Modelling the spectral energy distribution of ULIRGs - II. The energetic environment and the dense interstellar medium

Aims. By using the spectral energy distribution (SED) from the near-infrared to the radio of a statistically significant number of luminous infrared galaxies we determine important physical parameters for this population of objects. In particular we constrain the optical depth towards the luminosity source, the star formation rate, the star formation efficiency and the AGN fraction. Methods. We fit the near-infrared to radio spectral energy distributions of a sample of 30 luminous and ultra-luminous infrared galaxies with pure starburst models or models that include both starburst and AGN components. Results. We find that although about half of our sample have best-fit models that include an AGN component, only 30% (9/30) have an AGN that accounts for more than 10% of the infrared luminosity from 8 to 1000 mu m, whereas all have an energetically dominant starburst. Our derived AGN fractions are generally in good agreement with measurements of the mid-infrared line ratios, Ne[V]/Ne[II] and O[IV]/Ne[II] by Spitzer IRS, but much lower than those derived from PAH equivalent widths or the mid-infrared spectral slope. Our models determine the mass of dense molecular gas within which active star formation takes place via the extinction required to reproduce the infrared part of the SED. Assuming that this mass is that traced by the HCN molecule, we reproduce the observed linear relation between HCN flux and infrared luminosity found previously. We also find that the star formation efficiency, as defined by the current star formation rate per unit molecular gas mass, falls as the starburst ages. Conclusions. If the evolution of ULIRGs includes a phase in which an AGN contributes an important fraction to the infrared luminosity, this phase should last an order of magnitude less time than the starburst phase. However, we find no convincing evidence that an energetically important AGN is associated with a particular phase of the starburst. Because the mass of dense molecular gas that we derive is consistent with observations of the HCN molecule, it should be possible to estimate the mass of dense, star-forming molecular gas in such objects when molecular line data are not available.