Abstract text (incl. figure legends and references)

Energy-dispersive X-ray spectroscopy (EDX) maps are usually acquired by scanning with a regular raster of scan positions, where at every scan point the characteristic X-ray emission caused by the electron beam in the sample is detected. In this way, a three-dimensional dataset with two spatial dimensions (scan raster) and one energy dimension (one spectrum for each scan point) is created. After the acquisition, the spectrums can be evaluated and composition information about the specimen can be extracted by, e.g., the Cliff-Lorimer-method. This methodology of evaluation is well-established and largely automated. With modern EDX-detectors that cover an extremely large solid angle and have a high detective efficiency, atomic resolution of the composition maps can be achieved.

However, one of the most critical limiting factors for such EDX map acquisitions is beam damage. Because the X-ray gain per electron dose is relatively small, an electron dose that is magnitudes larger than required for, e.g., a HAADF-STEM image is needed. This can be a problem for beam-sensitive samples, which either deteriorate under the electron beam or have to be prepared in such a way, that a strong contamination occurs. Here, the critical dose, where the resulting images do not longer represent the original sample, is often reached before a sufficient X-ray signal has been acquired. As a result, the spectra are not evaluable at an atomic resolution.

While this dose problem generally is of fundamental nature and cannot completely be avoided, this work proposes a scheme to alleviate the issue. It is based on the observation that the respective excitation cross sections are relatively small and centered around the atoms: For a scan conducted with a sample oriented in a crystallographic direction, the EDX signal from beam positions on atomic columns is significantly stronger than from interstitial positions, if all scan positions receive the same dose – as is the case during a regular scan. The signal retrieved per dose is therefore advantageous directly on the atomic columns.

Given these deliberations, a scanning scheme, where the beam is quickly shifted from column to column, thereby spending little time in the interstitial area, promises an elevated EDX signal for the same specimen dose compared to a regular scan of the same time. Consequently, it should be possible to acquire better spectra per atomic column before the specimen deteriorates.

Quantitative inelastic multislice simulations are used to study the achievable EDX signal improvement for different electron probes and materials. Because any experimental realisation will not be able to target the center of each column flawlessly, the impact the positioning and atom column detection precision is investigated. Finally, possible experimental and instrumental realisations of the proposed technique are discussed.