Low-frequency oscillations in the absorption spectrum at 1020 nm, connected to the primary charge separation process in Rhodobacter sphaeroides, have been shown by Yakovlev et al. to be caused by rotational motion of an interstitial water molecule called "water-A". The same water molecule was shown by Potter et al. to increase the rate of charge separation by a factor of 8. We have carried out geometry optimization of water-A and its nearest atoms in the protein pocket, using density functional theory (DFT). There are strong hydrogen bonds to the axial imidazol group of the B part of the special pair (P = PAPB) and to the keto carbonyl group of ring V of the accessory chlorophyll (B-A). Rotation of water-A is thus impossible in the electronic ground state. We have tried to support our speculations on other possible mechanisms by calculations. The P+BA- charge transfer state is stabilized by proton transfer from water-A and simultaneous proton transfer from the axial group of P-B to water-A. After double proton transfer the hydrogen bond to the keto group disappears whereby a possibility opens up for almost free water rotation. The results therefore would explain the 32 cm(-1) oscillation of Yakovlev et al. The proposed mechanism assumes, however, that the general assumption that the activation energy disappears in the primary charge separation of bacterial photosynthesis, holds also for this special case.