3 The data

3.1 Observations

We obtained 12 hours of observing time in service mode (Period 63) and 2 nights in visitor mode (Period 64) at the VLT/ANTU with the ISAAC spectrograph. We have used the long-slit mode of ISAAC in its Short Wavelength configuration: characteristics are given in Table 2. Our spectral domain includes the first three or four (depending on the redshift of the galaxy) ¹²CO bandheads, around 2.3 $\mu$m.

We defined Observing Blocks (OB) of 14 (NDIT) exposures of 180 s (DIT) each, operated in noding mode: the objects in exposures "A'' and "B'' were centred on the first and last third of the NIR array respectively. This (classical) procedure allowed us to have an excellent sky subtraction, using A-B and B-A differential exposures as working frames for the reduction.

We have observed 4 galaxies in our sample of 12, namely NGC 1097, NGC 1365, NGC 1808 and NGC 5728. Details for each galaxy are given in Table 3. Each galaxy was originally supposed to be observed for a total of 10 080 s (2.8 hours = 4 OBs). However, as mentioned in Table 3, we had to discard a number of exposures due to technical problems mainly due to:

• Spatial drifts of the slit during the OB;
• Slit jumps which resulted in a tilted slit both with respect to the NIR array and the noding.

We also observed a set of stellar kinematical templates (typically G, K and M giants) to be used for the kinematical measurements, and solar type stars for the correction of telluric features (see Maiolino et al. 1996).

In our spectral domain, there are no OH lines, generally useful to perform a wavelength calibration of the exposures. We had thus to rely on independent arc lamp exposures to perform our wavelength calibration. In this context, we asked individual arc exposures during the night.

 

 

Table 2: Instrumental setup of ISAAC

ISAAC SW mode
Slit	$0\hbox{$.\!\!^{\prime\prime}$ }6\times 120$ $^{\prime\prime}$
Spatial sampling	$0\hbox{$.\!\!^{\prime\prime}$ }147$
Spectral sampling	1.19 Å
Spectral resolution	4478
Spectral FWHM	67 km$\,$s-1
Wavelength interval	1200 Å centred at 2.336 $\mu$m

 

 

Table 3: DEBCA- ISAAC data characteristics. The observation period is given in column "P''. Numbers of OBs are indicated as used/discarded respectively. "Exp'' is the total exposure time on target. $FWHM_{\star }$ corresponds to the mean seeing. The orientation of each slit with respect to the nuclear bar major-axis is given in Col. 3, the PA of the slit is given in Col. 4

Galaxy	P	axis	PA	# OB	Exp.	$FWHM_{\star }$
			[ $\hbox{$^\circ$ }$]		[min]	[ $\hbox{$^{\prime\prime}$ }$]
NGC 1097	63	//	29.5	4 / 0	168	0.8
		$\perp$	119.5	5 / 0	168	0.7
NGC 5728	63	//	264.5	4 / 0	168	0.6
		$\perp$	354.5	4 / 0	162	1.5
NGC 1365	64	//	45.5	4 / 0	162	1.0
		$\perp$	135.4	4 / 1	162	1.0
NGC 1808	64	//	335.5	5 / 0	210	0.7
		$\perp$	65.5	3 / 2	112	0.6

3.2 Data reduction

In the following paragraphs, we give a brief description of the reduction and analysis procedure we applied to our data. We emphasize some of the problems we encountered on the way, most of them linked with instrumental issues (all data were taken prior to the major overhaul in Feb. 2000). All the reduction processes were applied using the IRAF and MIDAS packages, as well as a few low level routines from the Eclipse package.

Since we observed in nodding mode, we used the differential comparison (A-B and B-A) to subtract the dark, bias and sky contribution from all exposures. The data were then flat-fielded using a previously prepared master flat field image: variations of up to 5% were measured on the flat fields during a night. We then corrected for the distortion along and perpendicular to the slit, using the star-trace exposures provided by ESO, and associated arc lamps. Systematic residual (low frequency) distortion were of the order of 0.2-0.3 pixel, not fully satisfactory, but sufficient in the context of our program. It seems that these residuals cannot be further damped, as the distortion pattern varied on a medium time range at the time of the observations (which means that the star-trace exposures were not stable enough).

The data were then wavelength calibrated. As already mentioned, there are no sufficiently bright OH lines in our spectral domain to allow any spectral calibration, and we had to rely on independent arc lamp exposures. Unfortunately, at the time of the observations, there was a (known) problem with the dispersor which seems to shift from one OB to the next, following an automatic software initialisation. We have indeed observed some significant shifts (typically a few tenths of a pixel) along the dispersion direction between successive OBs. This is critical for our program as we are looking for a velocity accuracy of <5 km$\,$s^-1, a third of a pixel. This problem was solved by using sky emission lines to correct for any residual zeroth order shift.

Individual exposures are then combined, after careful recentring, and corrected for telluric absorption using a solar type stellar template as described in Maiolino et al. (1996), and taking into account the difference in line depth (depending on e.g. the differential airmass). The result is illustrated in Fig. 1 for a K0 III star.

Figure 1: ISAAC (aperture) spectrum of HD 16492, a K0 giant, before (bottom) and after (top) correction for the telluric absorption. The main 12CO lines are identified

The present data reduction only provided us with a relative flux calibration, sufficient for kinematical purposes.

   
3.3 Kinematical analysis

The (stellar template and galaxy) spectra were finally rebinned in $\ln{(\lambda)}$ to be sampled with constant bins in velocity space. We first binned spectrally by a factor of 2 as this leads to a pixel of about 31 km$\,$s^-1, properly sampling the original spectral resolution of the data (see Table 2). We also binned the data spatially along the slit to ensure a minimum signal to noise ratio of 20, required to extract the stellar kinematics. We then performed a continuum subtraction using a low order polynomial. A refined version of the Fourier Correlation Quotient (Bender 1990) was used to derive the line-of-sight velocity distribution and to measure the first two velocity moments (V and $\sigma$): we used different templates and checked that the resulting kinematics were not significantly affected by template mismatching. Measurements of higher order Gauss-Hermite moments will wait for the building of optimal templates (Paper II). The central velocity value was assumed to be the systemic velocity and subtracted from each individual velocity profile.

We derived formal errors for the kinematics using a Monte Carlo approach. Fixing the signal to noise ratio and the velocity dispersion, we made 500 realisations of simulated broadened spectra, measured the kinematics via FCQ, and derived the resulting standard deviation for V and $\sigma$, S_V and $S_{\sigma}$respectively. S_V and $S_{\sigma}$ were tabulated for 5 values of $\sigma$ and 40 values of the signal to noise. We then derived the errors for individual data points via interpolation.