On OH line formation and oxygen abundances in metal-poor stars

The formation of the UV OH spectral lines has been investigated for a range of stellar parameters in the light of 3D hydrodynamical model atmospheres. The low atmospheric temperatures encountered at low metallicities compared with the radiative equilibrium values enforced in classical 1D hydrostatic model atmospheres have a profound impact on the OH line strengths. As a consequence, the derived O abundances using 3D models are found to be systematically lower by more than 0.6 dex at [Fe/H] = 3.0 compared with previous 1D analyses, casting doubts on the recent claims for a monotonic increase in [O/Fe] towards lower metallicities. In fact, taken at face value the resulting 3D LTE trend is in rough agreement with the conventional [O/Fe] plateau. Caution must, however, be exercised in view of the remaining assumptions in the 3D calculations. We have verified that the stellar parameters remain essentially unchanged with 3D model atmospheres provided that the infrared flux method (DeltaT(eff) less than or similar to 20 K), Hipparcos parallaxes (Delta log g less than or similar to 0.05) and Fe II lines (Delta [Fe/H] less than or similar to 0.1 dex) are utilised, leaving the 3D O abundances from OH lines largely intact (Delta [O/H] less than or similar to 0.05 dex). Greater concern stems from possible departures from LTE in both the line formation and the molecular equilibrium, which, if present, would increase the derived O abundances again. Non-LTE line formation calculations with 1D model atmospheres suggest no significant steepening of the [O/Fe] trend even if the abundance corrections amount to about 0.2 dex for all investigated stellar parameters. We note, however, that the 3D case may not necessarily be as metallicity-independent. The apparent lack of laboratory or theoretical rate coefficients at the relevant temperatures for the involved molecular reactions unfortunately prevents a quantitative discussion on the possible effects of non-equilibrium chemistry.