Proceedings of the Workshop
"The Magellanic Clouds and Other Dwarf Galaxies"
of the Bonn/Bochum-Graduiertenkolleg

Nearby Young Dwarf Galaxies

Trinh X. Thuan1 and Yuri I. Izotov2

1Astronomy Department, University of Virginia, Charlottesville, USA
2Main Astronomical Observatory, Kiev, Ukraine

Received 16th July 1998
Abstract. We develop the theme that the study of extremely metal-deficient Blue Compact Dwarf (BCD) galaxies is crucial for understanding galaxy formation and evolution and constraining cosmological parameters. We discuss in detail two BCDs with a metallicity less than 1/20 that of the sun, SBS 0335-052 (Zsun/41) and SBS 1415+437 (Zsun/21). We show that the two BCDs do not contain stars older than 100 Myr and thus appear to be truly young galaxies, undergoing now their first burst of star formation. The H I envelope in which SBS 0335-052 is embedded appears to be primordial, devoid of any heavy element. We then discuss the chemical abundances in young dwarf galaxies and argue that all galaxies with Z ≤ Zsun/20 are young, with ages not exceeding 100 Myr. We show how the study of the escape of Lyα photons in BCDs can shed light on the long-standing problem of the weakness or absence of Lyα emission in high-redshift star-forming galaxies. We also discuss the P Cygni profile phenomenon in extremely metal-deficient environments. We finally give a determination of the primordial helium abundance which is higher than previous measurements and more consistent with current measurements of the abundances of other primordial elements.

1. Introduction

The formation of galaxies is one of the most fundamental problems in astrophysics, and much effort has gone into the search for primeval galaxies (PG). A possible definition of a primeval galaxy is a young system undergoing its first major burst of star formation. It is now widely believed that the vast majority of galaxies underwent such a phase at redshifts ∼2 or greater. In most galaxy formation scenarios, young galaxies are predicted to show strong Lyα emission, associated with the cooling of the primordial gas and the subsequent formation of a large number of massive ionizing stars (Partridge & Peebles 1967; Charlot & Fall 1993). Yet, despite intensive searches, the predicted widespread population of Lyα primeval galaxies has remained elusive (Pritchett 1994).

Several objects have been put forward as possible PG candidates, ranging from high-redshift radio galaxies to Lyα emitters found around quasars and damped Lyα systems, mainly on the basis of very high luminosity and star formation activity. However, most of these candidate PGs already contain a substantial amount of heavy elements, as evidenced by the presence of strong P Cygni profiles and interstellar absorption in their spectra (Steidel et al. 1996; Yee et al. 1996). These spectra are very similar to those of nearby starburst galaxies known to contain old stellar populations (Leitherer et al. 1996). Thus high-redshift galaxies discovered thus far are not truly primeval. Moreover, even if true PGs are discovered at high-redshift, it is difficult to study them in detail, even with the largest existing telescopes, because of their extreme faintness and very compact angular size. We propose here to take a different approach to the PG problem. Instead of searching for very high-redshift galaxies in the process of forming, we look for nearby galaxies undergoing their first burst of star formation, and hence satisfying the above definition of a PG. The best candidates for such a search are blue compact dwarf galaxies (BCD).

BCDs are low-luminosity extragalactic objects with MB≥-18 where intense star formation is presently occuring, as evidenced by their blue UBV colors, and their optical spectra which show strong narrow emission lines superposed on a stellar continuum which is rising toward the blue, similar to spectra of H II regions. Star formation in BCDs cannot be continuous but must proceed by bursts because of several observational constraints:

  1. Gas is transformed into stars at the rate of approximately 1 Msun yr-1, so that the current burst cannot last more than about 108 yr before depleting the neutral gas supply of ∼108 Msun;
  2. Optical-infrared colors of BCDs give ages of about 107 yr; and
  3. Population synthesis of UV spectra of BCDs give invariably jumps in the stellar luminosity function, indicative of starbursts (see Thuan 1991 for a review).
Ever since their discovery, the question has arisen whether BCDs are truly young systems where star formation is occuring for the first time, or old galaxies with an old underlying stellar population on which the current starburst is superposed (Searle et al. 1973). Thuan (1983) carried out a near-infrared JHK survey of BCDs and concluded that all the objects in his sample possessed an old underlying stellar population of K and M giants. That result was not unambiguous as the JHK observations were centered on the star-forming regions and the near-infrared emission could be contaminated by light from young supergiant stars. The advent of CCD detectors allowed to look for the low-surface-brightness underlying component directly. Loose & Thuan (1985) undertook a CCD imaging survey of a large BCD sample and found that nearly all galaxies (≥95%) in their sample show an underlying extended low-surface-brightness component, on which are superposed the high-surface-brightness star-forming regions. Subsequent CCD surveys of BCDs have confirmed this initial result (Papaderos et al. 1996; Telles & Terlevich 1997). Thus, most BCDs are not necessarily young galaxies. However, there was a hint that extremely metal-deficient BCDs do not contain an old stellar population and can be primordial. Hubble Space Telescope (HST) imaging of I Zw 18, the most metal-deficient BCD known (Zsun/50, Searle & Sargent 1972), to V∼26 by Hunter & Thronson (1995) suggests that the stellar population is dominated by young stars and that the colors of the underlying diffuse component are consistent with those from a sea of unresolved B or early A stars, with no evidence for stars older than ∼107 yr.

For more than 20 years, I Zw 18 stood in a class by itself. The BCD metallicity distribution ranges from ∼Zsun/3 to ∼Zsun/50, peaking at ∼Zsun/10, and dropping off sharply for Z≤Zsun/10. Intensive searches have been carried out to look for low-metallicity BCDs but they have met until recently with limited success. For example, the majority of the BCDs in the Salzer (1989) and Terlevich et al. (1991) surveys have metallicities larger than Zsun/10. Several years ago, a new BCD sample has been assembled by Izotov et al. (1993) from objective prism survey plates obtained with the 1 m Schmidt telescope at the Byurakan Observatory of the Armenian Academy of Sciences during the Second Byurakan Survey (SBS). The most interesting feature of the SBS is its metallicity distribution (Izotov et al. 1992; Thuan et al. 1994): it contains significantly more low-metallicity BCDs than previous surveys. It has uncovered about a dozen BCDs with Z≤Zsun/15, more than doubling the number of such known low-metallicity BCDs and filling in the metallicity gap between I Zw 18 and previously known BCDs. We shall discuss here two examples of extremely metal-deficient SBS BCDs and the observational evidence that they may be young galaxies. In Sect. 2, we describe the properties of SBS 0335-052 which is the second most metal-deficient BCD known with a metallicity of only Zsun/41 (Izotov et al. 1990; Melnick et al. 1992). Section 3 discusses the properties of SBS 1415+437 (Thuan et al. 1998) which has a metallicity of Zsun/21. Section 4 is concerned with chemical abundances in young dwarf galaxies and the constraints they put not only on the evolutionary status of these galaxies but also on stellar nucleosynthesis models. We argue there that all BCDs with Z≤Zsun/20 are young, with ages not exceeding 100 Myr. The study of nearby young dwarf galaxies can also shed light on the long-standing problem of the weakness of Lyα emission of high-redshift star-forming galaxies. We discuss Lyα emission and absorption in BCDs in Sect. 5. Young BCDs, because of the primordial nature of their gas, are also excellent laboratories for measuring the abundances of primordial elements such as helium. Section 6 presents a new determination of the primordial helium abundance.

2. An extremely metal-deficient galaxy: SBS 0335-052

2.1. Age of the stellar component

HST WFPC2 images of the BCD show that most of the star formation in SBS 0335-052 (MB = -16.7, v = 4076 km s-1) occurs in 6 super-star clusters (SSCs) with -14.1≤MV≤-11.9, within a region ∼520 pc in size (Fig. 1, Thuan et al. 1997). Later processing by Papaderos et al. (1998) of the same HST images reveals several fainter clusters (not SSCs). The SSCs are roughly aligned in the SE-NW direction, and there is a systematic reddening of the V-I color of the SSCs away from the brightest one, with a flattening of the color of the clusters beyond 520 pc (Fig. 2). Some of the reddening may be due to dust which is clearly seen as white patches in the V-I color map (Fig. 3) and which is mixed spatially with the SSCs. It is interesting that even in a very metal-deficient interstellar medium (ISM) with only 1/40 of the Sun's metallicity, dust is clearly present. Thuan et al. (1997) and Papaderos et al. (1998) attribute however most of the color variation of the SSCs not to dust, but to an age variation resulting from sequential propagating star formation. The V-I colors are consistent with the picture that star formation started at the location of the most distant cluster, some 1.8 kpc away from the location of the brightest and bluest cluster at the South East end of the galaxy, at about 100 Myr ago and propagated through the ISM to the latter, whose age is only ∼4 Myr, with an average speed of ∼18 km s-1 (Thuan et al. 1997; Papaderos et al. 1998). Thus the star-forming clusters have ages between 4 and 100 Myr. Does SBS 0335-052 possess an underlying older stellar population? The extended underlying component is shown in Fig. 4 where the contrast has been adjusted to display very low surface-brightness features. The unusually blue colors of this underlying component (see the U-B color profile labeled E in Fig. 5) and its irregular, blotchy and filamentary structure suggest that a significant fraction of the light is of gaseous rather than stellar origin. A supershell of 380 pc radius can be seen delineating a large supernova cavity. However Izotov et al. (1997b) found that the Hβ equivalent width in the underlying component is ∼3 times lower than the value expected for pure gaseous emission, implying that two-thirds of the light comes from an underlying stellar population. Papaderos et al. (1998) have modeled the UBVRI colors of this stellar component, after removal of the ionized gas contamination. They found that the colors are consistent with an underlying stellar population not older than 100 Myr.

2.2. Age of the neutral gas component

Thus the stellar component in SBS 0355-052 is extremely young and the BCD is likely undergoing star formation for the first time. If this is the case, the neutral gas envelope surrounding the BCD must also be very metal-deficient. A 21 cm VLA map of the BCD (Pustilnik et al. 1998) has shown it to be embedded in an extraordinarily large H I cloud seen nearly edge-on, with dimensions some 64 by 24 kpc. This is to be compared with the typical size of H I envelopes around BCDs which is more like a few kiloparsecs in each dimension. We can use the BCD as a background light source shining through the H I envelope to probe the physical conditions of the neutral gas. The Lyα line seen in absorption would give the column density of atomic hydrogen, while the O I λ1302 line would give the column density of the most abundant heavy element that remains neutral in the H I cloud. This would allow us to set limits on the O/H abundance ratio in the neutral gas. Figure 6 shows the ultraviolet spectrum of SBS 0335-052 around the Lyα line obtained by Thuan & Izotov (1997) with the Goddard High Resolution Spectrograph aboard the Hubble Space Telescope (HST). A strong damped Lyα absorption line is seen along with several heavy element interstellar absoption lines such as O I λ1302, Si II λ1304 and S II λ1251, λ1254, and λ1259. The H I column derived by fitting the Lyα absorption profile is N(H I) = (7.0±0.5)·1021 cm-2, the highest derived thus far for a BCD, and ∼2 times larger than in I Zw 18 (Kunth et al. 1994). Comparison with high resolution quasar spectra implies that the O I λ1302 line along with other heavy element interstellar absorption lines such as Si II λ1304 and S II λ1251, λ1254, λ1259 are not saturated, which allow us to derive abundances. Assuming that these lines originate in the H I gas, we derive extremely low abundances of oxygen, silicon and sulfur, respectively 37000, 4000 and 116 times lower than the solar values. The oxygen abundance is a whole 37 times lower than in the neutral gas of I Zw 18. However, these highly discrepant deficiency factors between different elements suggest that the absorption lines are produced, not in the H I, but in the H II gas. Adopting that hypothesis, the derived abundance from the UV absorption lines are then consistent with that derived from the optical emission lines (Z∼Zsun/40). The conclusion that the heavy element absorption lines originate in the H II region is supported by the detection of several systems of blueshifted S II λ1259, Si II λ1260, O I λ1302, Si II λ1304, C II λ1335 absorption lines originating in fast-moving clouds with velocities up to ∼1500 km s-1, and also by the presence of heavy element absorption lines with excited lower levels. If this conclusion holds, then the H I cloud in SBS 0335-052 is truly primordial, unpolluted by heavy elements (Thuan & Izotov 1997).

In summary, all the known observational evidence suggests that SBS 0335-052 is truly a young galaxy.

[Click here to see Fig. 1-6!]

3. Another young galaxy: SBS 1415+437

3.1. Age from color-magnitude diagrams

Figure 7 shows the HST WFPC2 I image of SBS 1415+437 (MB = - 14.0, v = 607 km s-1) with the contrast level adjusted so as to show the low surface-brightness underlying extended component (Thuan et al. 1998). The galaxy has an elongated, comet-like shape with a bright H II region on its SW tip. Many point sources identified as luminous stars can be seen. To the SW of the brightest H II region, two stellar clusters with resolved stars are present. The luminous stars and the H II regions are aligned suggesting, just as in SBS 0335-052, propagating star formation (from the NE to the SW). The mode of star formation in the two BCDs is different however. While SBS 0335-052 makes stars in luminous super-star clusters, star-formation in SBS 1415+437 appears to be less extreme and is more similar to that in I Zw 18 (Hunter & Thronson 1995). The superior spatial resolution of the HST allows to resolve individual stars and construct color-magnitude diagrams to study the stellar populations in the BCD. To check the hypothesis of propagating star formation, we have derived stellar ages for 6 separate regions in the BCD, labeled from I to VI as shown in Fig. 7.

The (V-I) vs. I diagrams (uncorrected for extinction) for each region are shown in Fig. 8 together with stellar isochrones by Bertelli et al. (1994) for a heavy element abundance equal to 1/20 the solar value. Each isochrone is marked by the logarithm of the age in years. The region of the asymptotic giant branch (AGB) stars is shown by a dashed line while the observational limits are shown by dotted lines. We adopt V = 27.5 mag and I = 27 mag as completeness limits. Since we do not correct for extinction, the stars are really brighter and bluer so that the ages given by the isochrones are only upper limits. Inspection of Fig. 8 shows that there is a clear age gradient from region I to region VI. Region I is the youngest (about 5 Myr), containing only main-sequence stars. Region II is more evolved. It contains red supergiants and has an age of ≥10 Myr. Region III includes the brightest H II region and shows a mixture of stellar populations. The bulk of the stars in region III have ages ranging between ≤10 Myr and 100 Myr. The red stars with V-I between 1.0 and 1.8 are likely to be red supergiants rather than AGB stars. The stellar populations in regions IV - VI are similar to those in region III except for the fact that very young populations with age less 10 Myr are no more present. We emphasize that the properties of the stellar populations in SBS 1415+437 are quite different from those in other nearby low-metallicity dwarf galaxies with HST color-magnitude diagrams. In those dwarfs, very red AGB stars are present indicating a larger age (e.g. Dohm-Palmer et al. 1997; Schulte-Ladbeck et al. 1998). Two general conclusions can be obtained from the above color-magnitude analysis:

  1. star formation in SBS 1415+437 is propagating from the NE to the SW;
  2. there is no evidence for stars older than ∼100 Myr in the BCD.

[Click here to see Fig. 7-8!]

3.2. Age from spectral evolutionary synthesis models

The second conclusion is supported by evolutionary synthesis modeling of the spectrophotometric data of regions III to VI. As an example, we show in Fig. 9 the spectrum of region V. To illustrate the effect of extinction, we show in the upper panel the spectrum uncorrected for extinction, and in the lower panel the spectrum corrected for extinction as derived fom the Balmer decrement (C(Hβ) = 0.26 which corresponds to AV = 0.57 mag and AI = 0.35 mag). To estimate quantitatively the age of each region, we calculate a grid of spectral energy distributions (SED) for stellar populations with ages varying between 10 Myr and 20 Gyr and heavy element abundance Zsun/20, using isochrones from Bertelli et al. (1994) and the compilation of stellar atmosphere models from Lejeune et al. (1998). A Salpeter IMF with slope -2.35, an upper mass limit of 120 Msun and a lower mass limit of 0.6 Msun were adopted. For stellar populations with age less than 10 Myr we use theoretical spectral energy distributions by Schaerer & Vacca (1998) for a heavy element abundance Zsun/20 and a Salpeter IMF. The stellar emission in SBS 1415+437 is contaminated by emission of ionized gas from supergiant H II regions. Therefore, to study the stellar composition in the BCD, it is necessary to produce a synthetic SED which includes both stellar and ionized gaseous emission. We have chosen not to calculate the contribution of ionized gas emission from the model value for the Lyman continuum luminosity. Rather, we add the gaseous spectral energy distribution calculated from the observed line fluxes and equivalent widths to the calculated stellar spectral energy distribution. The contribution of the gaseous emission is scaled to the stellar emission by the ratio of the observed equivalent width of the Hβ emission line to the equivalent width of Hβ expected for pure gaseous emission. The contribution of bound-free, free-free, two-photon continuum emission has been taken into account for the spectral range from 0 to 5 µm. Emission lines with intensities derived from the spectra in the range λλ3700 - 7500Å are then added. The effect of gaseous emission is important for region III but it has a minor influence in other regions as indicated by the small Hβ emission line equivalent width (EW(Hβ) = 18 Å in region V). Therefore, we do not take into account gaseous emission in regions IV to VI. It is clear from Fig. 9, the best fit to the spectrum corrected for extinction (lower panel) gives an age between 30 and 100 Myr for region V (the models are labelled by the logarithm of the age in years). A similar analysis for the other regions give the same answer: none contains stellar populations older than 100 Myr. Thus SBS 1415+437, just like SBS 0335-052, is also a young galaxy.

[Click here to see Fig. 9!]

4. Heavy element abundances in blue compact galaxies and constraints on their evolutionary status

4.1. α elements

The study of the variations of one chemical element relative to another is crucial for our understanding of the chemical evolution of galaxies and for constraining models of stellar nucleosynthesis and the shape of the initial mass function. In the case of BCDs, it is particularly important for understanding their evolutionary status, whether they are young or old. Izotov & Thuan (1998c) have obtained very high-quality ground-based spectroscopic observations of 54 supergiant H II regions in 50 low-metallicity blue compact galaxies with oxygen abundances 12 + log O/H between 7.1 and 8.3 (Zsun/50≤Z≤Zsun/4). They use the data to determine abundances for the elements N, O, Ne, S, Ar and Fe. They also analyze Hubble Space Telescope (HST) Faint Object Spectrograph archival spectra of 10 supergiant H II regions to derive C and Si abundances in a subsample of 7 BCDs. The best studied and most easily observed element in BCDs is oxygen. Nucleosynthesis theory predicts it to be produced only by massive (M≥9 Msun) stars. We shall use it as the reference chemical element and consider the behavior of heavy element abundance ratios as a function of oxygen abundance. Figure 10 shows the dependence of the abundance ratios Ne/O, Si/O, S/O and Ar/O on the oxygen abundance. The elements neon, silicon, sulfur and argon are all products of α-processes during both hydrostatic and explosive nucleosynthesis in the same massive stars which make oxygen. Therefore, the Ne/O, Si/O, S/O and Ar/O ratios should be constant and show no dependence on the oxygen abundance. As predicted by stellar nucleosynthesis theory, none of the above heavy element-to-oxygen abundance ratios depend on oxygen abundance. The mean values of these element abundance ratios are directly related to the stellar yields and thus provide strong constraints on the theory of massive stellar nucleosynthesis (Izotov & Thuan 1998c).

[Click here to see Fig. 10!]

4.2. Carbon

Carbon is produced by both intermediate (3 Msun≤M≤8 Msun) and high-mass (M≥9 Msun) stars. Since C is a product of hydrostatic burning, the contributions of SNe Ia and SNe II are small. Therefore, the C/O abundance ratio is sensitive to the particular star formation history of the galaxy. It is expected that, in the earliest stages of galaxy evolution, when metallicity is still very low, carbon is mainly produced by massive stars, so that the C/O abundance ratio is independent of the oxygen abundance, as both C and O are primary elements. At later stages, at slightly higher metallicities, intermediate-mass stars add their carbon production, so that an increase in the C/O ratio is expected with increasing oxygen abundance.

Earlier studies did not conform to these expectations. Garnett et al. (1995) found a continuous increase of log C/O with increasing log O/H in their sample of metal-deficient galaxies, a relationship which could be fitted by a power law with slope 0.43. Subsequent HST FOS observations of I Zw 18 (Garnett et al. 1997) have complicated the situation even more. It was found that I Zw 18 bucks the trend shown by the other low-metallicity objects. Although it has the lowest metallicity known, it shows a rather high log C/O, significantly higher than those predicted by massive stellar nucleosynthesis theory. This led Garnett et al. (1997) to conclude that carbon in I Zw 18 has been enhanced by an earlier population of lower-mass stars and, hence, despite its very low metallicity, I Zw 18 is not a ``primeval'' galaxy. We have reanalyzed the data for I Zw 18. The use of new high signal-to-noise ratio MMT spectroscopic observations of I Zw 18 yields a much higher electron temperature (by ∼2000 K) within the FOS aperture and in a much lower C/O abundance ratio. Figure 11a shows log C/O against 12 + log O/H for the BCDs in the Izotov & Thuan (1998c) sample. It is clear that, in contrast to previous results, log C/O is constant in the extremely low-metallicity range, when 12 + log O/H varies between 7.1 and 7.6, as expected from the common origin of carbon and oxygen in massive stars. Furthermore, the dispersion of the points about the mean is very small: < log C/O > = -0.78±0.03. This mean value is in very good agreement with that of ∼-0.8 predicted by massive stellar nucleosynthesis theory (Woosley & Weaver 1995). Two models with Z = 0 and Z = 0.01 Zsun are shown by horizontal lines in Fig. 11a. They are in good agreement with the observations. At higher metallicities (12 + log O/H > 7.6), there is an increase in log C/O with log O/H and also more scatter at a given O/H, which is attributed to the carbon contribution of intermediate-mass stars in addition to that of massive stars.

[Click here to see Fig. 11!]

4.3. Nitrogen

The basic nucleosynthesis process is well understood - nitrogen results from CNO processing of oxygen and carbon during hydrogen burning - however the nature of the stars mainly responsible for the production of nitrogen remains uncertain. If oxygen and carbon are produced not in previous generation stars, but in the same stars prior to the CNO cycle, then the amount of nitrogen produced is independent of the initial heavy element abundance of the star, and its synthesis is said to be primary. On the other hand, if the ``seed'' oxygen and carbon are produced in previous generation stars and incorporated into a star at its formation and a constant mass fraction is processed, then the amount of nitrogen produced is proportional to the initial heavy element abundance, and the nitrogen synthesis is said to be secondary. In this case, the N/O ratio should increase linearly with the O abundance. This behavior is seen in high-metallicity H II regions in spiral galaxies with 12 + log O/H ≥ 8.4. As all of the BCDs discussed here are less metal-rich, only primary N concerns us here.

Figure 11b shows the behavior of the N/O abundance ratio as a function of the O/H ratio. Two remarkable facts can be seen. First, at low abundances (12 + log O/H ≤ 7.6), log N/O is constant (∼ -1.60) and with an extremely small scatter (±0.02 dex) at a given O abundance, implying that nitrogen is produced as a primary element by massive but not by intermediate-mass stars as commonly thought (Thuan et al. 1995; Izotov & Thuan 1998c). N production by intermediate-mass stars would introduce a time-delay as large as 5·108 yr with respect to the primary production of oxygen by massive stars, which would introduce a larger than observed scatter in N/O. Second, the value of log N/O increases above ∼ -1.60 along with the scatter at a given O abundance in BCDs with 7.6 < 12 + log O/H < 8.2. This increase in log N/O and its larger scatter is interpreted as due to the additional contribution of primary nitrogen produced by intermediate-mass stars, on top of the primary nitrogen produced by massive stars.

4.4. Iron

The iron abundance in BCDs was first discussed by Thuan et al. (1995). Their small sample of 7 galaxies has been considerably increased (38 BCDs) by Izotov & Thuan (1998c) who found that oxygen in these galaxies is overproduced relative to iron, as compared to the Sun: [O/Fe] = log (Fe/O)sun - log (Fe/O) = 0.40±0.14 (Fig. 11c). This value is in very good agreement with the [O/Fe] observed for Galactic halo stars, implying that the origin of iron in low-metallicity BCGs and in the Galaxy prior the formation of halo stars is similar, and supporting the scenario of an early chemical enrichment of the galactic halo by massive stars.

4.5. Chemical constraints on the age of BCDs

Because O, Ne, Si, S and Ar are made in the same high-mass stars, their abundance ratios with respect to O are constant and not sensitive to the age of the galaxy. By contrast, C, N and Fe can be produced by both high and lower-mass stars, and their abundance ratios with respect to O can give important information on the evolutionary status of BCDs.

The constancy of [O/Fe] for the BCDs and its high value compared to the Sun suggests that all iron was produced by massive stars, i.e. in SNe II only. Since the time delay between iron production from SNe II and SNe Ia is about 1-2 Gyr, it is likely that BCDs with oxygen abundance less than 12 + log O/H ∼ 8.2 are younger than 1-2 Gyr. The behavior of the C/O and N/O ratios as a function of oxygen abundance puts more stringent constraints on the age of BCDs (Izotov & Thuan 1998c). As we have seen, this behavior is very different whether a BCD has 12 + log O/H smaller or greater than 7.6 (Zsun/20). The C/O and N/O abundance ratios in BCGs with 12 + log O/H ≤ 7.6 are independent of the oxygen abundance and show a very small scatter about the mean value. This small scatter rules out any time-delay model in which O is produced first by massive stars and C and N are produced later by intermediate-mass stars, and supports a common origin of C, N and O in the same first-generation massive stars. Thus, it is very likely that the presently observed episode of star formation in BCDs with 12 + log O/H ≤ 7.6 is the first one in the history of the galaxy and the age of the oldest stars in it do not exceed ∼100 Myr (the lifetime of a 9 Msun star is ∼40 Myr). This conclusion is supported by the detailed analysis of some of these very metal-deficient galaxies, as discussed in Secs. 2 and 3.

The situation changes for BCDs with Z > Zsun/20. The scatter of the C/O and N/O ratios increases significantly at a given O abundance, which can be interpreted as due to the additional production of primary N by intermediate-mass stars, on top of the primary N production by high-mass stars. Thus, since it takes at least 500 Myr (the lifetime of a 2-3 Msun star) for C and N to be produced by intermediate-mass stars, BCDs with 12 + log O/H > 7.6 must have had several episodes of star formation before the present one and they must at least be older than ∼100 Myr. This conclusion is in agreement with photometric studies of these higher metallicity BCGs which, unlike their very low-metallicity counterparts, have a red old instead of a blue young underlying stellar component (Loose & Thuan 1985; Papaderos et al. 1996).

5. The escape of Lyα photons and P Cygny profiles in BCDs

5.1. Lyα emission and absorption

The study of nearby young galaxies can also shed light on a long standing problem concerning the Lyman-alpha emission of primeval galaxies. In any galaxy formation scenario, young galaxies are predicted to show strong Lyα emission (rest frame equivalent width of ∼100 Å) associated with the formation of a large number of massive ionizing stars (e.g. Charlot & Fall 1993). Yet, despite intensive searches, the predicted population of Lyα primeval galaxies remained elusive (Pritchet 1994). The absence of Lyα emission in high-redshift galaxies is reminiscent of the behavior of Lyα emission in nearby starburst and BCD galaxies, which is either absent or greatly diminished. The Lyα/Hβ line intensity ratio in those galaxies with detected Lyα emission does not exceed 10, significantly lower than the theoretical recombination ratio of 33. Some galaxies show strong Lyα absorption rather than emission. The favored explanation for such a reduction in Lyα emission from recombination values is redistribution of Lyα photons by multiple scattering in the H I envelope or absorption of these Lyα photons by dust in the star-forming region. The latter mechanism would imply increasing Lyα/Hβ line intensity ratios with decreasing metallicities, since presumably low-metallicity objects contain less dust, and hence suffer less destruction of Lyα photons (Terlevich et al. 1993). HST observations of the two most metal-deficient BCDs known, I Zw 18 (Kunth et al. 1994) and SBS 0335-052 (Thuan et al. 1997, Sect. 2), show Lyα not in emission but absorption, which goes against a Lyα strength-metallicity anticorrelation. Lequeux et al. (1995), in their study of the BCD Haro 2 (Zsun/3) which shows Lyα in emission, have argued that Lyα photons can escape when the neutral material where the absorption occurs is outflowing with a velocity of ≤200 km s-1 with respect to the star-forming region. The Lyα emission is redshifted with respect to both the H II region and the expanding absorbing shell, the motion of which is probably powered by stellar winds and supernovae. This explanation does not apply to the case of the BCD T1214-277 (Zsun/23) whose HST GHRS UV spectrum is shown in Fig. 12. It clearly shows Lyα in emission which makes it the lowest metallicity galaxy known with detected Lyα emission. Its Lyα-to-Hβ intensity ratio is ∼4, while its equivalent width is ∼70 Å, the highest found in a star-forming galaxy. Contrary to the case of Haro 2, Lyα emission is not redshifted with respect to the H II gas velocity. Thus the escape of Lyα photons in T1214-277 is not a consequence of the motions of the neutral H I envelope.

As for SBS 0335-052, dust extinction may play some role as it is directly seen in HST WFPC2 images (Thuan et al. 1997, Fig. 3). While there is evidence for fast gas motions in SBS 0335-052 with velocities up to ∼1500 km s-1, the H I gaseous envelope appears to be static with respect to the H II region, as the 21 cm and emission-line velocities are in good agreement. Thus, with its extremely large H I column density, the redistribution of Lyα photons in SBS 0335-052 by multiple scattering over the large volume of the H I cloud probably plays also an important role in diminishing the intensity of the Lyα line. The orientation of the H I cloud may also play a role. In the case of SBS 0335-052, the H I envelope is reasonably flattened (Pustilnik et al. 1998) suggesting it is seen nearly edge-on. In that case, Lyα photons escape more easily along directions perpendicular to the line of sight than along it. Moreover, as discussed by Giavalisco et al. (1996), the escape of Lyα photons may be controlled not only by the geometry but also by the porosity of the neutral gas.

In summary, there is no unique mechanism which controls the appearance of Lyα emission in nearby young dwarf galaxies. Dust extinction may play a role, but the velocity structure of the H I gas, the orientation of the H I cloud and its porosity may also play determining functions. The fact that some BCDs do not show Lyα in emission implies that Lyα searches for high-redshift galaxies will always be incomplete.

5.2. P Cygni profiles

A most interesting result concerning young galaxies is the detection in the spectrum of some of them of strong Si IV λ1394, λ1403 lines with P Cygni profiles. Figure 13 shows these profiles for SBS 0335-052 (Sect. 2). A P Cygni profile can also be seen in the spectrum of T1214-277 for the N V λ1240 line (Fig. 12). This makes these two BCDs the two most metal-deficient galaxies known with P Cygni profiles. These results are somewhat surprising because we do not expect to see P Cygni in such very low metallicity environments. A minimum amount of metals is needed to provide the necessary opacity to drive the stellar winds originating from massive O stars and responsible for the P Cygni profiles. A possible way out comes from the HST WFPC2 observation of SBS 0335-052 by Thuan et al. (1997) which shows that star formation in the BCD is self-propagating, resulting in a chain of 6 main super-star clusters roughly aligned in a northwest - southeast direction, and with age decreasing from ∼30 Myr to ∼4 Myr. Thus star formation in SBS 0335-052 has proceeded in 6 separate bursts of duration less than a few Myr. We can imagine a scenario where the stars born in the later bursts and responsible for the stellar winds, formed from gas already enriched in heavy elements by supernovae resulting from the previous bursts, although it is not clear whether the metals produced in supernovae have had enough time to mix with the pristine gas. Another possibility is to postulate that somehow the evolution of massive stars leads to an increase of their surface metallicity and hence to the onset of a stellar wind.

The terminal wind velocity as measured from the blue absorption edge of the P Cygni profiles is ∼2000 km s-1 for T1214-277, at the lower range of velocities (2000-4000 km s-1) obtained for the BCDs studied by Fanelli et al. (1988) and for massive stars in the Galaxy (Zsun) and the LMC (Zsun/3) (Prinja et al. 1990). As for SBS 0335-052, it has a substantially lower terminal velocity of ∼500 km s-1, below the velocity range (1200-1500 km s-1) observed for SMC stars (Zsun/8) (Puls et al. 1996). These results suggest a decrease of terminal wind velocities with decreasing metallicities.

[Click here to see Fig. 12-13!]

6. The primordial helium abundance

In the standard hot big bang model of nucleosynthesis (SBBN), four light isotopes, D, 3He, 4He and 7Li, were produced by nuclear reactions a few seconds after the birth of the Universe. Given the number of relativistic neutrino species and the neutron lifetime, the abundances of these light elements depend on one cosmological parameter only, the baryon-to-photon ratio η, which in turn is directly related to the density of ordinary baryonic matter Ωb. Thus precise abundance measurements of the four light elements can provide not only a stringest test of the consistency of SBBN, but also information about the mean density of ordinary matter in the Universe.

Very metal-deficient BCDs constitute an ideal laboratory in which to measure the primordial helium abundance. The primordial mass fraction Yp of 4He is usually derived by extrapolating the Y - O/H and Y - N/H correlations to O/H = N/H = 0. Here we use the data of Izotov et al. (1994, 1997a) and Izotov & Thuan (1998b) to construct a sample of 45 H II regions appropriate for the determination of Yp. Our sample constitutes one of the largest and most homogeneous (obtained, reduced and analyzed in the same way) data set now available for the determination of Yp. This sample is the same as the one used to analyze heavy element abundances in Sect. 4. Linear regressions of the Y - O/H and Y - N/H relations, with Ys determined from a self-consistent treatment of the five brightest optical He I emission lines, gives Yp = 0.244±0.001 (Fig. 14, Izotov & Thuan 1998b). This value agrees very well with the mean Y of the two most metal-deficient BCDs known [I Zw 18 (Yp = 0.242±0.009, Izotov & Thuan 1998a) and SBS 0335-052 (Yp = 0.249±0.004, Izotov & Thuan 1998b)], which is Ymean = 0.245±0.006. Values as low as Yp = 0.234 or Yp = 0.230, as those obtained by Olive et al. (1997) are excluded. Part of the difference comes from the fact that previous authors use in the determination of Yp a Y derived for the NW component of I Zw 18 which is subject to strong underlying He I stellar absorption and gives an erroneously low value. Izotov & Thuan (1998a) have argued that the Y value for the SE component of I Zw 18, where underlying stellar absorption is much less important, should be used instead.

Our higher primordial helium mass fraction is consistent with deuterium abundance measurements in high-redshift Lyα absorbing systems (Tytler et al. 1996 as corrected by turbulence effects in the clouds by Levshakov et al. 1998) and lithium abundance measurements in low-metallicity halo stars (Bonifacio & Molaro 1997). Figure 15 gives a baryon-to-photon ratio η = 4.0±0.5, which corresponds to a baryon mass fraction Ωb h250 = 0.058±0.007. We derive a slope dY/dZ = 2.3±1.0, considerably smaller than those derived before. With this smaller slope and taking into account the errors, chemical evolution models with an outflow of well-mixed material can be built for star-forming dwarf galaxies which satisfy all the observational constraints.

[Click here to see Fig. 14-15!]

References

Links (back/forward) to:
First version: 21stJuly,1998
Last update: 08thOctober,1998
Jochen M. Braun   &   Tom Richtler
 (E-Mail: jbraun|richtler@astro.uni-bonn.de)