hide
Free keywords:
Physics, Data Analysis, Statistics and Probability, physics.data-an, Condensed Matter, Materials Science, cond-mat.mtrl-sci, Physics, Computational Physics, physics.comp-ph
Abstract:
The electronic band structure (BS) of solid state materials imprints the
multidimensional and multi-valued functional relations between energy and
momenta of periodically confined electrons. Photoemission spectroscopy is a
powerful tool for its comprehensive characterization. A common task in
photoemission band mapping is to recover the underlying quasiparticle
dispersion, which we call band structure reconstruction. Traditional methods
often focus on specific regions of interests yet require extensive human
oversight. To cope with the growing size and scale of photoemission data, we
develop a generic machine-learning approach leveraging the information within
electronic structure calculations for this task. We demonstrate its capability
by reconstructing all fourteen valence bands of tungsten diselenide and
validate the accuracy on various synthetic data. The reconstruction uncovers
previously inaccessible momentum-space structural information on both global
and local scales in conjunction with theory, while realizing a path towards
integrating band mapping data into materials science databases.
