2025MNRAS.538..426V

In this paper, we further explore the properties of the HERschel Ultra Luminous Infrared Galaxy Survey (HERUS) sample of 42 local ultraluminous infrared galaxies (ULIRGs) with our recently developed Bayesian spectral energy distribution (SED) fitting code SMART (Spectral energy distributions Markov chain Analysis with Radiative Transfer models). SMART fits SEDs exclusively with multicomponent radiative transfer models. Mid-infrared spectroscopy can be included in the fitting at a spectral resolution matched to that of the radiative transfer models. We fit the SEDs of the HERUS ULIRGs with four different models for the active galactic nucleus (AGN) torus, a starburst and a spheroidal galaxy model, to put constraints on the AGN fraction of the galaxies and their star formation rate (SFR). Two of the AGN torus models we explored are smooth and two are two-phase. We find that, in most cases, a smooth tapered AGN torus provides the best fit to the data. We also find that solutions with other torus models may predict AGN and total luminosities up to an order of magnitude or more lower, but very rarely higher than the best-fitting model. In contrast, we find that, with minor exceptions, the predicted SFR and stellar mass of the ULIRGs are generally robustly estimated irrespective of the assumed torus model. This is despite the fact that one of the AGN torus models we use assumes fluffy grains with high emissivity in the far-infrared and submillimetre, which could potentially reduce the contribution of a starburst at those wavelengths and reduce the SFR.