The experiment described above establishes that the MF-burst and 2 auroral roar emissions are left-hand polarized, in contrast to auroral hiss which is right-hand polarized. For all three emissions, the mean power difference between alternate sweeps, during which the signals constructively and destructively sum, averages -10 dB. The instrumental error in this measurement due to phase inaccuracies in the polarization detection electronics is the order of -20 dB. The departure from circular polarization inferred from our measurements can in principle be interpreted either in terms of finite ellipticity of the polarization or in terms of off-zenith incidence of a perfectly circularly polarized wave. In practice, other factors limit the observed degree of polarization so that it is not possible to quantitatively infer more from the data than the sense of polarization. Foremost among these factors is the variability of the signals. High time resolution measurements show that both MF-burst [LaBelle, J. et al.,1997] and auroral roar [LaBelle, J. et al.,1995; Shepherd, S. G. et al.,1996] amplitudes vary on time scales far shorter than the 2 s time scale needed to make a polarization measurement. The fast time-scale variations effectively impart noise into the system which biases the result away from perfect circularity. Another possible factor is the effect of the component of the wave reflected from the conductive ground underlying the antenna.
As mentioned in the introduction, in the case of auroral roar the polarization measurement has significant implications for theories. Several authors have calculated that for realistic loss-cone distribution functions the X-mode cyclotron maser instability operating at F-region altitudes can have a growth rate exceeding electron-neutral collision frequencies [Weatherwax, A. T. et al.,1995; Yoon, P. H. et al.,1996]. However, Yoon, P. H. et al.,  point out that only the X-mode will reach the ground from this mechanism, because the O-mode remains trapped in the ionosphere. Hence this mechanism predicts that the waves should be right-hand elliptically polarized. The measurement of LP for auroral roar emissions excludes this mechanism. The cyclotron instability also excites trapped Z-mode waves with high growth rates at locations where the upper hybrid frequency matches the cyclotron harmonics [e.g., Kaufmann, R. L. 1980; Yoon, P. H. et al.,1997], and it has been suggested that these waves may convert by a variety of mechanisms to either L-O mode or R-X mode electromagnetic waves [Gough, M. P. and A. Urban 1983; Weatherwax, A. T. et al., 1995]. This mechanism therefore predicts either LP or RP radiation depending on the conversion mechansism and hence is not excluded by the polarization measurements presented above. Another suggested mechanism involves the interaction of upper hybrid waves near harmonics of the gyrofrequency [Willes, A. J. and S. D. Bale, 1998; Winglee, R. M. and G. A. Dulk 1986]. However, the polarization measurements place a constraint on the conversion mechanisms which may be considered in these theories.
This research was supported by National Science Foundation grant ATM-9316126 to Dartmouth College. The authors acknowledge helpful discussions with A.T. Weatherwax and R.A. Treumann.