New Opportunities in multi‐frame STEM Spectroscopy & Fractional Beam‐current EELS

Abstract

Electron energy‐loss spectroscopy (EELS) and energy‐dispersive x‐ray spectroscopy (EDX) are two of the most common means of chemical analysis in the scanning transmission electron microscope (STEM). While the instrumentation hardware has progressed markedly in recent years, the way that microscopists operate these spectrometers has changed very little. In general small areas are scanned just using coarse pixilation, slow scan‐speeds and high beam‐currents. Finally, once acquired these chemical signals are usually expressed in only relative or arbitrary units. This stands in stark contrast with modern best‐practice in STEM imaging, where wider fields of view are surveyed utilising multi‐frame acquisitions, faster scan speeds, finer pixel sampling and lower electron doses. More recently it has also become common place to express STEM image data in units of fractional‐beam‐current facilitating direct comparison with simulation. We will present EELS and EDX results where best‐practice techniques from STEM imaging have been repurposed to improve chemical map quality. Multi‐frame spectrum‐image data were recorded with simultaneous EELS and EDX spectra, before using non‐rigid data registration [1] in the spatial domain, Figure 1. This not only reduces scan‐distortion but also unlocks tools such as; digital super‐resolution, strain mapping performed directly on chemical maps, and the digital accumulation of weak signals such as those from monochromated EELS. We also present an ‘equal fixed‐dose’ fractionation study where sample damage was reduced drastically using a fast‐scanning multi frame approach compared with its single scan equivalent. EELS spectrum data fidelity was improved by energy‐drift tracking and correction [2,3]. After this, the limiting performance in multi‐frame spectra becomes the spectral‐noise and was found to depend strongly on the quality of the available dark‐reference, Figure 2. To mitigate this artefact, separate dark‐references and full‐beam gain‐references were recorded along with wholly un‐processed experimental spectra. Further improvement in the noise performance was released by stepping the EELS spectra in energy between each scan before un‐stepping them in post‐processing [4]. Dark‐correction was performed offline before dividing by the gain‐reference to finally express the EELS spectra in terms of fractional beam current [5]. It is expected that leveraging all these individually small improvements collectively will deliver more precise chemical maps while minimising sample damage and experimental time overheads.