This paper has been temporarily withdrawn by the authors. We have recently found that noise in the experiments is at the origin of the supposed back-and-forth motion which is discussed in the first version of the paper. As a consequence, figs 4 and 5 as well as their discussion are incorrect. Figure 1 and the general trend of fig. 2 are still valid. At this time, we are uncertain whether or not the short time behavior of cI, shown in fig. 3, is affected by measurement noise. We are working on a new version of the paper, using new techniques that allow us to correct for the experimental noise.
Download