Andrew Lupini1 2 Ondrej Dyck3 2 Xin Li3 2 Stephen Jesse3 2 Bethany Hudak1 2 Mark Oxley1 2 Sergei Kalinin3 2

1, Oak Ridge National Laboratory, Oak Ridge, Tennessee, United States
2, Oak Ridge National Laboratory, Oak Ridge, Tennessee, United States
3, Oak Ridge National Laboratory, Oak Ridge, Tennessee, United States

Scanning transmission electron microscopy (STEM) has traditionally involved recording a single, or very few, integrated signals as an electron probe is scanned across a sample. Advances in detector technology now allow the entire scattering distribution to be recorded, resulting in a four-dimensional (4D) dataset that includes two real-space dimensions and two reciprocal-space dimensions for every probe position. New fast cameras recently installed on low-voltage and monochromated aberration-corrected STEMs have greatly simplified the acquisition of these datasets. Each convergent beam diffraction pattern (CBED or Ronchigram) can contain as many pixels as a traditional TEM image, making these datasets extremely large. Thus the problem of how to extract the useful information remains a challenge.

The simplest approach is to mask regions of the data, simulating traditional STEM detectors. More advanced ptychographic methods allow details of the probe aberrations to be extracted, and potentially corrected, to improve the contrast or detectability of light atoms. Machine learning or deep data approaches will allow the local symmetry or changes in structure to be automatically identified and perhaps tracked in real-time.

Free-standing 2D materials, such as graphene, form an ideal test suite for new imaging modalities because every atom in the structure can be directly imaged. Moreover, they have some advantages over 3D samples, because unknown factors, such as background contributions from contamination or sample preparation and surface reconstructions, can be either imaged directly or ruled out. Two dimensional samples are also particularly amenable to simulations, since they contain a well-defined number of atoms. These patterns are scientifically interesting because they are recorded in diffraction-space, with a near atomic-sized probe, effectively generating diffraction from a single atom. Applications to the imaging and control of single dopant atoms will be discussed.

Work supported in part by US Department of Energy, Office of Science, Basic Energy Sciences, Materials Sciences and Engineering Division and in part by the Laboratory Directed Research and Development Program of Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U. S. Department of Energy.