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

3.2.2. Properties and Evolution of Starbursts

Dusty starburst galaxies constitute an important class of objects (cf Moorwood 1996). About 25% of the high-mass star formation within 10 Mpc occurs in just four starburst galaxies (Heckman 1998). Key open questions of present research relate to the form of the initial stellar mass function (IMF) in starbursts and to the burst evolution. ISO has contributed to these issues mainly by studying the content of the most massive stars in dusty starbursts through nebular spectroscopy. Following pilot studies by Fischer et al (1996), Kunze et al (1996), Lutz et al (1996a), Rigopoulou et al (1996a), Thornley et al (2000) carried out a survey with the SWS instrument (2.4 to 45 µm) of [NeIII] / [NeII] line emission in 27 starburst galaxies, with a range of luminosities from < 10⁸ to >10¹² L. The excitation potentials of Ne⁺ and Ne⁺⁺ are 22 and 41 eV, and the two lines are very close in wavelength and have similar critical densities. The Neon line ratio is a sensitive tracer of the excitation of HII regions and of the OB stars photoionizing them. The basic result for the starburst sample is that HII regions in dusty starburst galaxies on average have low excitation (left inset of Figure 7). The average [NeIII] / [NeII] ratio in starburst galaxies is a factor of 2 to 3 lower than in individual galactic compact HII regions. The SWS work establishes as a fact and puts on a safe statistical footing what was already suggested by earlier ground-based near-IR results (e.g. Rieke et al 1980, Doyon et al 1994a, Doherty et al 1995).

Figure 7. Left: SWS measurements of the [NeIII]/[NeII] ratio in starburst galaxies (filled rectangles) and nearby starburst templates (asterisks) (from Thornley et al 2000). Top right: [NeIII] / [NeII] line ratios (triangles) and [ArIII] / [ArII] ratios (open rectangles) as a function of galactocentric radius for HII regions in our Galaxy (Cox et al 1999). Bottom right: Models of [NeIII]/[NeII] ratio as a function of burst age, for different IMF upper mass cutoffs. For each Mup single cluster models for burst durations of 1, 5, and 20 Myr (top to bottom) are shown as continuous lines. The dashed lines are the corresponding 20 Myr models with a cluster mass function.

Thornley et al (2000) have modeled the HII regions as ionization-bounded gas clouds photoionized by central evolving star clusters. The cluster SED (as a function of IMF and evolutionary state) is used as the input for photoionization modeling of the nebular emission. The computed Ne-ratios (for a Salpeter IMF with different upper-mass cutoffs) as a function of burst age t_b and burst duration t_b are plotted in Figure 7 (lower right). The basic result is that on average stars more massive than about 35 to 40 M are not present in starburst galaxies, either because they were never formed (upper-mass cutoff) or because they already have disappeared as a result of aging effects.

An upper-mass cutoff in the intrinsic IMF is difficult to reconcile with a growing body of direct evidence for the presence of very massive stars ( 100 M) in nearby starburst templates (galactic center: Krabbe et al 1995, Najarro et al 1997, Serabyn et al 1998, Figer et al 1998, Ott et al 1999; R136 in 30 Doradus: Hunter et al 1995, Massey & Hunter 1998; NGC 3603: Drissen et al 1995, Eisenhauer et al 1998). In more distant HII region galaxies, the 4696 Å HeII emission line feature signals the presence of Wolf-Rayet stars of mass 60 M (e.g. Conti & Vacca 1994, Gonzalez-Delgado et al 1997). The low average excitation in starbursts is thus more plausibly caused by aging (Rieke et al 1993, Genzel et al 1994). This conclusion is also supported by the large spread and substantial overlap of the Galactic and extragalactic Ne-ratios (Figure 7). The overlap region of the Antennae and the near-nuclear knot in NGC 3690 have [NeIII] / [NeII] ~ 1, comparable to nearby HII regions. Precisely in these two regions most of the flux (Section 3.2.1) comes from a single compact (and probably young) region. Everywhere else in the Antennae the Ne-ratio is much lower (Vigroux et al 1996), and at the same time the stellar cluster data (Whitmore et al 1999) indicate greater ages. Similar results emerge if, instead of the Ne-ratio, far-infrared line ratios (Fischer et al 1996, Colbert et al 1999) or the ratio L_bol / L_Lyc are used as constraints (Thornley et al 2000).

If stars of masses 50-100 M are initially formed in most galaxies, the inferred burst durations must be less than 10 Myr (Figure 7). Such short-burst timescales are surprising especially for luminous, distant systems (Section 3.4.4). The dynamical time is a few million years or more. It appears that supernovae and stellar winds disrupt the ISM and prevent further star formation as soon as the first generation of O stars have formed and evolved. Starbursts must induce large negative feedback. Starburst models (Thornley et al 2000, Leitherer & Heckman 1995) as well as observations of galactic superwinds (Heckman et al 1990) indicate that 1% to 2% of the bolometric luminosity of starbursts emerges as mechanical energy in superwinds. If this mechanical energy input removes interstellar gas from the potential well with ~ 10% efficiency, the starburst is choked after the "dispersal" time t_d. For typical gas masses and starburst parameters, t_d is 10⁷ years. The burst durations thus appear to be significantly smaller than the gas consumption time scales, t_gc ~ 5 to 10 × 10⁷ yrs. Near-IR imaging spectroscopy in M82, IC 342, and NGC 253 indicates that in the evolution of a given starburst system, there are several episodes of short-burst activity (Rieke et al 1993, Satyapal et al 1997, Böker et al 1997, Engelbracht et al 1998, Förster-Schreiber et al 2000).

Although the scenario just given is qualitatively quite plausible, the small values of the burst timescales indicated by the modeling of Thornley et al (2000) seem unphysical. In reality, burst timescales are probably somewhat larger (~ 10⁷ years). The galactic center starburst region has been analyzed through a direct stellar census (Krabbe et al 1995, Najarro et al 1997). The Ne-ratio computed from the starburst model (t_b ~ 7 × 10⁶, t_b ~ 4 × 10⁶ years, log U = -1) is more than an order of magnitude greater than the observed one (Lutz et al 1996b). The models predict that in the galactic center, the ionizing UV continuum should be dominated by O stars and hot Wolf-Rayet stars. The data show, however, that more than half of the ionizing flux comes from a population of relatively cool (20,000-30,000 K) blue supergiants (Najarro et al 1997). It would thus appear that the tracks and stellar properties adopted for the galactic center are not appropriate (Lutz 1999). Likely culprits could be the effects of metallicity and of the dense stellar winds on the UV SEDs of the massive stars. The Ne-ratio depends strongly on metallicity, as is directly demonstrated by the much higher excitation of II Zw 40 and NGC 5253 (Z = 0.2-0.25 Z) and of 17 HII regions in the SMC / LMC observed by Vermeij & van der Hulst (1999). The average Ne-abundance in the starburst sample is ~ 1.7 times solar. However, the models of Thornley et al (2000) are for solar metallicity. Furthermore, there appears to be a galactocentric gradient in the Ne-ratios of Cox et al (1999), again signaling dependence on metallicity. A similar discrepancy between (high-excitation) current stellar models and (low-excitation) observed nebular emission lines has also been found for Wolf-Rayet stars (Crowther et al 1999a), suggesting that their UV-spectra are softer than considered so far (Hillier & Miller 1998). Another area of concern is the impact on the interpretation of spatial variations in the various tracers that cannot be properly taken into account by single aperture spectroscopy. Crowther et al (1999b) reported SWS and ground-based infrared spectroscopy of the low-metallicity starburst galaxy NGC 5253. Although [SIV] and Br come from a very compact (3") region centered on the nucleus, the [NeII] emission is much more extended. There is a high-excitation (T_eff > 38,000 K) young (t = 2.5 Myr) nuclear burst surrounded by a lower-excitation (T_eff ~ 35,000 K) older (5 Myr) one. Thornley et al have raised the possibility that some of the [NeII] flux in their starburst galaxies may come from a more diffuse, low-ionization parameter zone, thus lowering the effective Ne-ratio in the SWS aperture.