**8.3. Tensor degeneracy**

All of the above applies to models in which scalar modes
dominate. The possibility of a large tensor component
yields additional degeneracies, as shown in
figure 20. An *n* = 1 model with a large
tensor component can be made to resemble a zero-tensor model
with large blue tilt (*n* > 1) and high baryon content.
Efstathiou et al. (2002)
show that adding LSS
data does not remove this degeneracy; this is reasonable,
since LSS data only constrain
the baryon content weakly.
A better way of limiting the possible tensor contribution is to
look at the amplitude of mass fluctuations today: this normalization
of the scalar component is naturally lower if the CMB signal is
dominated by tensors. These issues are discussed further below.

Another way in which the remaining degeneracy may be lifted is through
polarization of the CMB fluctuations.
A nonzero polarization is inevitable because the
electrons at last scattering experience an anisotropic
radiation field. Thomson scattering from an anisotropic
source will yield polarization, and the practical size of
the fractional polarization *P* is of the order of the quadrupole
radiation anisotropy at last scattering:
*P* 1%.
This signal is expected to peak at
500, and the
effect was first seen by the DASI experiment
(Kovac et al. 2002).
Much more detailed polarization results were presented by
the WMAP satellite, including a critical detection of
large-scale polarization arising from secondary scattering at
low *z*, thus measuring the optical depth to last scattering
(Kogut et al. 2003).
On large scales, the polarization signature
of tensor perturbations differs from that of scalar perturbations (e.g.
Seljak 1997;
Hu & White 1997);
the different contributions to the total unpolarized
*C*_{} can
in principle be disentangled, allowing the inflationary
test to be carried out.