EDP Sciences
Free access
Volume 578, June 2015
Article Number L2
Number of page(s) 4
Section Letters
DOI http://dx.doi.org/10.1051/0004-6361/201425287
Published online 29 May 2015

© ESO, 2015

1. Introduction

A fundamental question in Astronomy is the question of the sources of the “First Light” ending the Cosmic Dark Ages and beginning the process of reionisation. The James Webb Space Telescope (JWST) promises direct access to this critical period via observations of the first star-forming galaxies at redshifts z ≳ 10. Dual Lyαλ1216 and He iiλ1640 emission is believed to be the main indicator of the first (Pop III) stars formed from pristine gas in these objects (e.g. Johnson et al. 2009). The reason is that only at extremely low metallicities (Z ≲ 10-5), young massive stars are believed to be hot enough to excite He ii in their surrounding H ii regions (e.g. Schaerer 2002, 2003). The investigation of star-forming galaxies with He ii emission at moderate redshifts, which are accessible with current ground-based instrumentation, is thus an important preparation for future studies of the first star-forming galaxies with the JWST.

Such a study has recently been performed by Cassata et al. (2013), who identified different groups of He ii emitters with narrow and broad emission lines in a sample of star-forming galaxies at redshifts 2 <z < 4.6. While the broad emission is usually attributed to stellar emission from Wolf-Rayet (WR) stars, Cassata et al. interpreted the narrow emission as being due to ionising radiation from Pop III stars that must have formed in pockets of residual pristine gas at relatively low redshift, or a stellar population that is rare at z ~ 0 but more common at z ~ 3.

In this letter we discuss the possibility that the narrow He ii emission might also be of stellar origin. To this purpose we discuss in Sect. 2 how the spectral properties of WR-type emission line stars are expected to change at low Z, estimate the expected line shapes and fluxes in Sect. 3, and compare our results with the observations of Cassata et al. in Sect. 4. Conclusions are drawn in Sect. 5.

2. He ii emission from WR stars at different metallicities Z

The emission-line spectra of WR stars are formed in their strong, optically thick stellar winds. As a result of the large optical depth of these winds, ions like He iii recombine already within the wind, leading to recombination cascades with strong emission lines between excited levels of the subordinate ions. Prominent examples are He iiλ1640 and C ivλ1550 in the UV, and He iiλ4686 and C ivλ5808 in the optical wavelength range. The width of the emission lines is mainly determined by the terminal wind speed ν of the stellar wind. To understand why we naturally expect two populations of WR stars with broad and narrow emission lines at low Z, we need to discuss the origin of the different types of WR-type stellar winds.

2.1. WR populations in the Galaxy and LMC

Comprehensive studies of the local population of WR stars of the nitrogen sequence (WN stars) have been performed for the Galaxy by Hamann et al. (2006) and for the Large Magellanic Cloud (LMC) by Hainich et al. (2014). Both works identified two different groups of WN stars. The first group are classical WN stars with H-poor/H-free surface compositions and hot temperatures up to 140 kK, in agreement with a core He-burning evolutionary stage. The second group are the so-called WNh stars with relatively H-rich surface compositions and cooler temperatures, in agreement with a core H-burning evolutionary stage. The WNh stars are further characterised by remarkably high luminosities in excess of ~105.9 L, while the group of classical WN stars has luminosities below this value.

The dividing luminosity at roughly 105.9 L suggests that the classical WN stars are core He-burning objects in a post-red supergiant (post-RSG) phase. This picture is further supported by the work of Sander et al. (2012), who found that the WR stars of the carbon sequence (WC stars), which are showing the products of He-burning at their surface, form a natural succession of the sequence of classical WN stars in the HR diagram (HRD).

In this work we are mainly interested in the group of WNh stars with luminosities in the range 105.9 LL ≲ 106.9 L (cf. Hamann et al. 2006; Crowther et al. 2010; Hainich et al. 2014; Bestenlehner et al. 2014). The HRD positions of these objects suggest that they are very massive stars (VMS) with initial masses in excess of ~100 M, and reaching up to 300 M (cf. Vink et al. 2015). In their LMC sample, Hainich et al. (2014) found 12% of all putatively single WN stars in the WNh stage. Combined with their high luminosities, this suggests that WNh stars have substantial impact on the integrated He ii emission of star-forming galaxies.

2.2. WR mass loss at low metallicities Z

Observational evidence for a Z-dependence of WR mass-loss has been found by comparisons of the sample of WC stars with known distances in the Galaxy and LMC by Gräfener et al. (1998) and Crowther et al. (2002). In the following, we invoke the results of theoretical studies to understand how the properties of WR stars are expected to change over a broad range of Z.

The winds of classical WR stars have been modelled by Gräfener & Hamann (2005, for early WC subtypes) and Vink & de Koter (2005, for late WC and WN subtypes). Although there are still problems to reproduce the winds of intermediate spectral subtypes (cf. Gräfener & Hamann 2005), these works suggest that the winds of WR stars are radiatively driven, predominantly by millions of spectral lines of the iron-group elements. As a consequence, the mass-loss rates of WR stars are expected to depend on the environment metallicity Z (Vink & de Koter 2005).

The winds of luminous WNh stars have been modelled by Gräfener & Hamann (2008) and Vink et al. (2011). These works suggest that the proximity to the Eddington limit plays a key role for the formation of WR-type stellar winds. This has been further supported in systematic studies of samples of VMS in young clusters by Gräfener et al. (2011) and Bestenlehner et al. (2014). According to Gräfener & Hamann (2008), Eddington factors of the order of Γe ≈ 0.5 are required to initiate WR-type stellar winds at solar metallicity Z (here Γe = (χeL)/(4πcGM) denotes the classical Eddington factor that is defined with respect to the opacity of free electrons χe). The required Eddington factors suggest that the masses of WNh stars are higher than those of core He-burning stars, meaning that WNh stars are indeed VMS in the phase of core H-burning.

How does the spectral appearance of the two groups of WR stars change at low metallicities? As a result of their dependence on the iron-group opacities, the winds of classical WR stars are expected to be weaker at low Z (Vink & de Koter 2005), resulting in significantly weaker emission lines (Crowther & Hadfield 2006). However, the situation is slightly different for WC stars because their surfaces are substantially self-enriched with the products of He-burning, namely C, O, and most likely Ne and Mg. According to Vink & de Koter (2005), these elements gain in importance for Z ≲ 0.1 Z, and may even dominate WR wind driving for Z ≲ 10-3 Z. WC stars may thus be the main source of broad He ii emission from classical WR stars at low Z. Because of the high carbon abundances of WC stars, this emission will most likely be accompanied by strong C iv emission.

For the group of very massive WNh stars, the dependence on the environment metallicity Z has been discussed by Gräfener & Hamann (2008). According to their models, the proximity to the Eddington limit is mandatory to support WR-type winds in this regime. The required Eddington factors increase for decreasing metallicity, from Γe ≈ 0.5 for Z = Z, to Γe ≈ 0.85 for Z = 0.01 Z. As a consequence, the resulting effective escape velocities decrease, and very low terminal wind velocities of only few 100 km s-1 are predicted for WNh stars in this regime. At the same time, the mass-loss rates do not change significantly, resulting in significantly higher wind densities and much stronger and narrower WR emission lines than for classical WR stars. Very massive WNh stars at low Z may thus be the sources of the narrow He ii emission observed by Cassata et al. (2013). In contrast to classical WC stars, they have very low carbon abundances and are not expected to show detectable C iv emission.

3. Expected He ii emission from populations of VMS

Because of their short lifetimes (2–3 Myr), VMS are predominantly found in young massive clusters. To estimate the line fluxes expected from populations of VMS at cosmological distances, we start with an examination of young massive clusters in our neighbourhood and their stellar He ii emission in Sect. 3.1 and extrapolate our results to low Z in Sect. 3.2.

3.1. VMS in young massive clusters

Wofford et al. (2014) recently found evidence for He ii emission from VMS in the nearby starburst galaxy NGC 3125 with roughly LMC metallicity. For the young cluster A1 they had to invoke a population of VMS to explain the large observed He ii equivalent width (EW(He ii) ~7 Å, cf. also Chandar et al. 2004) as the cluster is too young (3–4 Myr) to host evolved WR stars. Based on different UV extinction corrections, they determined an intrinsic He ii line luminosity in the range log (LHe ii) = 39.1 ... 40.0 (in erg/s) and a cluster mass of log (Mcl/M) = 5.1 ... 5.9.

To investigate whether VMS can account for the large EW(He ii), we also investigated R136 in the centre of 30 Dor of the LMC, the most prominent starburst-like cluster in our neighbourhood that is spatially resolved. Using the integrated IUE spectrum of the central 5 pc from Vacca et al. (1995), we measured a similarly high EW(He ii) ~5 Å for this object.

For the sum of the seven known WNh stars in this region (five WN5h, one WN6(h), and one WN9ha; cf. Crowther et al. 2010), we estimate in total log (LHe ii) ≈ 38.6. This estimate is based on a synthetic model for the LMC WN5h star VFTS 682 from Bestenlehner et al. (2011) with log (L/L) = 6.5 and log (LHe ii) = 37.5, which was scaled to the total luminosity of the seven objects (log (L/L) = 7.58 ± 0.03, as obtained from Crowther et al. 2010; Hainich et al. 2014; Bestenlehner et al. 2014). We note that this value is uncertain as for some of the stars different luminosities have been obtained in different works, and one of the stars, Mk 34, is a known binary.

To estimate the intrinsic continuum flux, we used Starburst99 models (Leitherer et al. 1999) that are similar to those of Wofford et al. (2014, cf. their Sect. 3.5) but for Z = 0.008. For a cluster mass of log (Mcl/M) ≈ 4.7 (Doran et al. 2013)1 and an age of 2–4 Myr, we obtain a continuum flux log (LUV) ≈ 37.74 (in erg s-1 Å-1). Together with our estimate of log (LHe ii) ≈ 38.6, this results in EW(He ii) ≈ 7 Å, in good agreement with the observations of R 136 and NGC 3125-A1. We note that the intrinsic continuum flux estimated by Vacca et al. (1995) is even lower (log (LUV) ≈ 37.33), which may be attributed to uncertainties in the UV extinction and the precise pointing of their observations.

We conclude that the He ii emission of very massive WNh stars at ZZLMC is probably strong enough to explain the observed strength of the He ii emission of NGC 3125-A1 and R136. In the following, we use the approximate average luminosity of the seven WNh stars in R 136 to define a representative luminosity of log (LWN5h/L) = 6.7. With the numbers above, we then obtain (1)where NWN5h denotes the equivalent number of WN5h stars.

The observed He ii line widths (FWHM) lie in the range of 1000–1400 km s-1 for NGC 3125-A1 (Chandar et al. 2004; Wofford et al. 2014) and ~2500 km s-1 for R 136. The individual WN5h stars in R 136 have a similar spectral morphology as VFTS 682, for which we estimate FWHM ~ 2000 km s-1. The variations in the observed FWHM may indicate that the real stellar populations are more heterogeneous than we assumed in our simplified approach.

3.2. He II emission at low Z

To estimate LHe ii for WNh stars at low Z, we computed synthetic UV spectra using the wind models of Gräfener & Hamann (2008). To obtain a more realistic FWHM, we used a line broadening velocity of 10 km s-1 instead of the artificially enhanced value of 100 km s-1 in the original work. In Fig. 1 we show three models with Z = Z, 0.1 Z, and 0.01 Z (cf. model grid 2 in Table 2 of Gräfener & Hamann 2008). For these models, Γe has been adjusted to reproduce very similar mass-loss rates for all three metallicities2. As discussed in Sect. 2.2, the models with low Z have lower terminal wind velocities ν and thus higher wind densities. This leads to stronger and narrower He ii emission at lower Z (cf. Fig. 1). The resulting line fluxes and FWHM are indicated in Fig. 1. We note that the wind models seem to underestimate ν at Z = Z (cf. the discussion in Gräfener & Hamann 2008). The important point is thus the predicted qualitative behaviour, that is, the strong decrease of ν vs. Z.

thumbnail Fig. 1

Synthetic UV spectra for different metallicities Z from wind models for WNh stars from Gräfener & Hamann (2008). The presented models are computed for a luminosity of log (L/L) = 6.3, a stellar temperature T = 44.7 kK, and have very similar mass-loss rates (~1.8 × 10-5M yr-1). The given He ii line luminosities have only been determined for the narrow-emission components of the He ii line profiles. Measurement errors for the given LHe ii, FWHM, and EW probably amount to ~±0.05 dex (or ±10%).

Open with DEXTER

The He ii line profiles in Fig. 1 consist of a narrow emission component, a comparatively weak blue-shifted P-Cygni absorption, and broad line wings that are caused by multiple electron scattering of line photons in the dense ionised wind. In addition to the iron forest between 1300 and 1500 Å, which is most prominent at high Z, the spectra show C iv, N iv, and N v features whose strength decreases for lower Z. We note that the models have been computed for an N-enriched surface composition, which will affect the relative strength of the C iv vs. N iv and N v features. Furthermore, the resulting He ii line fluxes will depend on the adopted stellar temperature T.

For low metallicities Z = 0.1 and 0.01, we find He ii line luminosities of the order of log (LHe ii) ≈ 37.5 for a given stellar luminosity of log (L/L = 6.3). Adopting the same representative stellar luminosity as in Eq. (1), we obtain (2)that is, a slightly higher value than the one that we obtained for LMC metallicity.

While LHe ii increases for lower Z, the strength of the observable metal lines decreases. Most notably, all metal lines are substantially weaker than He ii  and will not be detectable in the integrated spectra of He ii emitting galaxies, as observed by Cassata et al. (2013).

We conclude that WNh stars at low Z can produce narrow He ii emission with similar or slightly stronger line fluxes than those discussed in Sect. 3.1 if the high required Eddington factors Γe can be maintained. Theoretically, high effective Γe are indeed expected due to increased stellar rotation rates at low Z (e.g., Meynet & Maeder 2002). Adopting similar relative numbers and luminosities of WNh stars as in Sect. 3.1, we would thus expect similar or slightly higher line fluxes and EWs than those discussed in Sect. 3.1 for young starbursts at low Z.

We also note that rotating massive stars in extremely metal-poor environments (Z ~ 10-5... 10-8 Z) are expected to reveal primary nitrogen with mass fractions of about 1% at their surface (Yoon & Langer 2005; Meynet et al. 2006). Gräfener & Hamann (2008) showed that such self-enriched objects can develop very similar winds as WNh stars at low metallicity (roughly corresponding to Z ~ 0.02 Z). We thus expect that even very early stellar generations can contain WNh stars with similar spectral properties as discussed here. For Pop III stars with zero metal content, Ekström et al. (2008) showed that qualitatively different internal mixing properties can prevent strong self-enrichment, that is, their mass-loss properties may be different (but cf. also the discussion in Vink & de Koter 2005).

4. Narrow He ii emission in star-forming galaxies at moderate redshifts

A systematic study of star-forming galaxies with He ii emission at moderate redshifts has been performed by Cassata et al. (2013), based on data from the VIMOS VLT Deep Survey (VVDS; Le Fèvre et al. 2005). In the total VVDS sample of ~45 000 galaxy spectra, Cassata et al. found 277 star-forming galaxies with secure redshifts in the range 2 <z < 4.6, of which 39 show He iiλ1640 in emission. They divided this sample into four groups of 11 objects with narrow He ii lines, 13 objects with broad He ii lines, 12 possible He ii emitters, and 3 AGN, to distinguish between high-velocity outflows of WR stars, AGN, or supernova-driven winds, and the narrow nebular emission that is expected for Pop III stars.

Cassata et al. found that narrow and broad emitters show qualitatively different spectra. In particular, they found that only broad He ii emitters (with FWHM(He ii) ≥ 663 km s-1) show indications of underlying P-Cygni type C iv λ1550 emission. As we discussed in Sect. 2.2, this emission is most likely a signature of stellar-wind emission from classical WR stars, predominantly originating from WC subtypes at low Z. In the group of 11 narrow emitters, they only found two such cases, in line with our prediction of very weak C iv emission for WNh stars at low Z. Our predicted FWHM(He ii) for such objects in Sect. 3.2 is clearly within the limit for narrow emitters from Cassata et al., even if we take possible uncertainties (up to a factor of ~2) into account.

Excluding AGN, Cassata et al. found He ii line luminosities in the range log (LHe ii) ≈ 40 ... 41.5, and EWs of 1 ... 7 Å. They noted that, based on the predicted nebular He ii emission for Pop III objects from Schaerer (2003), the star-formation rates (SFRs) required to produce the observed line fluxes are about two orders of magnitude lower than those determined from the spectral energy distribution (SFRsed).

Here we furthermore note that the majority of narrow He ii emitters is found at the bottom of the He ii luminosity distribution in a very narrow range of log (LHe ii) ≈ 40 ... 40.3 (cf. Fig. 10 in Cassata et al. 2013). This low-luminosity group is additionally restricted to the lowest redshifts of z ≈ 2.0 ... 2.5. This implies that these objects are near the detection limit. Only three objects are located outside the low-luminosity group, with log (LHe ii) ≈ 41.0 ... 41.5, and z = 2.5 ... 4.0.

The observed line fluxes for the low-luminosity group are remarkably similar to our expectations for massive young super-clusters with Mcl ~ 106M at low Z. Clusters in this mass range are near the top end of the cluster initial mass function (~2 × 106M), as determined for instance by Bastian (2008) for interacting galaxies and luminous IR galaxies. They may thus be the most luminous He ii sources of this kind. For comparison, the observed continuum fluxes log (LUV) = log (LHe ii/EW(He ii)) ≈ 39.2 ... 40.3 are similar to or higher than the Starburst99 prediction of log (LUV) ≈ 39.04 for a cluster with 106M and an age of 2–4 Myr at low Z = 0.0004. Test computations for galaxies with continuous SFRs at the same metallicity suggest LUV ≲ 40.2 + log (SFR/ (M yr-1)), meaning that the host galaxies could provide a substantial contribution to the observed continuum.

In our scenario, the continuum of a star-forming host galaxy could easily become so bright that it would prevent the detection of He ii. The target selection by Cassata et al. may thus be responsible for picking the most extreme and recent starburst events at low Z that are located in host galaxies with continua weak enough to enable the detection of He ii. The latter point places strong constraints on the continuum flux, which may be the reason for the narrow range of observed i-band magnitudes (23.45 ... 24.63 magi) compared to the wide range of derived SFRsed ≈ 1 ... 103M yr-1 of Cassata et al. (cf. their Table 2).

5. Conclusions

We discussed a new scenario for the origin of narrow He ii emission in star-forming galaxies at moderate redshifts, based on predictions of strong WR-type stellar winds with low terminal wind speeds for VMS at low metallicities (Z ~ 0.01 Z). We estimated the expected He ii emission and found that it is in line with recent observations by Cassata et al. (2013). In our scenario the observed emission originates from a population of VMS in one or more young super-clusters located within a variety of star-forming galaxies at low Z. This scenario is in accordance with the observed line fluxes and EWs, the large observed spread of SFRs, and the different C iv profiles of the narrow and broad emitters.

If our interpretation is correct, the narrow and broad He ii emitters identified by Cassata et al. both originate from massive stellar populations at low Z in distinct evolutionary stages, and there is no need to invoke the existence of Pop III stars at moderate redshifts to explain the observed narrow He ii emission. The fact that the emission of VMS is largely neglected in current population synthesis models will generally affect the interpretation of the integrated spectra of young stellar populations (cf. also Wofford et al. 2014).

Finally, we note that the same argumentation will hold for rotating VMS at extremely low Z (~10-5... 10-8 Z), for which similarly strong winds are predicted as a result of the self-enrichment with primary nitrogen. The He ii emission from such very early stellar generations of VMS may be detectable in future studies of star-forming galaxies at high redshifts with the JWST.


Doran et al. used the same Kroupa (2001) IMF as in this work.


As the luminosities and surface abundances are kept fixed, this change in Γe translates into different stellar masses M for each object.


All Figures

thumbnail Fig. 1

Synthetic UV spectra for different metallicities Z from wind models for WNh stars from Gräfener & Hamann (2008). The presented models are computed for a luminosity of log (L/L) = 6.3, a stellar temperature T = 44.7 kK, and have very similar mass-loss rates (~1.8 × 10-5M yr-1). The given He ii line luminosities have only been determined for the narrow-emission components of the He ii line profiles. Measurement errors for the given LHe ii, FWHM, and EW probably amount to ~±0.05 dex (or ±10%).

Open with DEXTER
In the text