Introduction

It is well known that synchrotron radiation has an intrinsic linear polarization. Theory implies that there should also be a small component of circular polarization (e.g., Legg & Westfold 1968). The predicted degree of circular polarization, , is of order , where is the Lorentz factor of the emitting particle. The typical frequency radiated by a particle is , where is the cyclotron frequency. It follows that the degree of circular polarization should vary according to , where is the wavelength of the radiation. Through the 1970s there were a number of measurements of for extragalactic synchrotron sources (e.g., Biraud 1969; Conway et al.\ 1971; Berge & Seielstad 1972; Seaquist 1973; Roberts et al. 1975; Weiler & de Pater 1980), and also for the plerionic component of the Crab Nebula (Weiler 1975), although more recent data cast doubt on the significance of the reported value of (Wilson & Weiler 1997). In particular Roberts et al. (1975) detected circular polarization in eight quasars, all of which showed evidence of synchrotron self absorption, and some of which showed evidence of time variability in . The data on sources for which was measured at more than one frequency were not consistent with the predicted law for the intrinsic circular polarization of synchrotron radiation from a single source. Roberts et al. suggested that these data could be interpreted in terms of a multi-component model for the various sources, with at different frequencies being dominated by different subsources. The suggestion by Roberts et al. includes the possibility that some of the variation of with frequency can be attributed to the change in the sense of polarization as a source becomes self absorbed (Melrose 1972, 1980; Jones & O'Dell 1977a,b; Weiler & de Pater 1980). Nevertheless, neither the time variability nor the frequency dependence of is readily compatible with an interpretation in terms of the intrinsic circular polarization of synchrotron emission. The possibility that the observed circular polarization is due to a propagation effect needs to considered in detail.

In this paper we discuss an alternative way in which a circularly polarized component could arise as a propagation effect, referred to as circular repolarization by Pacholczyk (1973). The main emphasis in our discussion is, first, to explain in principle how this alternative mechanism can produce circular polarization, and, second, to make some rough estimates of the conditions under which the effect might be significant in cases such as plerions and compact extragalactic sources where there may be only relativistic particles and no cold plasma. [Pacholczyk (1973) discussed the case where the ambient medium consists of cold plasma with an admixture of highly relativistic electrons.] This investigation is partly motivated by a current program of observations of circular polarization with the Australia Telescope (R. Norris, D. Rayner, private communication 1997). In principle it is now possible to detect circular polarization with a much higher sensitivity than was available in the earlier observations reported by Roberts et al. (1975), and this paper is preliminary to a more detailed discussion of the various possible interpretations of circular polarization of synchrotron sources.

The physical basis for the mechanism considered here is propagation through a medium with elliptically (or linearly) polarized wave modes, as discussed in section 2. In section 3 we point out that a linearly polarized contribution to the natural modes of the plasma arises from the relativistic electrons (Sazonov 1969; Melrose 1997a). In section 4 we define a ``relativistic rotation measure'' that characterizes this effect and consider conditions under which it might be important.

  
Figure 1: The polarization ellipse for a wave propagating into the page.

© Copyright Astronomical Society of Australia 1997