Filtering in CMB data analysis with application to ACT DR4 and Planck
Erik Rosenberg 1,2,3, Steven Gratton 4,2,3, Anthony Challinor 2,3,4 · 2024
Read Filtering in CMB data analysis with application to ACT DR4 and Planck by Erik Rosenberg 1,2,3, Steven Gratton 4,2,3, Anthony Challinor 2,3,4 (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-13995.txt.
Introduction
Erik Rosenberg, Steven Gratton, and Anthony Challinor chase a mismatch between two maps of the universe's oldest light. A ground telescope's fourth data release and a satellite's final maps disagree in places. The gap is sharpest where temperature and polarization signals cross. The team asks whether data handling, not the sky, is to blame.
They focus on filtering steps and the masks used to hide bright point sources. Filtering stamps cross-shaped artifacts around those sources in the maps. The crosses stretch far past the masks meant to contain them. Left standing, they can bias cross-checks between data sets. After measuring the effect, the authors find it real but too small to explain the mismatch. The discrepancy survives their best suspect, which sharpens the puzzle.
Catalog record
| CREATOR | Erik Rosenberg 1,2,3, Steven Gratton 4,2,3, Anthony Challinor 2,3,4 |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-13995
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Filtering in CMB data analysis with application to ACT DR4 and Planck Erik Rosenberg 1,2,3, Steven Gratton 4,2,3, Anthony Challinor 2,3,4 Abstract Motivated by observed discrepancies between ACT DR4 and Planck 2018 cosmic microwave background (CMB) anisotropy power spectra, particularly in the cross-correlation of temperature and E-mode polarization, we investigate challenges that may be encountered in the comparison of satellite and ground-based CMB data. In particular, we focus on the effects of Fourier-space filtering and masking involving bright point sources. We show that the filtering operation generates bright cross-shaped artifacts in the map, which stretch far outside typical point-source masks. If not corrected, these artifacts can add bias or additional variance to cross-spectra, skewing results. However we find that the effect of this systematic is not large enough to explain the ACT- Planck differences presented with ACT DR4. I Introduction With the end of operations of the Planck satellite ( Planck Collaboration I 2014 ; Planck Collaboration I 2016 ; Planck Collaboration I 2020 ) in 2013 and its final data release in 2020 ( Planck Collaboration Int. LVII 2020 ) , it is clear that the next decade of CMB investigation will be conducted from the ground. Ongoing experiments like the Atacama Cosmology Telescope (ACT; Swetz et al. 2011 ; Sievers et al. 2013 ; Thornton et al. 2016 ; Louis et al. 2017 ; Mallaby-Kay et al. 2021 ) and the South Pole Telescope (SPT; Carlstrom et al. 2011 ; Reichardt et al. 2012 ; Sobrin et al. 2022 ; Dutcher et al. 2021 ; Balkenhol et al. 2023 ) have lower polarization noise and operate at higher resolution than Planck . These, and the forthcoming ground-based experiment Simons Observatory (SO; Ade et al. 2019 ) and the planned next-generation CMB-S4 ( Abazajian et al. 2019 ) , will significantly improve measurements of the primary CMB polarization as well as the temperature damping tail and small-scale secondary effects such as CMB lensing and the Sunyaev–Zel’dovich effects. However, all of these ground-based experiments are limited in their ability to measure large scales due to atmospheric noise. The atmosphere acts a source of strongly scale-dependent noise, with the noise power spectrum N_{\ell} rapidly increasing with multipole \ell as N_{\ell}\propto\ell^{-3} below a given ‘knee frequency’; in ACT intensity, for example, \ell_{\mathrm{knee}}\sim 2000 and 3000 at 98 GHz and 150 GHz, respectively ( Naess et al. 2020 ) . This noise is worse for higher frequencies, and for intensity measurements relative to polarization. This motivates in part the scale cuts implemented by ACT in Louis et al. 2017 [ \ell>500 (TT), \ell>350 (TE/EE)] and later in Choi et al. 2020 [ \ell>600 (TT), \ell>350 (TE/EE)]. To avoid losing information from the first two CMB peaks while remaining independent of Planck , ACT Data Release 4 (DR4) was therefore combined with WMAP data ( Bennett et al. 2013 ) at large and intermediate ( 2<\ell<1200 in TT and 24<\ell<800 in TE) scales for the baseline likelihood analysis in Aiola et al. 2020 . Similarly, baseline SO forecasts include Planck on large scales and areas of the sky not observed by SO. Higher Planck frequencies not accessible from the ground, such as 353 GHz, are also expected to be useful for constraining foreground properties like the spatial variation of the spectral index of Galactic dust ( Ade et al. 2019 ) . We anticipate that space-based measurements from Planck and WMAP , and in the future LiteBIRD ( Hazumi et al. 2020 ; LiteBIRD Collaboration et al. 2023 ) , will continue to be used in combination with ground-based data for the foreseeable future. It is therefore necessary to study in detail potential problems that may arise in combining these substantially different datasets. In this paper we investigate challenges in the combination of Planck and ACT data, drawing conclusions that are more widely applicable to other experiments as well. The principal artifacts studied here are created by strong point sources subject to a Fourier-space filter, applied to remove contamination from the ground. This effect is relatively well-known, mentioned in passing and accounted for in recent ACT and SPT papers, though not in all cases. One of our aims in this article is to highlight this effect as an example case of the care that should be used when applying such filters. We present the particular context of this study in Sec. II with an introduction to ACT DR4. In Sec. III we focus in particular on the filtering step of that analysis. In this section we also begin to explore the interaction of this filtering procedure with bright point sources and derive analytic results for the resulting artifacts and their impact on the power spectrum. We see in Sec. IV the impact this effect has on power spectra calculated from real data. Finally, some strategies for mitigation are discussed in Sec. V , and we conclude in Sec. VI . II ACT DR4 In this section we place ACT DR4 into context and review the aspects of the DR4 maps and power spectra of most relevance to this study. The early data releases from ACT consisted of temperature-only data from the 2008–2010 seasons ( Das et al. 2011 ; Das et al. 2014 ; Sievers et al. 2013 ) using the MBAC instrument ( Swetz et al. 2008 ; Swetz et al. 2011 ) . MBAC was later replaced by ACTPol ( Thornton et al. 2016 ) , measuring temperature and polarization using two arrays operating at 150 GHz and one dichroic array at both 98 GHz and 150 GHz. Data from the first ACTPol seasons 2013–14 have been previously released and analysed in Naess et al. 2014 and Louis et al. 2017 . Here we focus on ACT DR4, the 2020 release consisting of 2013–2016 ACTPol night-time data as described in Choi et al. 2020 , henceforth 12 , and Aiola et al. 2020 . As in the 12 comparison to Planck , we focus here on the patches of sky comprising most of the DR4 signal-to-noise, ‘BOSS-North’ (BN) and ‘Deep 56’ (D56), illustrated in Fig. 1 . These are 1837 and 565\,\mathrm{deg}^{2} patches with effective areas of 1400 and 340\,\mathrm{deg}^{2} used for power spectra after inverse-noise-variance weighting, and respective noise levels of 33–67 and 17–27 \,\mu\mathrm{K} -arcmin depending on array and frequency (see 12 Table 2). Figure 1: ACT DR4 sky coverage in equatorial coordinates. The right ascension (horizontal) and declination (vertical) are given in degrees. The colour scale shows map depth and the background is Planck 353 GHz intensity. In this work we focus on the BOSS and Deep 56 regions. Figure from Choi et al. 2020 , reproduced with permission. II.1 Ground pickup and filtering ACT data is contaminated by substantial emission from the ground, with an amplitude of up to \pm 20\,\mu\mathrm{K} in the polarization maps ( Naess et al. 2014 ) . Because ACT scans the sky horizontally at constant elevation this scan-synchronous ground pickup is projected into the maps as stripes of nearly-constant declination. This signal is partially removed in the map-making step, as detailed in Aiola et al. 2020 . Single-detector ‘azimuth pickup maps’ that are a function of azimuth only are produced as ground templates.