Do we need to correct the internal linear combination quadrupole?

The problem of high correlation of the CMB quadrupole with foregrounds is investigated in this paper. We present an analysis of the cross-correlation for the WMAP ILC(III) low multipoles and foregrounds and show that without correction of the sign for m = 0 harmonic of the quadrupole this correlation is very high. By analysis of 10^sup 4^ Monte Carlo simulations of the CMB-foreground separation, presented by Eriksen et al. [1], we show that practically for 50% realizations of the CMB quadrupole with the sign of the m = 0 mode equivalent to the sign of the foreground component, the input quadrupoles have the opposite sign of the output signal. Preserving the WMAP ILC(III) power spectrum of the quadrupole and changing the sign of the ILC (III) m = 0mode, we get the coefficient of cross-correlations |K( = 2)| < 10%, what decreases the level of cross-correlations by a factor of 5-6. We use themultipole vectors for the renormalized ILC(III) quadrupole and show that these vectors are very sensitive to the sign and amplitude of the m = 0 mode.[PUBLICATION ABSTRACT]