Correlative spectroscopic electron tomography

The combination of spectroscopy and tomography in the scanning transmission electron microscope (STEM)allows nanoscale 3D elemental mapping by energy-dispersive X-ray spectroscopy (EDXS) or electron energy-loss spectroscopy (EELS) [1,2], and EELS can additionally image chemical bonding information [3], surfaceplasmon [4,5] or phonon polaritons [6] in three dimensions. However, spectroscopic tomography tilt series aregenerally affected by noise and can be only acquired for a limited number and size of projections due to timeand dose constraints. In addition, especially for EELS the maximum sample diameter is limited in order toavoid plural scattering. Therefore, in order to get most out of the data, it is ideal to combine and correlate allavailable information.
One option for this purpose is the usage of regularized reconstruction algorithms, using joint constraints in the3D reconstruction of multiple datasets. Using the joint total generalized variation (TGV) reconstruction method,common edges and gradual transitions between different regions in all reconstructions are favored [7]. Thisimproves the visibility of sample features while reducing noise (Fig. 1). This method can be employed tocorrelate elemental maps from a single spectroscopic technique, but can also add other imaging modes suchas annular dark-field (ADF) imaging, or correlate multiple spectroscopic datasets (Fig. 2).
Potentially datasets can be also jointly reconstructed, which are acquired at a different resolution, over adifferent field of view or also locally within a larger area either by local tomography or before and after a finalsample preparation step. Spectroscopy can be done locally and for EELS also on a thinned sample, while aconventional tomography series can be acquired for a larger volume in order to extrapolate the localinformation to these larger regions.