Annu. Rev. Astron. Astrophys. 1998. 36: 189-231 Copyright © 1998 by Annual Reviews. All rights reserved

4.2. Dependence on Type and Environment

The star formation that takes place in the circumnuclear regions of galaxies also follows quite different patterns along the Hubble sequence, relative to the more extended star formation in disks. These distinctions are especially important in early-type galaxies, where the nuclear regions often dominate the global star formation in their parent galaxies.

4.2.1. HUBBLE TYPE     In contrast to the extended star formation in disks, which varies dramatically along the Hubble sequence, circumnuclear star formation is largely decoupled with Hubble type. Stauffer (1982), Keel (1983), Ho et al (1997a) investigated the dependence of nuclear H emission in star forming nuclei as a function of galaxy type. The detection frequency of HII region nuclei is a strong monotonic function of type, increasing from 0% in elliptical galaxies to 8% in SO, 22% in Sa, 51% in Sb, and 80% in Sc-Im galaxies (Ho et al 1997a), though these fractions may be influenced somewhat by AGN contamination. Among the galaxies with nuclear star formation, the H luminosities show the opposite trend; the average extinction-corrected luminosity of HII region nuclei in S0-Sbc galaxies is nine times higher than in Sc galaxies (Ho et al 1997a). Thus the bulk of the total nuclear star formation in galaxies is weighted toward the earlier Hubble types, even though the frequency of occurence is higher in the late types.

Similar trends are observed in 10-µm surveys of nearby galaxies (Rieke & Lebofsky 1978, Scoville et al 1983, Devereux 1987, Devereux et al 1987, Giuricin et al 1994). Interpreting the trends in nuclear 10-µm luminosities by themselves is less straightforward because the dust can be heated by active nuclei as well as by star formation, but one can reduce this problem by excluding known AGNs from the statistics. Devereux et al (1987) analyzed the properties of an optically selected sample of 191 bright spirals, chosen to lie within or near the distance of the Virgo cluster, and found that the average nuclear 10-µm flux was virtually independent of type and, if anything, decreased by 25-30% from types Sa-Sbc to Sc-Scd. An analysis of a larger sample by Giuricin et al (1994) shows that among galaxies with HII region nuclei (as classified from optical spectra), Sa-Sb nuclei are 1.7 times more luminous at 10 µm than Sc galaxies. By contrast, the disk SFRs are typically 5-10 times lower in the early-type spirals, so the fractional contribution of the nuclei to the total SFR increases dramatically in the early-type spirals. The nuclear SFRs in some early-type galaxies are comparable to the integrated SFRs of late-type spirals (e.g. Devereux 1987, Devereux & Hameed 1997). Thus while luminous nuclear starbursts may occur across the entire range of spiral host types (e.g. Rieke & Lebofsky 1978, Devereux 1987), the relative effect is much stronger for the early-type galaxies; most of the star formation in these galaxies occurs in the circumnuclear regions. Clearly the physical mechanisms that trigger these nuclear outbursts are largely decoupled from the global gas contents and SFRs of their parent galaxies.

4.2.2. BAR STRUCTURE     These same surveys show that nuclear star formation is strongly correlated with the presence of strong stellar bars in the parent galaxy. The first quantitative evidence came from the photographic work of Sérsic & Pastoriza (1967), who showed that 24% of nearby SB and SAB galaxies possessed detectable circumnuclear "hotspot" regions, now known to be bright HII regions and stellar associations. In contrast, none of the nonbarred galaxies studied showed evidence for hotspots. This work was followed up by Phillips (1993), who showed that the hotspots are found preferentially in early-type barred galaxies, a tendency noted already by Sérsic & Pastoriza.

The effects of bars on the H emission from HII region nuclei have been analyzed by Ho et al (1997b). They found that the incidence of nuclear star formation is higher among the barred galaxies, but the difference is marginally significant and no excess is seen among early-type barred galaxies. However, the distributions of H luminosities are markedly different, with the barred galaxies showing an extended tail of bright nuclei that is absent in samples of nonbarred galaxies. This tail extends over a range in H luminosities of 3-100 × 10⁴⁰ ergs s^-1, which corresponds to SFRs in the range 0.2-8 M year^-1. This tail is especially strong in the early-type barred galaxies (SB0/a-SBbc), where ~ 30% of the star forming nuclei have luminosities in this range.

Bars appear to play an especially strong role in triggering the strong IR-luminous starbursts that are found in early-type spiral galaxies. Hawarden et al (1986), Dressel (1988) found strong excess FIR emission in early-type barred spirals, based on IRAS observations, and hypothesized that this emission was associated with circumnuclear star-forming regions. This interpretation was directly confirmed by Devereux (1987), who detected strong nuclear 10-µm emission in 40% of the early-type barred spirals in his sample. Similar excesses are not seen in samples of late-type barred galaxies. These results have been confirmed in more extensive later studies by Giuricin et al (1994), Huang et al (1996). Although early-type barred galaxies frequently harbor a bright nuclear starburst, bars are not a necessary condition for such a starburst, as shown by Pompea & Rieke (1990).

The strong association of nuclear and circumnuclear star formation with bar structure, and the virtual absence of any other dependence on morphological type, contrasts sharply with the behavior of the disk SFRs. This implies that the evolution of the circumnuclear region is largely decoupled from that of the disk at larger radii. The strong distinctions between early-type and late-type barred galaxies appear to be associated with the structural and dynamical properties of the bars. Bars in bulge-dominated, early-type spirals tend to be very strong and efficient at transporting gas from the disk into the central regions, while bars in late-type galaxies are much weaker and are predicted to be much less efficient in transporting gas (e.g. Athanassoula 1992, Friedli & Benz 1995). All of the results are consistent with a general picture in which the circumnuclear SFRs of galaxies are determined largely by the rate of gas transport into the nuclear regions.

4.2.3. GALAXY INTERACTIONS AND MERGERS     Numerous observations have established a clear causal link between strong nuclear starbursts and tidal interactions and mergers of galaxies. Since this subject is reviewed in depth elsewhere (Heckman 1990, 1994, Barnes & Hernquist 1992, Sanders & Mirabel 1996, Kennicutt et al 1998), only the main results are summarized here.

The evidence for interaction-induced nuclear star formation comes from two types of studies, statistical comparisons of the SFRs in complete samples of interacting and noninteracting galaxies, and studies of the frequency of interactions and mergers among samples of luminous starburst galaxies. Keel et al (1985), Bushouse (1986) showed that the nuclear H emission in nearby samples of interacting spiral galaxies is three to four times stronger than that in a control sample of isolated spirals. Groundbased 10- to 20-µm observations of the nuclear regions of interacting and merging galaxies showed similar or stronger enhancements, depending on how the samples were selected (Lonsdale et al 1984, Cutri & McAlary 1985, Joseph & Wright 1985, Wright et al 1988). There is an enormous range of SFRs among individual objects. Spatially resolved data also show a stronger central concentration of the star formation in strongly interacting systems (e.g. Bushouse 1987, Kennicutt et al 1987, Wright et al 1988). Thus while the interactions tend to increase the SFR throughout galaxies, the effects in the nuclear regions are stronger. This radial concentration is consistent with the predictions of numerical simulations of interacting and merging systems (Barnes & Hernquist 1992, Mihos & Hernquist 1996, Kennicutt et al 1998).

The complementary approach is to measure the frequencies of interacting systems in samples of starburst galaxies. The most complete data of this kind come from IRAS, and they show that the importance of tidal triggering is a strong function of the strength of the starburst, with the fraction of interactions increasing from 20-30% for L_IR < 10¹⁰ L to 70-95% for L_IR > 10¹² L (Sanders et al 1988, Lawrence et al 1989, Gallimore & Keel 1993, Leech et al 1994, Sanders & Mirabel 1996). The relatively low fraction for the lower luminosity starbursts is understandable because the corresponding SFRs (< 1 M year^-1) can be sustained with relatively modest gas supplies and can be fed by internal mechanisms such as a strong bar. The most luminous starbursts, on the other hand, are associated almost exclusively with strong tidal interactions and mergers. SFRs larger than ~ 20 M year^-1 are rarely observed in isolated galaxies, though some possible exceptions have been identified by Leech et al (1994). In view of the enormous fueling requirements for such starbursts (Equations 5 and 6), however, it is perhaps not surprising that an event as violent as a merger is required. These results underscore the heterogeneous nature of the starburst galaxy population, and they suggest that several triggering mechanisms are involved in producing the population.