Lithium abundances of halo dwarfs based on excitation temperatures II. Non-local thermodynamic equilibrium

Context. The plateau in the abundance of Li-7 in metal-poor stars was initially interpreted as an observational indicator of the primordial lithium abundance. However, this observational value is in disagreement with that deduced from calculations of Big Bang nucleosynthesis (BBN), when using the Wilkinson microwave anisotropy probe (WMAP) baryon density measurements. One of the most important factors in determining the stellar lithium abundance is the effective temperature. In a previous study by the authors, new effective temperatures (T-eff) for sixteen metal-poor halo dwarfs were derived using a local thermodynamic equilibrium (LTE) description of the formation of Fe lines. This new T-eff scale reinforced the discrepancy. Aims. For six of the stars from our previous study we calculate revised temperatures using a non-local thermodynamic equilibrium (NLTE) approach. These are then used to derive a new mean primordial lithium abundance in an attempt to solve the lithium discrepancy. Methods. Using the code MULTI we calculate NLTE corrections to the LTE abundances for the Fe I lines measured in the six stars, and determine new T-eff's. We keep other physical parameters, i.e. log g, [Fe/H] and xi, constant at the values calculated in Paper I. With the revised T-eff scale we derive new Li abundances. We compare the NLTE values of T-eff with the photometric temperatures of Ryan et al. (1999, ApJ, 523, 654), the infrared flux method (IRFM) temperatures of Melendez & Ramarez (2004, ApJ, 615, L33), and the Balmer line wing temperatures of Asplund et al. (2006, ApJ, 644, 229). Results. We find that our temperatures are hotter than both the Ryan et al. and Asplund et al. temperatures by typically similar to 110-160 K, but are still cooler than the temperatures of Melendez & Ramarez by typically similar to 190 K. The temperatures imply a primordial Li abundance of 2.19 dex or 2.21 dex, depending on the magnitude of collisions with hydrogen in the calculations, still well below the value of 2.72 dex inferred from WMAP + BBN. We discuss the effects of collisions on trends of Li-7 abundances with [Fe/H] and T-eff, as well as the NLTE effects on the determination of log g through ionization equilibrium, which imply a collisional scaling factor S-H > 1 for collisions between Fe and H atoms.