Astron. Astrophys. 355, 891-899 (2000)

2. Observations and data reduction

Seventeen exposures of 3000 seconds each of the blue compact galaxy IZw 18 were obtained with the 3.6 m Canada-France-Hawaii Telescope during three successive nights between 1995 February 1 and 4 using the MOS spectrograph with the 2048 2088 Loral 3 thick CCD detector. A long slit (1.52 arcsec wide) was used with a position angle of 45 ^o covering the spectral range from 3700 to 6900 Å . The position of the slit is displayed on Fig. 1. The spatial resolution was 0.3145 arcsec/pix and the dispersion 1.58 Å/pix, leading to a spectral resolution of about 8.2 Å. The seeing was between 1 and 1.5 arcsec. The spectra were reduced using IRAF. The bias was removed using the overscan section from each frame. The pixel-to-pixel sensitivity correction and the illumination effects (vignetting) were corrected using dome flat field and sky flat images. The images were calibrated in wavelength using a combination of two exposures made during the second night with a Neon and a Helium lamp respectively. Five 50 seconds exposure images of the standard star Feige 34 were obtained in order to flux calibrate the spectra. To account for wavelength-dependent atmospheric refraction, we fitted a low-order polynomial along the stellar continuum in each frame and then realigned each spectrum before combination. The spatial profile on the seven frames obtained during the first night was different from those ten obtained during the two other nights. We assumed that the positioning of the slit was slightly different during the first and the two other nights. Nevertheless, as the offset was less than 1" (slit positionning error), we aligned and combined all the nights together in order to increase the S/N ratio.

Fig. 1. V-band image of IZw 18 (from Hunter & Thronson 1995) with the slit position overlaid.

After reduction, an abnormal "diffuse light" background in the blue part of the long exposure images appeared. The origin of this "light" is probably due to a slight increase in the temperature of the CCD with time or to light diffused in the instrument during long exposures. This feature was removed by the subtraction of the background (task BACKGROUND) and using the task APSCATTER which is especially designed for this kind of purpose. The residuals after correction were less than 0.5 of the continuum level. Three bad columns (579 to 581 i.e., 4600 and 4602 Å respectively) of the CCD were ignored. We applied a Doppler correction to shift the final spectrum to zero velocity.

Spectra were extracted by summing along the slit. The apertures used were 5 pixels (1.57") wide with 2 pixels (0.63") of overlap. In order to increase the signal to noise (especially for the [OIII]4363Å line), large aperture spectrum were extracted summing over 12 pixels (3.78") every 6 pixels (1.89") along the slit, but this did not allow extension of the region over which [OIII]4363Å could be measured. We also extracted a large aperture spectrum integrated over the whole galaxy (25 pixels centered on the maximum of the continuum emission) in order to compare our observations with the spectroscopic measurements (but with different PA) of Skillman & Kennicutt (1993). A small aperture spectrum integrated over 2 pixels (0.62") has also been extracted to match the aperture used by Izotov et al. (1997). The large aperture spectrum is displayed in Fig. 2 and results of line measurements (for both small and large aperture) are shown in Table 1. Their mean FWHM is around 8 Å, and the lines are unresolved.

Fig. 2. Large aperture spectrum (over 25 pix i.e., 7.85") of IZw 18, with zooms on the blue and red part to show the faintest lines. The most important lines are labelled.

Table 1. Observed line fluxes (without reddening correction) of the NW component of IZw 18 for the large aperture (LA) and the small aperture (SA) spectra.

Emission lines were measured automatically using the routine TWOFITLINES ¹. We compared the measurements with those made interactively with Gaussian fits through the IRAF task SPLOT and found no differences larger than two percent. A few weak lines in regions of low S/N high, for which no Gaussian could be fitted, were measured by direct integration. The errors bars were computed by summing in quadrature the effective photon noise on the line flux and the rms noise in the local continuum. An additional two percent error accounts for uncertainties in the flat-fielding and sky+diffuse light subtraction process.

© European Southern Observatory (ESO) 2000

Online publication: March 21, 2000
helpdesk.link@springer.de