Starburst galaxies are a complex superposition of individual starburst regions, embedded in gas and dust having highly irregular morphologies. Reading off the age-dating clock becomes far from trivial in such environments. A case in point is the center of the nearby starburst galaxy M83, as reproduced in Fig. 5 (Harris et al. 2001). Located within the main bar of M83, the center of the galaxy contains a well-defined nuclear subsystem. The optically bright nucleus is offset from the dynamical center opposite to the starburst ring. Dense molecular gas has been detected and is generally concentrated to the north of the starburst ring. The optically visible nucleus, as well as a second dust hidden nucleus, are dominated by evolved red giant stars. These observations suggest a complicated morphology, where starbursts have occurred in multiple bursts and in an inhomogeneous interstellar medium (ISM). Age-dating methods applied to such systems suffer from manifold age-metallicity-IMF-ISM degeneracies.
Figure 5. Multicolor image of the center of M83 constructed from HST/WFPC2 F300W, F547M, and F814W images. North is at the top, and east is at the left. Field size is 36" × 36", corresponding to 650 pc × 650 pc (Harris et al. 2001).
The age-metallicity degeneracy, while a major concern at old ages (Trager et al. 2000), is less of an issue for the age range relevant to starbursts. Metallicity effects in hot-star atmospheres are relatively minor because of reduced line blanketing at high Teff. The chemical composition does of course enter via the metal-dependent stellar evolution. However, any self-enrichment is generally small since the nucleosynthetic products of massive stars appear to require much longer than 107 yr to mix with the surrounding ISM (Kobulnicky & Skillman 1997). In any case, independent techniques are available for estimating the metallicity in young starburst, such as emission-line spectroscopy of H II regions.
Most age-dating methods rely on evaluating the luminosity weighted light contributions from stars having different ages. Since there exists an L M relation, mass can always be traded for luminosity, and an age-IMF degeneracy is introduced. Therefore, IMF variations can mimic age gradients. While the evidence for significant IMF variations in local star clusters is weak (Kroupa 2002), such variations cannot be fully excluded in the dense starburst environment. Figer et al. (1999) obtained a determination of the upper IMF for the Arches and Quintuplet clusters, two extraordinary young clusters near the Galactic center. They found an IMF slope that is significantly flatter than the average for young clusters elsewhere in the Galaxy. If IMF variability is a concern, combining spectral features that respond differently to age and IMF can help overcome the age-IMF degeneracy. In general, age indicators tracing stars in the red part of the HRD are less IMF- and more age-sensitive than their counterparts in the blue. There are two reasons for this behavior. (i) Since the most luminous blue stars are more luminous than the most luminous red stars, RSGs have descended from the main-sequence where the initial M - L relation is steeper than for more massive blue stars. As a result, RSGs sample a smaller mass range in the integrated spectrum and are less IMF sensitive. (ii) The evolutionary tracks are less degenerate in the red than in the blue. A particular HRD location in the blue may be occupied stars having vastly different current masses, whereas the mass distribution in the red is quite homogeneous. Taken together, both effects suppress the IMF sensitivity of RSG related spectral features.
Breaking the age-ISM degeneracy can be the most daunting challenge, as the following example of starbursts containing W-R stars demonstrates. W-R galaxies are a subset of the starburst class, whose observational characteristic is the broad emission bump around 4640 - 4690 Å (Conti 1991). The compilation by Schaerer et al. (1999) lists 139 members.
W-R galaxies are important because they permit age determinations via stellar spectral features, as opposed to indirect tracers based on gas and/or dust emission. Hot stars are notoriously elusive even in the strongest starbursts because their spectral signatures are too weak, coincide with nebular emission lines, or are in the satellite-UV. W-R stars are the only hot, massive stellar species detectable at optical wavelengths. This is because they have the strongest stellar winds, which in combination with their high temperatures produce broad (~ 1000 km s-1) emission lines not coinciding with emission from H II regions. Examples are N III 4640, C III 4650, He II 4686, C III 5696, and C IV 5808. The mere detection of such features proves the presence of stars with masses above 40 - 60 M and ages of 2-6 Myr (depending on chemical composition) since only stars this massive can evolve into the W-R phase. This powerful diagnostic can be used for, e.g., inferring a massive star population when the space-UV is inaccessible due to dust (IR-luminous galaxies), or when broad nebular lines veil the O stars (Seyfert2 galaxies).
We are currently involved in an optical+near-IR survey of luminous starburst galaxies (Leão, Leitherer, & Bresolin, in prep.). Our goal is to investigate the stellar content from purely stellar tracers. The starburst galaxies are drawn from an IRAS sample of Lehnert & Heckman (1995). The galaxies have 10 < log LIR / L < 11.5, warm IR colors, and are not AGN dominated. Detection of the W-R features, together with other standard diagnostics allows us to probe the age and IMF. In particular, we can search for or against evidence of a peculiar IMF at high metallicity, as indicated, e.g., by IR observations.
Thornley et al. (2000) carried out an ISO spectroscopic survey of 27 starburst galaxies with a range of luminosities from 108 to 1012 L. The [Ne III] 15.6 µm and [Ne II] 12.8 µm lines are particularly useful. The ionization potentials of neutral and ionized Ne are 22 and 41 eV, respectively. The two lines are very close in wavelength (12.8 and 15.6 µm), and they have similar critical densities. This makes the line ratio a sensitive probe for various star formation parameters, in particular the upper mass cutoff of the IMF (Mup) and the age and duration of the starburst. Photoionization models of Thornley et al. and Rigby & Rieke (2004) suggest that stars more massive than about 35 to 40 M are deficient in the observed sample. Either they never formed because of a peculiar IMF, or they have already disappeared due to aging effects. This result echos that obtained from ground-based near-IR spectroscopy: the strategic recombination lines are powered by a soft radiation field originating from stars less massive than ~ 40 M (Doyon et al. 1992). An upper-mass cutoff as low as 40 M, however, is difficult to reconcile with the ubiquitous evidence of very massive stars with masses of up to 100 M in many starburst regions (e.g., Leitherer et al. 1996; Massey & Hunter 1998; González Delgado et al. 2002). Therefore the alternative explanation, an aged starburst seems more plausible. Under this assumption, stars of masses 50-100 M are initially formed in most galaxies, but the starbursts are observed at an epoch when these stars are no longer present. This implies that the inferred burst durations must be less than a few Myr. Such short burst timescales are surprising, in particular for luminous, starburst galaxies whose dynamical timescales can exceed tens of Myr. Both the peculiar IMF or the short starburst timescales in dusty, IR-bright starbursts are quite unexpected and pose a challenge to conventional models in which the starburst is fed by gas inflow to the nucleus over tens of Myr as a result of angular momentum loss.
Early results from our W-R survey urge caution when relating nebular IMF tracers to the actual stellar content. Several metal-rich starburst galaxies exhibiting a soft radiation field do in fact have a substantial W-R population. Examples are the archetypal starburst galaxies NGC 1614, NGC 2798, or NGC 3690. We clearly detect the tell-tale signatures of W-R stars in these galaxies. Unless stellar evolution proceeds differently than in our Galaxy, massive O stars must be present as well. The fact that we do not directly observe these O stars is no contradiction: their spectral lines will be hidden by coinciding nebular emission lines. We should, however, detect their ionizing radiation in a proportion predicted by the measured number of W-R stars and the expected ratio of W-R/O stars. The deficit of radiation suggests that indirect age and IMF tracers, like nebular lines, still require careful calibration, in particular when applied to dusty, metal-rich starbursts.
Interpretation of the stellar W-R feature itself may sometimes be complicated by the inhomogeneous structure of the surrounding ISM. As part of a larger project to quantify the stellar and interstellar properties of local galaxies undergoing active star formation (Chandar et al. 2004), we have obtained HST STIS long-slit far- and near-UV spectra for 15 local starburst galaxies. The target galaxies were selected to cover a broad range of morphologies, chemical composition, and luminosity. Some of them are known W-R galaxies. The UV counterpart of the optical He II 4686 line is the 3 2 transition of He+ at 1640 Å. Broad He II 1640 emission is seen in the UV spectra of individual Galactic and Magellanic Cloud W-R stars (e.g., Conti & Morris 1990) but is not prevalent in the integrated spectra of galaxies because of the overwhelming light contribution from OB stars in the UV.
NGC 3125-1 has by far the largest He II 1640 equivalent width in the sample. The 1640 Å emission is quite prominent and appears significantly stronger than the C IV emission at 1550 Å. In addition to this feature, N IV 1488 and N IV 1720 emission are also detected, consistent with the interpretation that the strong He II 1640 emission arises in the winds of massive stars. 6100 WNL stars are estimated to reside in this starburst region, making it the most W-R-rich known example of an individual starburst cluster in the local universe. What makes this region extraordinary is the number of W-R stars relative to other massive stars. The W-R features are almost undiluted compared with those seen in single W-R stars. Therefore, most of the continuum light must be emitted by the very same W-R stars. Quantitative modeling leads to approximately equal W-R and O-star numbers. Such an extreme ratio is excluded by stellar evolution models, even for a 3-5 Myr old starburst with a most unusual IMF.
How can we reconcile these observations with our current understanding of W-R stars? The UV continuum slope is normally dominated by OB stars; however, in this case the large number of W-R stars also makes a significant contribution. One possible explanation is that we are seeing the W-R stars through a "hole" where their energetic winds have blown out the natal cocoon earlier than have the OB stars. The often large reddening derived in starburst regions is consistent with the generally accepted scenario in which very young clusters remain embedded in their natal material until energetic stellar winds from evolving massive stars blow out the surrounding gas and dust.
If W-R stars are preferentially less attenuated than OB stars in NGC 3125-1, the equivalent width of He II 1640 and other W-R lines is skewed towards artificially large values because for a standard stellar population surrounded by a homogeneous ISM, the continuum is emitted by the less massive OB stars. If the ISM is inhomogeneous, the equivalent widths of purely stellar lines and the associated age determination become reddening dependent.