deep21: a deep learning method for 21 cm foreground removal

Mäkinen, Timo; Lancaster, Lachlan; Villaescusa-Navarro, Francisco; Melchior, P.; Ho, Shirley; Perreault-Levasseur, Laurence; Spergel, David N.

doi:10.1088/1475-7516/2021/04/081

Cited by 48 publications

(40 citation statements)

References 83 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is important to bear in mind that all parameter constraints depend on our ability to mitigate the foregrounds and systematics in future 21-cm experiments [85][86][87]. However, looking at the reasonably good constraints even in the pessimistic foreground contamination case, we are hopeful that future 21-cm experiments will be able to determine the FDM particle mass with good accuracy.…”

Section: Discussionmentioning

confidence: 99%

Exploring delaying and heating effects on the 21-cm signature of fuzzy dark matter

Sarkar,

Flitter,

Kovetz

2022

Preprint

View full text Add to dashboard Cite

In the fuzzy dark matter (FDM) model, dark matter is composed of ultra-light particles with a de Broglie wavelength of ∼kpc, above which it behaves like cold dark matter (CDM). Due to this, FDM suppresses the growth of structure on small scales, which delays the onset of the cosmic dawn (CD) and the subsequent epoch of reionization (EoR). This leaves potential signatures in the sky averaged 21-cm signal (global), as well as in the 21-cm fluctuations, which can be sought for with ongoing and future 21-cm global and intensity mapping experiments. To do so reliably, it is crucial to include effects such as the dark-matter/baryon relative velocity and Lyman-Werner star-formation feedback, which also act as delaying mechanisms, as well as CMB and Lyman-α heating effects, which can significantly change the amplitude and timing of the signal, depending on the strength of X-ray heating sourced by the remnants of the first stars. Here we model the 21-cm signal in FDM cosmologies across CD and EoR using a modified version of the public code 21cmvFAST that accounts for all these additional effects, and is directly interfaced with the Boltzmann code CLASS so that degeneracies between cosmological and astrophysical parameters can be fully explored. We examine the prospects to distinguish between the CDM and FDM models and forecast joint astrophysical, cosmological and FDM parameter constraints achievable with intensity mapping experiments such as HERA and global signal experiments like EDGES. We find that HERA will be able to detect FDM particle masses up to mFDM ∼ 10 −19 eV − 10 −18 eV, depending on foreground assumptions, despite the mitigating effect of the delaying and heating mechanisms included in the analysis.

show abstract

Section: Discussionmentioning

confidence: 99%

Exploring delaying and heating effects on the 21-cm signature of fuzzy dark matter

Sarkar,

Flitter,

Kovetz

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…We also note that PCA represents arguably the most basic form of blind foreground cleaning. Many other more sophisticated methods haven been experimented with (Wolz et al 2014;Shaw et al 2015;Carucci et al 2020;Makinen et al 2021;Fonseca & Liguori 2021;Soares et al 2021b;Irfan & Bull 2021). Comparisons between many of these are presented in Spinelli et al (2021).…”

Section: Discussionmentioning

confidence: 99%

Detecting the power spectrum turnover with HI intensity mapping

Cunnington

2022

Preprint

View full text Add to dashboard Cite

A goal for pathfinder intensity mapping (IM) surveys will be detecting features in the neutral hydrogen (H ) power spectrum, which serve as conclusive evidence of cosmological signals. Observing such features at the expected scales in H IM auto-correlations, where contribution from systematics is uncertain, will provide a more convincing cosmological detection. We demonstrate how the turnover, i.e. the peak of the power spectrum at ultra-large scales, can be detected with H IM. We find that a MeerKAT 4,000 deg 2 survey using the UHF-band is capable of a 3.1𝜎 detection of the turnover, relative to a null model power spectrum with no turnover. This should exceed that capable from current galaxy surveys in optical and near-infrared. The detection significance falls to ∼1𝜎 in MeerKAT's L-band but can reach ∼13𝜎 with the SKAO, which should easily surpass the constraint capable from future Stage-IV-like spectroscopic galaxy surveys. We also propose a new model-independent methodology for constraining the precise turnover scale (𝑘 0 ) and our tests on UHF-band simulated data achieved a precision of 10%. This improved to 2.4% when using the full SKAO. We demonstrate how the results are robust to foreground contamination by using transfer functions, even when an incorrect cosmology has been assumed in their construction. Given that the turnover is related to the horizon scale at matter-radiation equality, a sufficiently precise constraint of 𝑘 0 presents the possibility for a novel probe of cosmology. We therefore present a potential methodology for constructing a standard-ruler-based distance measurement, independent of the sound horizon, using the turnover location in the H power spectrum.

show abstract

“…Machine learning has also been proposed as a way to confront the complexity of 21-cm neutral hydrogen data probing the epoch of reionization, with applications explored for parameter estimation [53][54][55][56][57] and signal extraction [58][59][60].…”

Section: Examples Of Science Cases 21 Cosmic Probesmentioning

confidence: 99%

Machine Learning and Cosmology

Dvorkin,

Mishra-Sharma,

Nord

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Methods based on machine learning have recently made substantial inroads in many corners of cosmology. Through this process, new computational tools, new perspectives on data collection, model development, analysis, and discovery, as well as new communities and educational pathways have emerged. Despite rapid progress, substantial potential at the intersection of cosmology and machine learning remains untapped. In this white paper, we summarize current and ongoing developments relating to the application of machine learning within cosmology and provide a set of recommendations aimed at maximizing the scientific impact of these burgeoning tools over the coming decade through both technical development as well as the fostering of emerging communities.

show abstract

deep21: a deep learning method for 21 cm foreground removal

Cited by 48 publications

References 83 publications

Exploring delaying and heating effects on the 21-cm signature of fuzzy dark matter

Exploring delaying and heating effects on the 21-cm signature of fuzzy dark matter

Detecting the power spectrum turnover with HI intensity mapping

Machine Learning and Cosmology

Contact Info

Product

Resources

About