2.1. VLA HI observations
The HI observations were made in two configurations. We used the VLA D-configuration (maximum baseline 1.03 km) on 8 November 1993 to observe the Arp 105 field. Two sessions in C-configuration (maximum baseline 3.4 km) were added, on 20 and 27 October 1994. The observational set-up is listed in Table 2. For a description of the VLA, see the article by Napier et al. (1983) . As always, one has to trade off spectral resolution against velocity coverage. Our aim was to get 10 km s-1 after Hanning smoothing the data, which normally is obtained by calculating a 64 channel spectrum over 3.125 MHz, corresponding to a velocity range of about 600 km s-1, and recording the signal in both right hand (R) as well as left hand (L) polarizations. Because the sensitivity drops off rapidly at the edges of the band, this set-up effectively covers a velocity range of 450 km s-1. As we were not entirely sure about the range over which we could expect HI emission, and 450 km s-1 would not have been sufficient to record line-free channels which are to be used to subtract the continuum emission, we decided to err on the safe side and use the capability of the VLA to tune the R and L polarizations to different frequencies (or central velocities). The band receiving R signal was centred at 8775 km s-1, the one recording L signal was at about 8622 km s-1. Thus, after deleting those channels which are near the edge of the passband, we covered the velocity range from 8400 km s-1 to 9000 km s-1 ; full sensitivity was obtained from about 8550 to 8850 km s-1, the remaining channels being less sensitive by a factor of as only one polarization is available (either R or L).
Table 2. VLA Observing Parameters
We used source 1328+307 (3C286) for absolute flux calibration and to determine the bandpass. We assumed a flux density of 15.06 Jy on the scale. Our secondary calibrator was 1153+317 which was observed on average every 30 minutes for 3 minutes. Its flux density was measured to be 3.05 Jy. The data were calibrated and mapped using the NRAO AIPS package.
Solar interference was visible in the D-array observations. As the data were taken before and after sunrise, part of the data were unaffected. As solar interference is concentrated on the shorter wavelengths, and as the uv- plane substantially oversamples these shorter spacings, we decided to edit out the shorter spacings (shortward of ) which were recorded after sunrise (or about of the run). This hardly increased the noise level in the maps and virtually eliminated the effects of the Sun.
The three sets of observations were merged in uv- space and Fourier transformed to yield maps of the HI distribution. A data cube was produced, using natural weighting, which has the lowest noise of 0.21 mJy beam-1 channel-1 at an angular resolution of (note that the noise value is for those channels where both R and L polarizations were measured). The data were blanked at the level and those features which were present in at least 3 consecutive channel maps were retained. Out of 64 useful channels (after merging the overlapping bands), channels 12 until 47 contain line emission. Except for channel 12, all these channels have both R and L polarization data. The remaining channels provided the continuum which was subtracted in the map plane. The continuum subtracted maps were cleaned down to a level of 0.5 mJy beam-1. Moment maps were calculated based on the blanked data.
2.2. 12 CO(1-0) PdB observations
The CO observations were centred on the redshifted restfrequency of 112.0022 GHz and were carried out between January and March 1994 with the IRAM 4 element interferometer under excellent weather conditions. We divided our time between two pointings, one towards the spiral and one towards the elliptical. A detailed description of this instrument was presented by Guilloteau et al. (1992). Table 3 summarizes the observing parameters. We used 5 configurations with baselines extending out to 280 m (BC configuration set). The observing sequence consisted of 4 min integrations on the calibrator 1156+295 followed by 16 min on the spiral and 4 min on the elliptical galaxy. Typical SSB system temperatures were 350 K. We configured the 6 correlator units to give 0.625 MHz channel spacing for the inner 220 MHz wide band and 2.5 MHz spacing for the inner 400 MHz band. The bands were overlapping to avoid the Gibbs phenomenon and edge effects. Amplitude and phase calibration was done against 1156+295, which itself was referred against 0316+413 (3C84), 0923+392, 1226+023 (3C273), and 1749+096, at least one of which was observed each day to calibrate the bandpass of the receiver. During the 3 observing months we noted no flux variation for 1156+295 at a limit of 20% and used a flux of 1.7 Jy for amplitude calibration.
Table 3. Plateau de Bure Observing Parameters
The data were calibrated and analyzed with the CLIC and GRAPHIC software packages developed at IRAM and Observatoire de Grenoble. For the spectral maps original channels were combined to achieve a 10 MHz spacing or 6.7 km s-1 velocity resolution. Map sizes were 128 x 128 cells of each. The synthesized beams were at position angle ( ) for both spiral and elliptical field centres. The conversion from flux density into main beam brightness temperature was calculated assuming the Rayleigh-Jeans approximation.
In order to superimpose the gas distributions on the optical image of the system, careful astrometry of the field had to be performed. This is particularly critical when making comparisons between CO and the optical because of their relatively high spatial resolution. The astrometric errors at 112 GHz are less than . The optical image used in this paper was taken during an observing run carried out in January 1992 at the Canada-France-Hawaii Telescope (Paper I). The pixel size was and the seeing . Unfortunately only 4 Guide Star Catalog objects, all extended, and no Proper Motion Stars were present in the x field of the CFHT image. In order to obtain an astrometric solution, corrected for distortion, secondary reference sources had to be used. These were found in a STScI Digital Sky Survey field, the astrometry of which had previously been determined on a larger field of x , using 34 GSC sources. The positions of 11 secondary stars were determined that way. A 3rd degree 2D polynomial transformation was applied to the original frame, using IRAF. The relative astrometric precision obtained that way was better than . The absolute astrometry was afterwards derived from the position of a QSO (QS1108+285) present in our field, the coordinates of which had been determined by Veron et al. (1976) , with a precision of .
© European Southern Observatory (ESO) 1997
Online publication: October 15, 1997