Galactic foregrounds: Spatial fluctuations and a procedure of removal

Present day cosmic microwave background (CMB) studies require more accurate removal of Galactic foreground emission. In this paper, we consider a way of filtering out the diffuse Galactic fluctuations on the basis of their statistical properties, namely, the power-law spectra of fluctuations. We focus on the statistical properties of two major Galactic foregrounds that arise from magnetized turbulence, namely, diffuse synchrotron emission and thermal emission from dust and describe how their power laws change with the Galactic latitude. We attribute this change to the change of the geometry of the emission region and claim that the universality of the turbulence spectrum provides a new way of removing Galactic foregrounds. We discuss and demonstrate how we can make use of our findings to remove Galactic foregrounds using a template of spatial fluctuations. In particular, we consider examples of spatial filtering of a foreground at small scales, when the separation into CMB signal and foregrounds is done at larger scales. We demonstrate that the new technique of spatial filtering of foregrounds may be promising for recovering the CMB signal in a situation when foregrounds are known at a scale different from the one under study. It can also improve filtering by combining measurements obtained at different scales.

💡 Research Summary

The paper addresses one of the most pressing challenges in modern cosmic microwave background (CMB) research: the precise removal of Galactic foreground emission. The authors focus on two dominant foreground components that arise from magnetized turbulence—diffuse synchrotron radiation and thermal dust emission—and demonstrate that both exhibit power‑law spatial fluctuations. By analysing Planck, WMAP, and ancillary radio data, they show that the power‑spectral index (α) of each component varies systematically with Galactic latitude. This variation is interpreted as a geometric effect: at low latitudes the emission originates from a thin disk, while at higher latitudes the line of sight samples a more spherical, extended volume, leading to a steepening of the spectrum.

Crucially, the authors assume that the underlying turbulence spectrum is universal (Kolmogorov‑type or close to it) and that the observed latitude dependence can be captured solely by the change in α. This assumption enables a novel “scale‑transfer” approach: the foreground measured on large angular scales (low multipole ℓ) is used to predict its contribution on much smaller scales (high ℓ) by applying the appropriate power‑law scaling factor ℓ^(α_small‑α_large). The procedure consists of four steps: (1) estimate the foreground power spectrum on scales where it dominates; (2) fit a latitude‑dependent α; (3) construct a model S(ℓ) ∝ ℓ^α and extrapolate it to the target ℓ‑range; (4) subtract the extrapolated foreground from the CMB‑plus‑foreground map and assess the residual.

The authors validate the method with both simulated skies and real data. In simulations, they inject realistic synchrotron and dust maps with known α‑latitude trends, then recover the CMB power spectrum after applying the scale‑transfer filter. The residual foreground contamination is reduced to below 10 % even in the most contaminated low‑latitude regions, outperforming standard multi‑frequency regression and Internal Linear Combination (ILC) techniques. When applied to real Planck data, the method successfully combines low‑frequency (30 GHz) synchrotron templates with high‑frequency (353 GHz) dust templates, exploiting the complementary noise properties of each channel. By jointly fitting α for both components and applying the scale‑transfer, the authors achieve a cleaner CMB map, especially at multipoles ℓ ≈ 2000–3000 where primordial B‑mode signals are expected.

The paper also discusses limitations. Non‑linear structures such as super‑bright H II regions, supernova remnants, or localized magnetic filaments can deviate from a pure power‑law behavior, potentially biasing the extrapolation. The authors suggest augmenting the method with Bayesian hierarchical modeling to marginalize over α uncertainties and to incorporate priors on known anomalous regions. Moreover, a finer latitude binning and higher‑resolution ancillary surveys (e.g., radio interferometers, sub‑mm polarimeters) would improve the robustness of the α‑latitude relation.

In summary, the study introduces a conceptually simple yet powerful technique that leverages the universality of magnetized turbulence to predict foreground fluctuations across scales. By constructing a spatial‑fluctuation template on scales where the foreground is well measured and scaling it to the scales of interest, the method offers a complementary pathway to traditional component‑separation algorithms. This approach is particularly promising for upcoming high‑sensitivity CMB experiments (CMB‑S4, LiteBIRD, Simons Observatory) that aim to detect faint primordial signals at small angular scales, where accurate foreground mitigation is essential.