SerialEM supports STEM imaging on JEOL microscopes through a Gatan DigiScan unit, and on FEI microscopes through either an FEI STEM unit or a DigiScan unit, although the latter has not been tested on an actual FEI scope. The following features are noteworthy, and the links are to help sections with more explanation and detail:
Some features do not work in STEM mode: beam autocentering, specimen cooking, assessing angular ranges, and checking autofocus.
The configuration and calibration of SerialEM for STEM mode is described in Setting Up STEM.
Autofocusing is done by analyzing the power spectra of a series of Focus images. It proceeds in two phases. In the first phase, the program does a simple analysis of the total power over a certain spatial frequency range, after subtracting a background power level. It changes focus in whatever direction increases power over that at the current defocus, and moves in relatively coarse steps until it reaches a focus where power falls off again. It then takes images at finer focus steps so that there are at least two images on each side of the image with the best focus. These five images allow the second phase of analysis, in which it derives a measure of beam size for each of the 4 adjacent images by comparing the shape of its power spectrum with that of the best focused image. This measure varies linearly with distance from focus, so it is possible to fit a pair of lines to the measured sizes and take focus as the intersection between the two lines. Because of this interpolation method, it is not necessary to search for and take an image at exact focus.
This method requires that there be enough features in the Focus images to give reliable power spectra for the analysis. The situation is more challenging than for autofocusing in TEM, where a few high-contrast particles on a featureless background can give an adequate cross-correlation for determining beam-tilt-induced image shift. It is definitely necessary to pay attention to whether the focus area has enough image detail, and it may be necessary to use larger images (center half instead of center quarter for FEI STEM, or with the magnification increased by 2 instead of 4 for DigiScan). Another limitation of the analysis is that the total power measure becomes insensitive for very blurred images, beyond ~5 microns of defocus. The first phase of the search may thus fail to find a peak in the power within a reasonable number of steps. When this occurs during tasks, tilt series, or scripts, the program stops with an error message. When it occurs after autofocus was started manually, you are given the choice of whether to stop (without modifying defocus) or to continue searching from the last defocus tested.
The development of these STEM capabilities was partially supported by contributions from owners of JEOL microscopes at the University of Texas, Dallas and the Naval Research Laboratory in Washington, D.C., and owners of an FEI microscope at the Weizmann Institute of Science in Rehovot, Israel. Assistance and access to STEM-equipped microscopes was provided by Chen Xu at Brandeis University, David Morgan at the University of Indiana, and Tomas Molina at the National Center for Microscopy and Imaging Research.