Here are some additional maps showing a simulation of extraction of
B modes using the well-supported-modes method. Unlike the maps in the paper,
these show just the magnitude of the polarization at each point, and the
cut has not been additionally added. Because we are using well supported
modes, the modes are essentially zero in the cut. Units of the magnitude
should be taken to be arbitary (the simulation uses A_t/A_s=0.2). The third
map shows what you get for the recovered well supported B modes from a simulation
with zero input B modes. The first three plots use l_max = 150, epsilon=0.001.
Using l_max=200, epsilon=0.001 the result is
(note: computing the diagonalization in this case took nearly three weeks on a slowish Itanium machine! However making the B map for this realization only took 30 seconds.)
A close up view:
These maps were made with HEALPix 1.2.
Antony Lewis July 2003.