2. The data
Table 1 above lists the main parameters and characteristics of the spectroscopic sample of the ESS (see de Lapparent et al. 1997). The redshifts are measured by cross-correlation (using the method developed by Tonry & Davis 1979) with galaxy templates which have been tested for the reliability of the redshift scale which they provide (Bellanger et al. 1995). For the of galaxies with emission lines, an "emission redshift" is also measured by fitting the detected lines (mostly [OII], H , and [OIII] at 4958 Å and 5007 Å). When the absorption and emission redshift agree, a weighted average is derived. The mean errors in the redshifts are given in Table 1. Detailed information on the acquisition and redshift measurements are described in Bellanger et al. (1995). The present spectral analysis is based on the subsample of 347 spectra having R(Cousins) 20.5, S/N 5, a reliable redshift measurement, and a spectro-photometric quality (see below). The remaining data to bring the redshift survey to completion are in the course of reduction. The only bias affecting the sub-sample used here is the tendency to observe the brightest galaxies in the R filter (see Fig. 14). There is no intended bias related to morphological type in the observing procedure. The full ESS spectroscopic sample is defined by only one criterion: ( is an estimate of aperture magnitudes in the R Cousins filter, using the Kron estimator (Arnouts et al. 1997).
Table 1. Characteristics of the ESO-Sculptor spectroscopic survey.
Because spectral classification techniques are sensitive to the continuum shape of the spectra, the flux-calibration is a crucial step which we now describe. This stage amounts to the calculation of the instrumental response curve, which depends on the telescope, instrument, and CCD combination, and is modulated by the variations in the transparency conditions at the moment of observation. We denote "calibrating curve" the product of these two independent functions. The calibrating curves for some of the different instrumental set-ups used for the observations of the ESS are shown in Fig. 1.
The instrumental response for each instrument is calculated from the observation of spectro-photometric standard stars and is the average ratio between the observed spectrum of the star and the reference spectrum, in good spectro-photometric conditions. For the ESS sample we used the standard stars LTT 377, LTT 7987 and Feige 21 (see Hamuy et al. 1992). Several standards (2-3) were observed each night or one standard was observed several times per night (2 to 3 times). The resulting r.m.s. variations in the calibrating curves during a night reported as photometric by the observer, and from one such night to another, are %. We therefore select from the available ESS spectroscopic sample all spectra obtained during these "stable" nights. We then correct each spectrum by the mean calibrating curve derived for the corresponding observing run. The resulting flux calibrations are only relative. An absolute calibration could be obtained using the photometric magnitudes (cf. Arnouts et al. 1997), but this is not necessary for the present analysis. The final subsample, which contains 347 spectra, represents 52% of the total of 669 galaxies with 20.5 (see Sect. 6.5 for completeness correction). Before the spectral analysis, the atmospheric absorption bands of the spectra, near 6900 Å and 7600 Å , are eliminated by linear interpolation from the surrounding continuum.
To assess quantitatively the spectro-photometric quality of the selected sample of 347 spectro-photometric calibrated spectra, two tests are performed: (1) the comparison of the spectra of the same galaxy, observed twice or more, and (2), the comparison of the photometric colors with the synthetic spectro-photometric colors. First, we found that 40 galaxies from the available spectral sample have 2 measured spectra. The r.m.s. variations in the ratios of the spectra for each pair are 7-10% when both are obtained in spectro-photometric conditions, and 10% when at least one spectrum of the pair is taken during a non-spectro-photometric night. This confirms that the spectro-photometric stability indicated by multiple observations of standard stars during each night is a reliable indicator of the spectro-photometric quality of the resulting calibrated spectra. Second, we calculate synthetic colors from the calibrated spectra, and compare the results with the standard colors obtained from the CCD photometric catalogue (see Arnouts et al. 1997). The photometric magnitude system is B(Johnson), V(Johnson), and R(Cousins). We compare colors rather than magnitudes in order to cancel out the unknown absolute flux calibration. We then fit a polynomial of degree 1 to the spectro-photometric versus photometric colors, for B-V and B-R. The slope is 0.952 0.07 and 0.905 0.08 for B-V and B-R, respectively (see Fig. 2). For a perfect correspondence, the slope should be 1.0. The dispersion around the fit are - ] = 0.17, and - = 0.19. These values are consistent with the dispersion resulting from the intrinsic photometric and spectrophotometric errors, which is , where 0.04 and 0.10 are the intrinsic errors of the photometric and spectro-photometric data, respectively. Therefore, there is a good agreement between the spectro-photometric and photometric B-V and B-R colors for spectra taken during photometric nights, and our estimate of 10% for the external uncertainty in our relative flux calibrations appears valid.
We examine one last effect which could bias our flux-calibrated spectra. The 1-D spectra are obtained from the 2-D spectra using the optimal extraction weight method (Robertson 1986). This method weights differently the wings and the central parts of the light distribution in such a way that the noisier parts of the spectrum (the outer regions of the galaxy) have a smaller weight than the high S/N part of the spectrum (the core of the object). Typically, the weight in the wings is 12 to 20% of the weight in the center. Due to well known color gradients in the surface photometry of individual galaxies, the extracted spectra can be affected differently for different wavelengths, in such a way that the extracted spectra are dominated by the stellar content of the center of the light distribution. However, comparison of the spectra obtained using the weighted and the un-weighted extraction for 27 objects, shows that the optimal extraction method does not change the shape of the spectra by more than 3% (for spectra with S/N 12). This is well inside the 10% spectro-photometric uncertainty in our flux-calibrated spectra.
© European Southern Observatory (ESO) 1998
Online publication: March 23, 1998