Fast Scanning Probe Microscopy via Machine Learning: Non-Rectangular Scans with Compressed Sensing and Gaussian Process Optimization

Kyle P. Kelley; Maxim Ziatdinov; Liam Collins; Michael A. Susner; Rama K. Vasudevan; Nina Balke; Sergei V. Kalinin; Stephen Jesse

doi:10.1002/smll.202002878

Fast Scanning Probe Microscopy via Machine Learning: Non-Rectangular Scans with Compressed Sensing and Gaussian Process Optimization

Kyle P. Kelley, Maxim Ziatdinov, Liam Collins, Michael A. Susner, Rama K. Vasudevan, Nina Balke, Sergei V. Kalinin, Stephen Jesse

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

42 Scopus citations

Abstract

Fast scanning probe microscopy enabled via machine learning allows for a broad range of nanoscale, temporally resolved physics to be uncovered. However, such examples for functional imaging are few in number. Here, using piezoresponse force microscopy (PFM) as a model application, a factor of 5.8 reduction in data collection using a combination of sparse spiral scanning with compressive sensing and Gaussian process regression reconstruction is demonstrated. It is found that even extremely sparse spiral scans offer strong reconstructions with less than 6% error for Gaussian process regression reconstructions. Further, the error associated with each reconstructive technique per reconstruction iteration is analyzed, finding the error is similar past ≈15 iterations, while at initial iterations Gaussian process regression outperforms compressive sensing. This study highlights the capabilities of reconstruction techniques when applied to sparse data, particularly sparse spiral PFM scans, with broad applications in scanning probe and electron microscopies.

Original language	English
Article number	2002878
Journal	Small
Volume	16
Issue number	37
DOIs	https://doi.org/10.1002/smll.202002878
State	Published - Sep 1 2020

Funding

The work was supported by the U.S. Department of Energy, Office of Science, Materials Sciences and Engineering Division (K.P.K., R.K.V., N.B.). The PFM and image reconstruction work was conducted at and supported by the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility (S.V.K, L.C., M.Z.). Partial support for sample synthesis was provided by the Laboratory Directed Research and Development program at the Oak Ridge National Laboratory (Michael A. McGuire). A portion of the writing of this manuscript was supported by AFOSR (LRIR # 19RXCOR052) and AFRL/RX (Lab Director's Funds). The work was supported by the U.S. Department of Energy, Office of Science, Materials Sciences and Engineering Division (K.P.K., R.K.V., N.B.). The PFM and image reconstruction work was conducted at and supported by the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility (S.V.K, L.C., M.Z.). Partial support for sample synthesis was provided by the Laboratory Directed Research and Development program at the Oak Ridge National Laboratory (Michael A. McGuire). A portion of the writing of this manuscript was supported by AFOSR (LRIR # 19RXCOR052) and AFRL/RX (Lab Director's Funds).

Funders	Funder number
AFRL/RX
Center for Nanophase Materials Sciences
LRIR	19RXCOR052
U.S. Department of Energy
Air Force Office of Scientific Research
Office of Science
Oak Ridge National Laboratory
Laboratory Directed Research and Development
Division of Materials Sciences and Engineering

Keywords

Gaussian process regression
atomic force microscopy
compressive sensing
ferroelectric heterostructures
piezoresponse force microscopy

Access to Document

10.1002/smll.202002878

Cite this

@article{53c894cd6d714ceb8a55d621ed71aabf,

title = "Fast Scanning Probe Microscopy via Machine Learning: Non-Rectangular Scans with Compressed Sensing and Gaussian Process Optimization",

abstract = "Fast scanning probe microscopy enabled via machine learning allows for a broad range of nanoscale, temporally resolved physics to be uncovered. However, such examples for functional imaging are few in number. Here, using piezoresponse force microscopy (PFM) as a model application, a factor of 5.8 reduction in data collection using a combination of sparse spiral scanning with compressive sensing and Gaussian process regression reconstruction is demonstrated. It is found that even extremely sparse spiral scans offer strong reconstructions with less than 6% error for Gaussian process regression reconstructions. Further, the error associated with each reconstructive technique per reconstruction iteration is analyzed, finding the error is similar past ≈15 iterations, while at initial iterations Gaussian process regression outperforms compressive sensing. This study highlights the capabilities of reconstruction techniques when applied to sparse data, particularly sparse spiral PFM scans, with broad applications in scanning probe and electron microscopies.",

keywords = "Gaussian process regression, atomic force microscopy, compressive sensing, ferroelectric heterostructures, piezoresponse force microscopy",

author = "Kelley, {Kyle P.} and Maxim Ziatdinov and Liam Collins and Susner, {Michael A.} and Vasudevan, {Rama K.} and Nina Balke and Kalinin, {Sergei V.} and Stephen Jesse",

note = "Publisher Copyright: {\textcopyright} 2020 Wiley-VCH GmbH",

year = "2020",

month = sep,

day = "1",

doi = "10.1002/smll.202002878",

language = "English",

volume = "16",

journal = "Small",

issn = "1613-6810",

publisher = "wiley",

number = "37",

}

TY - JOUR

T1 - Fast Scanning Probe Microscopy via Machine Learning

T2 - Non-Rectangular Scans with Compressed Sensing and Gaussian Process Optimization

AU - Kelley, Kyle P.

AU - Ziatdinov, Maxim

AU - Collins, Liam

AU - Susner, Michael A.

AU - Vasudevan, Rama K.

AU - Balke, Nina

AU - Kalinin, Sergei V.

AU - Jesse, Stephen

PY - 2020/9/1

Y1 - 2020/9/1

N2 - Fast scanning probe microscopy enabled via machine learning allows for a broad range of nanoscale, temporally resolved physics to be uncovered. However, such examples for functional imaging are few in number. Here, using piezoresponse force microscopy (PFM) as a model application, a factor of 5.8 reduction in data collection using a combination of sparse spiral scanning with compressive sensing and Gaussian process regression reconstruction is demonstrated. It is found that even extremely sparse spiral scans offer strong reconstructions with less than 6% error for Gaussian process regression reconstructions. Further, the error associated with each reconstructive technique per reconstruction iteration is analyzed, finding the error is similar past ≈15 iterations, while at initial iterations Gaussian process regression outperforms compressive sensing. This study highlights the capabilities of reconstruction techniques when applied to sparse data, particularly sparse spiral PFM scans, with broad applications in scanning probe and electron microscopies.

AB - Fast scanning probe microscopy enabled via machine learning allows for a broad range of nanoscale, temporally resolved physics to be uncovered. However, such examples for functional imaging are few in number. Here, using piezoresponse force microscopy (PFM) as a model application, a factor of 5.8 reduction in data collection using a combination of sparse spiral scanning with compressive sensing and Gaussian process regression reconstruction is demonstrated. It is found that even extremely sparse spiral scans offer strong reconstructions with less than 6% error for Gaussian process regression reconstructions. Further, the error associated with each reconstructive technique per reconstruction iteration is analyzed, finding the error is similar past ≈15 iterations, while at initial iterations Gaussian process regression outperforms compressive sensing. This study highlights the capabilities of reconstruction techniques when applied to sparse data, particularly sparse spiral PFM scans, with broad applications in scanning probe and electron microscopies.

KW - Gaussian process regression

KW - atomic force microscopy

KW - compressive sensing

KW - ferroelectric heterostructures

KW - piezoresponse force microscopy

UR - http://www.scopus.com/inward/record.url?scp=85089252557&partnerID=8YFLogxK

U2 - 10.1002/smll.202002878

DO - 10.1002/smll.202002878

M3 - Article

C2 - 32780947

AN - SCOPUS:85089252557

SN - 1613-6810

VL - 16

JO - Small

JF - Small

IS - 37

M1 - 2002878

ER -

Fast Scanning Probe Microscopy via Machine Learning: Non-Rectangular Scans with Compressed Sensing and Gaussian Process Optimization

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this