Thermal dust polarization is the workhorse of magnetic-field mapping: elongated grains align with the local field and emit preferentially polarized light, so the polarization fraction measures the alignment efficiency and the field's order along each sightline. Planck's all-sky maps delivered two surprises in opposite directions: a maximum polarization fraction near 20 percent, far higher than pre-Planck expectations and implying remarkably efficient alignment in the diffuse medium, and a systematic collapse of polarization with column density, the polarization holes, where dense clouds depolarize faster than any simple geometry predicts (Planck Collaboration 2015).
Radiative torque alignment, the standard theory, struggles with both ends at once. The high diffuse-sky maximum demands near-perfect alignment and field order; the steep dense-cloud decline then requires that alignment to fail rapidly with depth, and the natural RAT explanation, loss of the aligning photons inside opaque clouds, undershoots the observed steepness. The repairs invoke small-scale magnetic field tangling within the beam, or grain growth and coagulation changing the alignable population, neither independently constrained, both tuned to the polarization data they explain (Ritacco et al. 2025).
The standing carries the now-familiar foreground stakes: dust polarization is the dominant B-mode foreground, and a component whose polarization physics needs region-by-region repair propagates its uncertainty directly into primordial gravitational-wave searches.