Experimental data

Figure 1: Explorer-Nautilus coincidence events (upper plot) and background estimates (lower plot) as a function of the sidereal time in 1/2 hour bins.

This analysis has been performed on a data set of Explorer-Nautilus coincidences with an energy filter veto (i.e. requiring agreement between the event energies of the two antennae) and with a fixed time window of $\pm 0.5$ s. Data we are referring to are those obtained using runs longer than 12 hours. The data are grouped in half hour bins of sidereal time, as shown in Fig. 1. The upper plot of the figure reports the number of observed coincidences ($n_c$), while the lower plot gives the average number of the background events estimated by off-time techniques. It is worth remarking that the method we are going to use does not depend critically on the width of the bins, provided that the width is small enough to assure a good resolution of the antenna pattern. (To state it clearly, contrary to other methods in which some binning is required and the resulting significance depends dramatically on the choice of the binning, in our method we could have, virtually, bins of arbitrary small width. Rebinning does not spoils the quality of the information, as long as the binning is finer than the structures exhibited by the antenna pattern and there are no clustering of events within a bin. The latter possibility is excluded by inspecting the arrival time of the individual events, as shown in Ref. [1] for the events around 4:00.)

As far as the background is concerned, we recall that the random coincidence background is well described by a Poisson distribution [1], and that the sidereal hour fluctuations of the averages is compatible with the grand average over the 24 hours of $0.57\pm 0.03$ events/hour. For these reasons, we believe that the the value of $\lambda_B = 0.57$ is the most reasonable value to use as parameter of the Poisson distribution which models the background fluctuation in the 0.5-hour bins.