[Paper Review] BeyondPlanck XI. Bayesian CMB analysis with sample-based end-to-end error propagation

L. P. L. Colombo, J. R. Eskilt+37 more|arXiv (Cornell University)|Aug 30, 2022

Cosmology and Gravitation Theories4 citations

TL;DR

This paper presents a Bayesian, sample-based end-to-end error propagation framework for cosmic microwave background (CMB) analysis using Planck LFI and WMAP data within the BeyondPlanck framework. It achieves asymptotically exact Blackwell-Rao estimation of the CMB power spectrum up to 𝓁 ≤ 600, revealing a strong degeneracy between CMB temperature fluctuations and free-free emission at 400 ≤ 𝓁 ≤ 600, which is mitigated through model reduction and masking, resulting in a dipole amplitude of 3362.7 ± 1.4 μK in excellent agreement with prior results but with uncertainties derived purely from posterior sampling without ad hoc corrections.

ABSTRACT

We present posterior sample-based cosmic microwave background (CMB) constraints from Planck LFI and WMAP observations derived through global end-to-end Bayesian processing. We use these samples to study correlations between CMB, foreground, and instrumental parameters, and we identify a particularly strong degeneracy between CMB temperature fluctuations and free-free emission on intermediate angular scales, which is mitigated through model reduction, masking, and resampling. We compare our posterior-based CMB results with previous Planck products, and find generally good agreement, but with higher noise due to exclusion of HFI data. We find a best-fit CMB dipole amplitude of $3362.7\pm1.4μK$, in excellent agreement with previous Planck results. The quoted uncertainty is derived directly from the sampled posterior distribution, and does not involve any ad hoc contribution for systematic effects. Similarly, we find a temperature quadrupole amplitude of $σ^{TT}_2=229\pm97μK^2$, in good agreement with previous results in terms of the amplitude, but the uncertainty is an order of magnitude larger than the diagonal Fisher uncertainty. Relatedly, we find lower evidence for a possible alignment between $\ell = 2$ and $\ell = 3$ than previously reported due to a much larger scatter in the individual quadrupole coefficients, caused both by marginalizing over a more complete set of systematic effects, and by our more conservative analysis mask. For higher multipoles, we find that the angular temperature power spectrum is generally in good agreement with both Planck and WMAP. This is the first time the sample-based asymptotically exact Blackwell-Rao estimator has been successfully established for multipoles up to $\ell\le600$, and it now accounts for the majority of the cosmologically important information. Cosmological parameter constraints are presented in a companion paper. (Abriged)

Motivation & Objective

• To develop a fully Bayesian, end-to-end framework for CMB analysis that propagates all systematic and instrumental uncertainties through the full posterior distribution.
• To identify and mitigate strong degeneracies between CMB temperature fluctuations and astrophysical foregrounds, particularly free-free emission at intermediate multipoles (400 ≤ 𝓁 ≤ 600).
• To provide asymptotically exact power spectrum estimation using the Blackwell-Rao estimator on posterior samples, up to 𝓁 = 600.
• To reassess large-scale CMB anomalies—such as the quadrupole-octopole alignment—under a more conservative, fully propagated uncertainty model.
• To demonstrate the feasibility and advantages of the Bayesian approach for future high-sensitivity CMB experiments, especially for detecting subdominant primordial B-mode signals.

Proposed method

• The analysis employs a global Gibbs sampling scheme to jointly sample CMB sky maps, instrumental parameters, foreground components, and power spectra within a unified Bayesian hierarchical model.
• Posterior samples are used to compute the CMB angular power spectrum via the sample-based Blackwell-Rao estimator, which provides asymptotically exact uncertainty propagation.
• Model reduction, masking, and resampling techniques are applied to mitigate the strong degeneracy between CMB and free-free emission at 400 ≤ 𝓁 ≤ 600.
• The posterior distribution is used directly to infer uncertainties, eliminating the need for ad hoc corrections for Planck instrumental systematics.
• The framework integrates Planck LFI and WMAP data, with full propagation of both astrophysical and instrumental uncertainties through the sampling process.
• A conservative sky mask is applied to reduce contamination from the free-free degeneracy, ensuring robustness in low-ℓ power spectrum estimation.

Experimental results

Research questions

• RQ1 What is the impact of a strong degeneracy between CMB temperature fluctuations and free-free emission on intermediate angular scales (400 ≤ 𝓁 ≤ 600) in posterior inference?
• RQ2 How does end-to-end Bayesian error propagation affect the significance of large-scale CMB anomalies such as the quadrupole-octopole alignment?
• RQ3 To what extent do posterior-based uncertainties in the CMB dipole and quadrupole amplitudes differ from traditional Fisher-based estimates?
• RQ4 Can the sample-based Blackwell-Rao estimator provide asymptotically exact power spectrum uncertainty estimates up to 𝓁 = 600 in a realistic, full-sky analysis?
• RQ5 How does the fully Bayesian approach compare to conventional frequentist methods in terms of consistency and robustness for cosmological parameter inference?

Key findings

• The CMB dipole amplitude is measured as 3362.7 ± 1.4 μK, in excellent agreement with previous Planck results, with the uncertainty derived directly from the posterior distribution without additional ad hoc corrections.
• A strong degeneracy between CMB temperature fluctuations and free-free emission is identified at 400 ≤ 𝓁 ≤ 600, which is mitigated through model reduction, masking, and resampling.
• The temperature quadrupole amplitude is measured as σ²_TT = 229 ± 97 μK², which is consistent with prior results but has an uncertainty an order of magnitude larger than the naive diagonal Fisher estimate.
• The significance of the quadrupole-octopole alignment is reduced due to a larger scatter in the posterior distribution, caused by marginalizing over a broader set of systematic effects and a more conservative mask.
• The high-ℓ temperature power spectrum is in good agreement with both Planck and WMAP results, validating the robustness of the sample-based Blackwell-Rao estimator up to 𝓁 = 600.
• This work establishes the first successful application of the sample-based Blackwell-Rao estimator for multipoles up to 𝓁 ≤ 600, which now accounts for the majority of cosmologically relevant information.