4. GAS CONTENT

A substantial part of the SFDGs have high gas mass fractions. H I masses range from a few 10⁶ M at M_B ~ -10 (represented in the FIGGS sample) up to a few times 10⁹ M at M_B ~ -20. The relative H I mass fraction varies from about 5% of the dynamical mass at the high mass end to almost 100% at the low mass end (Begum et al. 2008, Schombert et al. 2001). The H I disk thickness increases with decreasing luminosity, reaching a mean axis ratio of <q> ~ 0.6 in the faintest dI galaxies (Roychowdhury et al. 2010).

In Fig 1 we show the relation between M_HI and M_B based on 199 galaxies loosely classified as Irregular, 315 LSB galaxies and 224 galaxies classified as BCGs or H II galaxies. The sample has been collected from various publications (van Zee et al. 1997, Lisenfeld and Ferrara 1998, Izotov and Thuan 1999, Burkholder et al. 2001, Hirashita et al. 2002, Salzer et al. 2002, Pustilnik et al. 2002, Geha et al. 2006, Begum et al. 2008, Bergvall et al., unpublished). The classification is therefore not homogeneous and the sample is not in any sense complete. But it displays interesting relations. There is a fairly tight linear relation between H I mass and luminosity for the different types. The M_HI / L_B varies with more than one magnitude over the luminosity range. The spread in luminosity is not significantly different for the different samples. We also see that the distribution of dIs nicely attaches to the faint end of the BCG distribution. It seems that BCGs are merely scaled up versions of dIs. It raises the question about the bursty character of the BCGs. How many galaxies classified as BCGs really differ significantly from 'normal' SFDGs in terms of SFRs and gas consumption rates? One could have expected that the BCGs, if they are bursting, would have an increased dispersion in luminosity since we would catch the galaxy in different stages of a transient event. But this is not the case. Is it telling s that the BCGs are not bursting but form stars at a sightly higher rate than LSBGs over a long period of time?

Figure 1. Left: Isodensity contours of total H I mass versus absolute B magnitudes for galaxies classified as LSBGs, BCGs and Irrs. The type definition is not strict but varies between different authors (see reference list in running text). Least-square fits to the data sets are shown in solid (LSBGs), dashed (BCGs) and dotted (Irrs). Centre: Isodensity contours for the full sample. Also shown are the previous fits to BCGs (blue) and LSBGs (red) and the positions of four well known BCGs. Right: log(MHI / LB) vs MB. Correlations as before.

There is an ongoing discussion about the relationships between the morphological types of SFDGs and one important question is how LSBGs and BCGs are related (e.g. Simpson and Gottesman 2000). Both have low metallicities and high gas mass fractions and reside in similar environments (Rosenbaum et al. 2009, Pustil'nik et al. 1995). If BCGs are bursting, clearly their progenitors must be either irregulars or LSBGs. As the galaxy enters the burst, the luminosity would increase, while the H I mass slowly would decrease. The galaxy would move to the right in Fig 1. As we see, the relation for the BCGs is shifted relative to the LSBGs with about one magnitude. Thus, it seems obvious that if LSBGs are the progenitors and the B luminosity roughly can be used as a measure of the SFR, they would increase the SFR with a factor of ~ 2 in the transition. This agrees with results from previous investigations (e.g., Papaderos et al. 1996). However, with such a small change, Occam's razor urges us to primarily consider the case where BCGs always had a higher SFR than the LSBGs. Thus, there is no strong support of starbursts in BCGs from these data. More surprise is found from the positions of three well known BCGs in the diagram, NGC 6456 (see below), I Zw 18, II Zw 70 (Kehrig et al. 2008) and ESO 338-04 (= Tol 1924-416, see below). The galaxies all cluster along the mean line. This is unexpected since they are regarded as strong starbursts and therefore should be separated from the distribution of 'normal' BCGs. We would expect strong starburst BCGs to deviate from their candidate progenitors with 2-3 magnitudes. But the shift is only ~ 1 mag. Extinction effects play a minor role since E(B - V) < 0.05 in all 4 galaxies (Lequeux et al. 1981, Oconnell et al. 1978, Bergvall 1985, McQuinn et al. 2010).

Another clue to the relation between LSBGs and BCGs can be obtained from a study of the spatial distribution of H I. The scalelength versus M_HI from small sizes up to a few 10 kpc follows that of a constant surface density. But the scalelength also correlates with surface brightness. In a comparison between BCGs, LSBGs and dI galaxies, van Zee et al. (2001) find that the scalelength of BCGs is systematically smaller than for LSBGs and dIs and roughly corresponds to the optical radius. The central surface density is about a factor of 5 higher. Thus, if LSBGs are progenitors of BCGs, they first must reduce the angular momentum of the gas during the transition phase, in order to increase the H I surface density and thereby the SFR. In a similar discussion, Staveley-Smith et al. (1992) concluded that LSBGs and BCGs as classes probably were unrelated.

An important issue concerns the H₂ content. Stars form out of molecular clouds and it is known that the relative H₂ mass can be substantial in SFGDs. H₂ lacks a dipole moment, making direct observations of H₂ extremely difficult. Instead indirect observations of the associated CO molecule is used. The H₂ mass is then calculated using the CO-to-H₂ conversion factor, _CO. The value of _CO is however very poorly constrained at low metallicities since the CO line becomes weaker and also because the radiation field, especially from starburst regions, tend to dissociate the molecule. The conditions under which this happens cannot be modelled to high accuracy and consequently much of our knowledge about the conversion factor is empirical. The problem becomes evident from observations of galaxies in the local group. Leroy et al. (2011) find that while M31, M33, LMC and the Milky Way have quite similar values, _CO can increase with a factor of 5-10 in SMC and NGC 6822. This has serious implications for models of chemical evolution (see below) and timescales of gas consumption, a central issue as we discuss starburst dwarfs. The transition from normal to high values of _CO seems to take place at a metallicity of 12 + log(O/H) ~ 8.2-8.4. Alternative means of estimating the H₂ mass have been exploited. Much H₂ is residing in photodissociation regions (PDRs). In fact PDRs may be the dominating phase in starburst nuclei (Stacey et al. 1991). PDRs contain ions of low ionisation potential, like C⁺, giving rise to the strong emission line [C II] 158µ. The emission is strongly correlated with H₂ (Crawford et al. 1985) and can be used to estimate the H₂ mass. As an example we can have a look at IC 10 where Madden et al. (1997) speculate that, as a consequence of the strong UV field and low metallicity, small CO cores are formed, surrounded by extensive [C II] emitting regions containing H₂. H₂ masses based on the CO flux would therefore be severely underestimated. A similar situation may be at hand in the luminous (M_B ~ -21) BCG Haro 11 (Bergvall et al. 2000). No doubt the Herschel observatory will provide us with data to further unveil the interplay between star formation and the ISM.

Although with some scatter, there is a clear one-to-one relation between SFR and molecular mass (Leroy et al. 2005, Krumholz et al. 2011). In more massive systems one may explain this relation as a coupling between the coolant CO (associated with H₂) and conditions for collapse of giant molecular clouds (GMC). In low mass systems though, a more likely explanation seems to be that the transition from predominantly atomic hydrogen to molecular occurs under the same conditions as a dramatic drop in gas temperature that destabilizes clouds and initiates collapse (Krumholz et al. 2011). While the mean ratio between H₂ and H I masses increases with metallicity, the scatter is large at low metallicities and there are cases when the H₂ mass is more than 10 times higher than the H I mass, like NGC 4630 (Leroy et al. 2005). Another case is perhaps Haro 11, discussed above. Perhaps there is a yet unknown mechanism that stimulates an overproduction of gas in molecular state, leading to a starburst event.