Ingenious techniques have been used to improve the accuracy of wavelength calibrations, such as placing gas cells at the entrance of the spectrograph (e.g., Deming & Plymate 1994), but it is rare to find such systems available and convenient for regular observations.
On occasion the available lamps are not very rich in lines in the spectral range of observation. In some circumstances, an external check of the final precision in the translation into wavelengths would be desirable. One method for tackling problems such as these is to use solar spectra as templates. Changes in the wavelengths of the lines in the integrated sunlight spectrum around the solar cycle have been proved to be very small, bellow some 15 m s-1 (Jiménez et al. 1980; Wallace et al. 1988; McMillan et al. 1993; Deming & Plymate 1994). At 5000 Å, this translates into 0.3 mÅ, so the solar spectrum does offer a very stable source. In most practical cases, the accuracy will be imposed by the spectral resolution achieved. During night-time observations, the solar flux spectrum is observable after reflection from the Moon.
Measurements of solar wavelengths in the integrated solar optical spectrum were published in 1929 by Burns and collaborators (Burns 1929; Burns & Kiess 1929; Burns & Meggers 1929), using photographic detectors and a grating spectrograph. The relatively recent solar flux FTS atlases offer a much higher quality spectrum of the Sun seen as a star.
As the solar spectrum is so intense, on some solar telescopes no calibration lamps are deemed necessary, and the wavelength scale is set using the solar spectrum itself. Reasonable precision can be reached using the spectrum at the centre of the disc to compare with previously measured disc centre wavelengths, thus avoiding differential shifts due to the limb effect. In this case, small scale motions have to be averaged out, integrating in time and/or space, in order to minimize errors.
The Kitt Peak Table of Photographic Solar Spectrum Wavelengths (Pierce & Breckinridge 1973) has been extensively used by solar observers to set up the wavelength scale on their spectra. These observations, made on photographic plates, have been superseded in quality by the more recent FTS observations at the centre of the disc.
To improve on the various sets of photographically based measurements (which date back to 1930 in the case of the solar flux spectrum), provide them in a homogeneous machine-readable format, use them to test spectral calibrations of very high resolution stellar spectra (e.g., Allende Prieto et al. 1995), and improve the accuracy of our own solar observations, we have determined the position of the central wavelengths of 4947 atomic lines in the optical solar spectrum. The employed source solar atlases, prepared from FTS data, and the fashion in which we performed the measurements is described in the succeeding sections.
Copyright The European Southern Observatory (ESO)