4 Discussion

The I-V phase $\ell -\nu$ diagram in the sodium D lines shows two dominant values: a positive one on the p-modes and a negative one on the so called "background''. The data analysed locally show the same I-V phase values in the $\ell -\nu$ diagram.

In the five minute band, the phase found with the local analysis corresponding to the high velocity power locations, is that found in the peaks of the p-modes, while the negative phase is found where the velocity power is low. As an analogy, we attribute the negative value obtained through the local analysis to the signature of the solar background. We investigated the spatial distribution of the background along the frequency domain in order to infer some characteristics of the source of the solar oscillations.

We summarise the obtained results and a possible interpretation. 1) The background locations, in the five-minute band, are associated with those points where the velocity power is low. This could mean that the p-modes are acting as a selective filter for a uniform background distribution over the disk. At this point, the correspondence to the magnetic oscillating points is not a proof of a physical relation between the magnetic field and the background, since the magnetic points usually correspond to low velocity power locations. For this reason, the spatial distribution of the background is studied at low frequencies, where the contamination due to p-modes is largely reduced. 2) At low frequencies, the probability of finding the background at the same location is much larger than expected. The slight increase in the area filled by the background is compatible with a 0.5 coverage of 4 $^{\prime \prime }$ border line of a 70 $^{\prime \prime }$ diameter region; the non-uniform spatial distribution is confirmed by the behaviour of the coverage, that seems to cluster around structures of the same order of magnitude. This is consistent with the observations of Chae et al. (1998), where a preferential occurrence of the events is reported. This suggests the presence of localised phenomena, whose temporal behaviour reproduce a phase difference that is constant in a $36^{\circ}$ interval between 1.6 and 2.6 mHz. 3) The autocorrelation of the phase coverage maps shows the rotation of the structures associated with the p-modes (at the five-minute band), but not of the background, whose characteristic scale is of the order of one pixel. This can be interpreted as a rotating subarcsec structure during the observing run, or to structures at the limit of the spatial resolution but lasting a period whose trace during the rotation at disk center is confined in one pixel, that is, less than 30 min. The delay between the intensity and velocity pulses can be accepted up to 60 s: this value causes a $20^{\circ}$ linear trend in the phase value in 1 mHz frequency interval and is considered as constant in an analysis performed to search for the co-spatial background events, as we described in Sect. 2 of the discussion.

These characteristics seem to match very well those of the downflows required to generate the background's characteristics in the $\ell -\nu$ diagram (according to the model of Skartlien & Rast 2000), even if the spatial resolution is low. This limitation could make the determination of the phase values uncertain, but we remark that the starting point of our discussion is that we are dealing with two distinct observed phases. No hypotheses were formulated regarding the phenomena that would reproduce these values, but we attributed them to the p-modes and the background signatures. Some processes are invoked to justify the spatial distribution of the background locations. The one we propose on the basis of some observational evidence is introduced below and discussed in the last section. Some questions remain unanswered. What are the border lines where the background locations seem to cluster? Are they related to the supergranular lanes, the magnetic network, or something else?

Any attempt to correlate the structures usually fails but, nevertheless, many coincidences in the shape are clearly visible (see Fig. 8) with the events associated with the bright points in chromosphere. So, if the "convective'' downflows are correlated to the intergranular lanes, where the magnetic cancellations are expected to occur, and if the latter process produces an observed upward plasma jet, this suggests that the photospheric downflows could be related to the downward counterpart.

Some observations confirm the perturbation at the surface induced by a downplume after a big flare (Kosovichev & Zarchova 1998) and the lack of co-spatiality suggests the non-vertical propagation of the plasma. Is the energy input from these events compatible with the observed balance of the p-modes during the solar cycle? It has been established that the magnetic field reduces the power of the p-modes (Chaplin et al. 2000; Braun & Lindsey 1999; Komm et al. 2000). Nevertheless, the energy input in the solar oscillations seems to be independent of the cycle (Komm et al. 2000; Chaplin et al. 2000). This evidence attributes a passive role to the magnetic field. On the contrary, stellar observations show a strong correlation between convection and the magnetic activity (Pettersen 1989). In fact, the turbulent motions stress the magnetic field lines resulting in a continuous flaring. The cycle is mainly related to the strong, stable magnetic structures, and the large and visible events in the maxima are rare and negligible in the energy balance. This could invoke two distinct processes coupling the oscillations and the magnetic field: the absorption due to the interaction with the long-term strong magnetic structures and the continuous energy input due the injection of downflowing plasma at small scales, caused by a flaring activity. For this reason, if the background in the I-V phase difference is believed to be the signature of the source of the solar oscillations, we suggest this source to be "magnetic'' instead of "convective'', even if the energy source is still of a kinetic nature.