3 Observations and data reduction

We used the TAURUS II camera in scanning Fabry-Perot mode, at the Cassegrain focus of the 4.2 m William Herschel Telescope at the ORM observatory La Palma, during the nights of March 26th and 27th 1996. The 125 $\mu$ etalon, and f/2.11 camera were employed, together with a TeK CCD, windowed to a size of $500 \times 500$ pixels, each pixel having an effective size of 0.56 $^{\prime \prime }$$\times$ 0.56 $^{\prime \prime }$. The nights were photometric, and the seeing had a mean value of 1.1 $^{\prime \prime }$.

The observations with TAURUS take the form of a series of images each in a narrow wavelength interval, the whole set covering the emission line in H$\alpha$ emitted by the galaxy. A narrow-band filter with the appropriate red-shift ( $\lambda_0 = 6589$ Å, $\Delta \lambda = 15$ Å) isolates the H$\alpha$ emission line; the wavelength was selected according to the recession velocity of the galaxy (1426 km s^-1 from RC3), and the filter serves not only to cut out contributions from the nearby lines of the [NII] doublet, but also to eliminate other orders of the interferometer. The whole "data cube" comprising 55 wavelength planes, with 140 s exposure time per plane, covered a full spectral range of 17.22 Å or the equivalent of 790 km s^-1. A "calibration cube" using a CuNe emission lamp was taken at the beginning of the night, and "calibration rings" before and after the data cube exposure on the galaxy. These were later used for phase and wavelength calibration of the data cube using the TAUCAL software package. After calibration the raw data cube, in which the surfaces of constant wavelength are paraboloids, was converted to a Cartesian cuboid of 55 planes, separated by 0.34 Å each of $500 \times 500$ pixels. After this we subtracted off the sky emission from each plane separately, and the planes were aligned using field stars. The next step was the astrometric relation of the positions of the H II regions with respect to these stars, using a high resolution H$\alpha$ image of NGC 6951 taken previously (Rozas et al. 1996a). The final reduced data cube had an effective angular resolution of 1.3 $^{\prime \prime }$.

With the reduced cube, we could apply MOMENTS, a set of tasks in the GIPSY suite of programmes, to produce moment maps which contain desired information about the intensity and velocity distributions implicit in the individual channel maps (i.e. the plane by plane information shown in Fig. 2). MOMENTS enabled us to calculate the zero, first and second order moments of the intensity in each pixel, which yield the intensity, velocity and velocity dispersion maps respectively. In performing these procedures we rejected as signal any bright pixel which was not reproduced in at least three adjacent wavelength planes. An initial inspection of the moment maps first obtained in this way revealed that the integration over the full velocity range had resulted in noise peaks, so that we had to revise the original data cube to find a method to minimize the effects of this noise. The procedure used for cleaning the cube is exactly that used for the equivalent data in NGC 3359, as described in detail in Rozas et al. (2000), using derived cubes at angular resolution 10 $^{\prime \prime }$, 6 $^{\prime \prime }$ and 3 $^{\prime \prime }$.

Figure 6: Integrated H$\alpha$ profile of NGC 6951.

The emission in H$\alpha$ as a function of wavelength channel, i.e. as a function of velocity, for the high resolution cube is shown in Fig. 2, as a set of "channel maps", following radioastronomical practice. The emission is presented in the central 40 $^{\prime \prime }$$\times$ 40 $^{\prime \prime }$ (4.2 kpc $\times$ 4.2 kpc). We detected significant H$\alpha$ emission in 33 channels, corresponding to a velocity range from 1214 to 1646 km s^-1. This range is comparable to that detected in CO and HCN by Kohno et al. (1999), although the H$\alpha$ emission is much stronger in the centre, i.e. in the circumnuclear zone, than the CO or HCN emission.

Figure 7: High resolution (1.3 $^{\prime \prime }$) velocity map for NGC 6951. The darker zone is the receding side (positive redshift) and the lighter part is the approaching side of the galaxy. The kinematical centre is marked with an asterisk.

The intensity, velocity, and velocity dispersion maps, obtained as moments of the data cube as outlined above, were the key results of the initial reduction. The high resolution (1.3 $^{\prime \prime }$) intensity and velocity maps are shown respectively in Figs. 3 and 7. There are in fact two possible approaches to produce these maps, one via the moments method, and the other by fitting the data within a given resolution element by a Gaussian in the velocity dimension. The resulting velocity maps were essentially the same for both methods, but there were notable differences in the intensity maps and above all in the velocity dispersion maps obtained using the two different techniques. This is because the moments method can lead to significant information loss in the line wings if these have low intensity and high velocity dispersion (van der Kruit & Shostak 1982). This is a result of the noise threshold, a fixed multiple of the measured rms noise, which leads to a cut-off level below which the information is deemed not to be significant. In the Gaussian profile method, the cut-off applied is a function of the signal strength itself, and is not a fixed level, which allows more sensitive detection of line wings, but may lead to a problem of treating noise as low level spurious signal in noisy images. In the present study we have chosen to use the moments method, knowing its limitations. A fuller analysis of the map of velocity dispersion, investigating the internal kinematics of the H II regions, and comparing both methods to obtain the full spectra of velocities over the full H II region population, together with the relation of the velocity dispersion in the principal component to the H II region luminosity will be presented in Relaño et al. (2001).