CMB component-separated power spectrum estimation by Spectral Internal Linear Combination (SpILC)

Jack Y. L. Kwok, William R. Coulton, Niall MacCrann, Fiona McCarthy, Blake D. Sherwin, Boris Bolliet · 2025

Read CMB component-separated power spectrum estimation by Spectral Internal Linear Combination (SpILC) by Jack Y. L. Kwok, William R. Coulton, Niall MacCrann, Fiona McCarthy, Blake D. Sherwin, Boris Bolliet (2025) 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-2509-10604.txt.

Introduction

Jack Y. L. Kwok and five co-authors clean the cosmic microwave background in a new way in this 2025 cosmology paper. Sky maps at different frequencies mix the ancient CMB light with galactic and extragalactic foregrounds. Standard internal linear combination methods blend the maps themselves, then measure the power spectrum of the result.

Their Spectral ILC instead combines the auto- and cross-spectra of the channels directly. That rearrangement yields a different optimization problem needing fewer constraint equations than map-level deprojection. In simulations modeled on the Simons Observatory, the spectral method's error bars shrink by more than a factor of seven at small scales compared with constrained map-level cleaning. The gain is largest where components are spatially uncorrelated and noise dominates, and it questions whether cleaned maps are the right route to precision measurements.

Catalog record

CREATORJack Y. L. Kwok, William R. Coulton, Niall MacCrann, Fiona McCarthy, Blake D. Sherwin, Boris Bolliet
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE74 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-10604

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

CMB component-separated power spectrum estimation by Spectral Internal Linear Combination (SpILC) Jack Y. L. Kwok, William R. Coulton, Niall MacCrann, Fiona McCarthy, Blake D. Sherwin, Boris Bolliet Abstract Component separation methods mitigate the cross-contamination between different extragalactic and galactic contributions to cosmic microwave background (CMB) data. This is often done by linearly combining CMB maps from different frequency channels using internal linear combination (ILC) methods. We demonstrate that deriving power spectrum estimators directly by linearly combining auto- and cross-spectra instead of maps allows us to obtain a different constrained-optimization problem that allows fewer (deprojection) constraint equations than combining at map level using the constrained ILC method. Through simulations, we show that our Spectral internal linear combination (SpILC) produces CMB power spectrum estimators with more than 7 times smaller errorbars than constrained ILC (with thermal Sunyaev-Zel’dovich and cosmic infrared background deprojections) at ℓ ≳ 4000 \ell\gtrsim 4000 for Simons Observatory-like observations. Spectral ILC outperforms constrained ILC methods when some modeled components are spatially uncorrelated, e.g. the primary CMB is uncorrelated with foregrounds, and the difference in performance is most significant at noise-dominated scales. More generally, our work shows that component-separated maps with foreground deprojections do not necessarily produce minimum-variance two-or-higher-point estimators. I Introduction Over the past decade, ground-based cosmic microwave background (CMB) experiments e.g. Atacama Cosmology Telescope (ACT) Henderson et al. 2015 ; Aiola et al. 2020 and South Pole Telescope (SPT) Carlstrom et al. 2011 ; Schaffer et al. 2011 ; Benson et al. 2014 have built upon space-based observations from Planck Planck Collaboration et al. 2020a , pushing the resolution of microwave observations to arcminute scales. This opens up the study of CMB secondary anisotropies —late-time perturbations to the primary CMB—including weak lensing of CMB temperature and polarization Smith et al. 2007 ; Ade et al. 2014 ; Planck Collaboration et al. 2020b ; Madhavacheril et al. 2024 ; Omori et al. 2023 , the thermal (tSZ) Sunyaev and Zeldovich 1972 ; Sunyaev and Zeldovich 1980 ; Planck Collaboration et al. 2016a ; Coulton et al. 2024a ; Hilton et al. 2018 ; Bleem et al. 2015 and kinetic Sunyaev-Zel’dovich (kSZ) effect Sunyaev and Zeldovich 1972 ; Sunyaev and Zeldovich 1980 ; Hand et al. 2012 ; Schaan et al. 2021 ; Reichardt et al. 2021 ; MacCrann et al. 2024a ; Raghunathan et al. 2024 , and the patchy screening effect Dvorkin and Smith 2009 ; Smith and Ferraro 2017 ; Coulton et al. 2024b . At the advent of the ground-based CMB experiment Simons Observatory (SO) Ade et al. 2019 , complemented by galaxy surveys including the Vera Rubin Observatory LSST Science Collaboration et al. 2009 , the Euclid Laureijs et al. 2011 and SPHEREx Crill et al. 2020 space telescopes, the Dark Energy Survey (DES) The Dark Energy Survey Collaboration 2005 and the Dark Energy Spectroscopic Instrument (DESI) DESI Collaboration et al. 2016 , the millimeter sky will continue to improve our understanding of the primordial universe and the growth of cosmic structures. In order to study the wealth of signals in intensity and polarization maps across frequency channels, one is motivated to isolate the signals from each other. This includes separating extragalactic signals (e.g. CMB, tSZ, kSZ, cosmic infrared background, radio point sources) from each other and from galactic (e.g. dust, synchrotron) emissions. Various component separation methods are devised for this purpose, whereby the different frequency dependencies ( spectral energy distributions , or SEDs) and spatial properties of components are exploited to produce a map of the desired component. These methods can be broadly separated into parametric and “blind” methods. Parameteric methods include Commander Eriksen et al. 2006 ; Eriksen et al. 2008 ; Planck Collaboration et al. 2016b , which samples the joint posterior distribution of spectral parameters and component maps using a Bayesian approach. “Blind” methods are model agnostic except for the component of interest. Examples are SEVEM Martínez-González et al. 2003 ; Leach et al. 2008 ; Fernández-Cobos et al. 2012 , which subtracts off a linear combination of internally-constructed foreground templates from a map to form a minimum-variance CMB map, and Internal Linear Combination (ILC) methods Eriksen et al. 2004 ; Tegmark et al. 2003 ; Delabrouille et al. 2009a ; McCarthy and Hill 2024a , which also assume that the signal SED is known, and again seek a minimum-variance component map by a linear combination of maps across frequency channels. The “semi-blind” constrained ILC (cILC) Remazeilles et al. 2011a method additionally assumes the SEDs of a few modeled foreground components. This paper focuses on the ILC methods (particularly the constrained ILC)—extensively used in WMAP Bennett et al. 2003 ; Delabrouille et al. 2009b , Planck Planck Collaboration et al. 2014 ; Planck Collaboration et al. 2016c ; Planck Collaboration et al. 2020c and ACT Madhavacheril et al. 2020 ; Coulton et al. 2024a —which have a minimal set of core assumptions: 1. known SEDs for the components to be modeled, 2. no spatial correlation between the modeled components and noise (where everything except the modelled signals are called noise), and 3. zero spatial decorrelation , meaning observed maps across frequency channels have the same underlying component maps. For example, consider a data model with two modeled components s_{p} and y_{p} , with other uncorrelated components and instrumental noise grouped into the noise term n_{p} . In equation form the data model would look as follows: \displaystyle x^{i}_{p} \displaystyle=a^{i}s_{p}+b^{i}y_{p}+n^{i}_{p}\;, (1) where x^{i}_{p} is the observed map at frequency channel i , s_{p} is the component map one wishes to recover, y_{p} is a modeled component map one wishes to explicitly remove, and n^{i}_{p} are noise maps at channel i . The label p denotes pixels, spherical harmonics or needlets for pixel-space, harmonic or needlet ILC, respectively. The SEDs a^{i} and b^{i} are assumed to be known exactly. 1 1 1 In the case of the CIB where the SED is not exactly known, the moment ILC method Chluba et al. 2017 is found to be an effective mitigation McCarthy and Hill 2024b (discussion in Sec. VI ). Under these core assumptions, the cILC method Remazeilles et al. 2011b ; Chen and Wright 2009 selects a set of weights w^{i} to linearly combine the observed maps x^{i}_{p} to form minimum-variance ILC map \hat{s}_{p}\equiv\sum_{i}w_{i}x^{i}_{p} , under the constraints that the maps have unit response to the component of interest, \sum_{i}w_{i}a^{i}=1 , and zero response to all other modeled components, \sum_{i}w_{i}b^{i}=0 ; this latter constraint is also called deprojection . The focus of this study is on extragalactic signals, which in many cases are deprojected rather than minimized together with other sources of noise since residuals may bias further analyses using the component-separated map or spectra Coulton et al. 2024a , whereas instrumental noise bias can be mitigated using data splits Hill and Spergel 2014 . While galactic foregrounds fall at small scales and can be mitigated by masking close to the galactic plane MacCrann et al. 2024b , their biases after mitigation e.g. on lensing measurements for SO need to be carefully studied Abril-Cabezas et al. 2025 . Our work intends to address the following question: what is the unbiased minimum-variance estimator for the foreground-cleaned power spectrum? Here we develop analytical solutions and

Read the complete text