Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy

Max Schwarzer; Jesse Farebrother; Joshua Greaves; Ekin Dogus Cubuk; Rishabh Agarwal; Aaron Courville; Marc G. Bellemare; Sergei Kalinin; Igor Mordatch; Pablo Samuel Castro; Kevin M. Roccapriore

doi:10.1002/admi.202400998

Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy

Max Schwarzer, Jesse Farebrother, Joshua Greaves, Ekin Dogus Cubuk, Rishabh Agarwal, Aaron Courville, Marc G. Bellemare, Sergei Kalinin, Igor Mordatch, Pablo Samuel Castro, Kevin M. Roccapriore

Data Nanoanalytics

Research output: Contribution to journal › Article › peer-review

Abstract

A machine learning approach is introduced to determine the transition dynamics of silicon atoms on a single layer of carbon atoms, when stimulated by the electron beam of a scanning transmission electron microscope (STEM). This method is data-centric, leveraging data collected on a STEM. The data samples are processed and filtered to produce symbolic representations, which is used to train a neural network to predict transition probabilities. These learned transition dynamics are then leveraged to guide a single silicon atom throughout the lattice to pre-determined target destinations. Empirical analyses are presented that demonstrate the efficacy and generality of the approach.

Original language	English
Article number	2400998
Journal	Advanced Materials Interfaces
Volume	12
Issue number	11
DOIs	https://doi.org/10.1002/admi.202400998
State	Published - Jun 9 2025

Funding

M.S. and J.F. contributed equally to this work. Autonomous STEM research was supported by the Center for Nanophase Materials Sciences (CNMS) proposal [CNMS2022-B-01647], which is a US Department of Energy, Office of Science User Facility at Oak Ridge National Laboratory. This work was supported (S.V.K.) by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences as part of the Energy Frontier Research Centers program: CSSAS-The Center for the Science of Synthesis Across Scales under award number DE-SC0019288. The authors would also like to thank their colleagues at Google DeepMind for their useful feedback on this work. M.S. and J.F. contributed equally to this work. Autonomous STEM research was supported by the Center for Nanophase Materials Sciences (CNMS) proposal [CNMS2022\u2010B\u201001647], which is a US Department of Energy, Office of Science User Facility at Oak Ridge National Laboratory. This work was supported (S.V.K.) by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences as part of the Energy Frontier Research Centers program: CSSAS\u2010The Center for the Science of Synthesis Across Scales under award number DE\u2010SC0019288. The authors would also like to thank their colleagues at Google DeepMind for their useful feedback on this work.

Keywords

atomic manipulation
graphene
machine learning
microscopy
scanning transmission electron microscopy

Access to Document

10.1002/admi.202400998

Cite this

Schwarzer, M., Farebrother, J., Greaves, J., Cubuk, E. D., Agarwal, R., Courville, A., Bellemare, M. G., Kalinin, S., Mordatch, I., Castro, P. S., & Roccapriore, K. M. (2025). Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy. Advanced Materials Interfaces, 12(11), Article 2400998. https://doi.org/10.1002/admi.202400998

@article{53dc1eb8279342d9954aade33cd073d5,

title = "Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy",

abstract = "A machine learning approach is introduced to determine the transition dynamics of silicon atoms on a single layer of carbon atoms, when stimulated by the electron beam of a scanning transmission electron microscope (STEM). This method is data-centric, leveraging data collected on a STEM. The data samples are processed and filtered to produce symbolic representations, which is used to train a neural network to predict transition probabilities. These learned transition dynamics are then leveraged to guide a single silicon atom throughout the lattice to pre-determined target destinations. Empirical analyses are presented that demonstrate the efficacy and generality of the approach.",

keywords = "atomic manipulation, graphene, machine learning, microscopy, scanning transmission electron microscopy",

author = "Max Schwarzer and Jesse Farebrother and Joshua Greaves and Cubuk, \{Ekin Dogus\} and Rishabh Agarwal and Aaron Courville and Bellemare, \{Marc G.\} and Sergei Kalinin and Igor Mordatch and Castro, \{Pablo Samuel\} and Roccapriore, \{Kevin M.\}",

note = "Publisher Copyright: {\textcopyright} 2025 Oak Ridge National Laboratory and The Author(s). Advanced Materials Interfaces published by Wiley-VCH GmbH.",

year = "2025",

month = jun,

day = "9",

doi = "10.1002/admi.202400998",

language = "English",

volume = "12",

journal = "Advanced Materials Interfaces",

issn = "2196-7350",

publisher = "Wiley Open Access",

number = "11",

}

TY - JOUR

T1 - Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy

AU - Schwarzer, Max

AU - Farebrother, Jesse

AU - Greaves, Joshua

AU - Cubuk, Ekin Dogus

AU - Agarwal, Rishabh

AU - Courville, Aaron

AU - Bellemare, Marc G.

AU - Kalinin, Sergei

AU - Mordatch, Igor

AU - Castro, Pablo Samuel

AU - Roccapriore, Kevin M.

PY - 2025/6/9

Y1 - 2025/6/9

N2 - A machine learning approach is introduced to determine the transition dynamics of silicon atoms on a single layer of carbon atoms, when stimulated by the electron beam of a scanning transmission electron microscope (STEM). This method is data-centric, leveraging data collected on a STEM. The data samples are processed and filtered to produce symbolic representations, which is used to train a neural network to predict transition probabilities. These learned transition dynamics are then leveraged to guide a single silicon atom throughout the lattice to pre-determined target destinations. Empirical analyses are presented that demonstrate the efficacy and generality of the approach.

AB - A machine learning approach is introduced to determine the transition dynamics of silicon atoms on a single layer of carbon atoms, when stimulated by the electron beam of a scanning transmission electron microscope (STEM). This method is data-centric, leveraging data collected on a STEM. The data samples are processed and filtered to produce symbolic representations, which is used to train a neural network to predict transition probabilities. These learned transition dynamics are then leveraged to guide a single silicon atom throughout the lattice to pre-determined target destinations. Empirical analyses are presented that demonstrate the efficacy and generality of the approach.

KW - atomic manipulation

KW - graphene

KW - machine learning

KW - microscopy

KW - scanning transmission electron microscopy

UR - http://www.scopus.com/inward/record.url?scp=105005776275&partnerID=8YFLogxK

U2 - 10.1002/admi.202400998

DO - 10.1002/admi.202400998

M3 - Article

AN - SCOPUS:105005776275

SN - 2196-7350

VL - 12

JO - Advanced Materials Interfaces

JF - Advanced Materials Interfaces

IS - 11

M1 - 2400998

ER -

Learning and Controlling Silicon Dopant Transitions in Graphene Using Scanning Transmission Electron Microscopy

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this