2 Observations and data reduction

2.1 Photometry

Direct images of SBS 0940+544 were acquired in two observing runs. Broad-band Johnson V and Cousins I images and narrow-band images in the H$\alpha$ line at $\lambda$6606 Å with a full width at half maximum (FWHM) through a passband of 78 Å and in the adjacent continuum at $\lambda$6520 Å through a passband with FWHM = 84 Å were obtained with the Kitt Peak 2.1 m telescope on April 18 and April 22, 1999, respectively. The telescope was equipped with a 1k Tektronix CCD detector operating at a gain of 3e^-ADU^-1, giving an instrumental scale of 0 $.\!\!^{\prime\prime}$305 pixel^-1 and field of view of 5$^\prime$. The total exposure of 40 and 60 min in V and I, respectively, was split into four subexposures, each being slightly offset from the others. Likewise, narrow-band exposures in the H$\alpha$ line and the adjacent continuum bluewards of H$\alpha$ were taken for 25 min each. The seeing during the broad- and narrow-band observations was FWHM 1 $.\!\!^{\prime\prime}$2 and 2 $.\!\!^{\prime\prime}$2, respectively.

Another series of exposures with a total integration time of 2.5 hours in Cousins R was taken with the 1.23 m telescope at Calar Alto in the period January 24 to February 22, 2000. The telescope was equipped with a 2k SITe detector at its Cassegrain focus giving a usable field of view of $\sim$11$^\prime$ and an instrumental scale of 0 $.\!\!^{\prime\prime}$5 pixel^-1. The FWHM of point sources in the coadded R exposures is $\approx$1 $.\!\!^{\prime\prime}$5.

Bias and flat-field frames were obtained during each night of both observing runs. Broad-band images were calibrated by observing different standard fields from Landolt (1992). Our calibration uncertainties are estimated to be well below 0.05mag in all bands. Standard reduction steps, including bias subtraction, flat-field correction and absolute flux calibration were carried out using the IRAF and ESO MIDAS.

   
2.2 Spectroscopy

High S/N long-slit spectroscopy of SBS 0940+544 was obtained with the 4.5 m Multiple Mirror Telescope (MMT) and with the 10 m Keck II telescope over a wavelength range 3600-7500 Å.

The first set of observations was carried out with the MMT at two slit positions (see Fig. 1b). Observations were made with the blue channel of the MMT spectrograph operated with a 3072 $\times$ 1024 CCD detector. A 1 $.\!\!^{\prime\prime}$5 $\times$ 180 $^{\prime \prime }$ slit was used. The spatial scale along the slit was 0 $.\!\!^{\prime\prime}$3 pixel^-1, with a spectral resolution of $\sim$7 Å (FWHM). The final spatial scale is 0 $.\!\!^{\prime\prime}$6 pixel^-1 after binning every two consecutive rows into one.

The first slit orientation with position angle ${\rm PA}=-$41$^{\circ }$ (MMT #1, cf. Fig. 1b), was centered on the brightest star-forming H II region (denoted as a in Fig. 1a) and oriented along the elongated main body of the galaxy close to the direction of the major axis. This 120-min spectrum was obtained on December 13, 1996 at an airmass of 1.2. The total exposure time was broken into four 30 min exposures.

The second slit orientation (MMT #2, cf. Fig. 1b) is in the N-S direction with PA = 0$^{\circ }$, centered on the second brightest region, denoted as c in Fig. 1a. These observations were carried out on December 12, 1996 with an exposure time of 120 min at an airmass of 1.2. The total exposure time was also broken into four 30 min exposures. The seeing during both sets of MMT observations was 0 $.\!\!^{\prime\prime}$8.

The Keck II observations were carried out on January 9, 2000 with the low-resolution imaging spectrograph (LRIS)(Oke et al. 1995), using the 300 groove mm^-1 grating which provides a dispersion of 2.52 Å pixel^-1 and a spectral resolution of about 8 Å in the first order. The slit was 1 $^{\prime \prime }$$\times$180 $^{\prime \prime }$, centered on the brightest H II region and oriented with a position angle PA = -41$^{\circ }$, the same as during the first set of MMT observations. No binning along the spatial axis has been done, yielding a spatial sampling of 0 $.\!\!^{\prime\prime}$2 pixel^-1. The total exposure time was 60 min, broken into three 20 min exposures. All exposures were taken at an airmass of 1.2. The seeing was 0 $.\!\!^{\prime\prime}$9.

Figure 2: a) V-I map of SBS 0940+544 displayed in the colour range -0.2 to +0.6 mag. The overlayed contours are from 21.0 to 25.5 V mag arcsec-2 in steps of 0.5 mag. b) V-R map of the same region of the BCD displayed in the colour interval 0.1 to 0.4 mag. The overlayed contours are from 21.0 to 25.0 R mag arcsec-2 in steps of 0.5 mag. The bluest region in either map coincides with the northwestern H II region a, with an average V-I and V-R of -0.5 mag and -0.2 mag, respectively. The angular resolution of each map is indicated by a circle, the diameter of which is equal to the FWHM of point sources in the corresponding smoothed input images used for its derivation (1 $.\!\!^{\prime\prime}$8 and 2 $.\!\!^{\prime\prime}$75 for the V-I and V-R map, respectively).

Signal-to-noise ratios S/N $\sim$ 50 and $\sim$100 were reached in the continuum near H$\beta$ for the brightest part of the galaxy during the MMT and Keck II observations. No correction for atmospheric dispersion was made because of the small airmass during all observations. This effect, although small, may introduce some uncertainties in the flux determination of the weaker lines. At an airmass of 1.2 and position angle -41$^{\circ }$ the shift of the emitting region at the [O II] $\lambda$3727 line relative to the same region at the [O  III] $\lambda$5007 line in the direction perpendicular to the slit is 0 $.\!\!^{\prime\prime}$4 (Filippenko 1982). The shift at [O III] $\lambda$4363 is much smaller, only 0 $.\!\!^{\prime\prime}$1. In Sect. 4 we estimate the uncertainties introduced by this effect on the abundance determination. Spectra of a He-Ne-Ar comparison lamp were obtained after each exposure during the MMT observations, and spectra of a Hg-Ne-Ar comparison lamp were obtained after each exposure during the Keck II observations. They were used for wavelength calibration. Two spectrophotometric standard stars, Feige 34 and HZ 44, were observed for flux calibration. The data reduction was made with the IRAF software package. The two-dimensional spectra were bias-subtracted and flat-field corrected. Cosmic-ray removal, wavelength calibration, night sky background subtraction, correction for atmospheric extinction and absolute flux calibration were then performed.

One-dimensional spectra for the bright H II region a of SBS 0940+544 were extracted from the Keck II and MMT data within large apertures of 1 $^{\prime \prime }$ $\times$ 4 $^{\prime \prime }$ and 1 $.\!\!^{\prime\prime}$5 $\times$ 3 $^{\prime \prime }$ respectively to minimize the effect of atmospheric dispersion (Fig. 5). In addition we extracted spectra showing hydrogen Balmer absorption lines for four regions along the main body of the galaxy (i.e. along the slit oriented at PA = -41$^{\circ }$). The selected regions, denoted 1 to 4 (Keck II observations) and 1a to 4a (MMT) in Tables 4 and 5 were at adjacent positions along the major axis of the BCD, with the origin taken to be at the center of region a. Regions 1 and 1a coincide with region c in Fig. 1a. The spatial extents of the 4 regions along the slit were 3 $.\!\!^{\prime\prime}$0, 2 $.\!\!^{\prime\prime}$4, 4 $.\!\!^{\prime\prime}$8 and 3 $.\!\!^{\prime\prime}$6 for the MMT observations and 3 $.\!\!^{\prime\prime}$0, 2 $.\!\!^{\prime\prime}$6, 3 $.\!\!^{\prime\prime}$6 and 4 $.\!\!^{\prime\prime}$8 for the Keck II observations. Additionally, spectra with apertures 10 $.\!\!^{\prime\prime}$8 and 7 $.\!\!^{\prime\prime}$2 centered on region c were extracted from the two-dimensional MMT spectrum obtained at the position angle PA = 0$^{\circ }$ (regions 5 and 6 in the Tables 4 and 5).