IMI Interdisciplinary Mathematics InstituteCollege of Arts and Sciences

Seeing Atoms on the Crossroads of Microscopy and Mathematics

  • Feb. 17, 2009
  • 3:30 p.m.
  • Sumwalt 102

Abstract

Aberration correction has revolutionized scanning transmission electron microscopy (STEM), for the first time allowing direct imaging of sub-angstrom atomic spacings. Spectroscopic sensitivity for electron energy loss spectra (EELS) has improved to the point when it’s possible to detect just one atom inside the material. Aberration correction also enabled 3D sensitivity by STEM, resulting in vertical resolution of a few nanometers in addition to sub- Ångström lateral resolution. New types of data call for improved methods of data analysis and statistical treatment. This seminar gives a few examples of how image analysis helps us convert images and spectra into information about material properties. When atomic configurations can be determined with precision of just a few tens of pm, the structural data can give us a direct link to the chemical and physical properties. We are now able to deduce coordination environment and acidity of surface species on catalysts and map ferroelectric polarization on a unit cell-level. EELS analysis of thin films and multilayers gives spatially resolved information on composition and electronic properties across the interfaces. By employing computational tools such as principal component analysis and artificial neural networks, we can get new insights into the interfacial phenomena. Depth sectioning in STEM is a promising method for 3D structure determination in materials, making it possible to locate quantities down to single atoms in three dimensions. However, while for single atom images in amorphous media the observed depth of field is quite close to theoretical (few nm), it can be considerably larger for finite-size particles due to channeling effects as well as “missing wedge” effect known from tomography. The problem can be mitigated by acquiring several focal series datasets at different tilt angles. To process this data, a multistep approach involving registration, deconvolution, and cross-correlation in three dimensions is required. Computational problems related to data analysis anticipated in the future will also be discussed.

© Interdisciplinary Mathematics Institute | The University of South Carolina Board of Trustees | Webmaster
USC