Convergence of rutile-TiO2(110) surface properties with the slab thickness, which imposes very thick slabs when all the atomic positions are optimized, is achieved more rapidly when one face of the slab is saturated. We report calculations using PW91 functional and plane-wave pseudopotential methodology showing that saturation by dissociated H2O of the 'bottom face' improves convergence for the estimation of the surface energy for the 'top face'. The description of the adsorption on top of a Ti site on this face is upgraded. The electronic band gap is enlarged, becoming closer to the value obtained on thick slabs by complete optimization. (C) 2012 Elsevier B.V. All rights reserved.