Constraining Data Mining with Physical Models: Voltage- and Oxygen Pressure-Dependent Transport in Multiferroic Nanostructures

Evgheni Strelcov, Alexei Belianinov, Ying Hui Hsieh, Ying-hao Chu, Sergei V. Kalinin*

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

22 Scopus citations


Development of new generation electronic devices necessitates understanding and controlling the electronic transport in ferroic, magnetic, and optical materials, which is hampered by two factors. First, the complications of working at the nanoscale, where interfaces, grain boundaries, defects, and so forth, dictate the macroscopic characteristics. Second, the convolution of the response signals stemming from the fact that several physical processes may be activated simultaneously. Here, we present a method of solving these challenges via a combination of atomic force microscopy and data mining analysis techniques. Rational selection of the latter allows application of physical constraints and enables direct interpretation of the statistically significant behaviors in the framework of the chosen physical model, thus distilling physical meaning out of raw data. We demonstrate our approach with an example of deconvolution of complex transport behavior in a bismuth ferrite-cobalt ferrite nanocomposite in ambient and ultrahigh vacuum environments. Measured signal is apportioned into four electronic transport patterns, showing different dependence on partial oxygen and water vapor pressure. These patterns are described in terms of Ohmic conductance and Schottky emission models in the light of surface electrochemistry. Furthermore, deep data analysis allows extraction of local dopant concentrations and barrier heights empowering our understanding of the underlying dynamic mechanisms of resistive switching.

Original languageEnglish
Pages (from-to)6650-6657
Number of pages8
JournalNano Letters
Issue number10
StatePublished - 14 Oct 2015


  • Bayesian linear unmixing
  • Bismuth ferrite
  • cobalt ferrite
  • multivariate analysis
  • oxide heterostructures

Fingerprint Dive into the research topics of 'Constraining Data Mining with Physical Models: Voltage- and Oxygen Pressure-Dependent Transport in Multiferroic Nanostructures'. Together they form a unique fingerprint.

Cite this