5. Toward a new star formation history for IZw 18
5.1. A previous star formation event
The oxygen abundance distribution in IZw 18 appears extremely homogeneous throughout the galaxy, indicating a thoroughly mixed interstellar medium. If the measured abundances result from the metals ejected by the massive stars involved in the current burst, as suggested by Kunth & Sargent (1986), this would imply efficient mixing of the ejecta on scales of at least 600 pc within a timescale comparable to the age of the present burst i.e., a few Myr (Hunter & Thronson 1995). However, dispersal of the heavy elements ejected by the massive stars can hardly be accomplished in less than yr on scales between 100 and 1000 pc (Roy & Kunth 1995); the timescale required for complete mixing is even longer (Tenorio-Tagle 1996). Thus the observed metals cannot arise from the material ejected by the stars formed in the current burst. It follows that the presently observed metals should have been formed in a previous star formation episode . The metals ejected in the current burst of star formation remain most probably hidden in a hot phase as suggested by Pantelaki & Clayton (1987) and more recently by Tenorio-Tagle (1996), Devost et al. (1997), Kobulnicky & Skillman (1997) and Pilyugin (1999). Bomans (1999) has shown on a deep pointing with the ROSAT HRI instrument that there is extended X-ray emission to the SW and maybe to the NE from the central bubble of IZw 18. This extended emission seems to trace the expanding loops, leading the author to conclude that it supports the picture of hot, metal-enriched gas streaming out of IZw 18. This gas would have been ejected into the halo where it would take long excursions while cooling and returning to the central galactic region to become available for future processing into stars (Tenorio-Tagle 1996). X-ray observations of the BCD VIIZw 403 (Papaderos et al. 1994) are also interpreted as hot material ejected by the present starburst. The availability of powerful X-ray observatories in the near future will allow the metallicity of this hot gas to be derived, allowing this scenario to be confirmed.
If the observed metals in IZw 18 were formed in a previous star formation episode, this would imply that the object is not a "young" galaxy undergoing its first star formation as suggested by Searle & Sargent (1972). Such a view is also supported by other studies, which independently lead to another scenario (Dufour et al. 1988, Dufour & Hester 1990, Hunter & Thronson 1995, Kunth et al. 1995, Garnett et al. 1996, Aloisi et al. 1999).
5.2. The dearth of low metallicity galaxies
The metal abundances measured in IZw 18 are the lowest known in the interstellar matter (but not in stars) of the local universe; this remains so despite extensive metallicity measurements in emission line galaxies (Terlevich 1982, Terlevich et al. 1991, Masegosa et al. 1994, Izotov et al. 1994, Terlevich et al. 1996). Because of the correlation between size, luminosity and metallicity in dwarf galaxies (Skillman et al. 1989), Masegosa et al. (1994) proposed that galaxies with very low metallicity are too faint to be "caught" in their sample. This raises the possibility that extremely metal deficient objects may be very faint. IZw 18 and other starburst galaxies (Roennback & Bergvall 1995) lie quite far away from the correlation established by Skillman et al. (1989) for dwarf irregular galaxies. This may reflect the fact that they are presently undergoing a strong star formation event which increases their luminosity. However, the galaxies used by Skillman et al. (1989) were selected from the H catalog of Kennicutt et al. (1989), thus the sample allows for current star formation! The origin of the correlation remains unclear (see also Skillman 1999).
It is easy to show that the present star formation rate in IZw 18 or in other starbursts cannot be sustained for a Hubble time without producing excessive chemical enrichment and a numerous stellar population. It is generally admitted that blue compact galaxies experience violent star formation events separated by long quiescent phases (Searle & Sargent 1972) during which they would appear as Low Surface Brightness Galaxies (LSBG) or quiescent dwarfs. However this population does not contain any objects more metal poor than IZw 18 (McGaugh & Bothun 1993, McGaugh 1994, Roennback & Bergvall 1995, Van Zee et al. 1997a,b). Does the metallicity of IZw 18 represent a lower limit for the abundance in the gas of local galaxies? If so, why?
5.3. The lack of HI clouds without optical counterpart
Different observing programs have been carried out to search for isolated intergalactic HI cloud, but without success so far (Briggs 1997). Most local so-called primeval HI clouds candidates turned out to be associated with stars (see for example Djorgovski 1990; Impey et al. 1990; McMahon et al. 1990; Salzer et al. 1991; Chengalur et al. 1995, for HI1225+01). Does this mean that such entities do not exist? If so, this would imply that all gas clouds (with a mass comparable to that of a dwarf galaxy) have formed stars . However, the detection limits for HI surveys remain quite high (), and the existence of very small primeval HI clouds cannot be ruled out. Nevertheless, if isolated dwarf galaxy progenitor HI clouds existed, they would have sizes and masses comparable to small galaxies, and they would present sufficient column densities to be detected by radio techniques. So far, non-detection indicates that if such clouds exist, they are very rare. This idea is reinforced by the presence of absorption line systems of high column densities in the spectra of quasars which seems to arise mainly from halos of bright galaxies and not from small HI clouds (Lanzetta et al. 1995, Tripp et al. 1997), indicating again that the latter are sparse. Furthermore, it has been shown that the diffuse cosmic UV background can ionize the extreme outer HI disks of spiral galaxies (Van Gorkom 1991, Corbelli et al. 1989, Maloney 1990, Corbelli & Salpeter 1993a,b), producing an abrupt fall in their HI column density. This effect could contribute to hide some primeval HI gas from the current surveys.
5.4. The temporal evolution of the metallicity
Absorption lines in Damped Lyman Alpha (DLA) systems are used to study the temporal evolution of the metallicity of the interstellar gas. Although the nature of the absorbing systems is still controversial (Tripp et al. 1997), it is generally admitted that the metallic lines are associated in some way with galaxies. The temporal evolution of the metallicity in the DLA systems reported by Lu et al. (1996) and more recently by Lu et al. (1998) and Prochaska & Wolfe (1999) is reproduced in Fig. 4.
One notices that the mean metallicity of the interstellar gas increases as one gets closer to local time. This is generally interpreted as the effect of cumulative enrichment by strong star formation events. However, a more intriguing feature is that the metallicity of the most underabundant systems seems also to increase with time! No extremely underabundant system has been found at low redshift. If the enrichment is solely the result of starburst events, we should find, locally, objects which have not undergone any burst (or very few of them); these objects would have a very low metallicity (comparable to what is observed at high redshift). Does the apparent increase in metallicity of the most underabundant systems indicate the existence of a minimal and continuous enrichment of the interstellar medium? The number of systems observed at low redshift is small (Meyer & York 1992, Steidel et al. 1995, Pettini & Bowen 1997, de la Varga & Reimers 1997, Boisse et al. 1998, Shull et al. 1998); if some unevolved systems exist, they must be very few. The non detection of such systems could arise from a selection effect rather than from their inexistence.
5.5. A new star formation regime
It is generally accepted that the metal enrichment of the ISM builds up mainly in bursts. Different studies have been carried out to model these bursts to reproduce the global properties of galaxies. In the case of IZw 18, Kunth et al. (1995) have shown that only one burst, with an intensity comparable to the present one, is enough to produce the observed abundances. As we have shown, this single burst cannot be the present one. Previous massive star formation has occurred. We cannot eliminate the possibility that this previous star formation event was a starburst.
However, starburst episodes must be separated by quiescent phases, during which these systems appear as quiescent dwarfs or Low Surface Brightness Galaxies (LSBG). Studies of the latter objects (Van Zee et al. 1997c) have revealed that, despite their low gas density, star formation occurs (with a weak efficiency), probably as a local process instead of a global event. The SFR between bursts is very low, but not zero, so the metallicity would increase slowly during these quiescent phases. Because these star formation rates are very weak, they are generally neglected in studies of star formation history of galaxies. However, in dealing with very low metallicity galaxies, they are capable of raising the metallicity levels up to values comparable to that of IZw 18 in less than a Hubble time.
For example the galaxy UGC 9128, studied by Van Zee et al. (1997b) presents a SFR of about for a HI mass of . If such a low SFR lasts 10 Gyr, it will form in stars, and no more than 5% of the initial mass of gas will have been transformed into stars. At this low continuous star formation rate, sustained during even a Hubble time, the fraction of gas still available at present epoch remains high (about 95%). Thus the existence of a continuous low star formation rate in dwarf galaxies is consistent with the large HI reservoirs generally observed in these objects.
where G is the fraction of gas presently available and y the yield in heavy elements. The uncertainties on this last parameter are large, but y is likely to be in the range 0.01 to 0.036 Maeder (1992). Using a mean value , and for the example above with , we estimate the metallicity of the gas resulting from this low SFR enrichment to be close to , that is 1/20th solar!
Consequently, a low continuous star formation rate cannot be neglected, especially when dealing with low metallicity galaxies; this may be the dominant star forming and metal enrichment process in dwarf galaxies.
We propose that in the most extreme objects, like IZw 18, a continuous low star formation regime can account for the observed abundances. We surmise that the present starburst is the first major one in the history of IZw 18, and that a mild star formation rate has been going on for several Gyr. Preliminary calculations strengthen this hypothesis (Legrand & Kunth 1998). If such a low regime is universal, we expect that all small systems have been forming stars and that their metallicity has increased slowly but steadily with time. This scenario explains the lack of local objects more underabundant than IZw 18, the absence of HI clouds without an optical counterpart and the evolution as a function of redshift of the most metal poor quasars absorption systems. Detailed modelling of this low star formation regime is presented in Legrand (2000).
© European Southern Observatory (ESO) 2000
Online publication: March 21, 2000