From the most extensive up-to-date measurements of solar wavelengths by Pierce & Breckenridge (1973), DLN extracted a list of 311 unblended FeI lines. After subtracting the gravitational redshift, they studied the trend of the shifts with the excitation potential of the lines and placed the bisectors measured in the solar atlas of Liège Atlas on an absolute scale. Their study demonstrated the power of line asymmetries and shifts in providing a deeper understanding of solar convection.
A version of the FTS disc-centre spectrum, which has been interpolated to a constant wavelength step, can be obtained as part of the KIS IDL library. We have measured on this atlas the location of the minima of 1446 FeI lines selected from the list of Thévenin (1989, 1990), thereby extending the existing data provided by DLN and improving its accuracy significantly. Thèvenin's list includes 2536 lines of neutral iron. Although many of the lines are blended, only those showing clear evidence for the blend to disturb the line centre were rejected. A fourth-order polynomial was fitted to the 50 mÅ wavelength interval around the line bottom to find the line centre as precisely as possible. Measured central wavelengths, together with their corresponding air values at rest, excitation potentials, transition probabilities and equivalent widths, are listed in Table 1. Since the measurement process was automatic, Table 1 may contain certain errors for the observed wavelengths.
The centres of those lines formed higher in the photosphere
show smaller blueshifts. It can
be seen in Fig. 2 (click here)a that even those lines located in a "plateau''
with equivalent widths (
) larger than 
100 mÅ seem to be somewhat
blueshifted, and there are no lines where the effect is negligible. The
mean
value for the shift of these lines is 539 m
s-1 (subtracting the gravitational redshift of
636 m s-1, the
minimum convective blueshift will be 97 m s-1) with a standard
deviation of 86 m s-1. No other source in the literature,
including spectra and wavelength measurements, is accurate enough to
perform a more reliable comparison discarding minor systematic effects
of the wavelength calibration.
Figure 2: a and b). Line shifts as measured in the
FTS disc-centre spectrum a)
and the FTS flux spectrum b) are plotted as a function of the
equivalent
width at the centre of the disc (Moore et al. 1966).
Known blends (marked with an asterisk in Table 1) have been rejected
To the best of our knowledge, the only existing measurements of line shifts from solar flux spectra are those of Burns, Meggers, and Kiess, published in 1929 (Burns 1929; Burns & Kiess 1929; Burns & Meggers 1929). Nonetheless, systematic analyses of shifts and asymmetries of spectral lines in the Sun seen as a star are of special importance as a reference standard for comparison with other stars.
The FTS flux spectrum is available from the
NOAO ftp
site. We have measured the wavelengths of the minima of the same 1446
FeI lines considered in the FTS disc-centre spectrum. Similarly to
the intensity case, a fourth-order polynomial was fitted to the 55 mÅ wavelength interval around the line centre. The results are included in the second
column of Table 1. This list provides the largest available set of
accurate solar wavelengths useful as a reference for comparison with
other solar-type stars.
The smoothing of the convective blueshift towards the solar limb, the
so-called limb effect, results in a smoothing of the convective shifts
in the integrated sunlight compared with the intensity spectrum. In
this atlas, the "plateau'' with equivalent widths larger that
200 mÅ is formed by lines distributed around the gravitational
redshift, as can be seen in Fig. 2 (click here)b. It makes sense that the
smoothing of the blue-shifts at the disc centre results in a plateau
closer to a null velocity shift. The scatter (standard deviation) in
these line shifts at the plateau is
58 m s-1, around a mean value
of 612 m s-1 (the subtraction of the gravitational redshift comes
with a mean of
24 m s-1 for the smaller convective blueshifts).
Wallace, Huang & Livingston (1988) studied the variability of the convective line shifts in the solar flux spectrum during the solar cycle measuring the relative shifts between a weak line, C I 
5380.3 Å ( = 26 mÅ ) and the stronger features Fe I 5379.6 Å ( = 67 mÅ ) and Ti II 5381.0 Å ( = 70 mÅ ). They found an upper limit of 5 m s-1 for the relative line shift and concluded with the possibility of detecting Jupiter, which would produce a 20 m s-1 amplitude
in the solar radial velocity, from an extra-solar system observer.
However, Fig. 2 (click here)b suggests that the 70 mÅ pair of lines they employed as reference wavelengths less affected by convection were not the ideal choice. Comparison between a much stronger feature
and the CI line will probably give a safer answer on this subject,
where different measurement techniques observe (Deming & Plymate 1994)
and deny (McMillan et al. 1993) the variations of the convective shifts.