Removing Point Sources from Cosmic Microwave Background Maps

Max Tegmark, A. de Oliveira‐CostaThe Astrophysical Journal 1998 · DOI

Abstract

For high-precision cosmic microwave background (CMB) experiments, contamination from extragalactic point sources is a major concern. It is therefore useful to be able to detect and discard point source contaminated pixels using the map itself. We show that the sensitivity with which this can be done can often be greatly improved (by factors between 2.5 and 18 for the upcoming Planck mission) by a customized hi-pass filtering that suppresses fluctuations due to CMB and diffuse galactic foregrounds. This means that point source contamination will not severely degrade the cleanest Planck channels unless current source count estimates are off by more than an order of magnitude. A catalog of around 40,000 far infra-red sources at 857 GHz may be a useful by-product of Planck.

Introduction

To extract precise cosmological parameters from the CMB, researchers must first "clean" the maps of foreground contamination. While diffuse foregrounds like galactic dust and synchrotron radiation are well-studied, extragalactic point sources (like distant galaxies) present a unique challenge. Their spectral indices vary wildly, they can be time-variable, and their abundance at CMB frequencies was poorly constrained at the time of publication.

The standard approach involved discarding pixels that deviate by more than $5\sigma$ from the mean. However, in high-sensitivity experiments, the "noise" ($\sigma$) is dominated by the CMB fluctuations themselves rather than detector electronics. This makes it difficult to distinguish a faint point source from a natural "hot spot" in the early universe.

Methodology: The Matched-Filter Approach

The authors propose a customized bandpass filter ($W$) designed to maximize the Signal-to-Noise Ratio (S/N) of point sources. Instead of analyzing the raw map, the data is convolved with this filter in the spherical harmonic (multipole) domain.

The filter $W$ is derived to be inversely proportional to the total power spectrum, which includes the CMB, detector noise, and diffuse foregrounds. By "zeroing out" the scales where the CMB and galactic dust are most prominent, the filter creates a "cleaned" map where point sources stand out against a suppressed background.

The power spectrum of various foregrounds and the CMB, showing how the filter $W_l^2$ targets the region where point sources are most detectable relative to the total noise.

Once the map is filtered, pixels exceeding a threshold (e.g., $5\sigma$) are identified. These contaminated pixels, along with a small buffer, are then masked in the original map before performing cosmological analysis. This method decouples point source detection from underlying CMB fluctuations, reducing selection bias where sources sitting on CMB "hot spots" were previously more likely to be detected.

Key Results and Sensitivity Gains

The filtering technique demonstrates significant improvements in sensitivity across various mission specifications, particularly for the Planck satellite. Unlike the MAP (later WMAP) mission, which was limited primarily by detector noise, Planck was limited by "confusion noise" from the CMB itself—the specific problem this filter addresses.

Performance metrics for the Planck mission channels, comparing standard detection (B) with the optimized filtering method (A). Note the significant GAIN factors and increased source counts $N(>5\sigma)$.

As shown in the performance table, the filtering method (Method A) provides sensitivity gains ranging from a factor of 2.5 to 18 compared to the standard method (Method B). In the 857 GHz channel, the number of detectable sources increases from approximately 300 to 40,000. Furthermore, the authors define a "safety margin" ($M$), representing the factor by which actual source counts would need to exceed current models to degrade the CMB data. For Planck, these margins are comfortably large, suggesting high robustness.

Conclusion

The introduction of optimized high-pass filtering significantly enhances the ability to identify and remove point source contamination in high-resolution CMB maps. While the results depend on the accuracy of current source count models (such as To98), the large safety margins provide confidence for the Planck mission. Future work may involve generalizing the method for asymmetric telescope beams and extending the spatial filtering to multi-frequency subtraction techniques to further isolate the primordial CMB signal.

Summary updated

Abstract

For high-precision cosmic microwave background (CMB) experiments, contamination from extragalactic point sources is a major concern. It is therefore useful to be able to detect and discard point source contaminated pixels using the map itself. We show that the sensitivity with which this can be done can often be greatly improved (by factors between 2.5 and 18 for the upcoming Planck mission) by a customized hi-pass filtering that suppresses fluctuations due to CMB and diffuse galactic foregrounds. This means that point source contamination will not severely degrade the cleanest Planck channels unless current source count estimates are off by more than an order of magnitude. A catalog of around 40,000 far infra-red sources at 857 GHz may be a useful by-product of Planck.
BibTeX
@article{tegmark1998removing,
  author = {Max Tegmark and A. de Oliveira‐Costa},
  title = {{Removing Point Sources from Cosmic Microwave Background Maps}},
  journal = {The Astrophysical Journal},
  year = {1998},
  doi = {10.1086/311410},
}

Loading related content...

Figure