The gallium problem in HgMn stars

Previous studies of visible-region high-excitation lines of Ga II in HgMn stars have usually concluded that these lines (lambda lambda 4251-4262 and lambda 6334) yield abundance estimates for gallium, similar to 1 dex greater than the UV resonance lines. Of the explanations proposed in the literature, we find that the presence of hyperfine structure (hfs) in the lines is the most likely. We analyse Lick Hamilton Echelle CCD spectra by spectrum synthesis with the code UCLSYN. Using the Bidelman & Corliss (1962) measurements of hfs in Ga 11, and the calculations given by Lanz et al. (1993), with the oscillator strengths by Ryabchikova & Smirnov (1994), we determine a difference in abundance (visual-UV) of only 0.2 dex. Of this difference, about 0.1 dex can be explained by the simplified approximation to the true hfs, which can be estimated theoretically, and the remaining 0.1 dex can probably be accounted for by the stratification of Ga found by Smith (1995, 1996b). We conclude that the visible-region and UV abundances are in agreement within the errors of the determinations, and that the anomaly previously found by several investigations is an artifact of the simplified atomic line structures assumed, or results from the use of very different gf-values from those adopted here.