Quality of stellar radial velocities

Spectra for approximately 48,000 Galactic stars of all common spectral types are available with DR3. Radial velocities (RVs) are stored as redshifts (multiply z by the speed of light to get the RV in km/s) and were measured by cross-correlation to a set of stellar templates. Also tablulated are measures of individual common spectral absorption and emission lines, such as Ca II K (3933), the Balmer series of Hydrogen, and Na (5896), and the offsets of these individual lines from their vacuum rest wavelengths may also be used to estimate a radial velocity (lines are tabulated with a significance indicator, nSigma, and generally speaking, only lines above a threshold, such as nSigma > 7, should be considered reliable). DR3 includes measures of the centers and depths of the Ca triplet lines (8500, 8544, 8664), useful for determining radial velocities of some type M stars. Note that the DR1 radial velocity measurements of stars had systematic errors which have been corrected in DR2/DR3.

The current DR3 dispersion solutions for individual objects have been demonstrated to be good to better than 5 km/s, including correction for such effects as flexure of the telescope. This dispersion error has been determined by examining multiple observations of the same astrophysical object on a variety of plates under a variety of observing conditions and showing that RVs were reproducible to this accuracy. The identity of the actual template star (which may be examined) are tabulated below. The 'Template Number' corresponding to the highest cross correlation peak match for any spectrum may be found under the keyword 'BESTTEMP' in the PDU header of a SDSS spSpec-$mjd-$plate-$fiberid.fit file. An estimate of the difference between the RV as determined from cross correlation with a template and that determined by the median offset of all strong (nSigma > 7) individual absorption lines is tabulated in the column RVshift (in km/s). This figure may be used as an estimate of the possible remaining systematics in Radial velocities for stars of a given Spectra Type. In particular, estimates for White Dwarf and spectral type M and L star Radial velocities are not accurately quoted for DR3. Interested scientists are encouraged to compute their own RVs off of the extracted spectra, keeping in mind that the wavelengths measured of individual lines measured are on a vacuum system. The absolute calibration of the wavelength system is based on to the positions of known night sky lines obtained with the spectra. This calibration has been checked by obtaining a special plate of known radial velocity standards in the cluster M67, and the velocities are in agreement with an absolute error of less than 4 km/s. The relative rms velocity error is magnitude (signal-to-noise) dependent, but repeated observations of the same stars have shown the errors to be about 5 km/s (one sigma rms) for stars brighter than about g = 18.2 mag, degrading to errors of 25 km/s for stars with g = 20.3 mag.