Machine Learning for EM

Machine learning for EELS data interpretation. Data analysis in EELS is powered by a dedicated Python machine learning framework, EELSfitter, developed in-house within our group. It makes it possible to access key properties of 2D nanomaterials such as modulation of the bandgap, dielectric function, dispersion relations, low-energy electronic states, excitons, and other collective excitations, and their direct correlation with structural features. This effort is carried out in collaboration with experts from Nikhef, the Dutch national institute for particle physics. For the complete list of publications from our lab please check here.

A toy model for the EEL spectrum and its derivative (in the inset)

Here we develop a machine learning method, inspired by particle physics techniques, to parametrize and subtract the zero-loss peak in electron energy-loss spectroscopy measurements. As an application, we determine the local bandgap of 2H/3R polytypic WS₂.

“Charting the low-loss region in electron energy loss spectroscopy with machine learning”, L. I. Roest, S. E. van Heijst, L. Maduro, J. Rojo, S. Conesa-Boj, Ultramicroscopy (2021) 222, 113202-113219

Schematic representation of the output of EELSfitter with a 2D map of the bandgap energy extracted from WS₂ nanostructures

Here we present a novel strategy based on machine learning techniques making possible the automated processing and interpretation of spectral images from electron energy loss spectroscopy for the determination of the bandgap and complex dielectric function in 2D materials achieving a spatial resolution down to a few nanometers.

“Spatially-resolved bandgap and dielectric function in 2D materials from Electron Energy Loss Spectroscopy“, A. Brokkelkamp, J. ter Hoeve, I. Postmes, S. E. van Heijst, L. Maduro, A. V. Davydov, S. Krylyuk, J. Rojo, and S. Conesa Boj (2021), under review.