Wn. Brandt et al., SEPARATION OF FOREGROUND RADIATION FROM COSMIC MICROWAVE BACKGROUND ANISOTROPY USING MULTIFREQUENCY MEASUREMENTS, The Astrophysical journal, 424(1), 1994, pp. 1-21
Future measurements of the isotropy of the cosmic microwave background
radiation (CBR) will be limited by confusing foreground radiation. We
examine the extent to which galactic foregrounds-synchrotron, bremsst
rahlung, and dust radiation-can be identified and removed using multif
requency data. We discuss the capabilities and limitations of this app
roach and investigate its effectiveness for various choices of observi
ng frequencies using simulations over 10-degrees x 10-degrees patches
of sky with 1-degree resolution. At this resolution galactic fluctuati
ons are at a minimum just below 100 GHz. In our simulations, observati
ons covering two or three octaves in frequency well below or well abov
e this minimum region perform much better than those in the minimum re
gion and allow accurate subtraction of foregrounds if the measurement
noise is sufficiently low. We are thus optimistic that Galactic foregr
ound fluctuations can be distinguished from CBR fluctuations in multif
requency data on 1-degree scales. The required noise level depends in
a highly nonlinear way on the frequencies observed, the model used to
fit the various signal components, and on (unknown) details of the for
eground emissions and CBR anisotropy. Even in favorable cases, however
, the required noise level is much below any achieved to date. Separat
ion of the CBR from foregrounds required at least three free parameter
s in the simulations, thus observations at a minimum of four frequenci
es are required for our method. We also discuss the usefulness of meas
urements of foreground components made at frequencies where those fore
grounds are strong, and we show that even if the functional form of th
e spectrum of the foregrounds is perfectly known, the absolute error t
hat can be tolerated in measurements at other frequencies is comparabl
e to the error that can be tolerated where the CBR is measured. Thus t
here is no easy way to subtract confusing foregrounds directly.