|Annu. Rev. Astron. Astrophys. 1994. 32:
Copyright © 1994 by Annual Reviews. All rights reserved
Penzias & Wilson (1965) set an upper limit of 10% on the polarization of the CMB, and there have been several subsequent upper limits published (e.g. Nanos 1979, Caderni et al 1978, Lubin & Smoot 1981, Lubin et al 1983, Partridge et al 1988, Wollack et al 1993), but so far none that have been low enough to be cosmologically interesting.
Any quadrupole anisotropy in the radiation field will give rise to a linear polarization when it is Thomson scattered. This was first suggested in the case of an anisotropic universe (Rees 1968, Anile 1974, Dautcourt & Rose 1978), but can also arise, generally of smaller amplitude, from any inhomogeneity, and can moreover be enhanced by reionization (Negroponte & Silk 1980, Basko & Polnarev 1980, Stark 1981, Tolman & Matzner 1984, Nasel'skii & Polnarev 1987, Harari & Zaldarriaga 1993, Ng & Ng 1993). A transfer equation similar to Equation (23) can be written for the polarization amplitude P (Bond & Efstathiou 1987). Solutions indicate that the level of anisotropy is generally small for standard recombination. Efstathiou (1988) has developed approximations for calculating the small angular scale autocorrelation function of P in reionization scenarios. Kaiser (1983) has shown that the rms polarization from density fluctuations is typically 10-20% of T / T for adiabatic fluctuations on arc-minute scales. This means that it is still well below the detection levels of present-day experiments. The problem has been studied numerically, including both the contribution of density fluctuations and gravitational waves in Crittenden et al (1993a), where it was shown that the additional contribution due to gravity wave modes was too small to be resolved with current sensitivities (see also Frewin et al 1994). However, in future, as these experiments become more sensitive and it becomes more difficult to distinguish real fluctuations from those caused by confusion with faint radio sources and diffuse emission from the Galaxy, it may be that polarization information could be used as a discriminant.
The polarization pattern can also be a useful test of the ionization history of the universe, since it is expected to be coherent over scales corresponding to the thickness of the last scattering surface (Hogan et al 1982). Hence coherence on scales >> 10 arc minutes would imply that reionization had indeed taken place. Note that all of the statements above are for linear polarization - circular polarization is not expected to be generated unless there are primordial magnetic fields or strong anisotropies at the scattering epoch.