University of Wisconsin–Madison

STEM Imaging

To support our research in crystalline defects and interfaces, we seek to push the boundaries in STEM imaging and microanalysis.  Some of the efforts in this area are dedicated projects, but everyone in the group contributes to to this research area as the problems arise and get solved.

(a) A single frame of GaN (b) the average of 512 images after non-rigid registration, the distribution of interatomic distances (c) along the scan direction and (d) perpendicular to it, both showing sub-pm precision.

One effort in this area is improving the precision in STEM images.  In imaging, precision quantifies our ability to determine the exact position of a column of atoms, once it has been resolved from its neighboring columns.  (Precision is different from accuracy.  So far, we have address precision only.)  Achieving high precision in STEM is difficult because the serial acquisition of the images means that instability in the probe, sample, or microscope shifts the apparent position of the atoms in the images.  We overcome this problem by acquiring hundreds of images of the same region of the sample.  The instabilities are random with zero mean, so they average out of the image series.  In collaboration with some applied mathematicians, we have developed a non-rigid registration scheme that allows averaging without loss of resolution.  The average image produced by this technique have reliably sub-picometer precision, the highest ever achieved in STEM.

Precision is important because it allows us to measure distortions of the crystal lattice associated with defects and interfaces. Picometer precision is easily sufficient to measure the strain fields from dislocations, many interfaces, and even polarizations in ferroelectric materials.  It may be good enough to detect the strain from point defects or point defect clusters.  We are also applying high precision imaging to measuring the distortions associated with surfaces, edges, and corners on nanoparticle catalysts.

Simulation of the propagation of a STEM probe through Si 110. The black dots mark the positions of Si columns. The probe starts on the left column, but eventually couples to the right column as a function of thickness.

Averaging after non-rigid registration also creates images with exceptionally good signal-to-noise.  Good signal to noise open the door to precise quantification of the image intensity through comparison to STEM image simulations.  Careful quantification requires knowing the experimental conditions, which poses its own set of challenges, but it may enable the detection of vacancies and anti-site defects purely from the image intensity.

Our other area of research associated with STEM imaging is working to understand the interaction of the probe beam of electrons with the crystal lattice and how that interaction results in the image contrast.  The probe-sample interaction is a complicated dynamical diffraction phenomena, so we address it primarily through simulations.  Simulations show, for example, that as the probe propagates, it does not necessarily stay on the atomic column where it was placed at the entrance surface of the sample.  Picometer-precision imaging makes it even more important to unwravel the physics connecting the STEM image to the underlying crystal structure.

For available positions in this area, please check the openings page.

Publications