Gas accreting onto a galaxy will be of low metallicity while halo gas due to a galactic fountain will be of near-solar metallicity .
We test these predictions by measuring the metal absorption line properties of halo gas 5 kpc above the plane of the edge-on galaxy NGC 891 , using observations taken with HST/STIS toward a bright background quasar .
Metal absorption lines of Fe ii , Mg ii and Mg i in the halo of NGC 891 are clearly seen , and when combined with recent deep H i observations , we are able to place constraints on the metallicity of the halo gas for the first time .
The H i line width defines the line broadening , from which we model opacity effects in these metal lines , assuming the absorbing gas is continuously distributed in the halo .
The gas-phase metallicities are [ Fe/H ] = -1.18 \pm 0.07 and [ Mg/H ] = -0.23 ~ { } +0.36 / -0.27 ( statistical errors ) and this difference is probably due to differential depletion onto grains .
When corrected for such depletion using Galactic gas as a guide , both elements have approximately solar or even supersolar abundances .
This suggests that the gas is from the galaxy disk , probably expelled into the halo by a galactic fountain , rather than from accretion of intergalactic gas , which would have a low metallicity .
The abundances would be raised by significant amounts if the absorbing gas lies in a few clouds with thermal widths smaller than the rotational velocity of the halo .
If this is the case , both the abundances and [ Mg/Fe ] would be supersolar .