VLTI/AMBER observations of cold giant stars: atmospheric structures and fundamental parameters

Aims. The main goal of this research is to determine the angular size and the atmospheric structures of cool giant stars (epsilon Oct, beta Peg, NU Pav, psi Peg, and gamma Hya) and to compare them with hydrostatic stellar model atmospheres, to estimate the fundamental parameters, and to obtain a better understanding of the circumstellar environment. Methods. We conducted spectro-interferometric observations of epsilon Oct, beta Peg, NU Pav, and psi Peg in the near-infrared K band (2.13-2.47 mu m), and gamma Hya (1.9-2.47 mu m) with the VLTI/AMBER instrument at medium spectral resolution (similar to 1500). To obtain the fundamental parameters, we compared our data with hydrostatic atmosphere models (PHOENIX). Results. We estimated the Rosseland angular diameters of epsilon Oct, beta Peg, NU Pav, psi Peg, and gamma Hya to be 11.66 +/- 1.50 mas, 16.87 +/- 1.00 mas, 13.03 +/- 1.75 mas, 6.31 +/- 0.35 mas, and 3.78 +/- 0.65 mas, respectively. Together with distances and bolometric fluxes (obtained from the literature), we estimated radii, effective temperatures, and luminosities of our targets. In the beta Peg visibility, we observed a molecular layer of CO with a size similar to that modeled with PHOENIX. However, there is an additional slope in absorption starting around 2.3 m. This slope is possibly due to a shell of H2O that is not modeled with PHOENIX (the size of the layer increases to about 5% with respect to the near-continuum level). The visibility of psi Peg shows a low increase in the CO bands, compatible with the modeling of the PHOENIX model. The visibility data of epsilon Oct, NU Pav, and gamma Hya show no increase in molecular bands. Conclusions. The spectra and visibilities predicted by the PHOENIX atmospheres agree with the spectra and the visibilities observed in our stars (except for beta Peg). This indicates that the opacity of the molecular bands is adequately included in the model, and the atmospheres of our targets have an extension similar to the modeled atmospheres. The atmosphere of beta Peg is more extended than that predicted by the model. The role of pulsations, if relevant in other cases and unmodeled by PHOENIX, therefore seems negligible for the atmospheric structures of our sample. The targets are located close to the red limits of the evolutionary tracks of the STAREVOL model, corresponding to masses between 1 M-circle dot and 3 M-circle dot. The STAREVOL model fits the position of our stars in the Hertzsprung-Russell (HR) diagram better than the Ekstrom model does. STAREVOL includes thermohaline mixing, unlike the Ekstrom model, and complements the latter for intermediate-mass stars.