1995A&A...301..797L

In the previous papers of this series we have described a new observational program of broadband linear polarimetry aimed at Ap stars. At the same time, we have established a canonical model, based on the oblique rotator geometry, which describes successfully the main features of the observed polarization: in some cases the linear polarization data, combined with the classical circular polarization measurements, allow one to determine the characteristic parameters which define the oblique dipolar rotator. However, we have also observed polarization diagrams that depart clearly from those predicted by the canonical model, which means that it is not always possible to rely on a pure dipolar model (nor on a combination of a dipole plus a linear quadrupole parallel to the dipole). Although an interpretation of the polarization peculiarities in terms of magnetic `anomalies' (i.e. deviations from the dipolar configuration) is quite natural, one must also take into account the possible influence of local abundance inhomogeneities. Therefore, we have first studied the sensitivity of the polarized signal (which is known to be due to the differential saturation of Zeeman components in spectral lines) to a variation of the metallic absorption spectrum. Then we have examined how a local enhancement (or reduction) of the polarization produced by a dipolar magnetic field affects the Fourier spectrum of the observed polarization signal. Finally, we have designed an inversion program making possible the recovery - under certain restrictions - of the spatial modulations of the polarization generated by a dipole, which are necessary to explain `odd' polarimetric data. This program has been applied to the data gathered from three stars (49 Cam, β CrB, HD 71866). As far as the last star is concerned, none of the spatial modulations considered was able to reproduce the observations. On the contrary, good solutions are found for the other two. However, if one interprets the variations of the polarization as the result of abundance variations, which must correspond to a modulation of the absorption spectrum, a contradiction arises, especially for β CrB, because the observed spectral variability of these stars is too small to account for our computed maps. Therefore, non-canonical polarization diagrams must essentially be interpreted in terms of magnetic anomalies, not of abundance anomalies: in other words, the peculiarities of the polarization diagrams are likely to result mainly from departures of the magnetic configuration from the pure dipolar configuration.