Astron. Astrophys. 336, 445-454 (1998)

1. Introduction

The understanding of the formation, evolution and present properties of the large-scale structure of the Universe is a key problem in modern cosmology (see Peebles 1980, 1993). One of the most important results of the first redshift surveys was the previously unexpected existence of coherent structures and voids at very large scales. Explaining these structures was a challenge for popular models of galaxy formation, but at the same time represented a problem for the interpretation of results obtained on small volumes which could not be representative of the Universe. Therefore, the need of a "fair sample" of the Universe, in order to understand the process of galaxy formation and evolution, led to an increasing number of deeper redshift surveys. Redshift surveys are now an "industry" with its own standards. Reduction of an ever growing number of data is based on software packages specially developed to this aim. The redshift , or, less rigorously, the "recession velocity" , is commonly determined using the wavelength shift of either absorption or emission lines appearing in the optical spectrum of a galaxy. Following the paper by Tonry & Davis (1979), most redshifts based on absorption lines are now obtained by cross-correlating galaxy spectra with one or more (or an average of) "template" spectra, while redshifts based on emission lines are measured by fitting the individual emission lines. Moreover, emission and absorption lines are produced in different environments. In normal galaxies, the former (such as the [OII] line) are generated in HII regions associated with recent star-formation, while the latter (such as the calcium Ca II K and H) are produced in stellar atmospheres and are related to the bulk of the star population. As a consequence, emission and absorption redshifts are not required to be exactly the same.

Despite the growing number of galaxy redshifts in the literature, most catalogues quote only the "best" estimate of the velocity of a galaxy, and take for granted the implicit and widespread assumption that, while for a given galaxy the absorption velocity and the emission velocity may differ, the average difference should be consistent with zero.

In the analysis of the ESO Slice Project (ESP; Vettolani et al. 1997; Zucca et al. 1997; Vettolani et al. 1998), we have devoted a particular effort to check the quality of our data, and in particular the precision of our absorption and emission redshift measurements which, as we have soon realized, present a puzzling discrepancy. Looking at the past and recent literature, we have also realized that this problem was not new, but was never discussed in a satisfactory way. We have therefore decided to study the effect in more detail, and we describe in this paper the results of our analysis and the possible explanations.

In Sect. 2 we discuss the evidence of discrepancies in   found in the past and in other surveys, and in Sect. 3 we present the discrepancy detected in the ESP data. In Sect. 4 we describe the tests we have performed on the ESP data, exploring instrumental and other effects which could in principle affect our results; in Sect. 5 we analyse in detail the biases on velocity measurements due to the choice of the template spectra; in Sect. 6 we discuss if such a discrepancy can be partly due to a real, physical effect; our conclusions are in Sect. 7.

© European Southern Observatory (ESO) 1998

Online publication: July 20, 1998
helpdesk.link@springer.de