4 Results

The atmospheric parameters and abundances derived for IS Vir and V851 Cen, with the three methods outlined in Sect. 3, are presented in Tables 6 and 7. The error bars were derived in several steps. First, the 1 $\sigma$ errors on each of the "individual'' diagnostics (e.g. average slope of Fe  I abundances versus excitation potentials relation) were calculated. They were then propagated to obtain the "individual'' errors on the atmospheric parameters and abundances. Finally, those "individual'' errors were quadratically summed.

Our concern with the first method was that Fe  I "low excitation potential'' transitions may be affected by non-LTE effects. This does not seem the case neither for IS Vir nor for V851 Cen since, in both cases, methods 1 and 3 (the first using Fe  I "low excitation potential'' iron transitions and the other not) give results in good agreement. Cooler or more evolved stars might, of course, exhibit stronger departure from LTE and would require some careful checking.

The B-V index leads to effective temperatures in very good agreement with those derived with the first method and 90 and 110 K lower, for IS Vir and V851 Cen respectively, than those derived with the third method. Given the error bars of the second and third methods (respectively 150 and 185 K for IS Vir and 150 and 160 K for V851 Cen), their results are compatible within less than one sigma error. As already presented in Sect. 3.6, the temperature derived for V851 Cen from the $V-I_{\rm c}$ index is about 230 K lower than the one derived from B-V. This is respectively 260 and 340 K lower than the temperatures obtained via the first and third methods. Neither IS Vir, nor V851 Cen display a significant departure from the B-V/temperature relation of non-active stars, whereas the V-I index deviate from the V-I/temperature calibration. This behaviour may be due to the presence of activity.

The third method exhibits fairly large error bars. This is in part due to its dependency on the knowledge of the calcium abundance, which relies on four lines (2 of them being on the "plateau'' of the curve of growth) and in part due to the propagation, on the final results, of the error bars on the surface gravity (derived by calcium wings fitting, see Sect. 3.7). Nonetheless it has the merit of relying on criteria which present few risks of being affected by non-LTE or activity effects, as well as being self-consistent. As discussed above, the results of methods 1 and 3 are consistent.