2. Observations and data reduction
This report is based on eight spectroscopic observing runs at ESO's La Silla observatory and also incorporates three sets of previous observations (Baade, 1991). The resolving powers range from 2 500 with the Boller&Chivens spectrograph at the ESO 1.52-m telescope (Sect. 2.2) over 20 000 with the Heidelberg FLASH and HEROS spectrographs and fibre link to the ESO 0.5 or 1.52 meter telescopes (Sect. 2.1) and finally to 60 000 with the Coudé Echelle Spectrometer (CES) and the ESO 1.4-m Coudé Auxiliary Telescope (CAT) (Sect. 2.3). A summarizing journal of the main observations and technical parameters characterizing the spectra is provided in Table 1. The particulars of the raw data and the reduction procedures applied to them are described in the following subsections. For the Boller&Chivens data the standard reduction and calibration techniques were used as implemented in ESO-MIDAS (1995). For the FLASH and HEROS data, a customized version of the ESO-MIDAS echelle context was developed (Stahl et al., 1995b). Note that throughout the entire paper Modified Julian Date (MJD JD-2 400 000.5) is used.
2.1. FLASH /HEROS observations
Our fiber-linked echelle spectrograph HEROS (H eidelberg E xtended R ange O ptical S pectrograph, Kaufer 1998) is an upgrade of the FLASH (F iber L inked A stro echelle S pectrograph of H eidelberg) described by Mandel (1994) and Stahl et al. (1995b). In HEROS, a dichroic beam splitter is used to divide the light beam into two channels after the echelle grating. Each channel has its own cross-disperser, camera, and detector. The blue one covers the range from 3450 Å to 5560 Å, and the red one from 5820 Å to 8620 Å. The spectral resolving power in both channels is . The link to the telescope was provided by a 10 m long glass fiber.
The stability of the red channel was assessed by Kaufer et al. (1997). They found the telluric water vapour lines in a time series of Ori of 84 spectra over 104 nights to be stable to within 0.35 rms, with a peak-to-peak range of 1.7 . The standard deviation corresponds to 1/10 of a 22 µ pixel and is caused by the shift of the lines of the ThAr-lamp between subsequent calibration exposures, which were typically taken every two hours. Schmutz et al. (1997) tested the blue channel using the interstellar Ca ii 3934 line in Cap. Since however Cap has a variable photospheric contribution to this line and the is not as high as in Ori, their rms value of 0.5 for the blue channel is to be regarded as an upper limit.
After the extraction and merger of the echelle orders, a wavelength dependent ripple became apparent at the order boundaries. It is noticeable at both ends of the blue channel, namely from 3450 Å to 3600 Å and from 5100 Å to 5560 Å where towards either end of the detector the peak-to-valley amplitude reaches 5 %. At all other wavelengths, including the whole red channel, the ripples are of the order of 1 % or less.
FLASH, the predecessor of HEROS, had the same spectral resolution, but only one channel, covering the wavelength range from 4050 Å to 6750 Å.
The signal-to-noise ratio ( ; for some typical values see Table 1) was measured for both the FLASH and the HEROS data as the inverse of the rms (in units of the continuum) in the wavelength range 4760-4800 Å. In May 1996 a more sensitive CCD was implemented in the blue channel of HEROS. This is reflected by the increase of the typical at comparable exposure times. For the later runs of 1997 a new fiber link was implemented. The improved throughput permitted from then on the exposure times to be shortened by one third without losing in .
The standard reduction, extraction, and visualization of the data is described in detail by Stahl et al. (1995b). To represent the large amount of data we display the spectra as dynamical spectra (cf. Fig. 2), i.e. time (or phase)-vs.-velocity grey-scale pictures around the rest wavelength of the spectral line.
2.2. Boller&Chivens observations
During 8 consecutive nights in 1995 June, low-resolution spectra were obtained with the Boller&Chivens spectrograph attached to the ESO 1.52-m telescope. On a 2048-pixel CCD, the spectra covered approximately the range from 3400 to 5100 Å. The resolving power as deduced from the helium and argon lines of the comparison spectra, which were taken 3 - 5 times per night, is 2 500. The typical rms of a third-order polynomial fitted to 35-40 HeAr comparison lines is 3 .
In order to approximately equalize the detected signal over the observed range in wavelength, a BG-24 filter was inserted into the beam. The primary purpose of the observations was the monitoring of rapid Balmer jump variations the results of which will be reported elsewhere (Stefl, Baade, Cuypers in preparation). Since at this short wavelength atmospheric dispersion would normally strongly compromise the spectrophotometric fidelity of the data, the telescope was drastically defocussed in order to scramble the atmospheric spectra on the entrance slit. The amount of defocussing was chosen such as to stay in the linear regime of the detector and to keep the exposure times in the range between 90 and 120 seconds where scintillation effects are of no importance. Only at times of poor sky transparency was it necessary to increase the exposure time (up to 6 minutes); this concerned 19% of all observations.
Since the observations were more nearly of the telescope pupil than of the stellar image, the resulting `long-slit' spectra featured two maxima along the slit axis. Both wavelength and flux calibration were performed preserving this 2-D character of the spectra. This permitted the extraction of two spectra (hereafter called `lower' and `upper') per exposure. Because the exposure levels were high and particle events were few in number, no signal-to-noise optimized extraction procedure was applied. The typical per extracted spectral pixel is between 250 and 350 and very nearly identical for both upper and lower spectra.
The total database accumulated in this way consists of 348 `upper' and `lower' spectra each. A statistical overview is given in Table 2.
Table 2. Modified Julian dates (MJD JD-2 400 000.5) and numbers of observations obtained in the observing nights with the Boller&Chivens spectrograph on the ESO 1.52-m telescope and the CAT/CES. One B&C observation yielded one `upper' and `lower' spectrum each (cf. Sect. 2.2)
2.3. CAT/CES observations
In 1995, the wavelength region from 4537-4563 Å was chosen to monitor the Si iii line at 4553 Å. At a resolving power of 60 000 the point spread function was sampled by 3.6 pixels. A second order polynomial fitted to the positions of 15-20 thorium lines had a standard deviation of 0.02-0.025 Å. An optimal extraction technique was applied to remove particle hits from the data.
However, this wavelength range turned out to be heavily contaminated by double peaked Fe ii emission lines which reached up to three percent of the continuum level. The average of the FLASH spectra from 1992, when the emission was weak even in the Balmer lines, was used to obtain the uncontaminated absorption line profiles and to improve the continuum rectification. By subtracting these profiles from the CAT/CES spectra it was possible to isolate the emission line spectrum. The emission lines proved to have been constant over the four days of observations to within the noise. Accordingly, the remaining variability is intrinsic to the photospheric Si iii 4553 line. This was confirmed by comparison with the variability of the residuals of the Si iii 4568 line, part of which was covered at the very edge of the wavelength range.
We also have at our disposal profiles of several lines that have been obtained from 1983 to 1990 (see also Baade 1991) with the CAT/CES equipped with a Reticon detector prior to 1990 and with a CCD from 1990 on. Besides the data already listed in Table 1, these are profiles of He i 6678 (March 1984, 8 spectra over 3 nights; Feb. 1987, 3/3; and March 1990, 8/3), He i 4471 (June 1983, 20/14; March 1984, 8/8; and July 1985, 5/2), Si iii 4553 (March 1984, 3/3), and H (June 1985, 4/15; Jan. 1986, 4/4; April 1986, 13/10; Feb. 1987, 3/3; and April 1987, 9/10). The usage of a Reticon resulted in a which in almost all cases exceeded 500, often quite substantially, but at exposure times of 15-20 minutes. The resolving power is 80 000-100 000.
© European Southern Observatory (ESO) 1998
Online publication: April 15, 1998