As an example of typical results, Fig. 1 displays the images ("cleaned'' and Hanning-smoothed in velocity and R.A.) for 3 adjacent declination strips of our survey The velocity range is indicated on the horizontal axis and the R.A. range (800 spectra) on the vertical axis. The strong positive and negative signals around 0 kms-1 are due to residuals of the Galactic HI emission. Some standing waves due to this emission entering the receiver are visible in the lower velocity band.
|Figure 1: A "clean'' spectrum, Hanning-smoothed in RA and velocity , of 3 adjacent strips of the HI drift-scan survey in the CVn constellation. The figure contains about 48 000 individual spectra. Eleven extragalactic signals from 9 galaxies are visible in these strips|
On the three images displayed here, eleven extragalactic signals from 9 galaxies can be identified. We inspected the images of the 17 strips of the survey carefully by eye and listed all candidates as well as all detections above the 4 threshold.
The galaxy candidates thus identified were run through the NED and LEDA databases for crossidentifications. The match in positions, velocity if available, morphology and orientation were taken into account.
The positions of detections identified in the driftscan survey were determined according to the line number: sequence in R.A. of the 800 contiguous spectra (cf., the 11 detections in Fig. 1) as ). The 16 s reflect the integration time per spectrum plus the 0.33 s per read-out period. The positions (R.A. and Dec) were interpolated if the signal appears in various lines or strips according to the strength of the signal. The positional accuracy turned out to be remarkably precise in R.A. over the whole R.A. range (cf., Table 2a) and reasonable in Dec if a candidate was evident in more than one strip.
If no clear-cut crossidentification could be made, the POSS I and II sky surveys were inspected. If this did not yield a likely counterpart, the five scans leading to the final images were inspected individually to decide whether the signal was produced by radio interference in an individual scan.
In this manner, 53 galaxy candidates were retained from the 17 Nançay strips. Of these galaxy candidates, 33 could be identified with an optical counterpart, 30 of which have published HI velocities in agreement with our driftscan measurements.
For the 20 galaxy candidates without a clear optical counterpart, pointed follow-up observations were made with the Nançay telescope. These follow-up observations are described in Sect. 3.2.2 and allow the establishment of a database of blind HI detections in the volume sampled, which is reliable down to the 4 level. Interestingly, none of the candidates could be confirmed. But, note that for the number of independent measurements that were obtained in this survey at, say 50 kms-1 resolution, one expects to obtain of the order of 15 positive and 15 negative deviations exceeding purely by chance.
The observing technique recorded the integrations every 16.3 s, during which time the beam drifted through the sky by a full Half-Power-Beam-Width. In comparison, a "pointed'' integration, with the source of interest located on the beam axis, would reach a sensitivity of about Jy for a given velocity resolution after s integration with dual polarization. This would lead to a detection limit for integral line fluxes Jy kms-1. However, sources that pass through the beam do not appear in the data at the full strength they have when observed at the peak of the beam for the full integration time. Instead, a source that traverses the HPBW from one half-power point to the other suffers a loss in signal-to-noise ratio of a factor of 0.81, and a source that chances to pass the centre of the beam at the boundary of two integrations suffers a factor of 0.74. (Had samples been taken at 8 s intervals, these factors would be 0.84 and 0.80.)
An additional loss of S/N occurs because the survey declination strips were spaced by a full beamwidth (22'). A source falling 11' from the beam centre is observed at half power. Some of this loss in sensitivity is recovered by averaging adjacent strips, so that these sources midway between the two strip centres then experience a net loss in S/N of relative to the strip centre.
Combining the loss factors incurred due to R.A. sampling and declination coverage shows that the S/N is reduced relative to pointed observations by factors ranging from 0.81 to . Thus there is a range to the detection limit. It is clear from the distribution of non-confirmable signals (cf., Fig. 2) that there is also non-Gaussian noise at work, such as that resulting from radio interference, leading to a few spurious signals with apparent significance of as much as .
The HI line flux and the profile velocity width are two measured quantities. Figure 2 shows the integral line flux for each galaxy plotted as a function of its velocity width. The detection limit (delineated in Fig. 2) rises as , since the signal I from a galaxy with flux density grows as , while the noise in measuring a flux density is distributed over is , so that the signal-to-noise ratio drops as a signal of constant I is spread over increasing width. The detection boundary would be a band for true fluxes due to the lower survey sensitivity for detections offset from the centre of the survey strip. The dashed line in the figure marks a region to the lower right that is not expected to be populated, since galaxies of large velocity width typically have large HI masses, and, to have such a low measured flux, these galaxies would lie beyond the end of the survey volume, which is limited by the bandwidth of the spectrometer.
|Figure 2: HI line flux for each galaxy plotted as a function of linewidth measured at the 50% level. Filled dots are confirmed detections; unfilled circles are false detections that were not confirmed in the "pointed'' follow-up observations. The solid line indicates the detection level. Detections are not expected below the diagonal dashed line at the lower right|
Within a declination strip, the HI parameters were measured from the sum of the lines in which a candidate was detected using the Spectral Line Analysis Package by Staveley-Smith (1985). We list here the heliocentric central line velocity using the optical convention , , the profile FWHM, W50, the profile width measured at 20% of the peak flux density level, W20, the integrated line flux, I, and the logarithm of the HI mass, . The latter should be regarded as preliminary as the HI mass was determined for the strongest appearance of the signal and is uncorrected for the reduction in observed flux (described in Sect. 3.1) due to offset from the centre of the beam. Moreover, the adopted distances were taken straight from the heliocentric HI velocity using a Hubble constant H0=100 kms-1 Mpc.
For each HI detection a possible optical identification is given. Optical redshifts are known for 17 optical candidates, though 30 do have independent HI velocities (cf., Table 4). The listed optical data are mainly from the NED and LEDA databases; D and d are the optical major and minor axis diameters, respectively, and is the heliocentric optical systemic velocity.
These follow-up observations have an rms noise of about 6 mJy on average (compared to 10 mJy for the survey). We obtained our observations in total power (position-switching) mode using consecutive pairs of two-minute on- and two-minute off-source integrations. The autocorrelator was divided into two pairs of cross-polarized (H and V) receiver banks, each with 512 channels and a 6.4 MHz wide bandpass. This yielded a velocity resolution of kms-1, which was smoothed to 10.5 kms-1, when required, during the data analysis. The centre frequencies of the two banks were tuned to the radial velocity of the galaxy candidate (cf., Table 2b).
The follow-up observations first were repeated on the optimized position from the driftscan detection. If not confirmed, a second round of pointed observations was made, offset by half a beam (11)to the north and south of the original position while the original R.A. position which can be determined to higher accuracy was kept constant. In a few cases, a further search at quarter beam offsets was done as well. The follow-up observations consisted of a number of cycles of 2 min on- and 2 min off-source integration. Off-source integrations were taken at approximately 25 east of the target position.
We reduced our follow-up pointed HI spectra using the standard Nançay spectral line reduction packages available at the Nançay site. With this software we subtracted baselines (generally third order polynomials), averaged the two receiver polarizations, and applied a declination-dependent conversion factor to convert from units of to flux density in mJy. The -to-mJy conversion factor is determined via a standard calibration relation established by the Nançay staff through regular monitoring of strong continuum sources. This procedure yields a calibration accuracy of .In addition, we applied a flux scaling factor of 1.25 to our spectra based on a statistical comparison (Matthews et al. 1998) of recent Nançay data with past observations. The derivation of this scaling factor was necessitated by the a posteriori discovery that a number of line calibration sources monitored at Nançay by other groups as a calibration normalization check (see Theureau et al. 1998) were quite extended compared with the telescope beam, and hence would be subject to large flux uncertainties.
None of the candiates were confirmed, despite the fact that some of them have a rather large mean HI flux density. Only 4 "candidates'' are below our threshold. But, as mentioned above, for the number of independent measurements that were obtained in this survey we expect of the order of 15 positive and 15 negative deviations exceeding purely by chance. Many of the "false'' detections have very low linewidths (around 30 kms-1) and could still be due to radio interference.
The most disappointing candidate was No. 50. This object was identified on two adajcent declination strips and on 2-3 adjacent spectra in R.A. with very consistent HI properties and a similar high S/N ratio (8 - 10) on both declination strips. It therefore seemed one of the most promising new gas-rich dwarf candidates in the CVn region.
Seven dwarf candidates of the optical BTS survey were not detected in the blind HI line survey. With the exception of BTS 160, these dwarfs were observed individually to a lower sensitivity as the driftscan survey, in the same manner as described in Sect. 3.2.2. We did not include BTS 160, because this dwarf had already been detected by Hoffman et al. (1989) at Arecibo with 0.9 Jy kms-1. With an average flux density of 26 mJy this dwarf was hence clearly below the threshold of our driftscan survey. Of the 6 remaining dwarfs, none were in fact detected, although the pointed observation of BTS 151 revealed a strong signal with a flux of 20 Jy kms-1 (cf., values in brackets in Table 3), i.e., a detection which should have popped up as about a 10 detection in the driftscan mode. However, this signal matches exactly the velocity of the nearby large spiral galaxy NGC 4656 (cf., Tables 2a and 4, object No. 32) which is 45 and 17 away in R.A. and Dec respectively, hence less than a beamwidth from the pointed observation. The detected signal thus clearly originates from the spiral NGC 4656, and not from the dwarf elliptical BTS 151.
The data of these observatons are summarized in Table 3. Most of these possible dwarf members of the CVn group were not detected in HI. This supports their classification as early type dwarfs.
Of the 33 reliable detections, 30 galaxies had HI velocities published in the literature. For the 3 galaxies without prior HI data we obtained follow-up observations with the Nançay radiotelescope. We also obtained pointed follow-up observations for 8 galaxies already detected before in the HI line, which seemed to merit an independent, new observation. The observations followed the procedures as described in Sect. 3.2.2.
The results from the driftscan images are summarized in Table 4 together with our new pointed observations and HI parameters from independent observations.
Overall, the agreement between the measurements obtained from the driftscan survey and independent pointed observations are very satisfactory, with a few discrepancies discussed below:
No. 12 (NGC 4173): The HI line parameters of this 5' diameter edge-on system measured at Arecibo by Williams & Rood (1986) and independently by Bothun et al. (1985) are quite different from the other available values.
No. 13 (NGC 4203): This lenticular galaxy has an optical inclination of about 20. Assuming that the gas rotates in the plane of the stellar disc, very high HI rotational velocities would be derived. However, radio synthesis imaging (van Driel et al. 1988) has shown that the outer gas rotates in a highly inclined ring at an inclination of about 60, the value adopted in correcting the profile widths (see Table 5).
For 4 galaxies (No. 33 = UGC 7916, No. 34 = UGCA 309, No. 35 = NGC 4861, No. 38 = CGCG 0999) the listed properties such as morphological type, diameters and magnitudes differ between the LEDA and NED. The values in Table 2b and Table 5 are from LEDA.
We have compared the systemic velocities, line widths, integrated line fluxes and HI masses of the 33 reliable survey detections with available pointed observations. It should be noted that the comparison data represent in no way a homogeneous set of measurements, as they were made with various radio telescopes throughout the years; especially, care should be taken that no HI flux was missed in observations with the round Arecibo beam.
A comparison of the difference between systemic velocities (actually, the centre velocities of the line profiles) measured at Nançay and elsewhere shows no significant dependence on radial velocity, with the exception of one data point for NGC 4173 (No. 12). Here we consider the radial velocity of 1020 kms-1 measured at Arecibo by Williams & Rood (1986) as spurious, seen the agreement between the 3 other measured values. The mean value and its standard deviation of the Nançay-others systemic velocity difference is kms-1 (for a velocity resolution of 10.2 kms-1 at Nançay).
A comparison of the difference between the W50 and W20 HI line widths measured at Nançay and elsewhere shows a good correlation as well; the largest discrepancy is found between the survey values for the 5' diameter edge-on system NGC 4173 (No. 12) and the Arecibo W50 and W20 measurements by Williams & Rood (1986) as well as by Bothun et al. (1985), while the other pointed observations of this object are consistent with ours.
A comparison of the integrated HI line fluxes measured in the Nançay survey and elsewhere shows a reasonable correlation. This indicates that the assumption of a constant (uncalibrated) system temperature of 50 K throughout the survey is justified. Three larger discrepancies occur; this concerns No. 13 = NGC 4203 for which the Nançay survey flux is considerably higher than the flux measured at Arecibo. As the galaxy's outer HI ring is much larger than the Arecibo beam (see van Driel et al. 1988) the Nançay measurement will reflect the total flux. For No. 06 = BTS59 and No. 14 = UGC 7300, the Nançay flux is also considerably higher than a flux measured at Arecibo by Hoffman et al. (1989), respectively Schneider et al. (1990); both are objects of about optical diameter only.
A comparison between the HI masses derived straight from the detections in the strip-images with pointed observations shows an astonishingly narrow correlation from the lowest to the highest HI masses detected in this survey. Overall, only a slight offset towards lower masses is noticeable for the driftscan results compared to pointed observations. This obsviously is due to the fact that the driftscan detects the galaxies at various offsets from the centre of the beam.
Copyright The European Southern Observatory (ESO)