This work enlarges the measurements of Allende Prieto & García López (1998) to many other species, with the aim of optimizing its use for wavelength calibration. The line list from Thèvenin (1989, 1990), including 6606 lines classified by Moore et al. (1966) as singly blended or unblended, has been chosen as a guide to select the features to be measured.
A fourth-order polynomial was fitted to the 50 mÅ wavelength interval around the line minimum to find the line centre as precisely as possible. In the electronic version of the atlases, this corresponds to 11 points for the solar flux spectrum and 25 points for the centre-of-the-disc spectrum. Errors in the wavelength determination were estimated by translating the standard deviation of the fit into the corresponding units of the wavelength axis, neglecting the extremely low photometric noise.
Table 1, available only in electronic form from the CDS, lists the central wavelengths in the flux and disc-centre solar spectra, the errors of the measurements, and the element identification, excitation potential and "solar" oscillator strengths (; from Thèvenin 1989, 1990) for 4947 lines between 3944 and 7960 Å included in Thèvenin's list. More than one line corresponding to the same feature is listed when the identification is not clear from the wavelengths published by Thèvenin.
A comparison between a limited sample of wavelengths (42 lines) listed by Pierce & Breckinridge (1973) and those we have measured in the atlas of the disc-centre shows that the absolute scales agree (difference: 79 6 10-5 Å) and that the relative differences are fully accounted for by a straightforward consideration of the expected errors in the photographic atlas ( 2.5 mÅ).
Errors of the wavelengths measured in the flux spectrum can be as large as 100 m s-1, as quoted by Kurucz et al. (1984). For the wavelengths in the spectrum of the disc centre, systematic errors are given from the comparison with the photographic atlas of Pierce & Breckinridge (1973), which has been used as reference. They estimate the absolute accuracy from comparison with the interferometrically determined wavelengths of Adam (1952, 1958) and Nichols & Clube (1958), arriving at 0.3 mÅ. Our systematic errors are then of the order of 0.9 mÅ, while FTS intrinsic errors are about 10 m s-1 (Neckel & Labs 1990), yielding a final precision of about 50 m s-1 at 5000 Å. These numbers allow us to claim that errors quoted in Table 1 (typically m s-1) are conservative.
The present database updates others which were previously available based on photographic spectra, and the catalogue is presented in an homogeneous machine-readable format available via internet from the CDS. The wavelengths measured conform a reference frame which can be used to calibrate or field test calibrations of astronomical spectra by comparing with the solar spectrum, and to set up an accurate wavelength scale for solar spectra, among other applications.
We thank H. Neckel for his help in dealing with his solar atlas. We wish to extend our gratitude to F. Thévenin, who has kindly lent a digital copy of his line list, M. Collados for fruitful discussions on the wavelength calibration of solar spectra, J.E. Beckman for careful reading of the draft and H.H.R. Kroll for performing the installation and maintenance of the KIS computer libraries at the IAC. NSO/Kitt Peak FTS data used here were produced by NSF/NOAO.
This work was partially supported by the Spanish DGES under projects PB92-0434-C02-01 and PB95-1132-C02-01.
Copyright The European Southern Observatory (ESO)