3 Reductions

Reduction of the data cubes were performed using the CIGALE/ADHOC software (Boulesteix 1993). The data reduction procedure has been extensively described in Amram et al. (1995, 1996, 1997) and references therein. The accuracy of the zero point for the wavelength calibration is a fraction of a channel width (<3 km s^-1) over the whole field. OH night sky lines passing through the filter were subtracted by determining the emission in the field outside the galaxies (Laval et al. 1987).

Since we used a high order Fabry-Perot interferometer, FP2, the FSR of the interferometer was of the same order of magnitude as the average H$\alpha$ line width ($\sim$ 100 km s^-1) of the target galaxies. This means that in general we have no emission line free continuum. Moreover, the wings of the emission lines are shifted by a FSR (+1 FSR for the blueshifted wing and -1 FSR for the redshifted wing) and superimposed on the central part of the line. However this does not affect the velocity of the line, as long as the velocity range of the galaxy is lower than the FSR. When this is not the case, a velocity jump close to one FSR could easily be detected and corrected for using the continuity of the isovelocities. The only problem is that monochromatic (H$\alpha$) emission could be mistaken for continuum, which in effect would lower the measured line intensity and over-estimate the continuum level. To separate the continuum from the body and wings of the line, we used a method based on fitting theoretical line profiles to the observed profiles. The instrumental line width (apparatus function) of the interferometer has a FWHM of $10~\pm~2$ km s^-1, where the range reflects the (known) variation of the FWHM over the field of view. Thus the observed widths of the profiles are dominated by the intrinsic H$\alpha$  line widths in the target galaxies. The convolution of the apparatus function with a theoretical H$\alpha$ profile yields something very close to a Gaussian. The function used to fit to the observed profiles was a "folded" Gaussian: its wings were cut (at $\pm$ one half FSR from the centre of the Gaussian), shifted (by $\pm$1 FSR) and added to the main body of the profile. The reduced datacube provide the centre of the central wavelength, and the sum of the continuum and monochromatic (H$\alpha$) emission. Thus only two parameters are unknown: the width and the amplitude of the Gaussian (i.e. the width and peak intensity of the H$\alpha$ line). These two parameters are determined by a least-squares fit and thereby both the continuum level and the H$\alpha$ intensity are determined for each pixel. To check the validity of this method, one galaxy (ESO 338-IG04) was observed with two different interferometers having different interference order: FP1 and FP2. Using FP1, the signal measured along the scanning sequence is separated into two parts: an almost constant level produced by the continuum light passing through the narrow band interference filter, and a varying part produced by the H$\alpha$ line emission. The continuum level is the mean of all channels which do not contain monochromatic (i.e. H$\alpha$) signal. First of all, we checked that without any "data cooking", the velocity fields have the same shape with both FPs. The agreement is even better if we degrade the high resolution FP2 data to the spectral resolution of FP1, see also Sect. 6.3. As a consequence of the lines having widths comparable to the FSR when using FP2, the continuum level and the true width of the lines are not known with accuracy; and therefore these quantities are not presented.

A rough flux calibration of the H$\alpha$ and continuum images of the galaxies could be made using the known instrumental sensitivity. For a narrow band filter having 100% transmission at the observed wavelength, one photo electron corresponds to a flux of $8.32 \ 10^{-17}$ Wm^-2 s^-1, with an uncertainty of $\pm 20\%$.Correcting for the transmission of the narrow band filter and the atmospheric absorption we could then determine the total (line plus continuum) flux detected. As explained above, the separation of continuum and H$\alpha$ emission is not trivial. However, in all cases the H$\alpha$ emission dominate the detected flux. In effect the total H$\alpha$ fluxes are known with reasonable accuracy, while the absolute continuum level is uncertain. Comparing with available R-band photometry (e.g. Bergvall & Olofsson 1986) we conclude that the contribution from the continuum to the detected flux is of the order of a few percent, which is slightly lower than the estimate form the continuum fit. In Table 5 the derived H$\alpha$ fluxes are presented. The quoted uncertainties represent the quadratic sum of the intrinsic uncertainty (20%) and the estimated uncertainty in determining the flux for each galaxy ($10-30\%$ due to uncertainty in the continuum level and the exact transmission of the filter at the wavelength of the redshifted H$\alpha$ line). The total derived H$\alpha$ fluxes can be used to estimate ${\cal M_{\rm ion}}$, the mass of the ionised H$\alpha$ emitting gas. For this we assume an electron density, $n_{\rm e}$, of 10 cm^-3, an electron temperature of 10000 K, a mean molecular weight, $\mu$, of 1.23 (to include the mass of helium) and use the H$\alpha$ effective recombination coefficient for case B as given by Osterbrock (1989). The assumed density is in the lower range of observed densities of the H II gas in BCGs, which usually lie in the range $10 < n_{\rm e} < 300$ cm^-3 (see e.g. Bergvall 1985; Masegosa et al. 1994) and thus our ${\cal M_{\rm ion}}$ estimate could be regarded as an upper limit, perhaps one order of magnitude too high.