Starburst and AGN activity in ultraluminous infrared galaxies

We examine the power source of 41 local ultraluminous infrared galaxies (ULIRGs) using archival infrared (IR) and optical photometry. We fit the observed spectral energy distributions with starburst and active galactic nucleus (AGN) components, each component being drawn from a family of templates. We find that all of the sample require a starburst, whereas only half require an AGN. In 90 per cent of the sample the starburst provides over half the IR emission, with a mean fractional luminosity of 82 per cent. When combined with other galaxy samples we find that starburst and AGN luminosities correlate over six decades in IR luminosity, suggesting that a common factor governs both luminosities, plausibly the gas masses in the nuclear regions. We find no trend for increasing fractional AGN luminosity with increasing total luminosity, contrary to previous claims. We find that the mid-IR F (7.7) /C (7.7) line-continuum ratio is no indication of the starburst luminosity, or the fractional AGN luminosity, and therefore that F (7.7) /C (7.7) is not a reliable diagnostic of the power source in ULIRGs. The radio flux correlates with the starburst luminosity, but shows no correlation with the AGN luminosity, in line with previous results. We propose that the scatter in this correlation is due to a skewed starburst initial mass function and/or relic relativistic electrons from a previous starburst, rather than contamination from an obscured AGN. We show that most ULIRGs undergo multiple starbursts during their lifetime, and by inference that mergers between more than two galaxies may be common amongst ULIRGs. Our results support the evolutionary model for ULIRGs proposed by Farrah et al., where they can follow many different evolutionary paths of starburst and AGN activity in transforming merging spiral galaxies into elliptical galaxies, but that most do not go through an optical quasi-stellar object phase. The lower level of AGN activity in our local sample compared with z similar to 1 hyperluminous infrared galaxies implies that the two samples are distinct populations. We postulate that different galaxy formation processes at high z are responsible for this difference.