The starburst-AGN connection. II. The nature of luminous infrared galaxies as revealed by VLBI, VLA, infrared, and optical observations

We present here detailed results of an 18 cm VLBI survey of 31 luminous (L-FIR > 10(11.25) L.), radio-compact (theta less than or similar to 0." 25) infrared galaxies (LIGs). High-resolution VLA maps at 15 and 22 GHz are presented for 14 of these galaxies that exhibit compact milliarcsecond-scale emission, providing information about radio structure of LIGs on scales from 0." 004 to 1." 0. We also present new optical spectrophotometric observations. Over half the sample galaxies show high brightness temperature radio emission from the VLBI data, with T-b > 10(5) K and structure on scales of 5-150 mas, as previously reported in Lonsdale, Smith, & Lonsdale. The median VLBI power for detected sources is log P-VLBI = 22.0 (W Hz(-1)), and the mean ratio of VLBI to total 1.6 GHz flux density is [S-VLBI/S-total] = 0.12. Further structure is observed on the larger VLA scales. No highly significant (P < 1%) statistical correlations are found between the presence or strength of the VLBI emission and other observed quantities, including total radio power, radio spectral index, IR luminosity and colors, radio-infrared ratio, molecular gas mass, and optical excitation. Statistical analysis does suggest that the infrared luminosity, molecular gas emission, and radio emission on VLA and VLBI scales are physically related. Previous work (Lonsdale, Smith, & Lonsdale) demonstrated that hidden (dust-enshrouded) active galactic nuclei (AGNs) are capable of powering LIGs and giving rise to the observed VLBI- and VLA-scale structures; here we investigate the complementary question of whether a starburst can completely explain the observed characteristics, including the high brightness temperature radio emission. Simple starburst models show that the far-infrared luminosity can be explained by starbursts in all cases except Mrk 231, although for some objects the constraints imposed on the initial mass function are severe. Using our starburst models we model the VLBI data for 11 galaxies with detailed radio structural information using complexes of radio supernovae. The required supernova rates are nu(sn) similar to 0.1-2 yr(-1), consistent with the rates derived from the starburst model to explain the observed far-infrared luminosities. However in all cases we require complexes of extremely luminous radio supernovae (RSNs) to explain the high-T-b emission. In some cases the RSN must have implausibly high radio powers, more than an order of magnitude larger than any previously reported RSNs; in our view these sources represent AGN radio cores. In most cases an acceptable fit requires that the RSN be clustered on parsec scales. Furthermore, only a few clumps may be active in the radio at a given time. Based on this analysis we conclude that 7/11 systems can be plausibly explained as starbursts. Four galaxies, UGC 2369, Mrk 231, UGC 5101, and NGC 7469, almost certainly house AGN radio cores. From our modeling, coupled with other recent VLBI and infrared evidence, we conclude that Arp 220 is dominated by a massive starburst at radio and infrared wavelengths.