3 Spectrosopic follow-up observations

Spectroscopic follow-up observations have been focussed on the candidates of high or medium priority. Most of the spectra were taken during five observing runs either with the 2.2 m telescope on Calar Alto or with the Tautenburg 2 m telescope. An overview of these observation runs is given in Table 2. In addition, three candidates with uncertain spectroscopic identification were re-observed in July 2001 with CAFOS on Calar Alto; this run is quoted as number 6 in Table 3. An additional 18 candidates of medium or low priority were proved to be foreground stars during several other campaigns with either the Tautenburg telescope or the Calar Alto 2.2 m telescope.

  
3.1 Multi-object spectroscopy with TAUMOK

The brighter candidates (B<18) were observed with TAUMOK in the Schmidt focus of the Tautenburg 2 m telescope. TAUMOK allows to obtain simultaneously spectra of up to 35 objects within an area of $2\hbox{$.\!\!^\circ$ }3$ diameter (see Meusinger & Brunzendorf 2001 for more details). The telescope was operated in a scanning mode prior to the spectroscopic observations in order to determine the most accurate positions of the fibres. The wavelength coverage is approximately 3800-9000 Å, the reciprocal linear dispersion is 400 Å mm^-1 corresponding to 9 Å per pixel.

The atmospheric conditions during the TAUMOK campaign were moderate. In five of the seven nights, spectra of QSO candidates were taken. Four different fibre configurations were necessary to cover the VPM field. Several 1800 s exposures were taken for each configuration. The total exposure time per field is between 1.5 and 3 hours. Since the number of bright high-priority candidates is much smaller than the total number of available object fibres, most of the fibres were positioned at candidates of lower priority or on non-priority objects. Five fibres were reserved for template spectra from known QSOs with redshifts beteween z=0.6and 2.5. Internal spectral lamps were used for the wavelength calibration prior and after the observation of a field. For the reduction of the TAUMOK spectra we applied a software package (Ball 2000) which is based on IRAF standard procedures for multi-object spectroscopy.

 

 

Table 2: Observation log for the major spectroscopic follow-up observation runs.

observing run	1	2	3	4	5
spectrograph	TAUMOK	CAFOS	CAFOS	CAFOS	CAFOS
year/month	1997/04	1998/04	1999/04	2000/04	2001/03
number of nights	7	7	5	3	3
number of objects	41	46	34	23	35

Figure 3: Field of the VPM survey centred on the globular cluster M 3. Star-like objects are shown as grey dots. The panel on the left hand side shows the distribution of the QSOs from the VPM search ($\bullet$) as well as the QSOs with B<19.7 from the NED ( $\blacklozenge$). The panel on the right hand side shows the distribution of the VPM candidates that were spectroscopically confirmed to be foreground stars ($\circ$) as well as the medium-priority candidates without follow-up spectra (+).

  
3.2 Single-object spectroscopy with CAFOS

The candidates fainter than $B\approx18$ were observed with the focal reducer and faint object spectrograph CAFOS at the Cassegrain-focus of the 2.2 m telescope of the German-Spanish Astronomical Centre on Calar Alto, Spain. The B 400 grism was used with a wavelength coverage of about 3000-9000 Å. The width of the entrance slit was adjusted to the seeing (typically 2-3 arcsec) resulting in an effective linear resolution of typically about 40-60 Å. Since the orientation of the long-slit was always North-South, some "slit-loss'' due to atmospheric dispersion was unavoidable for spectra taken at hour angles significantly different from zero.

The total integration time varied between 10 and 90 min, dependent primarily on the strength of the emission lines and on the weather conditions. Observations were made in a wide variety of atmospheric conditions. The weather was good in the 2001 observing campaign. In the previous runs, however, the fraction of observing time with good atmospheric transparency was rather low. Therefore, about half of all spectra have only a moderate signal-to-noise ratio. Several objects had to be observed in more than one run. Data reduction was performed using the long-slit spectroscopy package LONG of MIDAS. Wavelength calibration was done by means of calibration lamp spectra.

  
3.3 Overview of the spectroscopically observed objects

Figure 4: Absolute magnitude $M_{\rm B}$ versus redshift zfor all known QSOs/Sey1s in the field identified with objects from our basic sample. The redshifts are either from the present study ($\bullet$) or the NED ( $\blacklozenge$). The continuous line indicates the magnitude limit $B\le 19.7$ of the survey.

Spectra were taken for a total of 198 objects: 1.) In particular, all 54 high-priority candidates without classification in the NED were observed. 2.) Among the 95 medium-priority candidates, 17 objects have a measured redshift in the NED. For an additional 68 medium-priority objects spectra were taken in the framework of the present study. The remaining 10 objects have variabilities near the selection thresholds (Fig. 2d) and/or are located at the borders of the field (Fig. 3) where the contamination by foreground stars is obviously stronger (see below). The chance to find QSOs among these remaining 10 candidates is substantially lower than for the median-priority subsample as a whole. 3.) The number of low-priority candidates is obviously too large to reach a substantial completeness with regard to spectroscopic follow-up observations. We selected therefore from this priority class mainly the brighter objects ( 18< B <19.5) and/or the objects with the strongest indication for variability. Note that the high fraction of QSOs/Sey1s among the observed low-priority objects is thus not representative for the whole subsample. There may be some undetected QSOs among the remaining low-priority objects, but their number is expected to be small.

The statistics of the observations for the various priority classes are summarised in Table 1. Note that the total number of observed candidates listed there is smaller than the number of all observed objects given in Table 2. The reasons for this apparent discrepancy are the following: (a) the criteria for the definition of priorities have slightly changed during the survey. (b) Many of the brighter objects observed with TAUMOK are not candidates in the sense of Table 1, but were selected to allow a good positioning of the fibres. (c) Several objects were observed in more than one observing run. (d) Since a variability survey is expected to be biased against low-variability QSOs, we selected also a few objects with quasar-typical colours, but with variability indices slightly below the selection thresholds. For instance, four objects were observed because they are X-ray sources. (e) A few strongly variable objects with 19.7<B<19.9 have been observed as well.

3.4 Source classification

Figure 5: Normalised distribution of the redshifts (number of QSOs per 0.2 redshift bin) for a) the subsample from the present follow-up spectroscopy (Table3), and b) the (nearly) complete subsample of all 114 QSOs with $B\le 19.7$. For comparison, the dotted polygon gives the normalized z distribution for the QSOs from the SDSS Early Data Release (Schneider et al. 2002).

The spectral classification is based on the emission and absorption line properties. Three catagories are considered: (1.) redshifted broad emission lines and/or absorption lines, (2.) redshifted narrow emission lines, and (3.) unredshifted typical stellar absorption lines.

The first category comprises QSOs and Sey1s, which are discriminated by the usual luminosity threshold $M_{\rm B} = -23$. The absolute magnitude $M_{\rm B}$ is computed for H₀ = 50 km s^-1 Mpc ^-1,  q₀ = 0 and the $K_{\rm B}$-correction from Brunzendorf & Meusinger (2001). The data for the 69 QSOs and 5 Sey1s from the present study are listed in Table 3, and in the following, objects will be quoted with their number in this table. For most of these objects, redshifts were derived from several emission lines; in particular, strong narrow forbidden lines (e.g., [O III] $\lambda5007$) were used, if present. The wavelength of a single line was measured by Gaussian centroids.

When we selected the first QSO candidate list in 1996, this list has been checked against the NED (cf. Paper I) to reject all objects already catalogued with measured redshifts. A new check revealed that six QSOs from Table 3 are catalogued in the February 2002 release of the NED. For three of these objects (No. 16, 68, 70) redshifts were published later than 1996. (Number 68 is identical with FBQS J1348+2840, White et al. 2000.) The other three (No. 29, 36, 42) had uncertain positions in the 1996 NED and therewith too large position differences (>10 arcsec) for an unambiguous identification. Further, we note that the QSO No. 11 is identified with the radio source (without z in the NED) FIRST J133825.6+283637.

There are four narrow-emission line galaxies (NELGs) among the identified objects from the NED. An additional three NELGs were detected in the present study and are also listed in Table 4. The luminosities of the NELGs are clearly below the QSO-Sey1 threshold. In general, the class of the NELGs includes Seyfert 2s, narrow-line Seyfert 1s, LINERs, and H II galaxies. For this paper we have not attempted to separate the types of NELGs. One of the new NELGs (No. 43) is a high-priority QSO candidate, another one (No. 26) is of low priority, but with a high overall variability index $I_{\rm var} = 2.45$. The third one (No. 15) has a high proper motion index and is not a QSO candidate from the VPM survey, but is one of the X-ray sources observed for completing the QSO sample. Among all objects with z>0 from our basic sample, No. 15 is the only one with a proper motion index significantly larger than the selection threshold $I_{\rm pm}=4$ (Fig. 1), perhaps indicating a wrong spectral classification from a noisy spectrum. All 7 NELGs were classified as star-like objects on the Schmidt plates; the infered redshifts are between 0.137 and 0.433. In the frame of the VPM search in the M 92 field, a higher fraction of NELGs was detected due to a less stringent star-galaxy separation (Meusinger & Brunzendorf 2001). The high variability indices measured for the NELGs were explained by increased photometric errors for objects with image profiles deviating from stellar ones (Meusinger & Brunzendorf 2002).

Finally, an object is classified as a foreground star if its spectrum unambiguously shows typical un-redshifted stellar absorption lines. At a first glance, most of these objects are normal stars without unusual spectral features. Contrary to the QSOs, the classified stars show a remarkably inhomogeneous distribution over the field (Fig. 3): their strong concentration towards the outer parts and the corners of the field indicates an increase of the instrumental variability at large distances from the plate centre. Such an effect is in principle expected since we have not corrected for a position-dependence of the magnitude scale (Paper I). This interpretation implies that a substantial fraction of the selected stars are not really variables. In this context we note that most of the stars have lower variability indices than the QSOs (Fig. 2d).

To summarise, we have plausibly classified all 198 objects from our spectroscopic follow-up observations as either QSOs/Sey1s, NELGs, or foreground stars. There are new redshifts for 68 broad-lined objects and 3 narrow-lined objects. For an additional 6 already catalogued QSOs/Sey1s redshifts were confirmed.