Astron. Astrophys. 354, 17-27 (2000)

2. Observations and data analysis

We used most of the published QSO surveys (Condon et al. 1977; Foltz et al. 1989; MacAlpine et al. 1977; MacAlpine & Feldman 1982; Olsen 1970; Sargent et al. 1989; Schneider et al. 1991, 1992; Schmidt et al. 1987; Steidel et al. 1991; Wolfe et al. 1986) and also parts of the Hamburg QSO survey (Reimers et al. 1989; Hagen 1993) to compile an appropriate sample of high redshift quasars. The redshift range of 2.7 z 3.3 was chosen to asure that most of the diagnostic ultraviolet emission lines are shifted to the optical regime. Generally, the spectra of high redshift quasars are contaminated by absorption line systems like damped Ly systems and the Ly-forest. We selected quasars with no strong contamination since we are predominantly interested in the emission line flux and profiles. The quasars had to be brighter than 20 mag to achieve a sufficient S/N ratio of at least 20 for the continuum for reasonable integration times for observations with telescopes of the 3 m class. This original sample was complemented by four quasars which were observed with the VLT UT 1 Antu.

We observed the 16 quasars of our sample at Calar Alto Observatory/Spain (August 1993), McDonald Observatory/USA (July 1995), and at Paranal/Chile (Sept., Oct. , and December 1998). In Table 1 the quasars and details to the observations are listed. The name of the quasar is given in (1), the coordinates and denote column (2) and (3). In (4) the redshift of the quasars is listed. We measured the emission lines of Ly, SiIV1400, and CIV1549 to determine the redshift. A Gaussian profile was fitted to the upper part of the line profile characterized by more than 50 % of the peak intensity. The uncertainty of the measured redshift is of the order of rms(z) = 0.01. The apparent and absolute magnitude are given in (5) and (6), respectively. The date of the observation is given in column (7) and the integration time in (8). Finally, the site of the observations is listed in column (9).

Table 1. Observation log of the high redshift quasars.
Notes:
a) measured from the Ly, SiIV1400, CIV1549 emission lines
b) taken from Hewitt & Burbidge 1993
c) CA, 3.5 m telescope Calar Alto Observatory/Spain; McD, 2.7 m telescope McDonald Observatory/USA; Para, 8.2 m telescope VLT UT1 Antu Paranal/Chile

At Calar Alto Observatory we used the twin-spectrograph which was attached to the Cassegrain focus of the 3.5 m telescope. Tektronix CCD detectors with 1024  1024 pixels were mounted in both channels. Gratings with 72 Å /mm and 108 Å /mm were used in the blue and red channel of the twin-spectrograph, respectively. In the blue channel a wavelength range of 3700 - 5505 Å was covered while the red channel provided a coverage from 5465 - 8180 Å. The overlap of the spectra of the blue and red channel amounts to at least 20 Å around  = 5500 Å. The slit width was fixed to projected on the sky for both channels and a fixed position angle of PA = 0^o. For all exposures the hour angle of the quasars was less than 30 ^o to minimize the light loss caused by differential refraction. Furthermore, the effect was less than for the zenithal distance of the objects during the exposures.

Helium-argon spectra were taken after each object exposure for wavelength calibration. The relative uncertainty of the calibration amounts to =0.2 Å. Spectra of the standard stars BD+26^o 2606, BD+33^o 2642, BD+28^o 4211, and Feige110 were observed for flux calibration each night (Massey et al. 1988; Oke 1990; Oke & Gunn 1983; Turnshek 1990). The quasars and the standard stars observed for flux calibration can be taken as point sources. Therefore, the light loss due to the slit width used is nearly identical especially because the seeing during the observations was quite constant at . Furthermore, our main interest is relative line flux ratios and the corresponding emission line profiles hence we recorded the spectra with a slit width of only to obtain a spectral resolution which is sufficient to study the line ratio for different parts of the line profile.

The observations at the McDonald Observatory/USA were obtained with the 2.7 m telescope using the LCS longslit spectrograph. The measurements were recorded with the CC1 CCD with 10241024 pixel elements, pixel size 12µm². The position angle of the slit was set to PA = 0^o and a fix slit width of 2" was selected. The hour angle of the quasars was less than 30^o to ensure that the light loss due to differential refraction is minimized. To obtain a continuous spectral coverage of 3780 - 7820 Å we used three settings with grating # 43 (3780 - 5200 Å) and # 44 (5100 - 6500 Å, 6400 - 7820 Å ). Hence, the overlap of the orders amounts to 100 Å. Argon neon spectra were taken after each object exposure for wavelength calibration. Spectra of the standard stars BD+33^o 2642 and BD+28^o 4211 were observed for flux calibration each night.

In addition we used FORS 1 attached at the VLT UT 1 Antu during the commissioning and scientific observations in September, October, and December 1998 to observe four quasars with redshifts larger than z = 3.0 (Dietrich et al. 1999). Since the objective of these observations was to check and derive instrumental functions and parameters, very different observational conditions and relatively short exposure times were used. Two spectral wavelength ranges with different spectral resolutions were observed. With grism 150 I a wavelength range of 3300 - 11800 Å was covered and grism 600 B provided a spectral coverage of 3600 - 6000 Å. The position angle of the slit was selected with respect to minimize the influence of the differential refraction. The slit width was set to (Q0046-282, Q0103-294) and (Q0044-273, Q0103-260). To obtain the highest possible spectral resolution Q0103-260 were observed also with grism 600B and a slit width of . For wavelength calibration helium-mercury-cadmium spectra were taken. For flux calibrations the standard star LTT 9491 was observed with the same grisms.

The 2 D longslit spectra were reduced and analysed using standard MIDAS software. First, the spectra were bias subtracted and flatfield corrected. Cosmic-ray events on the CCD images were removed manually by comparing the individual spectra taken for each object. The night sky component of the 2 D-spectra was subtracted by fitting Legendre polynomials of third order perpendicular to the dispersion. These polynomials were fitted on each spatial row of the spectra using areas on both sides of the object spectrum which were not contaminated by the quasar or other objects.

The 1 D-spectra were derived using the Horne algorithm for optimal extraction (Horne 1986). The limits of the spatial extraction windows were chosen to 4 as derived from the seeing during the observation. was derived from a Gaussian fit which was applied to the spatial profile of the spectrum. The wavelength calibration frames were used to rebin the extracted spectra obtained at Calar Alto to linear wavelength steps of 1.76 Å and 2.64 Å for the blue and the red channel, respectively. To check the absolute error of the wavelength calibration night sky spectra were extracted from the two-dimensional object frames and wavelength calibrated in the same way as the object spectra. The night sky lines (HgI 4047 Å, HgI 4358 Å, [OI] 6300 Å, [OI] 6364 Å) give an absolute uncertainty of Å in the blue channel and Å in the red channel. The spectral resolution was determined by measuring the FWHM of these night sky lines, too. The spectral resolution of the spectra taken at Calar Alto Observatory amounts to = 4.0 Å and = 5.2 Å in the blue and the red channel, respectively (300 km s^-1). The achieved spectral resolution of the spectra obtained at McDonald Observatory was determined for the entire observed spectral range to be 5.4 Å. The quasars observed with FORS 1 were recorded with two different spectral resolutions which were determinated from the FWHM of several strong night sky emission lines. With grism 150 I a spectral resolution of 25 Å is achieved for a wavelength range of 3500 - 10000 Å and grism 600 B provide a spectral resolution of 5.4 Å for a slit width of 1". The spectrum of Q 0103-260 taken with grism 600 B and a slit width of yields a spectral resolution of 2.9 Å. The uncertainty of the wavelength calibration amounts to Å for grism 150 I and Å for grism 600 B, respectively.

We corrected each quasar spectrum for the atmospheric B-band at 6850 to 6900 Å, the A-band at 7590 to 7690 Å absorption which both are caused by O₂, and the atmospheric a-band 7170 to 7400 Å which is due to water vapor. The absorption features were linearly interpolated in the spectrum of the standard stars which were observed under comparable conditions. The ratio of the interpolated and the original spectrum gives the scaling factors for the correction of the atmospheric absorptions. However, for the four quasars recorded with FORS 1 no removal of the atmospheric absorption features was attempted. Therefore, the strong atmospheric A and B bands (at 7600 Å and 6900 Å) and are clearly visible in Fig. 1. On the other hand, as shown by Fig. 1, although affecting in some cases the outer wings of the broad emission lines, the atmospheric features had little effect on the line flux measurements.

Fig. 1. The spectra of the observed quasars. The flux is given in units of 10-15 erg s-1 cm-2 Å -1. For each quasar the corresponding power law fit of a nonstellar continuum is displayed.

For the correction of the atmospheric extinction we used the standard extinction curve of La Silla (Schwarz & Melnick 1993). The corresponding airmass was computed for the middle of each exposure. The interstellar extinction was corrected using the values of Burstein & Heiles (1982) and the extinction curve of Savage & Mathis (1979). Since the galactic latitude of the quasars observed with FORS 1 is less than b = -86^o no correction for interstellar extinction were applied to these spectra.

The spectra observed at Calar Alto Observatory and McDonald Observatory were corrected for losses due to the differential refraction (Filippenko 1982). This was done using the MIDAS application refraction/long which was designed for point sources. For one quasar, Q2231-0015, we obtained spectra at Calar Alto Observatory and McDonald Observatory as well. The spectra were identical within less than 6 % with respect to the overall continuum shape and emission-line strength.

The final step of the data reduction was the unweighted summation of the spectra for each quasar. Hence, the spectra of the three settings (McDonald Observatory) and two settings (Calar Alto Observatory) were rebinned to a uniform wavelength scale of 1 Å.

© European Southern Observatory (ESO) 2000

Online publication: January 31, 2000
helpdesk.link@springer.de