Cold dark matter models predict that low-mass systems were the first sites of star formation, possibly as early as at a redshift of 30 (e.g., Barkana & Loeb 2001). Since larger systems form through hierarchical merging of smaller systems, they should contain surviving populations of these early epochs of star formation. Furthermore, several models predict that small galaxies should have formed most of their stars prior to reionization, while reionization would have suppressed further star formation activity. In fact, galaxies less massive than 109 M should have lost their star-forming material during reionization (e.g., Susa & Umemura 2004). Hence one would expect that low-mass galaxies contain ancient populations, while star formation should have ceased after reionization.
These are predictions that can be tested in the dwarfs in our immediate surroundings. The least massive dwarfs, the dSphs, should have been most severely affected. Deep color-magnitude diagram data of these dwarfs that reach below the oldest main-sequence turn-offs permit us to carry out relative age dating of their old populations with internal accuracies of fractions of ~ 1 Gyr, the highest accuracy currently attainable for any method for old stars. Note that this method can only be applied to populations sufficiently numerous to form detectable main-sequence turn-offs. This holds only for old Population II stars, while potential Population III stars are far too few even in our Milky Way. The differential ages of old populations in dwarf galaxies - either field populations or globular clusters - can then be compared to the ages of the oldest Galactic globular clusters of similar composition.
Figure 1. Sketch indicating the approximate duration of star formation episodes in dSph galaxies (~ 107 M). The adopted beginning and end of the reionization epoch are based on results from WMAP and from the Sloan Digital Sky Survey. The predicted cessation of star formation after reionization is not observed. For details, see Grebel & Gallagher (2004).
Importantly, this method reveals that (1) old populations are ubiquitous (but their fractions vary) and (2) the oldest ages in all of the galaxies studied so far are indistinguishable within the measurement accuracy (see Grebel 2000 and Grebel & Gallagher 2004 for details). All nearby dwarf galaxies studied in sufficient detail were shown to contain ancient populations. Moreover, these nearby Milky Way companions and the Milky Way itself share a common epoch of star formation for their ancient Population II (within ~ 1 Gyr). These observations are consistent with the expectations from the building block scenario.
However, the predicted cessation of star formation after reionization, expected to have affected particularly dSphs owing to their low mass, is not observed (Grebel & Gallagher 2004). Instead, even dSphs entirely dominated by old populations show evidence for star formation extending over many Gyr (Harbeck et al. 2001; Ikuta & Arimoto 2002). In dSphs with a mixture of populations we usually find star formation episodes that lasted many Gyr without being interrupted by a pronounced, long hiatus to the extent that we can measure the duration of star formation (exception: Carina with its episodic star formation). This may mean that the above quoted cosmological models are incorrect and do not properly consider other effects that might prevent complete photoevaporation (e.g., Susa & Umemura 2004). On the other hand, photoionization is one plausible way to circumvent the substructure problem (e.g., Somerville 2002). Alternatively, it is conceivable that dSphs were once considerably more massive (by roughly a factor of 100), which could also have prevented photoionization squelching. In this case the galaxies observed today as dSphs must have undergone substantial mass loss.