Medidas de linhas de emissão com aprendizado de máquina

Abstract

Electronic transitions in nebula ions present in galaxies emit photons with characteristic energies, emerging from
the spectra of galaxies as emission lines. Measurements of emission line fluxes are fundamental in understanding
this physical system. For example, the equivalent width of the nebular spectral line Hα is linked to the specific
star formation rate of a galaxy and is also useful for quantifying the presence of ionized diffuse gas in galaxies.
Given that in astrophysics we work with cubes that contain thousands of spectra per galaxy, it is important
to evaluate the method used to extract the physical properties of the data, given the great computational
demand involved. To this end, the proposal of this work consisted of applying a convolutional neural network to
measure amplitude and flux obtained from Hα emission lines generated from a Gaussian function. The results
are promising and the learning is efficient. The perspective is to apply the method to data from MUSE (Multi
Unit Spectroscopic Explorer).