Towards a consistent framework of determining the active galactic nucleus contribution fraction and host galaxy properties

Abstract

Decomposing the contributions of active galactic nuclei (AGNs) from their host galaxies is essential for identifying AGN-dominated galaxies and accurately deriving key physical properties of both supermassive black holes (SMBHs) and their host galaxies, such as black hole accretion rates, stellar masses, and star formation rates. However, decomposing AGN contributions from multi-wavelength photometry remains challenging due to inherent parameter degeneracies in spectral energy distribution (SED) fitting. We establish a unified framework for estimating AGN contribution fractions (f_ AGN ) and host galaxy properties by combining complementary AGN diagnostics: SED decomposition from two independent SED-fitting codes (CIGALE and GRAHSP) and deep learning (DL)-based imaging decomposition. We applied SED decomposition to galaxies in the COSMOS-Web field using multi-wavelength photometry (ultraviolet to far-infrared) and calculated the AGN-to-host galaxy flux ratio (f_ AGN ^ SED ) in the JWST/NIRCam F150W filter. We then compared these SED-fitting derived AGN fractions with those obtained from image decomposition using DL (f_ AGN ^ DL ). Strong agreement is found for AGN-negligible galaxies across all methods: ∼82% are identified as such by both SED codes (f_ AGN ^ SED <0.1), and ∼87% of these also show minimal AGN contribution in the DL model. This high concordance demonstrates that current methods reliably identify pure galaxy systems in most cases. However, for potential AGN-dominant galaxies, large discrepancies emerge. Only 2.7% of galaxies show consistent significant AGN identification between SED codes, and merely ∼7.6% of these also exhibit a significant AGN fraction in DL estimation. Forcing DL AGN fraction constraints into SED fitting generally works for SED-significant but imaging-negligible cases, whereas it fails for around 30% of SED-negligible but imaging-significant cases. In summary, our results highlight the degeneracies inherent in current SED-fitting methods based on empirical observations or theoretical templates and demonstrate the power of incorporating independent morphological information to break these degeneracies. We present a more robust framework combining morphological information from high-fidelity imaging from surveys such as the James Webb Space Telescope (JWST) and , enabling improved AGN–host galaxy decomposition and more reliable measurements of both AGN and galaxy properties. Euclid