Knowledge Extraction from Atomically Resolved Images

Lukas Vlcek; Artem Maksov; Minghu Pan; Rama K. Vasudevan; Sergei V. Kalinin

doi:10.1021/acsnano.7b05036

Knowledge Extraction from Atomically Resolved Images

Lukas Vlcek, Artem Maksov, Minghu Pan, Rama K. Vasudevan, Sergei V. Kalinin

Data Nanoanalytics

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

30 Scopus citations

Abstract

Tremendous strides in experimental capabilities of scanning transmission electron microscopy and scanning tunneling microscopy (STM) over the past 30 years made atomically resolved imaging routine. However, consistent integration and use of atomically resolved data with generative models is unavailable, so information on local thermodynamics and other microscopic driving forces encoded in the observed atomic configurations remains hidden. Here, we present a framework based on statistical distance minimization to consistently utilize the information available from atomic configurations obtained from an atomically resolved image and extract meaningful physical interaction parameters. We illustrate the applicability of the framework on an STM image of a FeSe_xTe_1-x superconductor, with the segregation of the chalcogen atoms investigated using a nonideal interacting solid solution model. This universal method makes full use of the microscopic degrees of freedom sampled in an atomically resolved image and can be extended via Bayesian inference toward unbiased model selection with uncertainty quantification.

Original language	English
Pages (from-to)	10313-10320
Number of pages	8
Journal	ACS Nano
Volume	11
Issue number	10
DOIs	https://doi.org/10.1021/acsnano.7b05036
State	Published - Oct 24 2017

Funding

We wish to thank A. Sefat and B. Sales (ORNL) for the single crystal samples. This research was sponsored by the Division of Materials Sciences and Engineering, BES, DOE (R.K.V., S.V.K.). This research was conducted at the Center for Nanophase Materials Sciences, which is a U.S. DOE Office of Science User Facility. A portion of this research was supported by ORNL’s Laboratory Directed Research and Development Program, which is managed by UT-Battelle LLC for the U.S. DOE (L.V.). A.M. acknowledges fellowship support from the UT/ORNL Bredesen Center for Interdisciplinary Research and Graduate Education.

Funders	Funder number
ORNL’s Laboratory Directed Research and Development Program
UT/ORNL
U.S. Department of Energy
Basic Energy Sciences
Division of Materials Sciences and Engineering
UT-Battelle

Keywords

STM
image analysis
model
optimization
simulation
statistical distance

Access to Document

10.1021/acsnano.7b05036

Cite this

@article{f208322d2fac48e698e03e8250d88f30,

title = "Knowledge Extraction from Atomically Resolved Images",

abstract = "Tremendous strides in experimental capabilities of scanning transmission electron microscopy and scanning tunneling microscopy (STM) over the past 30 years made atomically resolved imaging routine. However, consistent integration and use of atomically resolved data with generative models is unavailable, so information on local thermodynamics and other microscopic driving forces encoded in the observed atomic configurations remains hidden. Here, we present a framework based on statistical distance minimization to consistently utilize the information available from atomic configurations obtained from an atomically resolved image and extract meaningful physical interaction parameters. We illustrate the applicability of the framework on an STM image of a FeSexTe1-x superconductor, with the segregation of the chalcogen atoms investigated using a nonideal interacting solid solution model. This universal method makes full use of the microscopic degrees of freedom sampled in an atomically resolved image and can be extended via Bayesian inference toward unbiased model selection with uncertainty quantification.",

keywords = "STM, image analysis, model, optimization, simulation, statistical distance",

author = "Lukas Vlcek and Artem Maksov and Minghu Pan and Vasudevan, {Rama K.} and Kalinin, {Sergei V.}",

note = "Publisher Copyright: {\textcopyright} 2017 American Chemical Society.",

year = "2017",

month = oct,

day = "24",

doi = "10.1021/acsnano.7b05036",

language = "English",

volume = "11",

pages = "10313--10320",

journal = "ACS Nano",

issn = "1936-0851",

publisher = "American Chemical Society",

number = "10",

}

TY - JOUR

T1 - Knowledge Extraction from Atomically Resolved Images

AU - Vlcek, Lukas

AU - Maksov, Artem

AU - Pan, Minghu

AU - Vasudevan, Rama K.

AU - Kalinin, Sergei V.

PY - 2017/10/24

Y1 - 2017/10/24

N2 - Tremendous strides in experimental capabilities of scanning transmission electron microscopy and scanning tunneling microscopy (STM) over the past 30 years made atomically resolved imaging routine. However, consistent integration and use of atomically resolved data with generative models is unavailable, so information on local thermodynamics and other microscopic driving forces encoded in the observed atomic configurations remains hidden. Here, we present a framework based on statistical distance minimization to consistently utilize the information available from atomic configurations obtained from an atomically resolved image and extract meaningful physical interaction parameters. We illustrate the applicability of the framework on an STM image of a FeSexTe1-x superconductor, with the segregation of the chalcogen atoms investigated using a nonideal interacting solid solution model. This universal method makes full use of the microscopic degrees of freedom sampled in an atomically resolved image and can be extended via Bayesian inference toward unbiased model selection with uncertainty quantification.

AB - Tremendous strides in experimental capabilities of scanning transmission electron microscopy and scanning tunneling microscopy (STM) over the past 30 years made atomically resolved imaging routine. However, consistent integration and use of atomically resolved data with generative models is unavailable, so information on local thermodynamics and other microscopic driving forces encoded in the observed atomic configurations remains hidden. Here, we present a framework based on statistical distance minimization to consistently utilize the information available from atomic configurations obtained from an atomically resolved image and extract meaningful physical interaction parameters. We illustrate the applicability of the framework on an STM image of a FeSexTe1-x superconductor, with the segregation of the chalcogen atoms investigated using a nonideal interacting solid solution model. This universal method makes full use of the microscopic degrees of freedom sampled in an atomically resolved image and can be extended via Bayesian inference toward unbiased model selection with uncertainty quantification.

KW - STM

KW - image analysis

KW - model

KW - optimization

KW - simulation

KW - statistical distance

UR - http://www.scopus.com/inward/record.url?scp=85033393679&partnerID=8YFLogxK

U2 - 10.1021/acsnano.7b05036

DO - 10.1021/acsnano.7b05036

M3 - Article

C2 - 28953356

AN - SCOPUS:85033393679

SN - 1936-0851

VL - 11

SP - 10313

EP - 10320

JO - ACS Nano

JF - ACS Nano

IS - 10

ER -

Knowledge Extraction from Atomically Resolved Images

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this