2. The data
Long-slit spectroscopy was obtained during two runs with the ESO 2.2m telescope in 1988 and with the ESO 1.52m telescope in 1991 using a Boller & Chivens spectrograph. Details on the observations are given in Table 1. For the 1988 run, the CCD used was #8, with a pixel size of 15 µm, corresponding to 0.9 arcsec on the sky; the grating was #26 (dispersion 59.5 Å /mm), giving a spectral resolution of 1.9 Å . For the 1991 run, the CCD used was #13, with a pixel size of 15 µm, corresponding to 0.68 arcsec on the sky; grating #23 was used (dispersion of 129 Å /mm), giving a spectral resolution of 4.8 Å . The slit width was 2 arcsec for both runs. Wavelength calibration lamps (He-Ar) were taken just before or just after each exposure.
Table 1. Journal of Observations.
A map of the [OIII] narrow band image from Durret & Bergeron (1987), with the slit positions superimposed, is displayed in Fig. 1. The exact positions of the slits on this image are derived assuming that they were centered on the maximum of the broad band emission, and by aligning our broad band images with the [OIII] image using two stars on the frames. Previously to any offsetting of the slit, a short exposure on the nucleus with the same position angle was performed, to allow the determination of the offset slit position.
The data reduction was performed, following the standard procedures for de-biasing, flat-fielding and calibrating the spectra, with the IHAP and MIDAS softwares for the 1988 and 1991 runs respectively.
The accuracy of the wavelength calibration was checked by measuring the position of the strong sky lines along each spectrum; the dispersion of the sky line wavelengths are always within the error bars of the calibration with the arc spectra. It is only for the [OIII] spectra along PA=122/0 that we found a shift of 54 km s-1, which was applied to the final data for this PA. Typical wavelength calibration errors give velocity uncertainties of 20 and 60 km s-1 for the 1988 and 1991 runs respectively.
Since the illumination of the slit in the considered spatial region is uniform, a 2D-sky could be determined by averaging two strips on either side off the galaxy and subtracted to the galaxy spectrum.
The ionized gas velocity field along each slit is obtained by applying a cross-correlation method (Tonry & Davis 1979) to the emission lines by using a program developed by J. Perea within the FIGARO software; this allows to calculate relative velocities with respect to a reference cross-section chosen to have a high signal to noise ratio. The redshift corresponding to this cross-section is measured by fitting gaussian profiles to the different emission lines. We checked that the cross sections corresponding to the nucleus on the various slits had the same velocity within the error bars. The systemic velocity deduced is 2330 km s-1, in agreement with the stellar value measured by Nelson & Whittle (1995). Since the cross-correlation method takes into account all the emission features available in a given wavelength range, the velocities obtained are weighted averages, and are therefore more representative of those of the strongest lines, and/or strongest component in a line. Thus, as velocities in the H -[OIII] range are dominated by [OIII], the high ionization component, we made an attempt to derive [OIII] and H velocities separately. Unfortunately, H is only bright enough close to the nucleus. We therefore decided to derive the velocity field in [OIII] on one hand (hereafter the high excitation domain), and in H -[NII] on the other (low excitation domain). In any case, these are not peak velocities. We plot the [OIII] (squares) and H (triangles) velocities in Figs. 2 - 3 - 4.
The use of cross-correlation methods gives smaller uncertainties for the relative velocity distributions than the typical wavelength calibration errors. This can be seen in Figs. 2 - 3 - 4, where the error bars correspond only to the velocity accuracy with respect to the reference cross section. Therefore, realistic error bars are somewhat larger.
Along PA= offset by 7 arcsec to the northeast (Fig. 3d), the signal to noise was not sufficient and the cross-correlation method could not be used. So we binned spatially the data over a few cross sections and fitted gaussians to the line profiles to estimate velocities.
Since the H and H domains were not observed simultaneously and the nights were not photometric, it is not possible to draw excitation maps.
Notice that since the velocities derived from the cross correlation are not peak velocities, the presence of various components in a single line or different line shapes due to dust contamination (producing asymmetric emission lines) can produce artificial kinematical shifts (see below).
© European Southern Observatory (ESO) 1998
Online publication: April 20, 1998