paper

Robust AGN and host-galaxy decomposition in optical spectral fitting

arXiv:2604.27783 · doi:10.1051/0004-6361/202558027

Abstract

Unraveling the growth of supermassive black holes and their connection to host galaxies requires disentangling the Active Galactic Nuclei (AGN) emission from that of the stellar populations. When an AGN spectrum is observed at different activity phases, if the spectral decomposition properly recognizes the nuclear and stellar components, key physical properties - such as black-hole mass, stellar mass, and stellar velocity dispersion - should remain consistent. We present a novel optical spectral-fitting approach that combines pPXF and PyQSOFit to robustly decompose spectra into stellar and AGN components. We apply this technique to three SDSS samples with repeated optical spectra of the same objects at z<0.55: 32 changing-look AGN in bright and dim states, and 15 quasars and 15 galaxies with three single-epoch and one stacked spectrum each. To compare with the literature, we use SDSS spectra and photometric data from AGN in the eFEDS field, as well as Gemini and VLT observations of some of our changing-look AGN. We evaluate the reliability of stellar mass, velocity dispersion, and black-hole mass measurements, especially in relation to the AGN-to-total continuum contribution (fAGN). For host-derived properties, especially when fAGN<0.8, our method yields consistent results. For single-epoch black-hole mass estimates from Ha and Hb, 3-sigma confidence in the broad-line flux and FWHM provides effective criteria for selecting reliable measurements. After applying these quality cuts, measurements across different epochs agree within uncertainties, and their reliability is confirmed by the alignment with previously established scaling relations. Many changing-look AGN in our sample exhibit "breathing" broad-line regions, as determined from Ha analysis, while some deviate significantly, suggesting non-virialized systems across the spectral transition.

21 pages, 20 figures, accepted by A&A

References in corpus (62)