1Advanced Microscopy Laboratory (CRANN), Trinity College Dublin, Dublin, Ireland
2School of Physics, Trinity College Dublin, Dublin, Ireland
3Department of Physics and Astronomy, University of Uppsala, Uppsala, Sweden
4EMAT, University of Antwerp, Antwerp, Belgium
5Department of Materials, University of Oxford, Oxford, UK
The scanning transmission electron microscope (STEM) yields many imaging and spectroscopic signals fully simultaneously and at very high-resolution. With the aberration corrected STEM now readily delivering atomic resolutions in these signals, the new frontier becomes sensitivity – the ability to detect more reliably and at lower imaging doses than before. The enemy of sensitivity is uncertainty, and in a scanned instrument this can come from spatial-uncertainty (scanning distortions); or from signal-uncertainty, the finite signal-to-noise which limits the data.
Previously we have shown how scanning distortions in annular dark-field (ADF) images can be compensated for by using non-rigid registration across a series of individual scans . Going further, by analytically expressing this scanning-uncertainty along with the Poisson-limited uncertainty  and with the probability of introducing sample-damage, the experiment design can be optimised by minimising the combined uncertainty a function of electron-dose.
Shifting to a multi-frame regime, in practice, introduces many new experiment design parameters and these will be discussed. If the imaging conditions of these scans is not changed, then this process intrinsically increases the total sample exposure many times over. We can instead reduce the dwell-time and maintain a fixed total-dose, or we can reduce the beam current and reduce also the dose and dose-rate. Where the inherent cross-section is small for spectroscopic mapping, then multi-frame approaches can be used to accumulate signal over many frames while allowing time for heat or charge to dissipate leading to an overall reduction in damage . However, left unchecked this spawns a new problem of unmanageable data-rates in the recording and approaches to mitigate this will be presented.
Lastly, we will present a new method to form ADF images using digital (electron counting) approach . This pushes down the noise-floor to zero, improves the ADF MTF and DQE and returns images natively calibrated in an absolute greyscale units of electrons.
 L. Jones et al., Adv. Struct. Chem. Imaging 1, 8 (2015).
 A. De Backer et al, Ultramicroscopy 151, 56 (2015).
 L. Jones et al., Microscopy 67, Supp. 1, (2018).
 R. Ishikawa et al., Microsc. Microanal. 20, 99 (2014).
Trinity College Dublin