|Annu. Rev. Astron. Astrophys. 2000. 38: 761-814 |
Copyright © 2000 by Annual Reviews. All rights reserved
3.4.2. ULIRGs as a Class: The Starburst-AGN Connection
Lutz et al (1998a), Rigopoulou et al (1999a), and Tran et al (1999) extended this study to a larger sample by exploiting that the UIB (or PAH) strength criterion alone may be sufficient to distinguish between AGNs and starbursts (Figure 5a). The strong 6 to 11 µm UIB features can be detected in much fainter and more distant sources (up to z = 0.4 for the ISOCAM sample of Tran et al 2000). Note, however, that the UIB strength criterion rests heavily on the assumption that mid-IR features as well as continua are affected by the same extinction; otherwise, a more detailed discussion is necessary (see Sections 3.3.1, 3.4.3). The average spectrum of all 60 ULIRGs in the ISOPHOT-S sample is starburst-like (Figure 12, see color insert; Lutz et al 1998a). UIB strength and the ratio of mid-IR (5.9 µm) and far-IR (60 µm) flux densities are anticorrelated. "Warm" ULIRGs (S25 / S60 0.2, ~ 20% of the sample) are AGN dominated, and "cold" ULIRGs are star formation dominated, in agreement with Sanders et al (1988b). The data are consistent with a model in which the UIB/FIR ratio is constant and mid-infrared extinction is one to a few magnitudes.
Figure 12. Mid-IR SEDs of ULIRGs. Left: Average spectrum of 60 ULIRGs with z < 0.3 (lower inset), compared to average spectra of starburst templates (upper left) and AGNs (upper right). Individual spectra were shifted to the proper rest-wavelength and normalized to give all sources equal weight. The dotted curves in the upper insets show the impact of 50 magnitudes of extinction on the SED shape (from ISOPHOT-S data of Lutz et al 1998a). Right: UIB strength as a function of IR-luminosity of ULIRGs. In addition to the Lutz et al sample (see also Rigopoulou et al 1999), the figure contains 15ULIRGs with z< 0.4 from Tran et al (2000) and two hyperluminous ULIRGs from Taniguchi et al (1997b), Aussel et al (1998), all observed with the CVF of ISOCAM. In the text it is (arbitrarily) assumed that sources above the dotted line (UIB strength ~ 1) are dominated by star formation, and those below the line by AGNs.
The UIB (PAH) strength as a qualitative tracer of AGN versus starburst nature can be exploited further to study the AGN-starburst connection as a function of luminosity and merger evolution. Figure 12 (right panel) shows the UIB strength as a function of infrared luminosity, for all ULIRGs of the ISOPHOT-S and ISOCAM samples, plus the two hyperluminous (L ~ 1013 L) ULIRGs: FSC15307+3252 (z = 0.9; Aussel et al 1998) and P09104+4109 (z = 0.44; Taniguchi et al 1997b). The AGN fraction appears to increase with luminosity, from ~ 20% for log L = 11.97-12.3, to ~ 50% for log L = 12.3-13.2. Optical classifications lead to similar results. Kim et al (1998) classify as Seyfert 1/2s 26% of their ULIRGs for log LIR = 12-12.3, but 45% for log L = 12.3-12.9. The ISO and optical data thus indicate that starbursts have an upper limit of 1013 L, consistent with the luminosities of a gas-rich galaxy whose entire molecular gas mass is converted to radiation in a dynamical time and with the nucleosynthesis efficiency ~ 0.001-0.007.
Lutz et al (1999a) found for 48 ULIRGs with ISOPHOT-S/ISOCAM data and good quality optical spectra that the optical and the ISO classifications of the individual galaxies agree very well if optical HII-galaxies and LINERs are both identified as starbursts. All but one of 23 ISO star formation dominated ULIRGs are HII/LINER galaxies, and 11 of 16 ISO AGNs are Seyferts. All but one of the Seyfert 1/2s are ISO AGNs. Identification of infrared-selected, optical LINERs as starbursts is quite plausible. LINER spectra in ULIRGs are likely caused by shock excitation in large-scale superwinds, and not by circumnuclear gas photoionized by a central AGN (Heckman et al 1987, Armus et al 1990).
The evolutionary scenario proposed by Sanders et al (1988a) postulates that interaction and merging of the ULIRG parent galaxies triggers starburst activity that later subsides, while the AGN increasingly dominates the luminosity and expels the obscuring dust. An implication of this scenario is that advanced mergers should, on average, be more AGN-like than in earlier stages when the interacting galaxies are still well separated. However, Rigopoulou et al (1999a), Lutz et al (1998a) find that powerful starbursts are present at the smallest nuclear separations. There is no obvious trend for AGNs to dominate with decreasing separation. This suggests that local and short-term factors, in addition to the global state of the merger, play a role in determining which of the energy sources dominates the total energy output. Relevant processes include the accretion rate onto the central massive black hole(s), the radiation efficiency of the accretion flow, and the compression of the interstellar gas on scales of a few tens of parsecs or less.