Mitigating point-source contamination in CMB polarization: a Generalized Point Spread Function fitting approach

Yiming Wang, Wen-Zheng Chen, Yang Liu, Siyu Li, Hong LiJournal of Cosmology and Astroparticle Physics 2026 · DOI

TL;DR

This paper introduces Generalized Point Spread Function Fitting (GPSF), a novel method to remove contamination from extragalactic point sources in Cosmic Microwave Background (CMB) polarization maps. GPSF significantly reduces biases in measurements of the tensor-to-scalar ratio ($r$) – a key parameter for understanding cosmic inflation – while minimally impacting the CMB signal and its statistical properties.

What Problem Does This Paper Address?

Detecting the faint B-mode polarization signal from primordial gravitational waves in the CMB is a primary goal of modern cosmology. However, this signal is easily contaminated by various astrophysical sources. While diffuse foregrounds (like Galactic synchrotron and dust emission) have been extensively studied and mitigated, extragalactic point sources are emerging as a significant challenge. These sources can mimic the primordial B-mode signal, leading to erroneous estimates of $r$, especially for future experiments aiming for $r \sim 10^{-3}$. Traditional methods like masking and inpainting have limitations, particularly for small-aperture telescopes with large beams where sources can blend. This paper addresses the need for a more robust and accurate method to mitigate point-source contamination.

Methods

The core of the paper is the Generalized Point Spread Function (GPSF) fitting method. GPSF models point sources using their known positions and a parameterized beam-convolved template. What makes GPSF novel is its use of the full pixel-domain covariance matrix, including off-diagonal terms, which accounts for correlations between pixels due to the CMB and diffuse foregrounds. It also explicitly models overlapping sources, a crucial feature for crowded sky regions and large telescope beams.

The fitting process involves minimizing a $\chi^2$ statistic within a defined fitting region. For polarization maps (Q and U), it uses a standard co-polar approximation and accounts for the coordinate-dependent nature of these Stokes parameters. The method is extended to multi-source fitting to handle blended sources simultaneously.

The analysis pipeline involves:

  1. Point-source mitigation using GPSF.
  2. Map preprocessing: smoothing to a common resolution and correcting for EB leakage.
  3. Foreground cleaning using the Needlet Internal Linear Combination (NILC) method.
  4. Power spectrum estimation and likelihood analysis to constrain $r$.

The authors also compare GPSF against standard masking and inpainting techniques.

Datasets

The paper uses simulated sky maps for evaluation. These simulations incorporate:

  • Cosmic Microwave Background (CMB) signal.
  • Diffuse foreground emissions (synchrotron and thermal dust) from the Planck Sky Model (PSM).
  • Extragalactic radio and far-infrared sources, modeled using data from various surveys (NVSS, SUMSS, GB6, PMN, IRAS).
  • Instrumental noise, emulating specifications of a future ground-based SAT CMB experiment.

No new observational datasets are introduced. The simulations are designed to mimic realistic conditions for future CMB experiments.

Key Insights and Ideas

  • Full Covariance is Crucial: Accounting for the full pixel-domain covariance, including off-diagonal terms, is essential for accurately modeling the background and its correlations with point sources. This is a key improvement over simpler methods.
  • Handling Overlapping Sources: Explicitly modeling multiple, potentially blended, sources simultaneously is vital for accurate flux estimation, especially in regions with high source density or when using large telescope beams.
  • GPSF Balances Bias and Variance: GPSF achieves a remarkable balance between reducing point-source bias and maintaining a low increase in the variance of the CMB signal, outperforming traditional methods like masking and inpainting.
  • Impact on $r$ Constraint is Significant: The reduction in bias for the tensor-to-scalar ratio ($r$) achieved by GPSF is substantial, making it a promising tool for future experiments aiming for high precision measurements.

Key Results

  • Map Level: GPSF effectively suppresses point-source residuals while preserving the CMB signal, producing maps that closely resemble the "No PS" (no point sources) case. Inpainting introduces noticeable distortions, and masking reduces usable sky area.
  • Power Spectrum Level: GPSF yields lower residuals in the B-mode power spectra compared to the unmitigated case, particularly at low and high multipoles. It maintains standard deviations comparable to the ideal case.
  • Tensor-to-Scalar Ratio ($r$) Constraint:
    • The presence of point sources introduces a bias of $1.67 \times 10^{-3}$ in $r$.
  • GPSF reduces this bias to $2.9 \times 10^{-4}$ (an 82% reduction) with only a 2% increase in standard deviation.
  • Inpainting yields higher bias and standard deviation than the "No PS" case.
  • Masking reduces bias but significantly increases standard deviation due to sky area reduction.

Limitations and Future Work

  • Computational Cost: While GPSF is effective, it is computationally more demanding than masking. The authors note that the dominant cost is associated with the pixel covariance matrix, but suggest optimizations like precomputing templates.
  • Atmospheric Filtering Impact: The paper discusses the impact of atmospheric filtering on point-source recovery, showing that certain filtering methods (like masking the source region during filtering) are crucial for preserving GPSF's effectiveness.
  • Future Work: The authors suggest that GPSF's ability to suppress residuals at high multipoles could be beneficial for CMB lensing reconstruction, which is sensitive to small-scale contamination. A detailed investigation into this application is left for future work.

Summary updated

Abstract

Abstract Observations of Cosmic Microwave Background (CMB) B-mode polarization provide a way to probe primordial gravitational waves and test inflationary predictions. Extragalactic point sources become a major source of contamination after foreground cleaning and can bias estimates of the tensor-to-scalar ratio r at the 10 -3 level. We introduce Generalized Point Spread Function Fitting (GPSF), a method for removing point-source contamination in polarization maps. GPSF uses the full pixel-domain covariance, including off-diagonal terms, and models overlapping sources. This allows accurate flux estimation under realistic conditions, particularly for small-aperture telescopes with large beams that are more susceptible to source blending. We test GPSF on simulated sky maps, apply foreground cleaning using the Needlet Internal Linear Combination (NILC) method, and compare its performance with standard masking and inpainting. The results show GPSF reduces point-source contamination without significantly affecting the background signal, as seen in both the maps and their power spectra. For the constraint on r , GPSF reduces the bias from 1.67 × 10 -3 to 2.9 × 10 -4 , with only a 2% increase in standard deviation. Compared to inpainting and masking, GPSF yields lower bias while maintaining comparable variance. This suggests that it may serve as a promising method for future CMB experiments targeting measurements of r ∼ 10 -3 .
BibTeX
@article{wang2026mitigating,
  author = {Yiming Wang and Wen-Zheng Chen and Yang Liu and Siyu Li and Hong Li},
  title = {{Mitigating point-source contamination in CMB polarization: a Generalized Point Spread Function fitting approach}},
  journal = {Journal of Cosmology and Astroparticle Physics},
  year = {2026},
  doi = {10.1088/1475-7516/2026/03/042},
}
Figure