0.2-0.3 nm'; this hand-selected value controls which sites are blasted in directional growth.
  • FPGA spiral scan size = 2-5 Å
    Chosen to concentrate dose at the center with radial symmetry; not derived.
  • image rescaling pixel size = 0.085 Å/pixel
    Used uniformly in the U-net training; a manual preprocessing choice.
  • triangular mask geometry = not specified numerically
    The mask defines the seeded pore shape and is chosen by the operator, not derived from data.
assumptions (4)
  • domain assumption HAADF STEM image intensity is a monotonic function of atomic number (Z-contrast) for Mo and S in MoS2.
    The entire atom identification and classification pipeline is built on the assumption that Mo and S atoms can be distinguished by their scattered intensity; this is standard in STEM but not verified in this paper.
  • domain assumption 60 kV electron beam preferentially ejects sulfur atoms from MoS2, with molybdenum displacement requiring higher energy.
    The fabrication strategy relies on this selective sputtering; the paper cites refs 51-52 for support instead of proving it anew.
  • domain assumption The CVD-grown monolayer MoS2 is clean and stable enough for many imaging and beam-exposure iterations, with only correctable drift.
    The workflow assumes that repeated HAADF imaging and targeted exposure will produce the intended changes without uncontrolled damage or contamination.
  • domain assumption The structures identified by the U-net and HAADF contrast as MoS-NW are indeed metallic MoS6S6 nanowires with the expected properties.
    No direct verification such as EELS or high-resolution STEM imaging is presented in this work; identity rests on image contrast and prior literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Autonomous Fabrication of Tailored Defect Structures in 2D Materials using Machine Learning-enabled Scanning Transmission Electron Microscopy." pith.science (2026). https://pith.science/paper/D243BCVM

@misc{pith2026250701194,
  author       = {Pith},
  title        = {Pith review of: Autonomous Fabrication of Tailored Defect Structures in 2D Materials using Machine Learning-enabled Scanning Transmission Electron Microscopy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D243BCVM}},
  note         = {Machine review of arXiv:2507.01194}
}
abstract

Materials with tailored quantum properties can be engineered from atomic scale assembly techniques, but existing methods often lack the agility and accuracy to precisely and intelligently control the manufacturing process. Here we demonstrate a fully autonomous approach for fabricating atomic-level defects using electron beams in scanning transmission electron microscopy (STEM) that combines advanced machine learning and automated beam control. As a proof of concept, we achieved controlled fabrication of MoS-nanowire (MoS-NW) edge structures by iterative and targeted exposure of $MoS_2$ monolayer to a focused electron beam to selectively eject sulfur atoms, utilizing high-angle annular dark-field (HAADF) imaging for feedback-controlled monitoring structural evolution of defects. A machine learning framework combining a random forest model and convolutional neural networks (CNN) was developed to decode the HAADF image and accurately identify atomic positions and species. This atomic-level information was then integrated into an autonomous decision-making platform, which applied predefined fabrication strategies to instruct beam control about atomic sites to be ejected. The selected sites were subsequently exposed to localized electron beam using an FPGA-controlled scan routine with precise control over beam positioning and duration. While the MoS-NW edge structures produced exhibit promising mechanical and electronic properties, the proposed autonomous fabrication framework is material-agnostic and can be extended to other 2D materials for the creation of diverse defect structures and heterostructures beyond $MoS_2$.

Figures

Figures reproduced from arXiv: 2507.01194 by the authors.

Figure 1
Figure 1. Experimental workflow of the Automated Fabrication Controller (AFC) on the Nion UltraSTEM microscopes: (a) Signal from the ADF detectors on the microscope are fed into atom identification models through either the Nion Swift controller or our FPGA (1) which is shown in (b) as raw data through a live data stream (2). (c) The raw data is then run through the machine learning labeling workflow (3) to first (i) identify… view at source ↗
Figure 2
Figure 2. Schematics for details of the machine learning aided feature classification models (described in the Methods section): (a) Overview of the classification workflow; (b) Example instances of the support set curated for Mo, S, and SVL atom sites and their neighborhoods; (c) Vectorization scheme, where the local surroundings around an atom site is described by the pixel intensity difference (Δ𝐼) and distance between the… view at source ↗
Figure 3
Figure 3. Uncertainty quantification of random forest classification between Mo, S, and SVL sites. (a-b) an example of (a) raw HAADF and (b) fully labeled image at a certain iteration in our fabrication workflow, with green, yellow, cyan, and pink representing Mo, S, SVL, and NW sites. (c-e) Predicted probability of the atom belong to (c) Mo, (d) S, and (e) SVL site by “voting” of trees in the random forest classifier. (f) Fo… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: MoS-NW site identification and selection: (a-d) Initial seeding of nanopore from pristine MoS2 monolayer, regardless of the desired MoS-NW structure to be fabricated, with (a,b) first attempt to eject S within the center triangular mask and (c,d) continuing enlargement…
Figure 5
Figure 5. Figure 5: A selection of results from different MoS-NW fabrication strategies: (a) Example of the technique used to form a MoS-NW along a pre-selected edge of a nanopore using a seed hole and then preferential S￾site targeting. (b) Example of the technique to make a freestanding…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 63 canonical work pages

  1. [1]

    and Y.-W

    Pan, H. and Y.-W. Zhang, Edge-dependent structural, electronic and magnetic properties of MoS2 nanoribbons. Journal of Materials Chemistry, 2012. 22(15): p. 7280-7290

    work page 2012

  2. [2]

    Solid State Communications, 2014

    Davelou, D., et al., MoS2 nanostructures: Semiconductors with metallic edges. Solid State Communications, 2014. 192: p. 42-46

    work page 2014

  3. [3]

    Nano Letters, 2017

    Tinoco, M., et al., Strain-Dependent Edge Structures in MoS2 Layers. Nano Letters, 2017. 17(11): p. 7021-7026

    work page 2017

  4. [4]

    Nature Reviews Materials, 2019

    Dyck, O., et al., Atom-by-atom fabrication with electron beams. Nature Reviews Materials, 2019. 4(7): p. 497-507

    work page 2019

  5. [5]

    Rodriguez, and F

    Mathew, P.T., B.J. Rodriguez, and F. Fang, Atomic and Close-to-Atomic Scale Manufacturing: A Review on Atomic Layer Removal Methods Using Atomic Force Microscopy. Nanomanufacturing and Metrology, 2020. 3(3): p. 167-186

    work page 2020

  6. [6]

    International Journal of Extreme Manufacturing, 2022

    Gao, J., et al., Fundamentals of atomic and close-to-atomic scale manufacturing: a review. International Journal of Extreme Manufacturing, 2022. 4(1): p. 012001

    work page 2022

  7. [7]

    Fang, and N

    Chen, Y., F. Fang, and N. Zhang, Advance in additive manufacturing of 2D materials at the atomic and close-to-atomic scale. npj 2D Materials and Applications, 2024. 8(1): p. 17

    work page 2024

  8. [8]

    ACS Central Science, 2021

    Yu, C.-J., et al., A Molecular Approach to Quantum Sensing. ACS Central Science, 2021. 7(5): p. 712-723

    work page 2021

Show all 63 references
  1. [9]

    Nature Nanotechnology, 2024

    Esat, T., et al., A quantum sensor for atomic-scale electric and magnetic fields. Nature Nanotechnology, 2024. 19(10): p. 1466-1471

    2024

  2. [10]

    Advanced Quantum Technologies, 2022

    Vajner, D.A., et al., Quantum Communication Using Semiconductor Quantum Dots. Advanced Quantum Technologies, 2022. 5(7): p. 2100116

    2022

  3. [11]

    Nature, 2011

    Specht, H.P., et al., A single-atom quantum memory. Nature, 2011. 473(7346): p. 190-193

    2011

  4. [12]

    ACS Nano, 2018

    Hudak, B.M., et al., Directed Atom-by-Atom Assembly of Dopants in Silicon. ACS Nano, 2018. 12(6): p. 5873-5879

    2018

  5. [13]

    Nanotechnology, 2018

    Jesse, S., et al., Direct atomic fabrication and dopant positioning in Si using electron beams with active real-time image-based feedback. Nanotechnology, 2018. 29(25): p. 255303

    2018

  6. [14]

    ACS Nano, 2013

    Xu, Q., et al., Controllable Atomic Scale Patterning of Freestanding Monolayer Graphene at Elevated Temperature. ACS Nano, 2013. 7(2): p. 1566-1572

    2013

  7. [15]

    2D Materials, 2019

    Clark, N., et al., Nanometre electron beam sculpting of suspended graphene and hexagonal boron nitride heterostructures. 2D Materials, 2019. 6(2): p. 025032

    2019

  8. [16]

    ACS Nano, 2016

    Lin, J., et al., Structural Flexibility and Alloying in Ultrathin Transition-Metal Chalcogenide Nanowires. ACS Nano, 2016. 10(2): p. 2782-2790

    2016

  9. [17]

    Nature Nanotechnology, 2014

    Lin, J., et al., Flexible metallic nanowires with self-adaptive contacts to semiconducting transition-metal dichalcogenide monolayers. Nature Nanotechnology, 2014. 9(6): p. 436-442

    2014

  10. [18]

    MRS Bulletin, 2017

    Zhao, X., et al., Engineering and modifying two-dimensional materials by electron beams. MRS Bulletin, 2017. 42(9): p. 667-676

    2017

  11. [19]

    Advanced Materials, 2018

    Zhao, X., et al., Atom-by-Atom Fabrication of Monolayer Molybdenum Membranes. Advanced Materials, 2018. 30(23): p. 1707281

    2018

  12. [20]

    Nature Communications, 2018

    Sang, X., et al., In situ edge engineering in two-dimensional transition metal dichalcogenides. Nature Communications, 2018. 9(1): p. 2051

    2018

  13. [21]

    Meyer, and J

    Susi, T., J.C. Meyer, and J. Kotakoski, Manipulating low-dimensional materials down to the level of single atoms with electron irradiation. Ultramicroscopy, 2017. 180: p. 163-172

    2017

  14. [22]

    Nano Letters,

    Tripathi, M., et al., Electron-Beam Manipulation of Silicon Dopants in Graphene. Nano Letters,

  15. [23]

    Advanced Functional Materials, 2019

    Mustonen, K., et al., Electron-Beam Manipulation of Silicon Impurities in Single-Walled Carbon Nanotubes. Advanced Functional Materials, 2019. 29(52): p. 1901327

    2019

  16. [24]

    Small, 2018

    Dyck, O., et al., Building Structures Atom by Atom via Electron Beam Manipulation. Small, 2018. 14(38): p. 1801771

    2018

  17. [25]

    Carbon, 2020

    Dyck, O., et al., Electron-beam introduction of heteroatomic Pt–Si structures in graphene. Carbon, 2020. 161: p. 750-757

    2020

  18. [26]

    ACS Applied Nano Materials, 2020

    Dyck, O., et al., Doping of Cr in Graphene Using Electron Beam Manipulation for Functional Defect Engineering. ACS Applied Nano Materials, 2020. 3(11): p. 10855-10863

    2020

  19. [27]

    Carbon, 2021

    Dyck, O., et al., Doping transition-metal atoms in graphene for atomic-scale tailoring of electronic, magnetic, and quantum topological properties. Carbon, 2021. 173: p. 205-214

    2021

  20. [28]

    Advanced Materials Interfaces, 2025

    Schwarzer, M., et al., Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy. Advanced Materials Interfaces, 2025. n/a(n/a): p. 2400998

    2025

  21. [29]

    Scientific Reports, 2017

    Sang, X., et al., Precision controlled atomic resolution scanning transmission electron microscopy using spiral scan pathways. Scientific Reports, 2017. 7(1): p. 43585

    2017

  22. [30]

    Advanced Structural and Chemical Imaging, 2016

    Sang, X., et al., Dynamic scan control in STEM: spiral scans. Advanced Structural and Chemical Imaging, 2016. 2(1): p. 6

    2016

  23. [31]

    Small, 2015

    Jesse, S., et al., Atomic-Level Sculpting of Crystalline Oxides: Toward Bulk Nanofabrication with Single Atomic Plane Precision. Small, 2015. 11(44): p. 5895-5900

    2015

  24. [32]

    Advanced Materials, 2023

    Boebinger, M.G., et al., The Atomic Drill Bit: Precision Controlled Atomic Fabrication of 2D Materials. Advanced Materials, 2023. 35(14): p. 2210116

    2023

  25. [33]

    ACS Nano, 2022

    Roccapriore, K.M., et al., Probing Electron Beam Induced Transformations on a Single-Defect Level via Automated Scanning Transmission Electron Microscopy. ACS Nano, 2022. 16(10): p. 17116-17127

    2022

  26. [34]

    Nat Nanotechnol, 2014

    Lin, J., et al., Flexible metallic nanowires with self-adaptive contacts to semiconducting transition-metal dichalcogenide monolayers. Nat Nanotechnol, 2014. 9(6): p. 436-42

    2014

  27. [35]

    npj Computational Materials, 2021

    Ghosh, A., et al., Ensemble learning-iterative training machine learning for uncertainty quantification and automated experiment in atom-resolved microscopy. npj Computational Materials, 2021. 7(1): p. 100

    2021

  28. [36]

    Science Advances

    Roccapriore, K.M., et al., Dynamic STEM-EELS for single-atom and defect measurement during electron beam transformations. Science Advances. 10(29): p. eadn5899

  29. [37]

    npj Computational Materials, 2024

    Boebinger, M.G., et al., Exploring electron-beam induced modifications of materials with machine-learning assisted high temporal resolution electron microscopy. npj Computational Materials, 2024. 10(1): p. 260

    2024

  30. [38]

    Microscopy and Microanalysis, 2022

    Boebinger, M.G., et al., Atomic-scale Fabrication of 1D-2D Nano Hetero-structures within 2D Materials through Automated Tracking and Electron Beam Control. Microscopy and Microanalysis, 2022. 28(S1): p. 2178-2180

    2022

  31. [39]

    Nature Reviews Methods Primers, 2022

    Kalinin, S.V., et al., Machine learning in scanning transmission electron microscopy. Nature Reviews Methods Primers, 2022. 2(1): p. 11

    2022

  32. [40]

    npj Computational Materials, 2023

    Kalinin, S.V., et al., Machine learning for automated experimentation in scanning transmission electron microscopy. npj Computational Materials, 2023. 9(1): p. 227

    2023

  33. [41]

    ACS Nano,

    Ziatdinov, M., et al., Deep Learning of Atomically Resolved Scanning Transmission Electron Microscopy Images: Chemical Identification and Tracking Local Transformations. ACS Nano,

  34. [42]

    Computational Materials Science, 2025

    Abebe, W., et al., SAM-I-Am: Semantic boosting for zero-shot atomic-scale electron micrograph segmentation. Computational Materials Science, 2025. 246: p. 113400

    2025

  35. [43]

    npj Computational Materials, 2021

    Akers, S., et al., Rapid and flexible segmentation of electron microscopy data using few-shot machine learning. npj Computational Materials, 2021. 7(1): p. 187

    2021

  36. [44]

    Small, 2023

    Prifti, E., et al., Variational Convolutional Autoencoders for Anomaly Detection in Scanning Transmission Electron Microscopy. Small, 2023. 19(16): p. 2205977. 20

    2023

  37. [45]

    Microscopy and Microanalysis, 2021

    Sadre, R., et al., Deep Learning Segmentation of Complex Features in Atomic-Resolution Phase- Contrast Transmission Electron Microscopy Images. Microscopy and Microanalysis, 2021. 27(4): p. 804-814

    2021

  38. [46]

    Nano Letters, 2022

    Lee, K., et al., STEM Image Analysis Based on Deep Learning: Identification of Vacancy Defects and Polymorphs of MoS2. Nano Letters, 2022. 22(12): p. 4677-4685

    2022

  39. [47]

    Unsupervised segmentation of irradiation$\unicode{x2010}$induced order$\unicode{x2010}$disorder phase transitions in electron microscopy

    Ter-Petrosyan, A.H., et al. Unsupervised segmentation of irradiation$\unicode{x2010}$induced order$\unicode{x2010}$disorder phase transitions in electron microscopy. 2023. arXiv:2311.08585 DOI: 10.48550/arXiv.2311.08585

    2023 doi arXiv

  40. [48]

    Physical Review B, 2013

    Komsa, H.-P., et al., From point to extended defects in two-dimensional MoS${}_{2}$: Evolution of atomic structure under electron irradiation. Physical Review B, 2013. 88(3): p. 035301

    2013

  41. [49]

    ACS Nano, 2016

    Wang, S., et al., Detailed Atomic Reconstruction of Extended Line Defects in Monolayer MoS2. ACS Nano, 2016. 10(5): p. 5419-5430

    2016

  42. [50]

    Small Methods, 2024

    Liu, Y., et al., AEcroscopy: A Software–Hardware Framework Empowering Microscopy Toward Automated and Autonomous Experimentation. Small Methods, 2024. 8(10): p. 2301740

    2024

  43. [51]

    Nano Letters, 2020

    Kretschmer, S., et al., Formation of Defects in Two-Dimensional MoS2 in the Transmission Electron Microscope at Electron Energies below the Knock-on Threshold: The Role of Electronic Excitations. Nano Letters, 2020. 20(4): p. 2865-2870

    2020

  44. [52]

    Nanoscale, 2023

    Yoshimura, A., et al., Quantum theory of electronic excitation and sputtering by transmission electron microscopy. Nanoscale, 2023. 15(3): p. 1053-1067

    2023

  45. [53]

    1996, AAAI Press: Portland, Oregon

    Ester, M., et al., A density-based algorithm for discovering clusters in large spatial databases with noise, in Proceedings of the Second International Conference on Knowledge Discovery and Data Mining. 1996, AAAI Press: Portland, Oregon. p. 226–231

    1996

  46. [54]

    Nature Communications, 2022

    Chen, I.J., et al., Precise atom manipulation through deep reinforcement learning. Nature Communications, 2022. 13(1): p. 7499

    2022

  47. [55]

    Pedregosa, F., et al., Scikit-learn: Machine Learning in Python. J. Mach. Learn. Res., 2011. 12(null): p. 2825–2830

    2011

  48. [56]

    Kondor, and G

    Bartók, A.P., R. Kondor, and G. Csányi, On representing chemical environments. Physical Review B, 2013. 87(18): p. 184115

    2013

  49. [57]

    Physical Review B, 2019

    Drautz, R., Atomic cluster expansion for accurate and transferable interatomic potentials. Physical Review B, 2019. 99(1): p. 014104

    2019

  50. [58]

    Fischer, and T

    Ronneberger, O., P. Fischer, and T. Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation. in Medical Image Computing and Computer-Assisted Intervention – MICCAI

  51. [59]

    Nature Machine Intelligence, 2022

    Ziatdinov, M., et al., AtomAI framework for deep learning analysis of image and spectroscopy data in electron and scanning probe microscopy. Nature Machine Intelligence, 2022. 4(12): p. 1101-1112

    2022

  52. [60]

    2019, Curran Associates Inc

    Paszke, A., et al., PyTorch: an imperative style, high-performance deep learning library, in Proceedings of the 33rd International Conference on Neural Information Processing Systems. 2019, Curran Associates Inc. p. Article 721

    2019

  53. [2015]

    Cham: Springer International Publishing

    2015. Cham: Springer International Publishing

    2015

  54. [2017]

    12742-12752

    11(12): p. 12742-12752

  55. [2018]

    5319-5323

    18(8): p. 5319-5323. 19

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.