Annu. Rev. Astron. Astrophys. 1991. 29: 581-625 Copyright © 1991 by Annual Reviews. All rights reserved

9.3 Ultraluminous Infrared Galaxies

Perhaps most dramatic in starburst activity are the ultraluminous (L_IR > 10¹² L) galaxies highlighted by the IRAS survey. At the highest luminosities, one sees a preponderance of galaxies with double nuclei and/or extended tidal tails indicative of galactic interactions or the merging of two galaxies (Sanders et al 1988c). The optical spectra of the ultraluminous galaxies are dominated by emission line ratios characteristic of a narrow line AGN or Seyfert nucleus rather than the thermal, H II region, line ratios seen in lower luminosity galaxies. This qualitative assessment of the optical data strongly suggests that the highest luminosities are initiated by galactic collisions, and the energy is supplied probably by the combination of an extremely energetic starburst and an embedded, nonthermal AGN.

Virtually all of the luminous IRAS galaxies have been shown to be extremely rich in interstellar gas, predominantly molecular hydrogen (see Figure 9). This gas is also highly concentrated in the nuclei of the galaxies. Over the last four years, aperture synthesis maps have been made of the CO emission in approximately 25 of the luminous and ultraluminous infrared galaxies (Scoville et al 1986b, 1989, 1991; Sargent et al 1987, Sargent & Scoville 1991, Sanders et al 1988a, Meixner et al 1990, Ishizuki et al 1990). The results for 18 of these systems are summarized in Scoville et al (1991). In most cases the size of the CO-emitting region is 1 kpc in radius, and the masses of molecular gas in the nuclear sources are 10⁹-4 x 10¹⁰ M (assuming the Galactic CO-to-H₂ conversion factor). The galaxies span the luminosity range 2 x 10¹⁰ L to 3 x 10¹² L (for = 8 - 1000 µm), and the global infrared luminosity-to-total-molecular mass ratios range from 4 L / M, like the Milky Way, to 200 L / M for Mrk 231.

Scoville et al (1991) computed the gas mass fraction (M_H2 / M_dyn) for those galaxies in which the nuclear mass concentration is resolved and for which the velocity dispersion can be estimated from the CO line-width. For the six cases in which this ratio has been evaluated, the interstellar gas constitutes a significant fraction of the total mass in the nucleus. This is consistent with numerical simulations of merging galaxies (Hernquist 1989) including a gas component. The simulations show that the gas sinks to the center of the merged system more readily than the stellar component because gas is much more dissipative than the stars.

Virtually all of the very luminous infrared galaxies show evidence of a significant galactic interaction (cf. Sanders 1990). The interstellar matter can play a central role in the dynamics of a galactic interaction, since the gas is dissipative and hence responds irreversibly to the perturbation, sinking toward the center of the potential well. The H₂ masses for the luminous infrared galaxies are large but, in most cases, not more than would be found in two galaxies such as M51 or M83 (M_H2 ~ (1-3) x 10¹⁰ M). In most cases, it is therefore plausible that the luminous infrared galaxies result from the merging of two gas-rich spiral galaxies, and that the disturbed dynamics in the merging system results in dissipation of kinetic energy (and outward transport of angular momentum) in the ISM, thus leading to the deposition of a significant fraction of the original interstellar matter in the central region of the system. There, binary processes such as cloud-cloud collisions or stimulated star formation can enhance the overall efficiency and rate of conversion of interstellar gas into young stars.

Arp 220, the prototype ultraluminous infrared galaxy, has an infrared luminosity at = 8-1000 µm of 1.5 x 10¹² L, exceeding that in the visual by nearly 2 orders of magnitude and placing it in the luminosity regime of quasars. Single dish measurements show the CO emission extending over a velocity range of 900 km s^-1, and the derived H₂ mass is 3.5 x 10¹⁰ M, approximately a factor of 15 greater than that of the Galaxy (Solomon et al 1990). At optical wavelengths, the galaxy appears approximately spherical with a central dust lane and tidal tails, both characteristics of galactic merging, extending up to 70 kpc away (Sanders et al 1988c). The CO emission, observed at 2" resolution (Scoville et al 1991), can be separated into two components: a core 1.4 x 1.9" and an extended component 7 x 15" containing 2/3 and 1/3 of the flux, respectively. For the adopted distance of 77 Mpc (for H₀ = 75 km s^-1 Mpc^-1) the H₂ masses in the core and extended components are 1.8 x 10¹⁰ M and 9 x 10⁹ M, respectively. The mean diameter of the core component (~ 1.7") corresponds to a radius of 315 pc, and the H₂ density, smoothed out over the volume, is 2900 cm^-3. Solomon et al (1990) deduced a higher density (10⁴-10⁵ cm^-3) based on CS measurements. Within the core source of radius 315 pc, the dynamical mass is 2.5 x 10¹⁰ M, i.e. almost precisely equal to the total gas mass (H₂ + He) derived from the CO line flux for the core component.