Spectroscopic properties of granulation in K-type dwarf stars

The presence of surface convection in K-type dwarfs is revealed in very high quality spectra (R ≃ 180, 000, S/N &300) of nine bright stars. Relative radial velocities between pairs of spectra of the same object are determined with a mean accuracy of 12 m s−1 , which is necessary for achieving high S/N without distorting the spectral line shapes by coadding individual frames. The observed asymmetries and wavelength shifts of the Fe i absorption line profiles are mainly due to granulation. The bisectors of the strongest Fe i lines have a span of about 100 m s−1 and the central wavelengths of the weakest Fe i lines are shifted by up to −200 m s−1 . The blueshifts decrease for stronger Fe i lines, but they become independent of line strength for equivalent widths larger than about 100 m˚A. The detection of this “plateau” in the velocity shifts of the strongest Fe i lines is necessary to remove the uncertainty introduced by granulation, which is of the order of hundreds of meters per second, in the determination of absolute radial velocities. Line profiles computed using a 3D model atmosphere accurately reproduce the observations. Statistical tests show that the theoretical line asymmetries and wavelength shifts correspond to the observed ones at a 90–95% confidence level, thus validating the 3D model for spectroscopic studies of abundances and fundamental parameters of K-dwarfs. We find that 3D effects reduce the difference in the iron abundance determined separately from Fe ii and Fe i lines, which is about 0.15 dex for 1D models, by two thirds, thus alleviating significantly the iron ionization imbalance problem in K-dwarfs. However, the 3D iron abundances from Fe i lines show a small dependence with excitation potential, similar to the 1D case, possibly due to non-LTE effects that have not been taken into account. We also find that the 3D correction to the effective temperatures of solar metallicity K-dwarfs derived with the infrared flux method is about +30 K. Finally, we show that the 3D spectrum synthesis of molecular bands greatly improves the agreement with the observational data compared to the 1D analysis, which overestimates the abundances derived from molecular features by a factor of 2.