Evidence for Line-of-Sight Frequency Decorrelation of Polarized Dust Emission in $Planck$ Data

If a single line of sight (LOS) intercepts multiple dust clouds of different spectral energy distributions and magnetic field orientations, the frequency scaling of each of the Stokes $Q$ and $U$ parameters of thermal dust emission may be different ("LOS frequency decorrelation"). We present first evidence for LOS frequency decorrelation in $Planck$ data. We use independent, neutral-hydrogen--measurements of the number of clouds per LOS and the magnetic field orientation in each cloud to select two sets of sightlines: (i) a target sample (pixels likely to exhibit LOS frequency decorrelation); (ii) a control sample (pixels lacking complex LOS structure). We test the null hypothesis that LOS frequency decorrelation is not detectable in $Planck$ 353 and 217~GHz polarization data at high Galactic latitudes. The data reject this hypothesis at high significance. The detection is robust against choice of CMB map and map-making pipeline. The observed change in polarization angle due to LOS frequency decorrelation is detectable above the $Planck$ noise level. The probability that the detected effect is due to noise alone ranges from $5\times 10^{-2}$ to $4\times 10^{-7}$, depending on the CMB subtraction algorithm and treatment of residual systematics; correcting for residual systematics increases the significance of the effect. The LOS decorrelation effect is stronger for sightlines with more misaligned magnetic fields, as expected. We estimate that an intrinsic variation of $\sim15\%$ in the ratio of 353 to 217~GHz polarized emission between clouds is sufficient to reproduce the measured effect. Our finding underlines the importance of ongoing studies to map the 3D structure of the magnetized dusty ISM that could help component separation methods to account for frequency decorrelation effects in CMB polarization studies.