ARlogo Annu. Rev. Astron. Astrophys. 2000. 38: 761-814
Copyright © 2000 by Annual Reviews. All rights reserved

Next Contents Previous

3.4.3. Hidden AGNs: The Role of Dust Obscuration in ULIRGs

The active regions of ULIRGs are veiled by thick layers of dust. Fine structure and recombination line ratios imply equivalent "screen" dust extinctions in ULIRGs (and LIRGs) between AV ~ 5 and 50 (Genzel et al 1998). ISOPHOT-S and ground-based data confirm this evidence for high dust extinction (Lutz et al 1998a, Dudley & Wynn-Williams 1997, Dudley 1999). The most extreme case is Arp 220, where the ISO SWS data indicate AV(screen) ~ 50 (A25µm ~ 1), or an equivalent mixed model extinction of 500 to 1000 mag in the V-band 8 (Sturm et al 1996). Smith et al (1989), Soifer et al (1999) found similar values from the depth of the silicate feature and from mid-IR imaging. These large extinctions combined with a mixed extinction model solve the puzzle that starburst models based on near-IR/optical data cannot account for the far-IR luminosities (AV leq 10; Armus et al 1995, Goldader et al 1997a, b). If the emission line data are instead corrected for the (much larger) ISO-derived extinctions, the derived LIR / LLyc ratios are in reasonable agreement with starburst models (Section 3.4.4).

Submm/mm CO and dust observations imply yet larger column densities than the mid-IR data (geq 1024 cm-2, or AV geq 500; Rigopoulou et al 1996b, Solomon et al 1997, Scoville et al 1997, Downes and Solomon 1998). Is it possible, therefore, that most ULIRGs contain powerful central AGNs that are missed by the mid-IR data? Hard X-rays penetrate to column densities geq1024 cm-2. ASCA has observed a small sample of ULIRGs in the 2-10 keV band. About a dozen sources are common between ISO and ASCA (Brandt et al 1997, Kii et al 1997, Nakagawa et al 2000, Misaki et al 1999). In Mrk 273, 05189-2524, NGC 6240, and 230605+0505, ASCA finds evidence for a hard X-ray source with < LX / LIR > geq 10-3. BeppoSAX observations show that the AGN in NGC 6240 is attenuated by a Compton thick (NH ~ 2 × 1024 cm-2) absorber (Vignati et al 1999). After correction for this absorption, and depending on its filling factor, the ratio of intrinsic AGN X-ray to IR luminosity is 2 to 6 × 10-2. In Mrk 231, a hard X-ray source is seen but is weak (< LX / LIR > ~ 10-3.5). ISO finds evidence for significant AGN activity in all of these sources as well. In sources classified by ISO as starburst dominated (Arp 220, UGC 5101, 17208-0014, 20551-4250, 23128-5919), ASCA also finds no hard X-ray source. The limit to the hard X-ray emission in Arp 220 corresponds to leq 10-4 of the infrared luminosity. For comparison, in Seyfert 1 and Seyfert 2 galaxies < LX / LIR > is 10-1 and 10-2, respectively (Boller et al 1997, Awaki et al 1991). For radio-quiet QSOs, the sample averaged SEDs of Sanders et al (1989), Elvis et al (1994), giving LX / LIR = 0.2 and LX / Lbol ~ 0.05. Although the statistics are still relatively poor at this point, the hard X-ray data do not present evidence for powerful AGNs that are completely missed by the mid-IR observations. Still, there are exceptions to this reasonable agreement between IR and X-ray data. The nearby galaxy NGC 4945 fulfills all criteria of a pure starburst at optical to mid-IR wavelengths (Moorwood et al 1996b; in Figure 5a, NGC 4945 is the starburst to the bottom right of Arp 220 and top left of M82). There is no evidence for a narrow line region or any other AGN indicator at these wavelengths. The [NeIII] / [NeII] line ratio is small and indicates that the starburst is aging (Spoon, private communication). Yet ASCA and BeppoSAX data show that at its center lurks a powerful AGN, attenuated by a Compton thick foreground absorber (Iwasawa et al 1993). As in NGC 6240, both the AGN (from the X-ray data) as well as the starburst (from the optical to IR data) can account for the entire bolometric luminosity of NGC 4945. Although NGC 4945 is much less luminous than a ULIRG, the case is puzzling and requires further study.

If mid-infrared continuum and UIB features suffer different obscurations, as in Seyfert 2 galaxies (Section 3.3.1; Clavel et al 1998), the UIB strength criterion5 loses its meaning. Instead it is necessary to directly compare the ratio of UIB luminosity to total far-IR (60 + 100 µm) luminosity, even if this ratio depends on mid-infrared extinction. The UIB/FIR ratio in ULIRGs (Figure 10) is on average half of that in starbursts (and Seyferts). Following the discussion in Section 3.3.2, this suggests that at least half of the luminosity in the average ULIRG comes from star formation if the same UIB/FIR ratio holds as in other galaxies. Correction for extinction increases this fraction. If the average mid-IR extinction is AV(screen) ~ 15 and A7.7 / AV ~ 0.04, the UIB/FIR ratio is fully consistent with that in starburst galaxies.

Hence, despite the large extinctions present in most ULIRGs, optical/near-IR emission line diagnostics remain useful qualitative diagnostic tools in the majority of sources. For these objects, mid-IR emission lines/UIB features can be used as quantitative tools for estimating the relative contributions of AGN and star formation. The reason is that optical and IR emission line diagnostics only rely on penetrating through dust in the disk to the narrow line region, and not through a high column density circumnuclear torus to the central AGN itself. The large-scale (geq 100 pc) obscuring material is likely very patchy and arranged in thin (and self-gravitating) disks (as in Arp 220; Scoville et al 1998). Radiation and outflows from AGNs punch rapidly through the clumpy obscuring screen, at least in certain directions.

8 In the screen extinction model the dust is in a homogeneous screen in front of the source and the attenuation at lambda is given by exp(-taud(lambda)) where taud(lambda) is the optical depth of the dust screen at lambda. An alternative, and probably more plausible scenario is that obscuring dust clouds and emitting HII regions are completely spatially mixed throughout an extended region. In that case the attenuation at lambda is taud(lambda) / (1 - exp(-taud(lambda))) which changes much more slowly with wavelength. Back.

Next Contents Previous