Layered dust / B-modes — three-model matrix

Preserved original model output. Check claims against the independent audit →

optimist · 16.2 KB

GPT-5.6 Solcritic Fable 5.1optimist GPT-5.6 Sol

1. Strongest physical case

The strongest argument begins with the complex linear polarization,

\mathcal P_\nu(\hat n)\equiv Q_\nu+iU_\nu =\sum_{c} A_c(\hat n)\, s_c(\nu)\, e^{2i\psi_c(\hat n)},

where $c$ labels dust clouds or magnetically coherent layers, $A_c$ contains dust column, alignment efficiency and field-inclination factors, $s_c(\nu)$ is the layer’s normalized polarized SED, and $\psi_c$ is its polarization angle.

This equation isolates the necessary physics:

Multiple clouds do not by themselves cause frequency decorrelation. If every layer has the same $s_c(\nu)$ , it factors out and the polarization pattern scales coherently with frequency despite field tangling.
Different SEDs do not by themselves rotate the polarization angle if all layers have the same $\psi_c$ .
LOS-induced decorrelation becomes important when layers have both different polarized SEDs and non-parallel polarization vectors. Their relative weights then change with frequency, rotating the net angle and changing the degree of cancellation in $Q$ and $U$ .

That is the two-cloud mechanism identified by Tassis and Pavlidou: the 353-GHz polarization map can cease to predict the 150-GHz map when the polarized-intensity ratio of two clouds changes with frequency and their fields are substantially misaligned. Tassis & Pavlidou 2015

The role of the proposed information is therefore precise:

Three-dimensional dust tomography constrains $A_c$ and the number and location of layers. Present products are principally differential extinction or reddening maps, not direct 3D thermal-emission maps. They reveal whether a sightline contains one dominant cloud or several comparable columns and provide distance-localized column priors. The Gaia-based Edenhofer et al. reconstruction, for example, reaches 1.25 kpc with angular resolution up to $14'$ and parsec-scale radial information, although its effective resolution and completeness vary strongly with position. Edenhofer et al.
Grain-size, composition and alignment information constrains $s_c(\nu)$ and $A_c$ , rather than determining them exactly. The polarized emissivity is weighted by

\int da\,n_c(a)\,f_{{\rm align},c}(a)\, C_{{\rm pol}}(a,\nu)\,B_\nu[T_c(a)] .

Thus the relevant quantity is not merely the total size distribution $n(a)$ , but its combination with composition, shape, temperature and size-dependent alignment. At CMB frequencies polarized thermal emission is mainly a large, equilibrium-grain signal; PAHs and other nanoparticles are not a generic explanation for thermal-dust decorrelation. Competing grain models also differ: the astrodust+PAH model argues that a common large-grain composition naturally explains the observed weak frequency dependence of the FIR polarization fraction. Hensley & Draine Grain physics should therefore supply a family of priors and spectral derivatives, not a supposedly known per-cloud $\beta_d$ .

A distance- or velocity-resolved plane-of-sky field supplies $\psi_c$ . A single projected-field map integrated through the Galaxy is insufficient. What matters is the relative angle of the individual emitting layers. Gaia-distance-resolved starlight polarization is the cleanest tomographic observable; H I filament orientations in velocity channels provide much wider coverage but position–position–velocity structure is not identical to physical distance.

The empirical case is already nontrivial. Pelgrims et al. selected high-latitude sightlines using independent H I cloud counts and field orientations, then found excess 217–353 GHz angle changes and depolarization precisely where multiple, misaligned components were expected. About 15% cloud-to-cloud variation in the 353/217 polarized-intensity ratio reproduced the observed signal in that selected sample. This is evidence for the mechanism, not a measurement of a universal all-sky decorrelation amplitude. Pelgrims et al. 2021

The defensible claim is therefore not that ancillary maps uniquely reconstruct the microwave sky. It is that they constrain the latent variables responsible for non-factorizable polarized emission, identify where additional spectral freedom is warranted, and reduce a blind high-dimensional foreground fit to a spatially structured, testable model.

2. Additional mechanisms and observational handles

Several real levers would strengthen the case.

Gaia extinction tomography: differential reddening supplies cloud distances, columns and posterior uncertainties. Its most immediate use is cloud segmentation and a prior on relative layer weights, not direct prediction of polarized emissivity.
H I4PI, GALFA-H I and other spectroscopic H I surveys: narrow-channel filament orientations trace magnetically coherent atomic structures. Clark and Hensley obtained H I-derived $Q,U$ maps correlating at $r>0.75$ with Planck 353-GHz polarization and reproduced the large-scale dust $E/B$ asymmetry. The result supports H I morphology as a field-orientation prior, while leaving the velocity-to-distance association uncertain. Clark & Hensley 2019
Gaia plus optical starlight polarimetry: polarization toward stars at successive distances measures cumulative foreground Stokes parameters. Differencing statistically between distance brackets can recover the orientation and polarizing efficiency of individual layers. PASIPHAE-like surveys directly target the high-latitude stellar density and accuracy needed for this application.
High-frequency polarization: Planck 353 GHz remains the all-sky amplitude and angle anchor. Additional polarized channels above the CMB-dominated bands are essential for separating temperature, opacity and angle effects. Far-IR/submillimetre spectropolarimetry can also test whether one aligned large-grain population or several populations better describe the polarized SED.
Gas-phase separation: H I should be supplemented by CO and molecular-cloud information where necessary, while intermediate- and high-velocity H I components should remain explicit “distance-uncertain” layers rather than being forced into a Gaia-derived distance bin.
Radiation-field and temperature priors: FIR intensity and colour data constrain the heating field and temperature distribution. These data must be treated carefully because temperature–opacity degeneracy, CIB anisotropy and beam averaging can bias intensity-derived $T_d$ and $\beta_d$ .
Polarization-ratio and angle statistics: jointly analyze

\Delta\psi_{\nu_1\nu_2},\qquad P_{\nu_1}/P_{\nu_2},\qquad \mathcal R_\ell^{EE,BB}(\nu_1,\nu_2),

rather than compressing all complexity into one power-spectrum decorrelation coefficient. Angle rotation and frequency-dependent depolarization are especially diagnostic of LOS mixing.

Complex spin moments: expand $\mathcal P_\nu=Q+iU$ , not only scalar intensity. Complex moments of $\beta_d$ , $T_d$ and other spectral parameters naturally encode correlations between SED variations and polarization angle. Recent work shows that this representation can compress multilayer polarized skies and expose which kinds of 3D complexity affect different component-separation methods. Vacher et al. 2025
Morphological and parity statistics: dust $TE$ , $TB$ , $EB$ , $EE/BB$ , filament–field alignment and H I coherence provide independent tests of whether a model has the correct geometry. They should be validation observables, not rigid assumptions imposed on the CMB solution.
Target/control tests: single-cloud, magnetically coherent sightlines form a control sample; comparable-column, strongly misaligned sightlines form a high-risk sample. A credible model should predict a larger $\Delta\psi$ , stronger frequency-dependent depolarization and lower correlation ratios in the latter without being tuned on the tested frequency pair.

3. Why this could matter for $r$

Experiments targeting $\sigma(r)\sim10^{-3}$ cannot tolerate dust residuals merely because they are a small percentage of the original dust map. What matters is their residual $BB$ power around the reionization and recombination scales. Representative forecasts require foreground residuals near $D_\ell^{BB}\sim10^{-5}\,\mu{\rm K}^2$ on $\ell\lesssim100$ , although the exact requirement depends on sky region, delensing, frequency coverage and likelihood construction. McBride, Bull & Hensley 2023

The 3D approach could help through three distinct routes:

Bias reduction: it replaces a false rank-one dust mixing law with the few layer SEDs actually supported by ancillary data. This addresses the dangerous case in which $Q$ and $U$ have different effective scalings.
Variance control: external information fixes or narrows layer columns and angles, so the CMB fit need not estimate an independent amplitude, angle, temperature and index for every layer. That is the principal advantage over an unconstrained multi-MBB model.
Spatial adaptation: most pixels need not carry maximum foreground complexity. Cloud counts and magnetic misalignment can determine where to use a single MBB, where to activate extra moments or layers, and where masking or down-weighting is preferable.

The scale of the potential problem is demonstrated, not universally forecast, by existing simulations. In one LiteBIRD-like study, spatially varying dust temperature and index produced an approximately $21\sigma_r$ bias under an inadequate MBB likelihood; a joint $\beta_d$ - and $T_d$ -moment treatment recovered an unbiased result with $\sigma_r=8.8\times10^{-4}$ . Those numbers depend on the simulated sky, but they establish that modest-looking SED complexity can matter at next-generation sensitivity. Vacher et al. 2022

The realistic goal is not to prove that 3D information alone reaches a particular $\Delta r$ . It is to demonstrate, through end-to-end injections, that a 3D-informed likelihood reduces both foreground bias and the number of freely deprojected modes relative to a blind moment expansion. Success should be reported directly as residual $BB$ , $\Delta r$ , $\sigma(r)$ , and robustness across grain and cloud-association models.

4. Pipeline sketch

Assemble compatible inputs. Bring the 3D extinction posterior, H I channel maps, starlight polarization, high-frequency $Q,U$ , gas tracers and CMB-frequency data to controlled beams and pixelization. Integrate all sky models through the actual experimental bandpasses.
Construct a probabilistic layer catalogue. Cluster distance bins into a small number of dust-column peaks. Associate H I velocity components where the match is supported, while retaining ambiguous and distant H I components as separate latent layers. Propagate extinction-map distance and column uncertainties.
Infer per-layer field orientations. Use differential starlight polarization where stellar sampling permits it; use H I filament orientation elsewhere. Treat the angle as a distribution, not an exact map. Require the vector sum of the layers to be consistent with high-frequency Planck $Q,U$ .
Assign emissivity priors. For each layer, infer priors on $T_d$ , polarized opacity, $\beta_d^P$ , alignment efficiency and permitted spectral curvature from the local environment and a small set of competing grain models. Absorb the poorly known field inclination and intrinsic alignment efficiency into calibrated amplitude nuisance parameters unless independent information genuinely constrains them.
Forward-model polarized emission.

\mathcal P_\nu^{\,d}(\hat n)= \sum_c A_c^P(\hat n)\, s_\nu\!\left[T_c,\beta_c,\theta_{{\rm grain},c}\right]\, e^{2i\psi_c(\hat n)} +\mathcal P_\nu^{\,{\rm unresolved}} .

The final term is a low-order complex moment expansion for sub-beam structure, imperfect cloud segmentation and residual grain variability.

Fit jointly with the CMB and other foregrounds. Use a hierarchical pixel-, needlet- or harmonic-space likelihood. The CMB has its known thermodynamic spectrum; synchrotron receives its own spatially varying spectrum and angle, without being forced to follow the neutral-dust field. Fit layer amplitudes and small spectral corrections under ancillary-data priors.
Deliver the model in the form the cosmology analysis needs. Depending on the pipeline, this may be a set of frequency-dependent dust templates, a mixing matrix with prior covariance, constrained moment maps, or predicted multi-frequency dust cross-spectra plus a residual covariance term.
Validate out of sample. Predict withheld frequency maps, angle rotations and polarization ratios; compare target and control sightlines; test $EE$ , $BB$ , $TE$ , $TB$ , $EB$ and frequency-correlation ratios; repeat across data splits and bandpass perturbations.
Propagate uncertainty into $r$ . Marginalize over cloud association, grain family, field angle, opacity conversion and unresolved moments. Only then form the cleaned $BB$ likelihood. A rigid best-fit physical template without this uncertainty propagation would be unsafe.

5. What to keep from the draft

The distinction between sky-to-sky spectral variation and genuine LOS mixing is essential.
The explanation that sums of MBBs are generally not another MBB is correct and central.
The emphasis that LOS angle rotation requires both SED differences and magnetic misalignment is the right physical core.
The three ingredients are correctly presented as complementary: dust tomography supplies layers, grain physics supplies emissivity priors, and magnetic tracers supply polarization geometry.
The use of H I-selected target and control samples is an unusually strong observational argument because the selection is external to the tested Planck frequency maps.
The proposed hybrid of a physical forward model and a residual moment expansion is much more credible than treating either as exact.
The limitations concerning extinction-to-emission conversion, velocity–distance ambiguity, unknown field inclination, unresolved turbulence, grain-model uncertainty and instrument systematics should remain.
The concluding posture—use the 3D construction as an informative prior and diagnostic, not yet as a turnkey replacement for multifrequency fitting—is scientifically sound.

6. Upgrade list