Despite the challenges of correctly interpreting astrophysical clocks in complex starbursts, the stellar modeling is rather reliable over most of the HRD (see, e.g., the comprehensive review of Bruzual 2003). One remaining nagging issue concerns RSGs. I will elaborate on this trouble spot using blue compact dwarfs (BCDs) as an example. BCDs are thought to be strongly starbursting galaxies powered by ionizing stars and with an underlying population of red stars whose precise age is still under discussion (e.g., Aloisi, Tosi, & Greggio 1999; Thuan, Izotov, & Foltz 1999). There is, however, convincing evidence for a significant number of RSGs in those objects with well-established CMDs (e.g., Schulte-Ladbeck et al. 2001).
Vázquez & Leitherer (2004) collected near-IR photometry of a sample of BCDs from the literature and compared them to synthesis models. The large-aperture photometry is sensitive to the young starburst, the surrounding field consisting of potential earlier starburst episodes, and the older underlying population. Therefore any RSGs present in these galaxies will affect or even dominate the photometry. The observations collected from the literature are plotted and compared to models in Fig. 6. Both reddening and line emission are negligible at IR wavelengths. The synthetic colors predicted for a SSP and for continuous star formation are in the left and right panels of Fig. 6, respectively. Fig. 6 (left) suggests reasonable agreement between the bulk of the data and the computed colors of a SSP with solar chemical composition.
Figure 6. Color-color diagrams of four samples of BCDs compared with starburst models of Z = 0.001 (solid line), 0.004 (long-dashed), and 0.02 (short-dashed line) based on the Geneva tracks. Filled symbols indicate the ages of the models. Open symbols show the data: triangles (Doublier et al. 2001), squares (Noeske et al. 2003), starred circles (Thuan 1983), and four-legged stars (Telles 2004). The error bars are the dispersion value in each sample. The solid vector indicates the reddening correction for AV = 0.25. Left: models for a SSP; right: continuous star formation (Vázquez & Leitherer 2004).
The synthetic models for SSPs were terminated at an age of 100 Myr. Higher ages are unrealistic, as BCDs are defined via their emission lines and blue colors. 100 Myr old starbursts would not be classified as BCDs anymore. The details of the star formation during the first tens of Myr, however, is a subject of debate. Dwarf galaxies are known to have had rather complex star-formation histories, with periods of quiescence and intermittent bursts of star formation (e.g., Greggio et al. 1998). Depending on the burst frequency, the effective star formation may mimic a steady-state situation. This is addressed in the right panel of Fig. 6. As expected, the imprint of the RSGs is more diluted, and the colors are more degenerate than for a SSP. If the star-formation history in BCDs were constant, the comparison between the data and the models would force one to postulate ages far in excess of 100 Myr. While this would not necessarily be in conflict with observational selections (ionizing stars are continuously replenished), the associated gas consumption would become entirely unreasonable. BCDs do not constantly form stars over 1 Gyr. Therefore the appropriate star-formation scenario is between the extremes plotted in Fig. 6, but most likely much closer to the SSP case in the left panel.
Does this suggest consistency between the observed and synthetic colors? The only track in Fig. 6 (left) that matches the data points is the one at solar chemical composition. The other tracks at lower abundance are significantly bluer and fail to reproduce the observed colors regardless of the assumed age and reddening correction. Only the solar models produce RSGs in large enough numbers and with sufficiently low Teff to reach the colors covered by the data points. Yet, the approximate average oxygen abundance of the sample is 20% solar. Therefore the assumption of solar composition is invalid, and the applicable models are those with Z = 0.004. Once forced to compare the data to the Z = 0.004 tracks, one arrives at the inescapable conclusion that the predicted colors of evolutionary models for metal-poor populations with a significant RSG component are incorrect. This conclusion is unchanged for either the Padova or Geneva tracks. Our results echo those of Massey & Olsen (2003), who demonstrated that both the Geneva and Padova evolution models fail to predict the location of RSGs and the Large and Small Magellanic Clouds. The evidence of failure at solar chemical composition is much weaker, if present at all.