The search for the primordial B-modes polarization in the cosmic microwave background (CMB) radiation,
carrying the signature of the primordial gravitational waves from the inflation epoch, motivated a significant
technological progress enabling the next generation of CMB instruments (e.g. CMB-S4, LiteBIRD)
to reach an unprecedented sensitivity. However, such a challenging detection demands a very high control
of the instrumental systematics and CMB foreground emissions.
Among those, the galactic dust polarized emission spectral dependence, not yet fully
characterized, could leave a high level of uncertainty in the cosmological polarization data
producing an ambiguous detection of the CMB B-modes.
Characterizing the dust spectral energy distribution (SED) spatial variations became one of
the most critical issues in the quest for primordial B-modes.
In the work that I will present we have used the release of the Planck satellite HFI data
obtained with the software Sroll2 (Delouis+2019, A&A 629, A38), in order to characterize
and compare the SEDs for polarization and total intensity.
The mean SEDs for dust polarization and total intensity from 353 to 100 GHz are confirmed
to be remarkably close. However, the data show evidence for spatial variations of the
polarization SED. These variations are correlated with variations of dust temperature
measured on total intensity data but the correlation is tight only in the Galactic plane.
At higher latitudes, by considering 90% of useful sky fraction and less, the amplitude of the dust
emission residuals in polarization suggests that an additional contribution, coming from
variations of the polarization angle, becomes dominant. Current models, which extrapolate
the SED spatial variations from total intensity to polarization, would be therefore grossly
simplifying and underestimating the foreground signal to CMB polarization.