Spectral Imaging with QUBIC: building astrophysical components from Time-Ordered-Data using Bolometric Interferometry

TL;DR

Primordial B-mode detection is challenged by foregrounds and instrumental systematics. The authors propose a time-ordered data (TOD) based component separation that exploits QUBIC's spectral imaging via a frequency-dependent synthesized beam, using a linear forward model with $\vec{H}$ and $\vec{A}$ to reconstruct sky components directly. External Planck data are integrated to regularize edge effects, enabling end-to-end analysis and estimation of the tensor-to-scalar ratio $r$ with precision $\sigma(r) \approx 0.023$ under realistic foregrounds. Both parametric (constant and varying $\beta$) and blind foreground modeling are explored, with the blind approach reducing dust decorrelation biases in $r$ and improving robustness. The results demonstrate a viable TOD-based pipeline for QUBIC that can be extended to handle additional systematics and more extensive external data.

Abstract

The detection of B-modes in the CMB polarization pattern is a major issue in modern cosmology and must therefore be handled with analytical methods that produce reliable results. We describe a method that uses the frequency dependency of the QUBIC synthesized beam to perform component separation at the map-making stage, to obtain more precise results. We aim to demonstrate the feasibility of component separation during the map-making stage in time domain space. This new technique leads to a more accurate description of the data and reduces the biases in cosmological analysis. The method uses a library for highly parallel computation which facilitates the programming and permits the description of experiments as easily manipulated operators. These operators can be combined to obtain a joint analysis using several experiments leading to maximized precision. The results show that the method works well and permits end-to-end analysis for the CMB experiments, and in particular, for QUBIC. The method includes astrophysical foregrounds, and also systematic effects like gain variation in the detectors. We developed a software pipeline that produces uncertainties on tensor-to-scalar ratio at the level of $σ(r) \sim 0.023$ using only QUBIC simulated data.

Figures (13)

• Figure 1: Illustration of the Components Map-Making.
• Figure 2: Reconstructed gain at 150 GHz (red) and at 220 GHz (blue) for a single realization of CMB + dust (model d0) sky model.
• Figure 3: Reconstruction of the Q Stokes parameter, in $\mu\K_\text{CMB}$, for both components on the QUBIC patch (centered on $15^{\circ}$ radius sky patch at $\text{RA}=0^{\circ}$, $\text{DEC}=-57^{\circ}$) assuming constant spectral index across the sky. The assumed data are QUBIC $150$ GHz + $220$ GHz + Planck HFI. Each row represents a component, and columns show the input, output, and residuals from left to right.
• Figure 4: Convergence of the spectral index as a function of the number of iterations. Each color shows different noise realizations using different starting points. Note that despite the widely spread initial values, the algorithm converges. The bottom plot shows the residual with respect to the input value of the simulation defined by $\Delta = |\beta_{\text{input}} - \beta_{\text{output}}|$.
• Figure 5: Posterior likelihood on tensor-to-scalar ratio $r$ assuming d0 model.