3 Weak nebular emission and stellar absorption lines

One of main results (see Fig. 1) is the detection of numerous weak permitted and forbidden emission lines, which are - to the best of our knowledge - detected for the first time in the spectrum of an extra-galactic giant H  II region. Lines with fluxes down to $\sim$0.1% of HH$\beta$ are clearly detected. For weaker lines the identifications are tentative and the measured fluxes are fairly uncertain, with relative errors $\sim$100%, while the relative errors of the fluxes for the strongest lines are $\sim$1%. Thirty unidentified emission features with fluxes 0.02%-0.12% that of H$\beta$ are also shown in Table 1 and labeled by "?''.

Some permitted lines, e.g. O I $\lambda$6046, $\lambda$7002, $\lambda$7254, Si II $\lambda$5958, $\lambda$5979, $\lambda$6347, are likely of fluorescent origin and produced by absorption of the intense UV radiation (Esteban et al. 1998). Other lines (broad features at $\lambda$4620 and $\lambda$5700-5850) can be stellar in origin and likely produced by Wolf-Rayet stars. Weak He II $\lambda$4200 absorption is also detected, which is produced in the atmospheres of O stars. Although our detection level is comparable to that in deep spectra of Orion and planetary nebulae, optical recombination lines found in those objects (e.g. Esteban et al. 1998; Liu et al. 2000) are not detected here. This is likely due to the much lower metallicity of SBS 0335-052.

Figure 3: a) Distribution along the slit of the equivalent widths of H$\alpha$ (solid line), H$\beta$ (dashed line) and [O III] $\lambda$5007 (dotted line) emission lines. The slit is oriented at position angle -30$^\circ$. Thick solid line shows the range of the reliable EW(H$\alpha$). Outside this range the continuum is very weak, corresponging to R band surface brightness fainter 25 mag arcsec-2. b) Same as a) but slit is oriented at position angle 80$^\circ$.

Several emission lines indicate the presence of intense hard radiation at energies above 54 eV ($\lambda <$ 228 Å) and possibly even above 75-99 eV. The well-known He II $\lambda$4686 line and also He II $\lambda$5411 are detected. We confirm the presence of the forbidden [Fe V] $\lambda$4227 line, previously discovered in SBS 0335-052 and another low-metallicity BCD (Tol 1214-277) by Fricke et al. (2001). In addition several weak [Fe VI] and possibly [Fe VII] emission lines are detected in our spectrum. The ionization potentials corresponding to these Fe lines are 54.8, 75.0, and 99 eV respectively; the He⁺ emitting region is therefore expected to be associated with the emission of [Fe  V-VII]. The [Ar  V] $\lambda$7006 line is likely detected. However, it is blended with O  I $\lambda$7002 emission line. The possible identification of [Ar  V] $\lambda$6435 line by Fricke et al. (2001) appears very uncertain from our spectrum.

The origin of hard radiation in giant H II regions has been debated for many years. Possible sources for this radiation include shocks, X-ray binaries, or hot Wolf-Rayet stars (cf. Garnett et al. 1991; Schaerer 1996). The spatial distribution of He II $\lambda$4686, [O III] $\lambda$4363, H$\beta$, and the adjacent continuum (see Fig. 2a) show some evidence for the presence of shocks, at least in a limited area of SBS 0335-052. Indeed, in the NW direction He  II $\lambda$4686/HH$\beta$ remains relatively strong out to the position of the ionized gas shell seen in HST images at a distance $\sim$6 arcsec from the two central clusters 1 and 5 of Thuan et al. (1997; cf. also Papaderos et al. 1998). Although expected, the other high excitation lines of [Fe  V-VII] are too faint to be detected away from the main clusters. Furthermore, an increase of the electron temperature is observed away from the main clusters towards the shell (Fig. 2b), which may result from additional heating due to shocks. Both the extent of the region with He  II $\lambda$4686/HH$\beta$ $\sim$ 0.06 and the increase of $T_{\rm e}$ at large distance from the main ionizing clusters, where geometric dilution will greatly reduce the local ionization parameter, are suggestive of shocks in the area within and out to the shell. On the other hand, in the region centered on the main clusters we cannot distinguish between shocked and photoionized gas, as nearby stellar objects with hard ionizing spectra cannot be excluded. In passing we note that if shocks are present, the oxygen abundance gradient derived assuming a pure photoionized H  II region model (Fig. 2b) may not be real. Observations of the spatial distribution of various high excitation lines (including [Fe  V-VII] and [Ne  V] $\lambda$3426 which is expected to be strong from shock models; e.g. Dopita & Sutherland 1996) as well as a detailed modeling are required to establish more firmly the importance of shocks on the spectrum of SBS 0335-052.