A&A 459, 71-84 (2006)
DOI: 10.1051/0004-6361:20065622

VLT/GIRAFFE spectroscopic observations of the metal-poor blue compact dwarf galaxy SBS 0335-052E[*],[*]

Y. I. Izotov1 - D. Schaerer2,3 - A. Blecha2 - F. Royer4 - N. G. Guseva1 - P. North5


1 - Main Astronomical Observatory, Ukrainian National Academy of Sciences, Zabolotnoho 27, Kyiv 03680, Ukraine
2 - Observatoire de Genève, 51 Ch. des Maillettes, 1290 Sauverny, Switzerland
3 - Laboratoire d'Astrophysique Toulouse-Tarbes, UMR 5572, 14 Av. E. Belin, 31400 Toulouse, France
4 - GEPI, CNRS UMR 8111, Observatoire de Paris, 5 place Janssen, 92195 Meudon Cedex, France
5 - Laboratoire d'Astrophysique, École Polytechnique Fédérale de Lausanne (EPFL), Observatoire, 1290 Sauverny, Switzerland

Received 17 May 2006 / Accepted 24 July 2006

Abstract
Aims. We present two-dimensional spectroscopy of the extremely metal-deficient blue compact dwarf (BCD) galaxy SBS 0335-052E to study physical conditions, element abundances and kinematical properties of the ionised gas in this galaxy.
Methods. Observations were obtained in the spectral range $\lambda $3620-9400 Å with the imaging spectrograph GIRAFFE on the UT2 of the Very Large Telescope (VLT). These observations are the first ones carried out so far with GIRAFFE in the ARGUS mode which allows one to simultaneously obtain 308 spectra covering a 11 $.\!\!^{\prime\prime}$4$\times$7 $.\!\!^{\prime\prime}$3 region.
Results. We produced images of SBS 0335-052E in the continuum and in emission lines of different stages of excitation. While the maximum of emission in the majority of lines, including the strong lines H$\beta $ 4861 Å, H$\alpha $ 6563 Å, [O III] 4363,5007 Å, [O II] 3726,3729 Å, coincides with the youngest south-eastern star clusters 1 and 2, the emission of He II 4686 Å line is offset to the more evolved north-west clusters 4, 5. This suggests that hard ionising radiation responsible for the He II $\lambda $4686 Å emission is not related to the most massive youngest stars, but rather is related to fast radiative shocks. This conclusion is supported by the kinematical properties of the ionised gas from the different emission lines as the velocity dispersion in the He II $\lambda $4686 Å line is systematically higher, by $\sim$50-100%, than that in other lines. The variations of the emission line profiles suggest the presence of an ionised gas outflow in the direction perpendicular to the galaxy disk. We find a relatively high electron number density $N_{\rm e}$ of several hundred cm-3 in the brightest part of SBS 0335-052E. There is a small gradient of the electron temperature $T_{\rm e}$ and oxygen abundance from the East to the West with systematically higher $T_{\rm e}$ and lower 12+log O/H in the western part of the galaxy. The oxygen abundances for the whole H II region and its brightest part are 12 + log O/H = $7.29 \pm 0.02$ and $7.31 \pm 0.01$, respectively. We derive the He mass fraction taking into account all systematic effects. The He mass fraction $Y = 0.2463 \pm 0.0030$, derived from the emission of the whole H II region, is consistent with the primordial value predicted by the standard Big Bang nucleosynthesis model. We confirm the presence of Wolf-Rayet stars in cluster 3.

Key words: galaxies: fundamental parameters - galaxies: starburst - galaxies: ISM - galaxies: abundances - galaxies: indvidual: SBS 0335-052E

1 Introduction

The blue compact dwarf (BCD) galaxy SBS 0335-052E is an excellent nearby laboratory for studying star formation in low-metallicity environments. Since its discovery as one of the most metal-deficient star-forming galaxies known (Izotov et al. 1990), with oxygen abundance 12 + log O/H $\sim$ 7.30 (Thuan & Izotov 2005; Melnick et al. 1992; Izotov et al. 1997b,1999), SBS 0335-052E has often been proposed as a nearby young dwarf galaxy (Papaderos et al. 1998; Izotov et al. 1997b; Thuan et al. 1997; Izotov et al. 1990; Pustilnik et al. 2004). Thuan et al. (1997) and Papaderos et al. (1998), using the same Hubble Space Telescope (HST) images, have found several luminous clusters. The brightest clusters are labelled in Fig. 1 which represents the highest spatial resolution archival UV HST/Advanced Camera for Surveys (ACS) image of SBS 0335-052E obtained by Kunth et al. (2003). Some of the clusters are very young and produce extended regions of ionised gas (Melnick et al. 1992; Papaderos et al. 1998; Izotov et al. 1997b; Pustilnik et al. 2004). In particular, Izotov et al. (2001b), using deep long-slit spectra of SBS 0335-052E, have shown that extended H$\alpha $ emission is detected over $\sim$6-8 kpc, suggesting that hot ionised gas is spread out far away from the central part of the galaxy.

Thuan & Izotov (1997) using HST/GHRS UV spectrum of SBS 0335-052E have discovered a very broad Ly$\alpha $ line in absorption suggesting that this galaxy is embedded in a large envelope of neutral gas. The column density of N(H I) = $7\times10^{21}$ cm-2 in SBS 0335-052E derived by Thuan & Izotov (1997) is the largest one known for the BCDs. Later, Pustilnik et al. (2001) using Very Large Array (VLA) observations in the line H I $\lambda $21 cm have detected a large neutral gas cloud around SBS 0335-052E with a size 66 by 22 kpc elongated in the east-west direction and with two maxima separated by 22 kpc. The first maximum in H I distribution is connected to SBS 0335-052E, and the second one to the companion dwarf galaxy SBS 0335-052W discovered by Pustilnik et al. (1997). The latter galaxy is shown by Izotov et al. (2005) to be the lowest metallicity emission-line galaxy known with 12 + log O/H = $7.12 \pm 0.03$. Thuan et al. (2005) using Far Ultraviolet Spectroscopic Explorer (FUSE) observations have found that the oxygen abundance of the neutral gas around SBS 0335-052E of 12 + log O/H $\sim$ 7.0 is only slightly lower than that of the ionised gas, implying that this galaxy was not formed from pristine gas.

Despite the low metallicity of SBS 0335-052E, an appreciable amount of dust has been detected in it. Izotov et al. (1997b) and Thuan et al. (1997) have found variations of extinction in this galaxy from the optical spectroscopic and photometric observations. Later, Thuan et al. (1999) and Houck et al. (2004) using Infrared Space Observatory (ISO) and Spitzer mid-infrared observations have found an emission of an appreciable amount of warm dust with a characteristic temperature of $\sim$100 K. Even hotter dust with a temperature of several hundred degrees is expected to be present in SBS 0335-052E which is indicated in the infrared spectra at shorter wavelengths, namely 2-4 $\mu$m (Vanzi et al. 2000; Hunt et al. 2001).

Since the discovery paper by Izotov et al. (1990) it has been known that the ionising radiation in SBS 0335-052E is hard. The presence of the high-ionisation He II $\lambda $4686 Å emission line (e.g. Melnick et al. 1992; Izotov et al. 1997b,1990) and of the [Fe V] $\lambda $4227 Å emission line (Izotov et al. 2001b; Fricke et al. 2001) suggests that radiation with photon energies greater than 4 Rydberg is intense and could not be explained by stellar emission. Furthermore, Izotov et al. (2001b) and Thuan & Izotov (2005) have found [Fe VII] and [Ne V] emission lines which require intense radiation with photon energies above 7 Rydberg. SBS 0335-052E has also been detected in the X-ray range (Thuan et al. 2004). The origin of the hard ionising radiation remains unclear. Several mechanisms such as radiation from most massive main-sequence stars, Wolf-Rayet stars, high-mass X-ray binaries and radiative shocks have been discussed by e.g. Garnett et al. (1991), Schaerer (1996), Izotov et al. (2001b), Izotov et al. (2004) and Thuan & Izotov (2005). The most recent investigations have shown that although the stellar origin of hard radiation is not completely excluded, the most likely source of hard radiation is fast radiative shocks.


  \begin{figure}
\par\hspace*{1.0cm}\psfig{figure=5622f1.ps,angle=0,width=6.0cm} %
\end{figure} Figure 1: Archival HST/ACS UV image of SBS 0335-052E with thelabelled compact clusters. North is up and East is to the left.
Open with DEXTER

Table 1: Journal of observations.

Despite the efforts of different groups in studying the properties of SBS 0335-052E and its evolutionary status many problems remain unsolved. Since this galaxy is possibly a young galaxy it could be considered as a local counterpart of the high-redshift young dwarf galaxies. Therefore the continuation of its studies is important for cosmological applications. In this paper we present a two-dimensional spectroscopic study of SBS 0335-052E with the VLT/GIRAFFE. These are the first observations carried out so far in the ARGUS mode. Two new features are characteristic of these new observations, which were not present in previous spectroscopic studies of this galaxy. First, new observations allow us to map the whole galaxy in different emission lines and in the continuum. This gives integrated characteristics of different emission lines for the whole H II region, such as the line luminosities, which are necessary input parameters to build up the model of the H II region. Second, the spectral resolution of new observations is by one order of magnitude better than in all previous spectroscopic observations of SBS 0335-052E and it is enough to obtain the intrinsic profiles of emission lines. This allows us to study the kinematics of the H II region, and to make a comparison of kinematic characteristics in regions of different ionisation stages. In particular, it is important to compare the He II $\lambda $4686 Å line profiles with the profiles of other emission lines and, hence, to draw conclusions concerning the origin of hard radiation in SBS 0335-052E.

In Sect. 2 we describe the observations and data reduction. Morphology of SBS 0335-052E in different emission lines and continuum is considered in Sect. 3. Kinematic properties are discussed in Sect. 4. Heavy element abundances and helium abundance are derived in Sect. 5. The Wolf-Rayet stellar population in SBS 0335-052E is discussed in Sect. 6. Our conclusions are summarised in Sect. 7.

   
2 Observations and data reduction

Observations of SBS 0335-052E with the VLT/GIRAFFE spectrograph were performed during the nights 19-22 November, 2003 in the entire visible range. GIRAFFE is equipped with a 2K$\times$4K EEV CCD. The size of the CCD pixels is $15~\mu {\rm m} \times15$ $\mu$m. The spatial scale is 0 $.\!\!^{\prime\prime}$52/pixel in the ARGUS direct mode used during our observations. The ARGUS array is a rectangular array of 22 by 14 microlenses which is fixed at the center of one positioner arm. We used a spatial scale with a sampling of 0 $.\!\!^{\prime\prime}$52 per microlens and a total aperture of $11\hbox{$.\!\!^{\prime\prime}$ }4\times7\hbox{$.\!\!^{\prime\prime}$ }3$. The major axis of the array was directed along the major axis of SBS 0335-052E at a position angle PA = -30$^{\circ}$ and centered on the cluster 1 (Fig. 1). The low-resolution mode with the 600 lines/mm grating was used. Since the wavelength coverage with this grating ranges from 400 Å to 1200 Å depending on the central wavelength, eight exposures were done with setups LR1 - LR8 (Table 1) to obtain panoramic spectra in the wavelength range $\lambda $3620-9400 Å and a spectral resolution of 0.5-1 Å. Each exposure was split in two subexposures for the removal of cosmic ray hits. The journal of observations is shown in Table 1. Additionally, for the same setups the spectra of two standard stars, Feige 110 and HD 49798, were obtained for flux calibration. During the days the exposures of bias, Nasmyth screen flats and comparison lamps for the wavelength calibration have been obtained. The description of GIRAFFE may be found in Pasquini et al. (2002).

The spectra were extracted and calibrated using the standard Python version of BLDRS - Baseline Data Reduction Software (girbldrs-1.12) available from http://girbldrs.sourceforge.net. Basic description of BLDRS is given in Blecha et al. (2000) and Royer et al. (2002). The processing includes the bias subtraction, correction for the pixel sensitivity variations, localisation, optimal extraction, rebinning to linear wavelength scale and night sky subtraction.

The flux calibration was done using IRAF[*]. The flux-calibrated and redshift-corrected spectrum of the brightest rectangular region delineated by the thick solid line in the center of Fig. 7a is shown in Fig. 2. The region includes clusters 1 and 2. Many strong permitted and forbidden emission lines are seen in the spectrum of this region. One important feature of this spectrum is that its spectral resolution is the highest among all other spectra of SBS 0335-052E obtained so far. In particular, several blends are resolved for the first time for this object, most notably, [O II] $\lambda $3726 and [O II] $\lambda $3729, He I $\lambda $3965, [Ne III] $\lambda $3967 and H7 $\lambda $3970, [Ar IV] $\lambda $4711 and He I $\lambda $4713. The spectrum contains emission lines of ions of a broad range of ionisation stages. In particular, [Fe II], [Fe III], [Fe IV], [Fe V], [Fe VI] and [Fe VII] emission lines are detected.


  \begin{figure}
\par\mbox{\psfig{figure=5622f2a.ps,angle=0,width=9.0cm}\psfig{fig...
...0,width=9.0cm}\psfig{figure=5622f2d.ps,angle=0,width=9.0cm,clip=} }
\end{figure} Figure 2: Spectrum of the brightest part of SBS 0335-052E shown in Fig. 7a as a rectangular region delineated by the thick solid line.
Open with DEXTER

The spectrum of another region centered on the clusters 4 and 5 and delineated by the thick dashed line in Fig. 7a is shown in Fig. 3. The level of continuum in this spectrum is comparable to that in the spectrum of the brightest region (Fig. 2), but emission lines are weaker.

The emission line fluxes and widths were measured in each of the 22$\times$14 lens array using the routine SPLOT in IRAF. Flux errors were derived from the photon statistics using non-flux-calibrated spectra. These errors were propagated in the determination of the electron temperatures, electron number densities and element abundances.


  \begin{figure}
\par\mbox{\psfig{figure=5622f3a.ps,angle=0,width=9.0cm}\psfig{fig...
...0,width=9.0cm}\psfig{figure=5622f3d.ps,angle=0,width=9.0cm,clip=} }
\end{figure} Figure 3: Spectrum of a region centered on clusters 4 and 5 and shown in Fig. 7a as a square region delineated by a thick dashed line.
Open with DEXTER


  \begin{figure}
\mbox{\psfig{figure=5622f4a.ps,angle=0,width=3.5cm}\psfig{figure=...
...0,width=3.5cm}\psfig{figure=5622f4i.ps,angle=0,width=3.5cm,clip=} }
\end{figure} Figure 4: Images in the continuum near H$\beta $ emission line a) and in emission lines [O II] $\lambda $3727 Å b), [O III] $\lambda $4363 Å c), He II $\lambda $4686 Å d), [Ar IV] $\lambda $4740 Å e), H$\beta $ $\lambda $4861 Å f), [O III] $\lambda $5007 Å g), He I $\lambda $5876 Å h) and H$\alpha $ $\lambda $6563 Å i). All images are shown in the logarithmic flux scale. Positions of stellar clusters are labelled in panel (a). White cross in each panel shows the location of the region with maximum flux of the H$\alpha $ $\lambda $6563 Å emission line.
Open with DEXTER


  \begin{figure}
\par\mbox{\psfig{figure=5622f5a.ps,angle=0,width=6.0cm}\hspace*{1.0cm}\psfig{figure=5622f5b.ps,angle=0,width=6.0cm,clip=} }
\end{figure} Figure 5: Distribution of the H$\beta $ $\lambda $4861 Å equivalent width a) and of the H$\alpha $ $\lambda $6563 Å equivalent width b) in Å, with errors indicated in parentheses. Each square region corresponds to $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$. Only regions in which both EW(H$\beta $) and EW(H$\alpha $) could be measured are shown. Dark squares indicate the locations of clusters 1+2 with the maximum flux of the H$\alpha $ emission line, grey squares the locations of clusters 4+5.
Open with DEXTER

   
3 Morphology in continuum and emission lines

One of the advantages of the SBS 0335-052E panoramic observations with GIRAFFE/ARGUS is that the spectra of each region with an angular size of $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$ within an aperture $11\hbox{$.\!\!^{\prime\prime}$ }4\times7\hbox{$.\!\!^{\prime\prime}$ }3$ were obtained. This allows us to study the morphology of the galaxy in the continuum and individual emission lines and to construct a model of its H II region.

The central part of SBS 0335-052E containing the brightest clusters has an angular size $\la$2 $^{\prime\prime}$ (Fig. 1) which is only $\la$4 times larger than the angular size of 0 $.\!\!^{\prime\prime}$52 of each ARGUS lens. Therefore, for better viewing we rebinned the images, splitting each pixel in 9 from 0 $.\!\!^{\prime\prime}$52 pixels to 0 $.\!\!^{\prime\prime}$17 pixels linearly interpolating flux values in adjacent 0 $.\!\!^{\prime\prime}$52 pixels. In Fig. 4 we show the rebinned images of SBS 0335-052E in the continuum near H$\beta $ $\lambda $4861 Å (a), and in the emission lines [O II] $\lambda $3727 Å (b), [O III] $\lambda $4363 Å (c), He II $\lambda $4686 Å (d), [Ar IV] $\lambda $4740 Å (e), H$\beta $ $\lambda $4861 Å (f), [O III] $\lambda $5007 Å (g), He I $\lambda $5876 Å (h) and H$\alpha $ $\lambda $6563 Å (i). In all panels white crosses denote the pixel with the largest flux of the H$\alpha $ $\lambda $6563 Å emission line which is coincident with the location of clusters 1+2 in Fig. 1.

The image in the continuum (Fig. 4a) with labelled clusters resembles the HST UV image (Fig. 1) despite the much lower angular resolution determined by a seeing of $\sim$1 $^{\prime\prime}$ (Table 1). Several clusters are seen. However, the angular resolution in the GIRAFFE data is not enough to separate clusters 1 and 2, and 4 and 5.

The images in all emission lines (except for the He II $\lambda $4686 Å line) are very similar. They show very bright emission in the region of clusters 1, 2 and much fainter emission in the direction of other clusters. Furthermore, the equivalent widths of H$\beta $ $\lambda $4861 Å and H$\alpha $ $\lambda $6563 Å emission lines in the clusters 1 and 2 (dark squares in Fig. 5) and in the regions around these clusters are high. These facts suggest that clusters 1 and 2 are young, with an age 3-4 Myr, and contain numerous hot and massive ionising main-sequence stars. It is likely that clusters 7 and probably 3 are also young because EW(H$\beta $) and EW(H$\alpha $) are high. However, the number of ionising massive stars in those clusters is much lower than in clusters 1 and 2. On the other hand, clusters 4, 5 and 6 are probably more evolved as suggested by their lower equivalent widths of the H$\beta $ $\lambda $4861 Å and H$\alpha $ $\lambda $6563 Å emission lines. In particular, the H$\beta $ equivalent width EW(H$\beta $) = 46 Å for clusters 4, 5 (grey square in Fig. 5) and 87 Å for the square region delineated by the thick dashed line correspond to an age of $\sim$6-8 Myr and $\sim$5 Myr adopting a heavy element mass fraction Z = 0.001 for stars (Schaerer & Vacca 1998). The age for clusters 4, 5 is larger if a lower Z = 0.0004 corresponding to the metallicity of the ionised gas is adopted. The older age of clusters 4 and 5 is supported by their weak P$\alpha $emission as compared to that of clusters 1 and 2 (Thompson et al. 2006). The different ages of clusters 1+2and 4+5 could in principle explain why the brightness of clusters 4 and 5 in the UV range is greater than that of clusters 1 and 2 (Fig. 1), why their brightness is comparable in the optical continuum (Fig. 4a) and why clusters 4 and 5 are fainter in the NIR (Thompson et al. 2006). This is because the relative contribution of the ionised gas emission is increased from the UV to the NIR. The effect is stronger for clusters 1 and 2 because of the higher EW(H$\beta $) and hence of higher contribution of the ionised gas to the total emission. In addition, the interstellar extinction may play role if it is higher for clusters 1 and 2.

The morphology of SBS 0335-052E in the He II $\lambda $4686 Å emission line (Fig. 4d) significantly differs from that in other emission lines. The emission of this line in the direction of the clusters 3 and 4+5is stronger than that in the direction of the clusters 1 and 2. This offset of the He II $\lambda $4686 Å emission line relative to other nebular lines was noted by Izotov et al. (1997b) and Izotov et al. (2001b). Thus, it is evident that the hard ionising radiation responsible for the He II $\lambda $4686 Å emission is not related to the young main-sequence stars, but rather related to the post-main-sequence stars or their remnants. This effect is more clearly seen in Fig. 6 where we show the distribution of the relative flux He II $\lambda $4686/H$\beta $. In the direction on the clusters 1 and 2 the relative He II $\lambda $4686/H$\beta $ flux is $\sim$1-2% while in north-west regions it increases to $\sim$6-7%. Such high relative fluxes of the He II $\lambda $4686 emission line is difficult to explain by the ionising stellar radiation (e.g. Thuan & Izotov 2005; Izotov et al. 2004). Although a small number of Wolf-Rayet stars are found in cluster 3 (see Papaderos et al. 2006, and this paper), other mechanisms such as radiative shocks need to be invoked.


  \begin{figure}
\par\hspace*{1.0cm}\psfig{figure=5622f6.ps,angle=0,width=6.0cm,clip=}\end{figure} Figure 6: He II $\lambda $4686/H$\beta $ relative intensity distribution (in percent). Only regions with a strong enough He II $\lambda $4686 emission line are shown. The dark square is the location of clusters 1+2, while the grey square is the location of clusters 4+5.
Open with DEXTER

   
4 Kinematics of the ionised gas

The second advantage of the present GIRAFFE spectra for SBS 0335-052E is that they are obtained with sufficient spectral resolution. Therefore, the panoramic spectroscopic data can be used to study the kinematics of the H II regions in this galaxy.

In Fig. 7 we show the profiles of the H$\alpha $ $\lambda $6563 Å (a), He  II $\lambda $4686 Å (b), [O II] $\lambda $3726, 3729 Å (c) and [O III] $\lambda $4363 Å (d) emission lines in each pixel of the ARGUS array. Dotted lines show the wavelengths of emission lines adopting the average redshift derived from the observed wavelengths of all strong emission lines in the spectrum of the brightest rectangular region delineated by a thick solid line in Fig. 7a. H$\alpha $ $\lambda $6563 Å is the strongest line in all ARGUS array spectra and thus allows us to study the kinematical properties in the low-intensity extended regions of SBS 0335-052E. Other lines originate in different zones of the H II region: He II $\lambda $4686 Å is characteristic of the highest ionisation zone, [O II] $\lambda $3726, 3729 Å are characteristic of the lowest ionisation zone, and [O III] $\lambda $4363 Å emission corresponds to the intermediate zone which is overlapped with the highest- and lowest-ionisation zones.


  \begin{figure}
\par\mbox{\psfig{figure=5622f7a.ps,angle=0,width=4.4cm}\hspace*{0...
...ce*{0.1cm}\psfig{figure=5622f7d.ps,angle=0,width=4.4cm,clip=} }
\par\end{figure} Figure 7: Profiles of the emission lines: a) H$\alpha $ $\lambda $6563 Å, b) He II $\lambda $4686 Å, c) [O II] $\lambda $3726,3729 Å and d) [O III] $\lambda $4363 Å. All $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$regions are shown. Dark squares mark the location of clusters 1+2, grey squares the location of clusters 4+5. Thick solid and thick dashed lines in (a) delineate the brightest rectangular region and the second brightest square region for which the spectra are shown in Figs. 2 and 3.
Open with DEXTER

H$\alpha $ emission is seen almost in the whole region observed with GIRAFFE (Fig. 7a). The total aperture 1 $1\hbox{$.\!\!^{\prime\prime}$ }4\times7\hbox{$.\!\!^{\prime\prime}$ }3$ of ARGUS corresponds to linear size $\sim$3.1 kpc$~\times~$1.8 kpc adopting the distance to SBS 0335-052E of 54.3 Mpc (Izotov et al. 1997b). Thus, the observed region is only a part of a much larger H II region with a size of $\sim$6-8 kpc detected by Izotov et al. (2001b). In the brightest central region and in the slice oriented west-east the H$\alpha $ $\lambda $6563 line is narrow and no systematic offset of the line profile from the dotted line is seen. Thus, no evidence is present for the rotation of the observed part of the galaxy, since the west-east orientation of the region with narrow profiles is close to that of the disk-like H I cloud seen edge-on (Pustilnik et al. 2001). On the other hand, in the north-south direction the H$\alpha $ profiles are much broader and with more complex structure, suggesting an outflow of ionised gas in the direction perpendicular to the H I disk. The schematic H$\alpha $ kinematic model is shown in Fig. 8 where the grey rectangular region is the region with narrow H$\alpha $profiles oriented approximately along the H I cloud detected by Pustilnik et al. (2001). Two regions with the ionised gas outflow are shown to the north and south of the region with the narrow H$\alpha $ line. The double-peaked H$\alpha $ profiles in the northern and southern parts of Fig. 7a suggest the presence of expanding shells of ionised hydrogen with radial components of velocities of $\sim$50 km s-1. This finding is consistent with the presence of a shell of ionised gas at an angular distance of $\sim$5 $^{\prime\prime}$ to the north of cluster 1 (Thuan et al. 1997). The width of H$\alpha $ in clusters 4, 5 (grey square in Fig. 7a) is $\sim$2 times larger than that in clusters 1 and 2 (dark square in Fig. 7a) implying higher dynamic activity of the interstellar medium in the former, older clusters. This higher activity is probably due to supernova explosions.

All observed profiles were fit with a single Gaussian profile. In Fig. 9a we show FWHMs of the H$\alpha $ emission line in km s-1 due to the macroscopic motion only. Errors are given in parentheses. The widths are obtained after correction for the instrumental profiles and for the thermal motion, following the formula

\begin{displaymath}{\it FWHM} ^2_{\rm tur}={\it FWHM}^2_{\rm obs}-{\it FWHM} ^2_{\rm i}-{\it FWHM} ^2_{\rm th},
\end{displaymath} (1)

where FWHM $_{\rm tur}$, FWHM $_{\rm obs}$, FWHM$_{\rm i}$ and FWHM$_{\rm th}$ are the width of the macroscopic motion, the observed width, the width of the instrumental profile and the width of the thermal motion, respectively. The FWHM$_{\rm i}$ is obtained from the profiles of night sky emission lines and it is $\sim$25 km s-1. The width of the thermal profile of 30 km s-1 is derived for the electron temperature $T_{\rm e} = 20~000$ K and it is applied to the H$\alpha $ $\lambda $6563 Å, He II $\lambda $4686 Å and [O III] $\lambda $4363 Å emission lines, while the width of 26 km s-1 is obtained for $T_{\rm e} = 15~000$ K and it is applied to the [O II] $\lambda $3729 Å emission line.

In the brightest part of SBS 0335-052E the FWHM(H$\alpha $) in and around the clusters 1 and 2 is $\sim$40 km s-1. This value is lower than the FWHM(H$\alpha $) of 83 km s-1 and 62 km s-1 derived by Petrosian et al. (1997) for the NW and SE components of I Zw 18. Apparently, the dispersion of macroscopic motion is strongly dependent on the age of the starburst and it is larger in the evolved starbursts where the SNe activity is higher. Indeed, the difference between clusters 1 and 2 in SBS 0335-052E and the NW and SE components of I Zw 18 is that the clusters 1 and 2 are younger as the equivalent width of the H$\beta $ emission line there, EW(H$\beta $) $\sim$ 300-400 Å (Fig. 5a), is much larger than the EW(H$\beta $) $\la$ 100 Å in the NW and SE components of I Zw 18. It is probable that SNe have not yet appeared or their number is small in clusters 1 and 2.

On the other hand, the macroscopic motion in and around the older clusters 4 and 5 is significantly larger with FWHM(H$\alpha $) $\sim$ 100 km s-1, likely due to a high SN activity. Similar FWHM values and a similar spatial behaviour are found for the [O II] $\lambda $3729 Å and [O III] $\lambda $4363 Å emission lines. (Figs. 9c and 9d). The tendency of higher FWHM in more evolved clusters is also retained for the He II $\lambda $4686 Å emission line (Fig. 9b). However, the macroscopic velocity in the regions of He II emission is significantly higher than that in the emission regions of other lines. This difference may be an additional indication that the source of hard radiation is connected to fast radiative shocks.

   
5 Physical conditions and element abundances

The large aperture ( $11\hbox{$.\!\!^{\prime\prime}$ }4\times7\hbox{$.\!\!^{\prime\prime}$ }3$), high enough spectral resolution and large wavelength coverage of the ARGUS observations allow the detailed study of physical conditions (electron temperature and electron number density) and element abundances in the H II region. However, there are some limitations to this study. First, observations in different wavelength ranges have been done not in single but in separate exposures over several nights (see Table 1). Therefore, due to the varying weather conditions (seeing, transparency), effects of atmospheric refraction and non-perfect pointing during different exposures, the spectra in small apertures such as in a single pixel of $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$ and in different wavelength ranges are not well adjusted since they represent slightly different regions of the galaxy. These effects tend to be lower with increasing aperture. Therefore, depending on the adopted aperture, we will follow different approaches in the determination of element abundances. We consider the element abundance determination from the spectra obtained in apertures with three different sizes. First, for the spectra obtained within the largest available aperture of $11\hbox{$.\!\!^{\prime\prime}$ }4\times7\hbox{$.\!\!^{\prime\prime}$ }3$ there is no need to adjust the different wavelength ranges. This allows us to consistently correct the spectra for interstellar extinction using the observed decrement of hydrogen Balmer lines and then derive element abundances. Second, we consider the spectrum of the brightest region of SBS 0335-052E obtained within an aperture $1\hbox{$.\!\!^{\prime\prime}$ }56\times1\hbox{$.\!\!^{\prime\prime}$ }04$(the rectangular region delineated by a thick solid line in Fig. 7a). This spectrum is shown in Fig. 2. Since the aperture for this spectrum is relatively small, some adjustment of adjacent wavelength ranges is needed. However, thanks to the high brightness of this region many bright emission lines are seen in its spectrum. Therefore we use the same brightest emission lines in the overlapping wavelength ranges (where this is possible) to scale the spectra in the adjacent wavelength ranges. These lines are H7 $\lambda $3970 Å, He I $\lambda $4471 Å, [O III] $\lambda $4959 Å, [O I] $\lambda $6300 Å and He I $\lambda $7065 Å. In the remaining two overlapping wavelength ranges $\lambda $5650-5750 and $\lambda $8100-8200 where no strong emission lines are seen we used the continuum levels to adjust the adjacent spectra. Thus, in the spectrum of the brightest region the determination of the interstellar extinction is still possible, which was used to correct the spectra. Third, in the case of smallest apertures of $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$, in general the signal-to-noise of the spectra is not enough to use the same emission lines in the overlapping wavelength ranges. Therefore, for these apertures we adjust spectra in different wavelength ranges assuming that the ratios of hydrogen Balmer lines correspond to the theoretical values at the electron temperature of $T_{\rm e} = 20~000$ K. Hence, no determination of the interstellar extinction is possible for the smallest apertures and not all wavelength ranges could be adjusted. Fortunately, it is possible to adjust wavelength ranges containing the [O II] $\lambda $3726, 3729 Å, [O III] $\lambda $4363, 4959, 5007 Å, [S  II] $\lambda $6717, 6731 Å emission lines, therefore at least the determination of the electron temperature $T_{\rm e}$(O III), the electron number density $N_{\rm e}$(S II) and the oxygen abundance is possible.

To derive $T_{\rm e}$, $N_{\rm e}$ and heavy element abundances we follow the prescriptions by Izotov et al. (2006a). Where possible, the coefficient of interstellar extinction C(H$\beta $) and the equivalent width of absorption hydrogen lines EW $_{\rm abs}$ are derived from the observed hydrogen Balmer decrement. In this procedure we assume that EW $_{\rm abs}$ is the same for all hydrogen lines. Then the fluxes of emission lines were corrected for interstellar extinction and underlying stellar absorption (where this is possible).

We adopt the three zone model of the H II region. The electron temperature $T_{\rm e}$(O III) in the high-ionisation zone is derived from the [O III] $\lambda $4363/($\lambda $4959+$\lambda $5007) flux ratio. This temperature is used to derive abundances of ions O2+ and Ne2+. Since He  II $\lambda $4686 Å emission is present in the SBS 0335-052E spectrum, the O3+ abundance is derived following Izotov et al. (2006a) and adopting $T_{\rm e}$(O III). Since the O3+ abundance is significantly lower than the O2+ abundance, the uncertainties in the temperature for the zone where the O3+ ion is present introduce only a small uncertainty in the total oxygen abundance. Some other emission lines of high-ionisation ions Ar3+, Cl3+, Fe3+ - Fe6+ are seen in the spectrum of SBS 0335-052E in Fig. 2. In general these ions are present in the inner part of the H II region with a temperature higher than $T_{\rm e}$(O III).


  \begin{figure}
\par\hspace*{1.5cm}\psfig{figure=5622f8.ps,angle=0,width=5.3cm} %
\end{figure} Figure 8: Schematic H$\alpha $ kinematic model of SBS 0335-052E. Clusters 1, 2 and 4, 5 are shown by dark and grey circles respectively. The region with narrow H$\alpha $profiles is shown by a grey rectangle. Its orientation is close to the orientation of the H I cloud discussed by Pustilnik et al. (2001). Two regions with the ionised gas outflow are located to the north and south of the region with a narrow H$\alpha $ emission line.
Open with DEXTER


  \begin{figure}
\par\mbox{\psfig{figure=5622f9a.ps,angle=0,width=5.0cm}\hspace*{0...
...hspace*{0.1cm}\psfig{figure=5622f9d.ps,angle=0,width=5.0cm,clip=} }
\end{figure} Figure 9: Macroscopic/turbulent velocity dispersion at the FWHM (in km s-1) derived from the a) H$\alpha $ $\lambda $6563 Å, b) He II $\lambda $4686 Å, (c) [O II] $\lambda $3729 Å and d) [O III] $\lambda $4363 Å emission lines. The errors of the velocity dispersion are shown in parentheses. Only regions with the detected lines are shown. The dark square shows the location of clusters 1+2 with a maximum flux of the H$\alpha $ 6563 Å emission line. The grey square shows the location of clusters 4+5.
Open with DEXTER

However, since there is no temperature constraint from observations for these ions and atomic data are not well known for some of them, we decided not to use these ions for the abundance determination. For the intermediate-ionisation zone we adopt the electron temperature $T_{\rm e}$(S  III) which was derived using the relation between $T_{\rm e}$(S  III) and $T_{\rm e}$(O III) from Izotov et al. (2006a). The temperature $T_{\rm e}$(S III) is used to derive the abundances of ions S2+, Ar2+ and Cl2+. Finally, for the low-ionisation zone we adopt the electron temperature $T_{\rm e}$(O II) which was derived using the relation between $T_{\rm e}$(O II) and $T_{\rm e}$(O III) from Izotov et al. (2006a). This temperature is used to derive the abundances of ions O+, N+, S+ and Fe2+. Emission lines of some other low-ionisation ions and neutral species are present in the low-ionisation zone, namely, Fe+, O0, N0. We decided not to use them for the abundance determination since they reside both in the neutral and ionised gas and their abundances in the H II region are subject to large uncertainties.

The electron number density $N_{\rm e}$ is derived using the [O  II] $\lambda $3726/$\lambda $3729, [Ar IV] $\lambda $4711/$\lambda $4740 and [S II] $\lambda $6717/$\lambda $6731 emission line fluxes.

The total heavy element abundances are obtained with the use of ionisation correction factors ICFs from Izotov et al. (2006a) for each element. ICFs take into account the abundances of ions in unseen stages of ionisation.

5.1 Heavy element abundances in the brightest region

We first consider the chemical composition in the brightest part of SBS 0335-052E. Observed emission line fluxes relative to the H$\beta $ flux, F($\lambda $)/F(H$\beta $), emission line fluxes relative to the H$\beta $ flux, corrected for interstellar extinction and underlying stellar absorption, I($\lambda $)/I(H$\beta $), and the equivalent widths of emission lines EW($\lambda $) are shown in Table 4. The extinction coefficient C(H$\beta $), the equivalent width EW $_{\rm abs}$of the absorption hydrogen Balmer lines and the observed flux F(H$\beta $) of the H$\beta $ emission line are shown at the end of Table 4. Electron temperatures, electron number densities, ionic and total heavy element abundances for the brightest region are shown in the second column of Table 2. In general, the derived parameters are consistent with previous determinations e.g. by Izotov et al. (1997b), Izotov et al. (1999), Thuan & Izotov (2005). In particular, the electron temperature $T_{\rm e}$(O III) in all measurements is high and is close to 20 000 K. It was found in previous studies that the H II region in SBS 0335-052E is relatively dense. We confirm this finding. The electron number density, which we derive from the [S II] $\lambda $6717/$\lambda $6731 flux ratio in the brightest region, is $\sim$150 cm-3. A similar value is obtained from the [O II] $\lambda $3726/$\lambda $3729 flux ratio. On the other hand, the electron number density derived from the [Ar IV] $\lambda $4711/$\lambda $4740 flux ratio, using the data from Aller (1984), is much larger, $\sim$6000 cm-3. Thus, it appears that the high-ionisation regions are much denser than the low-ionisation regions.

Table 2: Heavy element abundances.

The oxygen abundance 12 + log O/H = 7.31 $\pm$ 0.01 is in perfect agreement with recent determinations by Izotov et al. (1997b), Izotov et al. (1999) and Thuan & Izotov (2005). The Ne/O, S/O, Cl/O and Ar/O abundance ratios are very close to the average values found by, e.g., Izotov et al. (2006a) for the large sample of low-metallicity emission-line galaxies. On the other hand, the N/O abundance ratio -1.37 appears higher than the mean value of -1.5 to -1.6 for the most metal-deficient BCDs (Izotov et al. 2006a; Izotov & Thuan 1999). Since only N+ lines are observed in the optical spectrum of SBS 0335-052E the total nitrogen abundance is derived as N/H = ICF(N)$\times$N+/H+, where ICF(N) $\sim$ (O3++O2++O+)/O+. Inspection of Table 4 shows that the relative flux [O II] $\lambda $3727/H$\beta $ is 0.2, or 30% lower than that in some other observations of SBS 0335-052E (e.g. Papaderos et al. 2006; Izotov et al. 1999) resulting in high ICF(N). The lower [O II] $\lambda $3727 flux in the bright region (Table 4) could be due to observational uncertainties (slightly different pointings of SBS 0335-052E during observations with different setups, effect of differential refraction, variable seeing, etc.). Adopting an [O II] $\lambda $3727 flux $\sim$30% higher will result in log N/O $\sim$ -1.5, in much better agreement with other determinations. Such increase of the [O II] $\lambda $3727 flux will also slightly decrease by $\sim$0.1 dex the iron abundance, while the abundances of other heavy elements will remain almost unchanged. In particular, the oxygen abundance 12 + log O/H will be increased only by 0.01 dex. The Fe/O abundance ratio is high and is typical for the extremely metal-deficient BCDs (Izotov et al. 2006a). This fact suggests that the depletion of iron onto dust in SBS 0335-052E is small.

5.2 Heavy element abundances from the integrated spectrum of SBS 0335-052E

We use panoramic VLT/GIRAFFE data to obtain the integrated spectrum of SBS 0335-052E by summing each $22\times14$ spectra in the whole aperture. The resulting spectrum is significantly more noisy than the spectrum of the brightest region because many spectra of low-brightness regions were co-added to the spectrum of the brightest region. However, the integrated spectrum is not subject to the observational uncertainties which are much more important for the spectra obtained with the smaller apertures (non-perfect pointing, variable seeing). Additionally, it allows us to obtain integrated characteristics such as the luminosity of the galaxy in individual lines.

In Table 5 we show the measured absolute fluxes F($\lambda $) of the emission lines, the absolute fluxes I($\lambda $) corrected for the interstellar extinction and underlying stellar absorption, the respective fluxes relative to the H$\beta $ $\lambda $4861 flux, F($\lambda $)/F(H$\beta $) and I($\lambda $)/I(H$\beta $), the equivalent widths EW($\lambda $) of the emission lines, the interstellar extinction coefficient C(H$\beta $) and equivalent width of hydrogen Balmer absorption lines EW $_{\rm abs}$. The absolute measured flux of H$\alpha $ emission line F(H$\alpha $) = $3.45\times10 ^{-13}$ erg s-1 cm-2 is consistent with the value $3.23\times10^{-13}$ erg s-1 cm-2 obtained by Pustilnik et al. (2004). The luminosities of the H$\beta $ and H$\alpha $ emission lines corrected for interstellar extinction and underlying stellar absorption are equal to L(H$\beta $) = $6.4\times 10 ^{40}$ erg s-1 and L(H$\alpha $) = $1.8\times 10 ^{41}$ erg s-1, corresponding to the equivalent number of O7 V stars N(O7 V) = $1.3\times10^4$. The major fraction of these stars ($\ga$90%) is located in the two compact clusters 1 and 2 as it is shown by the high-resolution spatial distribution of the P$\alpha $ emission in SBS 0335-052E obtained by Thompson et al. (2006) from the HST observations. To the best of our knowledge, these two clusters (most likely, cluster 1) are among the richest super-star clusters, hosting a very large number of O stars within a region of angular size $\la$0 $.\!\!^{\prime\prime}$1-0 $.\!\!^{\prime\prime}$2, corresponding to a linear size $\la$25-50 pc.

Electron temperatures, electron number densities $N_{\rm e}$(S II) and $N_{\rm e}$(O II), ionic and total element abundances derived from the integrated spectrum are shown in the third column of Table 2. They are similar to the parameters derived for the brightest region despite the fact that the statistical errors for the parameters derived from the integrated spectrum are higher. Note that the N/O abundance ratio derived from the integrated spectrum is lower than that derived from the spectrum of the brightest region and is consistent with the average values of N/O obtained for most metal-deficient galaxies (Izotov et al. 2006a; Izotov & Thuan 1999).


  \begin{figure}
\par\mbox{\psfig{figure=5622f10a.ps,angle=0,width=5.0cm}\hspace*{...
...psfig{figure=5622f10e.ps,angle=0,width=5.0cm,clip=}\hspace*{1.5cm}}
\end{figure} Figure 10: a) Electron temperature distribution (in 104 K) from the [O III] $\lambda $4363/($\lambda $4959+$\lambda $5007) line ratio. b) Electron number density distribution (in cm-3) from the [S II] $\lambda $6717/$\lambda $6731 line ratio. c) Oxygen abundance 12+log O/H distribution. d) Distribution of the nitrogen-to-oxygen abundance ratio. e) Distribution of the neon-to-oxygen abundance ratio. The dark region is the region with maximum intensity of H$\alpha $ 6563 Å emission line coincident with the location of clusters 1 and 2. The grey region corresponds to clusters 4 and 5.
Open with DEXTER

5.3 Distribution of the physical parameters and oxygen abundance

The VLT/GIRAFFE panoramic spectra allow us to study the distribution of the electron temperature $T_{\rm e}$(O III), the electron number density $N_{\rm e}$(S II) and heavy element abundances in the H II region. For this we use the spectra obtained for the $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$ apertures. We took into consideration only spectra in which at least the following lines of heavy elements are detected: [O II] $\lambda $3726, 3729 Å, [O III] $\lambda $4363, 4959, 5007 Å. This allows us to derive the electron temperature $T_{\rm e}$(O III) and oxygen abundance. From these spectra we excluded those, where the oxygen abundance 12 + log O/H is derived with an error greater than 0.1 dex.

In Fig. 10a we show the distribution of the electron temperature $T_{\rm e}$(O III). It is seen that the H II region is hot in all small apertures and has a characteristic temperature of $\sim$20 000 K. There is a slight spatial trend of the electron temperature with $T_{\rm e}$(O III) being slightly higher in the western part and slightly lower in the eastern part. The electron number density derived from the [S II] $\lambda $6717, 6731 Å emission lines is high, of several hundred particles per cm3 (Fig. 10b). However, the errors in the determination of $N_{\rm e}$ are large and are caused by the low intensity of the [S II] emission lines. Similar number densities are derived from the [O II] $\lambda $3726/$\lambda $3729 flux ratio. Although [O II] $\lambda $3726, 3729 Å emission lines are brighter than [S II] $\lambda $6717, 6731 Å, the low signal-to-noise ratio of the spectrum containing [O II] lines (Fig. 2) due to lower sensitivity of the GIRAFFE detector in that wavelength range prevents the determination of the electron number density from the [O II] lines in a region larger than that from the [S II] emission lines. As already mentioned, other emission lines detected in our spectra can in principle be used to determine the electron density. [Ar IV] $\lambda $4711 and 4740 Å are strong enough only in the brightest region of SBS 0335-052E, where they consistently indicate an electron number density in the range 103-104 cm-3. [Cl III] $\lambda $5517 and 5537 Å are too weak even in the brightest region (Fig. 2) and are therefore not used.

The oxygen abundance 12 + log O/H distribution is shown in Fig. 10c. There are some variations of the oxygen abundance in the range 7.00-7.42 with a slight trend of a decreasing 12 + log O/H from the East to the West. In particular, it appears that the oxygen abundance in cluster 7 of $\sim$7.2 (western side of Fig. 10c) is slightly lower than in other clusters, confirming the finding by Papaderos et al. (2006). Thus, there is some evidence for a possible self-enrichment by heavy elements (cf. Izotov et al. 1997b,1999) or for the presence of "initial'' abundance variations in the gas. However, the errors in the electron temperature, electron number density and oxygen abundance include only errors derived based on the photon statistics of non-flux-calibrated emission line fluxes and they do not take into account uncertainties in pointing, variable seeing, differential refraction, etc., which a difficult to estimate. Therefore, variations in the oxygen abundance may not be statistically significant.

In Fig. 10d we show variations of N/O abundance ratio. The log N/O varies in the range from -1.13 to -1.65 with relatively small errors. However, the real errors might be much higher because of the limitations introduced by the small apertures of $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$ for each spectrum. The same is true for the distribution of the Ne/O abundance ratio (Fig. 10e).

5.4 Helium abundance from the integrated spectrum

SBS 0335-052E, being one of the most metal-deficient BCDs, plays an important role in the determination of the primordial He mass fraction $Y_{\rm p}$ and, thus, in the determination of the baryonic mass fraction of the Universe. Since the precision in the determination of $Y_{\rm p}$ should be better than $\sim$1% to place useful constraints on the cosmological models, high signal-to-noise spectra are needed for this. Additionally, several systematic effects should be taken into account, and spectra and emission line fluxes should be corrected for them (see for details Izotov et al. 2006b; Izotov & Thuan 2004). These are the corrections for (1) interstellar extinction, (2) ionisation structure, (3) collisional excitation of helium lines, (4) fluorescence in helium lines, (5) temperature variations, (6) underlying stellar absorption, (7) collisional excitation of hydrogen lines. All these corrections are at a level of a few percent and, apart from (2) influence each other in a complicated way. The case of SBS 0335-052E is particularly complicated, because its H II region is dense, hot and optically thick in some He I emission lines (Izotov et al. 1999). Therefore, effects (3), (4) and (7) are strong in the H II region of SBS 0335-052E.

To derive the He abundance we use the integrated spectrum of SBS 0335-052E because it is least dependent on the observational parameters discussed above. The He+ abundance y+ which is derived from the He I emission lines depends on the adopted He I emissivities. We adopt the new He I emissivities by Porter et al. (2005). In this paper, following Izotov et al. (1997a,1994), Izotov & Thuan (1998,2004) and Izotov et al. (2006b) we use the five strongest He I $\lambda $3889, $\lambda $4471, $\lambda $5876, $\lambda $6678 and $\lambda $7065 emission lines to derive $N_{\rm e}$(He+) and $\tau$($\lambda $3889). The He I $\lambda $3889 and $\lambda $7065 lines play an important role because they are particularly sensitive to both quantities. Since the He I $\lambda $3889 line is blended with the H8 $\lambda $3889 line, we have subtracted the latter, assuming its intensity to be equal to 0.107 I(H$\beta $) (Aller 1984).

Besides the emissivities the derived $y^+ \equiv {\rm He}^+$/H+ abundance depends on several other parameters: collisional excitation of hydrogen emission lines, electron number density $N_{\rm e}$(He+) and electron temperature $T_{\rm e}$(He+), equivalent widths EW($\lambda $3889), EW($\lambda $4471), EW($\lambda $5876), EW($\lambda $6678) and EW($\lambda $7065) of He I stellar absorption lines, optical depth $\tau$($\lambda $3889) of the He I $\lambda $3889 emission line. We use Monte Carlo simulations for the y+ determination, randomly varying the parameters in their ranges. First, we subtract the fractions of the H$\alpha $ and H$\beta $observed fluxes due to the collisional excitation. We adopt that the fraction $\Delta I_{\rm coll}$(H$\alpha $)/I(H$\alpha $) of the H$\alpha $ flux due to the collisional excitation varying in the range 0%-5% of the total flux. This fraction is randomly generated 100 times in the adopted range. The fraction of the H$\beta $ emission line flux due to the collisional excitation is adopted to be three times less than that of the H$\alpha $ flux. For each generated fraction of the H$\alpha $ and H$\beta $ they are subtracted from the total observed fluxes and then all emission line fluxes are corrected for the interstellar extinction and abundances of elements are calculated.

To calculate y+ we simultaneously and randomly vary $N_{\rm e}$(He+), $T_{\rm e}$(He+) and $\tau$($\lambda $3889). The total number of such realizations is 105 for each value of $\Delta I_{\rm coll}$(H$\alpha $)/I(H$\alpha $). Thus, the total number of Monte Carlo realizations is 100 $\times$ 105 = 107. As for the He I underlying stellar absorption, we keep values of EW($\lambda $3889), EW($\lambda $4471), EW($\lambda $5876), EW($\lambda $6678) and EW($\lambda $7065) constant during simulations. In our spectrum, other He I emission lines, namely He I $\lambda $3820, $\lambda $4388, $\lambda $4026, $\lambda $4921, and $\lambda $7281 are seen. However, we do not attempt to use these lines for He abundance determination because they are much weaker than the five brightest lines, and hence have larger uncertainties.

We solve the problem by minimization of the expression

 \begin{displaymath}\chi^2=\sum_i^n\frac{(y^+_i-y^+_{\rm mean})^2}{\sigma^2(y^+_i)}
,
\end{displaymath} (2)

where y+i is the He+ abundance derived from the flux of the He I emission line with label i, $\sigma(y^+_i)$ is the statistical error of the He abundance. The quantity $y^+_{\rm mean}$ is the weighted mean of the He+abundance as derived from the equation

 \begin{displaymath}y^+_{\rm mean}=\frac{\sum_i^k{y^+_i/\sigma^2(y^+_i)}}
{\sum_i^k{1/\sigma^2(y^+_i)}}
\cdot
\end{displaymath} (3)

We use all five He I emission lines to calculate $\chi^2$ (i.e., n = 5), but only three lines, He I $\lambda $4471, He I $\lambda $5876 and He I $\lambda $6678 to calculate $y^+_{\rm mean}$ (k = 3). This is because the fluxes of He I $\lambda $3889 and He I $\lambda $7065 emission lines are more uncertain than the other three He I emission lines.

The best solution for $y^+_{\rm mean}$ is found from the minimum of $\chi^2$, the systematic error $\sigma_{\rm sys}$ is obtained from the dispersion of $y^+_{\rm mean}$ in the range of $\chi^2$ between $\chi^2_{\rm min}$ and $\chi^2_{\rm min}$ + 1. Then the total error for $y^+_{\rm mean}$ is derived from $\sigma^2_{\rm tot}$ = $\sigma^2_{\rm stat}$ + $\sigma^2_{\rm sys}$.

Additionally, since the nebular He II $\lambda $4686 Å emission line was detected, we have added the abundance of doubly ionised helium $y^{2+} \equiv {\rm He}^{2+}$/H+ to y+. Although the He2+ zone is hotter than the He+ zone, we adopt $T_{\rm e}$(He2+) = $T_{\rm e}$(He+). This last assumption introduces little change in the y value, because the value of y2+ is small ($\sim$4% of y+). Finally, the ionisation correction factor ICF(He) is taken into account from Izotov et al. (2006b) to convert y++y2+ to the total He abundance $y \equiv {\rm He/H}$.

The derived parameters and He abundances for each He I emission line are shown in Table 3. Here, $\Delta I_{\rm coll}$(H$\alpha $)/I(H$\alpha $) is the fraction of H$\alpha $flux due to the collisional excitation. The electron number density $N_{\rm e}$ in the He+ zone is high, 295 cm-3, and is consistent with the number density derived from the [S  II] emission lines. The electron temperature $T_{\rm e}$(O  III) in Table 3 slightly differs from that in Table 2 (right column). This is because in Table 3 collisional excitation of hydrogen lines is taken into account resulting in smaller C(H$\beta $) than that in Table 2. The electron temperature $T_{\rm e}$(He+) is only slightly lower than $T_{\rm e}$(O  III) suggesting that temperature fluctuations in the H II region of SBS 0335-052E are small. The optical depth $\tau$($\lambda $3889) of 2.9 is high compared to that in other BCDs (see e.g. Izotov & Thuan 2004) implying an important contribution of the fluorescent enhancement of He I emission lines.

The derived weighted mean He mass fraction in SBS 0335-052E, $Y = 0.2463 \pm 0.0030$, is slightly lower (but consistent within the errors) than the primordial He mass fraction $Y_{\rm p} = 0.24815 \pm 0.0003 \pm 0.0006$(syst.) from the 3-year data of the WMAP experiment (Spergel et al. 2006) and the D abundance, supporting the standard cosmological model of the primordial nucleosynthesis.

   
6 Wolf-Rayet stars

Table 3: Helium abundance


  \begin{figure}
\par\hspace*{0.0cm}\psfig{figure=5622f11.ps,angle=0,width=8.5cm} %
\end{figure} Figure 11: Part of the spectrum of cluster 3 showing the probable broad Wolf-Rayet emission lines N III $\lambda $4640 and C IV $\lambda $4658 labelled "WR''.
Open with DEXTER

The search for Wolf-Rayet (WR) stars in extremely low-metallicity dwarf galaxies is of great interest for constraining stellar evolution models. However, such studies are difficult since the strength of WR emission lines is significantly reduced with decreasing metallicity. Therefore, very high signal-to-noise ratio spectra are required to detect weak WR features. For a long time no WR galaxies with an oxygen abundance 12 + log O/H < 7.9 were known (Masegosa et al. 1991). Later, Izotov et al. (1997c) and Legrand et al. (1997) discovered WR stars in I Zw 18, at that time the most metal-deficient emission-line galaxy, with the oxygen abundance 12 + log O/H = 7.17. Thus, it appears that WR stars could be found in any other dwarf emission-line galaxy with active star formation, including SBS 0335-052E. However, the strength of WR emission features depends not only on metallicity but also on the age of a starburst. In starbursts with the metallicity of SBS 0335-052E, the WR stage is expected to be very short, typically less than 1 Myr (Schaerer & Vacca 1998; de Mello et al. 1998). Therefore, not all young clusters ionising the interstellar medium in SBS 0335-052E may be expected to contain WR stars. Recently, Papaderos et al. (2006) found that WR stars of the early carbon sequence (WC4 stars) are present in cluster 3 of SBS 0335-052E.

GIRAFFE/ARGUS observations allow in principle a more detailed search for WR stars in SBS 0335-052E, a more precise localisation in the galaxy, and also to resolve the $\lambda $4650 WR bump into N III $\lambda $4640 and C IV $\lambda $4658 broad features, thus allowing the detection of both late nitrogen WR stars (WNL stars) and early carbon WR stars (WCE stars). Previous observations by Papaderos et al. (2006) had too low spectral resolution to definitely distinguish between two types of WR stars. However, there are some limitations of GIRAFFE/ARGUS observations that make such a study more difficult than that of Papaderos et al. (2006). Although Papaderos et al. (2006) have observed with the smaller 3.6 m ESO telescope, their spectrum has a higher S/N ratio because of the $\sim$3 times longer exposure and $\sim$10 times lower spectral resolution.

We checked all $0\hbox{$.\!\!^{\prime\prime}$ }52\times0\hbox{$.\!\!^{\prime\prime}$ }52$ spectra obtained with GIRAFFE and found that broad WR features near $\lambda $4650 are likely present only in one spectrum associated with cluster 3. This spectrum is shown in Fig. 11. The S/N ratio of $\sim$5 of this spectrum is not high, but broad WR features are clearly seen. Thus, we confirm the finding by Papaderos et al. (2006) that WR stars appear to be present in cluster 3. However, we find that the WR feature consists in fact of two lines: N  III $\lambda $4640 and C IV $\lambda $4658. The latter line is blended with the much narrower nebular [Fe III] $\lambda $4658 emission line. We find that, after subtraction of the [Fe III] line from the $\lambda $4658 blend, the fluxes of the N  III $\lambda $4640 and C IV $\lambda $4658 lines and their FWHMs are similar, $\sim$( $4\pm1)\times10^{-17}$ erg s-1 cm-2 and 6.5 Å, respectively. Thus, the total flux of the $\lambda $4650 bump (N III $\lambda $4640 + C IV $\lambda $4658) non-corrected for extinction is $\sim$ $8\times10^{-17}$ erg s-1 cm-2, or 2/3 that measured by Papaderos et al. (2006) in a larger aperture. The total equivalent width of this bump is EW($\lambda $4650) $\sim$ 9 Å. The C  IV $\lambda $4658 emission line should be accompanied by the C  IV $\lambda $5808 emission line with a comparable flux. Unfortunately, the redshifted C  IV $\lambda $5808 emission line in SBS 0335-052E coincides with the night sky Na  I $\lambda $5890, 5895 emission lines. Therefore, the imperfect night sky subtraction hinders the detection of the WR line.

The observed (i.e. not extinction corrected) N III $\lambda $4640 and C IV $\lambda $4658 emission line luminosities of cluster 3 are equal to L(N III $\lambda $4640) = L(C IV $\lambda $4658) = $1.4\times10^{37}$ erg s-1 and correspond to the number of WNL and stars N(WNL) $\sim$ $35\pm8$ and to the number of WC4 stars N(WC4) $\sim$$3\pm 1$ if we adopt the "standard'' WR line luminosities computed by Schaerer & Vacca (1998), i.e. assuming N III $\lambda $4640 and C IV $\lambda $4658 emission line luminosities of $4\times10^{35}$ erg s-1 and $5\times 10 ^{36}$ erg s-1 respectively for a single WNL and WC4 star. The derived values of WNL and WC4 stars are not reliable for various reasons. First these are likely lower limits because of the neglected reddening, and because WR stars at low metallicity may have lower intrinsic line luminosities (e.g. Crowther & Hadfield 2006). Second, it is not necessarily clear that the observed lines are indeed due to WN and WC stars, as their strengths and widths are somewhat unusual. Their relatively narrow widths could be due to lower mass loss rates and/or smaller wind velocities in WR stars at such low metallicity (Crowther et al. 2002; Crowther & Hadfield 2006; Vink & de Koter 2005). Alternatively, from their relative strength and the narrow widths, the observed lines could also correspond to very late WNL stars (WN 10h-11h) (cf. Crowther & Smith 1997), if most of the C IV $\lambda $4658 was nebular. Given the faintness of these spectral signatures and the lack of known individual WR stars at such low metallicities as comparison objects, it is difficult to draw firm conclusions on the WR content in this cluster.

   
7 Conclusions

In this paper we present panoramic spectroscopic observations with the VLT/GIRAFFE in the spectral range $\lambda $3620-9400 Å of one of the most metal-deficient blue compact dwarf (BCD) galaxies, SBS 0335-052E. Our findings can be summarized as follows:

1.
The morphology of the galaxy in different lines is very similar, except for the He II $\lambda $4686 Å line, suggesting that the main ionising source in this galaxy is the compact super-star clusters (SSC) 1+2 in the SE part of the galaxy, most likely the cluster 1. The equivalent number of O7 V ionising stars in this cluster is $\sim$13 000, making it the most powerful SSC known so far.

2.
The maximum emission of the He II $\lambda $4686 Å line is offset towards the older compact clusters 4 and 5. Additionally, its width is 1.5-2 times greater than the width of other lines. Furthermore, the width of this and other emission lines is much higher in clusters 4, 5 than in cluster 1. These findings imply that the hard ionising radiation responsible for He II emission is most likely produced by fast radiative shocks. However, some contribution from WR stars located in cluster 3 cannot be excluded.

3.
The analysis of kinematical properties of the ionised gas in SBS 0335-052E suggests the presence of gas outflow in the direction perpendicular to the galaxy disk.

4.
We derive physical parameters in the H II region of SBS 0335-052E, suggesting that the ionised gas is hot ($\sim$20 000 K) and dense ($\ga$several hundred particles per cm-3). The oxygen abundances 12 + log O/H in the brightest region and in the integrated spectrum of SBS 0335-052E are respectively $7.31 \pm 0.01$and $7.29 \pm 0.02$, and they are consistent with previous determinations. The other heavy element-to-oxygen abundance ratios are consistent with the average values derived for other most metal-deficient galaxies. There is a slight decrease of oxygen abundance from the East to the West suggesting some self-enrichment of the ionised gas by heavy elements.

5.
We derive the He mass fraction Y from the integrated spectrum taking into account all possible systematic effects in the He abundance determination. Our value $Y = 0.2463 \pm 0.0030$ is slightly lower but consistent within the errors with the primordial He mass fraction determined from the 3 year WMAP data and from primordial deuterium measurements.

6.
We confirm the presence of Wolf-Rayet stars in cluster 3 found previously by Papaderos et al. (2006). The lower limits of the WNL and WC4 star numbers are $\sim$$35\pm8$ and $\sim$$3\pm 1$, respectively if standard WR line luminosities are assumed.

Acknowledgements
Y.I.I. and N.G.G. thank the Observatoire de Genève for hospitality. D.S. thanks Paul Crowther for useful comments on WR stars and the Swiss National Science Foundation for support.

References

 

  
Online Material

Table 4: Emission line fluxes and equivalent widths in the spectrum of the brightest region.

Table 5: Emission line fluxes and equivalent widths in the integrated spectrum of the H II region.



Copyright ESO 2006