1 Introduction

A more complete knowledge of atomic transition parameters has always been crucial for progress in astronomical spectroscopy and for the modelling of stellar atmospheres. Although stellar structure models have benefited from the development of the OP and OPAL projects (see e.g. Hummer & Mihalas 1988; Iglesias & Rogers 1996), more reliable line opacities and thus more accurate and more complete atomic data are still a key requirement for the study of the surface layers of stars (cf. Baschek 1995). Hence, stellar spectroscopists had to create the line lists for their work from inhomogeneous and constantly changing data sets making the quality ranking of these data an important issue.

The "Vienna Atomic Line Data Base'' (VALD, Piskunov et al. 1995, Paperi) has originally been created in order to merge such data sets for the computation of opacity distribution functions (cf. Piskunov & Kupka 1999). The computation of line blanketing effects involves the handling of bulk data with emphasis on completeness rather than on accuracy. On the other hand, stellar spectroscopy usually requires the compilation of a relatively small amount of highly accurate atomic data. Driven by the research interests of our group, VALD was designed to satisfy both requirements. Public access to the data base through an automatic e-mail interface (VALD-EMS, see Paperi) has attracted a continuously growing community of users working in astrophysics, plasma physics, and laboratory spectroscopy. Today VALD-EMS is the main user interface to our data base: at the time of writing of this paper more than 240 clients from all around the world are using it. The feedback from users and our own experience has provided an important input for the improvement of the software and line data.

VALD-2 represents the extended and improved version of VALD. We added new line lists for 34 atoms (Resp. 51 ions, see Fig.1) with the best available measured or computed line parameters, in many cases verified by using astrophysical abundance analyses. The new referencing system helps to achieve adequate acknowledgements for researchers who are providing atomic data and to stimulate their collaboration. Major changes to the VALD software include new extraction options, while the extraction became more efficient. Compressed format of data storage allows to more than double the amount of information stored on the same disk. The data base is portable and two mirror sites are now processing VALD requests.

In this paper, we describe improvements to the software, the user interface and extraction capabilities, as well as extensions and corrections to the data archive. These changes are significant enough to call the current implementation VALD-2. The version described in Paperi will be referred to as VALD-1.

In Sect.2 we provide a description of the structure of VALD-2 as well as a description of the information now stored stored for each transition. Section3 discusses the VALD-2 software that provides efficient handling of over 50 million lines, offers flexible extraction modes, ensures its portability and supports mirror sites. We also describe the new and more complete referencing system, the new WWW interface and certain provisions made for a future release (VALD-3). Section 4 is dedicated to the new atomic data in VALD with the emphasis on quality checks and ranking of the new lists. The concluding Sect.5 summarizes the changes introduced with VALD-2 and describes our plans for the future VALD-3 which will also contain molecular lines.

  

Figure 1: Within the VALD project data for about 23000 lines from 58 different ions have been collected, compared, and tested in order to extend the Kurucz (1993bc) compilations. We have marked the spectrum number (from neutral, i, to doubly ionized, III) for each element for which improvements have been achieved either in VALD-1 or VALD-2 (this paper) or both. Work to include data on C, N, O was finished while this paper was being refereed, while work on S is still in progress. Note that all line lists for rare earth elements also have been supplemented by Landé factors as far as possible (not specifically marked here)