Expanded Generalized Needlet Internal Linear Combination (eGNILC) Framework for the 21-cm Foreground Removal

(2) Stellenbosch University); Wei-Ming Dai (1); Yin-Zhe Ma (2) ((1) Ningbo University

REVIEW 3 major objections 4 minor 69 references

Expanded Generalized Needlet Internal Linear Combination (eGNILC) Framework for the 21-cm Foreground Removal

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read eGNILC recovers the 21-cm power spectrum in simulated SKA-MID and BINGO skies with 10-20 percent power loss.

desk verdict A credible incremental extension of GNILC for 21-cm foreground cleaning, but the headline power-loss numbers rest on an unvalidated model-selection step. read the letter →

arxiv 2411.16899 v2 pith:G2BTKNHH submitted 2024-11-25 astro-ph.CO astro-ph.IM

Wei-Ming Dai (1) , Yin-Zhe Ma (2) ((1) Ningbo University , (2) Stellenbosch University) This is my paper · ORCID

classification astro-ph.COastro-ph.IM

keywords 21-cmintensitymappingforegroundremovalGNILCneedletinternallinearcombinationdiscretecosinetransformAkaikeinformationcriterionrobustprincipalcomponentanalysisSKA-MID

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper expands the Generalized Needlet Internal Linear Combination method into eGNILC for 21-cm intensity mapping, where foregrounds outshine the signal by roughly four orders of magnitude. The key moves are a discrete cosine transform along the frequency axis, a bias-corrected Akaike criterion for choosing the foreground degrees of freedom, and an optional robust principal component analysis step that makes the method blind. In noiseless SKA-MID-like and BINGO-like simulations with a varying Airy beam, eGNILC recovers the 21-cm auto- and cross-power spectra with about 10-20 percent power loss over multipoles $\ell \in [20,250]$ and $\ell \in [20,300]$, respectively. The paper also derives an eGNILC bias that depends on the averaging domain size and foreground degrees of freedom rather than on the 21-cm signal itself, and shows that the frequency transform reduces power loss specifically at low multipoles. A reader should care because successful foreground removal is the main obstacle to measuring the low-redshift 21-cm power spectrum with single-dish telescopes.

What carries the argument

The central machinery is the needlet-space internal linear combination with mixing matrix $S = $R_s^{{1/2}}$U_s$, where $R_s$ is the 21-cm signal covariance and $U_s$ is the signal subspace obtained from the whitened data covariance. The discrete cosine transform of the frequency axis makes the foreground contribution low-rank; the modified AIC, $\mathrm{AIC}(m, N_p) = 2m + n_{ch}\left[1/(1-2m/N_p) + \ln(1 - 2m/N_p)\right] + \sum_{i=1}^{n_{ch}-m}[\mu_i - \ln\mu_i - 1]$, selects the foreground degrees of freedom $m$; and robust principal component analysis splits the data covariance into a low-rank foreground part plus a sparse signal part to supply a blind estimate of the mixing matrix.

What would settle it

Run eGNILC on the same SKA-MID mock twice, once with the heuristic $N_p$ multiplied by 0.5 and once by 2; if the recovered power spectrum moves by more than the quoted 10-20 percent, the result depends on an unvalidated $N_p$ and the AIC step is the fragile link.

Watch

Extended reading notes

Core claim

eGNILC performs the internal linear combination in needlet space but first transforms the frequency axis with a discrete cosine transform, so that spectrally smooth foregrounds occupy only a few low-order modes. The mixing matrix is built from the 21-cm signal subspace, and the number of foreground modes $m$ is selected by a modified Akaike Information Criterion that includes the eGNILC bias $1 - 2m/N_p$, where $N_p$ is the effective number of independent pixels in the covariance-estimation domain. In simulations, the recovered 21-cm auto- and cross-power spectra match the input to within roughly 20 percent for SKA-MID and 10 percent for BINGO over the quoted multipole ranges when no instrumental noise is added; the frequency-dependent Airy beam causes serious errors only at large multipoles. The paper additionally shows that when adjacent frequency channels are highly correlated, the algorithm must be applied to decimated subsets of channels, and that the eGNILC bias is negligible for simple power-law foregrounds outside the Galactic plane when no beam is present.

Load-bearing premise

The recovery claim rests on the AIC correctly identifying the foreground degrees of freedom $m$, which requires knowing the effective number of independent pixels $N_p$; the paper sets $N_p$ through a heuristic beam-based estimate and never validates it against the original AIC.

Editorial extensions

If this is right

For simple power-law foregrounds with no beam, eGNILC leaves the 21-cm power spectrum essentially unbiased outside the Galactic plane.
With a realistic frequency-dependent Airy beam and no thermal noise, eGNILC recovers SKA-MID power spectra within about 10-20 percent over multipoles $\ell \in [20,250]$ and BINGO spectra within about 10 percent over $\ell \in [20,300]$.
Applying the discrete cosine transform along frequency reduces the power loss at low multipoles, where plain GNILC lacks accuracy.
When adjacent frequency channels are highly correlated, eGNILC must be run on decimated uncorrelated subsets; BINGO-like channel spacing avoids this need.
The usable multipole range is limited at low $\ell$ by the number of independent samples available for covariance estimation and at high $\ell$ by the frequency-varying beam.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper leaves thermal and $1/f$ noise untreated, so a natural extension is to add instrumental noise to the SKA-MID and BINGO mocks and test whether the RPCA sparse component absorbs the noise as the text suggests.
Because high adjacent-channel correlation forced the SKA-MID analysis to split 500 channels into 10 subsets, eGNILC's performance at very fine frequency resolution will depend on how that grouping is chosen; a formal criterion for choosing the number of subsets would settle this dependence.
The modified AIC is never compared with the original AIC on the same mocks; an explicit comparison would show whether the $N_p$ term actually changes the selected foreground degrees of freedom and the recovered spectrum.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Worth a look if you work on 21-cm intensity mapping foreground cleaning. The genuinely new pieces are the DCT-before-needlets step, the bias-corrected AIC in Eq. (48), and the RPCA-embedded mixing-matrix estimation. On reading, the ILC math is standard, the bias derivation in Appendix B is internally consistent, and the demonstration plots do show the DCT step helping at low multipoles. Credit where it is due: the paper tests on SKA-MID-like and BINGO-like mocks with varying Airy beams and masks, reports auto and cross spectra, and explicitly warns that the modified AIC should not be combined with RPCA. That is honest and keeps the headline experiments tied to the tested pipeline.

The soft spots are real but not fatal. The main one is that Eq. (48) depends on Np, the effective independent pixel count, which is never concretely defined. The paper gestures at an effective theta_FWHM and says the method is insensitive to the domain shape, but it shows no sensitivity test for Np and never compares the modified AIC with the original AIC from Olivari et al. (2016) on the same simulations. Since m sets the ILC constraint, any error in Np moves the recovered spectra, and the force of the 10-20% power-loss claims depends on this step not being tuned. Second, the abstract's "less than 20%" claim is contradicted by the 21.3% entry in Table 2 (SKA-MID 1069.7 MHz, Extrapol+Unresolved, B=2.5). That is a small inconsistency, but it undercuts the headline and should be fixed. Third, the recovered spectra are from single realizations; there are no error bars on Table 2, so the differences between B=2.5 and B=1.7 are hard to judge. Lack of code is a practical hindrance for a methods paper, and using CORA-generated signal covariances to clean CORA-generated signal maps makes the validation partly self-referential. That last point is common in this literature, but it should be stated.

Bottom line: this is a competent, incremental methods paper. The central idea is not flawed, but the headline claims are conditional on the unvalidated AIC/Np step. A serious referee would want Np specified or tested, a direct AIC comparison, error bars on the power-loss numbers, and ideally code. The likely readers are people building single-dish 21-cm pipelines for BINGO, MeerKAT, or SKA-MID.

Recommendation: yes, send it to peer review. It is the kind of paper that improves with referee input and should be in the literature, but not with the current over-broad abstract.

Referee Report

3 major / 4 minor

Summary. The paper presents an extension of the GNILC foreground-removal method for 21-cm intensity mapping, called eGNILC. The extensions are (i) a Discrete Cosine Transform along the frequency axis before needlet-space ILC, (ii) a modified AIC criterion for selecting the number of foreground degrees of freedom, derived from an analytic calculation of the ILC bias in Appendix B, and (iii) an RPCA-based blind estimate of the 21-cm covariance that replaces the simulated-signal prior. The method is demonstrated on simulated SKA-MID-like and BINGO-like noiseless skies with a frequency-dependent Airy beam, with reported power losses of roughly 10-20% over the quoted multipole ranges. The paper claims that the DCT step reduces power loss at low multipoles and that the eGNILC bias depends on the averaging-domain size and foreground dof but not on the 21-cm signal itself.

Significance. If validated, the analytic bias result in Appendix B is a useful contribution: it gives a closed-form correction to the ILC bias that depends on m and Np rather than on the assumed signal, and it motivates a modified AIC. The DCT improvement at low multipoles is clearly visible in Fig. 3, and the RPCA-embedded variant is shown to work in the noiseless demonstrations of Fig. 2. The paper also includes a useful stability check against cosmic variance in Fig. 4. However, the headline SKA-MID and BINGO claims rest on a model-selection step whose key input, Np, is only heuristic, and the validation is partially self-referential because the signal prior is generated with the same CORA code used for the simulated input maps. No public code is mentioned, so the derivations are not machine-checked by the reader, but the algebraic steps in Appendix B are internally consistent.

major comments (3)

[Sec. 3.1.1 and Sec. 3.1.2, Eq. (48)] The modified AIC is the only criterion used to set the foreground degrees of freedom m in the SKA-MID and BINGO runs, so every recovered spectrum in Figs. 6-7 and Table 2 inherits this m via Eqs. (32)-(35). However, Eq. (48) depends on Np, the effective number of independent pixels, which is introduced only through the qualitative "effective theta_FWHM" prescription in Sec. 3.1.1. No formula or algorithm for Np is given, no sensitivity test to Np is reported, and the modified AIC is never compared with the original AIC of Eqs. (36)-(37) on the same simulations. A mis-estimated Np shifts m and changes all of the quoted power-loss numbers, and the paper's own caveat that the modified AIC should not be combined with RPCA (Sec. 3.2, Sec. 6) means the headline results rely on this AIC step. Please add an explicit Np prescription, an m-selection audit, and a comparison of Eq. (48) with the original AIC.
[Table 2 and Abstract] The abstract states that SKA-MID exhibits ≲20% power loss, but Table 2 lists 21.3% for the 1069.7 MHz auto-spectrum with B=2.5 and the "Extrapol+Unresol" foregrounds over 20<ell<250. This is an internal contradiction in the central quantitative claim. In addition, Table 2's header gives the BINGO multipole range as 30<ell<300, while the abstract and Sec. 6 state [20,300]; these numbers must be reconciled, and the abstract should report either the cell-wise values or the maximum rather than a bound that excludes one of the listed cells.
[Sec. 3.1.1, Secs. 5.1-5.2] The validation of the headline SKA-MID and BINGO results is partially self-referential. The mixing matrix is built from the simulated 21-cm covariance prior generated with the same CORA code that produced the input maps (Sec. 2.1), the tests omit instrumental noise, and no independent signal simulator is used. Because the RPCA variant is explicitly not used with the modified AIC in these runs, the claimed applicability to real data rests on a prior that is derived from the same simulation machinery as the input. Please quantify the sensitivity of the recovered spectra to the assumed Hi power-spectrum amplitude and shape, or test with an independent signal generator, before drawing conclusions about real BINGO and SKA-MID data.

minor comments (4)

[Sec. 5 vs Sec. 6] The text in Sec. 5 says the beam is not deconvolved, but Sec. 6 states that the Airy-disk beam was applied and then "deconvolved after foreground removal"; this contradiction should be resolved because it changes the interpretation of the high-ell residuals.
[Eq. (1)] The DCT definition sums over n=1 to N-1, which appears to omit the n=0 term, and the scaling factor f combines constants in an unconventional way; please check this against the standard type-II DCT convention used by the scipy implementation named in the text.
[Sec. 3.3] The phrase "imaginary mixing matrix" is confusing, since the matrix S is real; a term such as "effective mixing matrix" or "frequency-basis mixing matrix" would be clearer.
[Table 2] The BINGO rows for the 1257.5 x 1252.5 MHz cross-spectrum are empty even though Fig. 7 shows this cross-power spectrum; please either report the values or state explicitly why they are omitted.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the eGNILC bias and modified AIC are derived analytically, and the recovery tests, while using a simulated 21-cm prior from the same CORA code as the input maps, are not forced by construction.

full rationale

The paper's derivation chain is self-contained. The mixing matrix in Eq. (35) is obtained from the eigen-decomposition of the whitened data covariance, and the modified AIC in Eq. (48) follows from the analytic bias calculation in Appendix B (Csδ = -(m/Np)Rs; Eq. B15), with no parameter fitted to the output power spectra. The recovery claims are validated against simulated SKA-MID and BINGO maps where the 21-cm prior is generated with the same CORA code that produced the input signal; this makes the test partially in-sample, but the paper states this dependence explicitly (Sec. 3.3: 'With the help of simulated 21-cm signals, this goal is achievable...'), and the recovered spectrum is not equal to the input by construction because the foreground dof m, needlet windows, masks, and beams all affect the result. The self-citations (e.g., Yohana et al. 2021 for 1/f noise) are not load-bearing for the central derivation. The paper's own caveat that the modified AIC should not be combined with RPCA (Sec. 6) is a scope limitation, not a circular step. The heuristic Np in Sec. 3.1.1 is a validation and robustness gap, not a circularity.

Assumptions & free parameters 5 free parameters · 7 assumptions · 0 invented entities

The ledger shows the method is largely a re-derivation and recombination of published machinery (GNILC, ILC bias, RPCA). The genuinely new items are the DCT basis choice and the AIC correction, but both rest on hand-chosen quantities (B, Np, n, RPCA weight) that are not fully specified.

free parameters (5)

Needlet spectral window parameter B = B = 1.7 and B = 2.5
Controls the needlet window family (Eqs. A6-A9). Table 2 shows the recovered power loss changes by several percent when B changes, and the paper does not provide a principled choice.
Effective number of independent pixels Np = not specified
Enters the bias formula (Eq. B16) and modified AIC (Eq. 48). Defined only via an effective theta_FWHM (Sec. 3.1.1) with no concrete recipe, so the AIC result depends on an unstated quantity.
Foreground degrees of freedom m (per needlet scale) = selected by AIC (Eq. 48)
The mixing matrix S (Eq. 35) is built from m, chosen by minimizing AIC. The recovered power spectrum therefore depends on this data-driven choice; no sensitivity scan of m is shown.
Channel grouping factor n for SKA-MID = n = 10
In Sec. 5.1 the 500 frequency channels are split into 10 subsets to break frequency correlation; n is chosen by hand without a sensitivity study.
RPCA regularization weight = not stated
The low-rank/sparse split in Sec. 3.2 requires a regularization parameter; the value used for the Fig. 2 RPCA results is not reported, hampering exact reproduction.

assumptions (7)

domain assumption The 21-cm signal and foregrounds can be treated as zero-mean Gaussian random fields in the needlet domain.
Sec. 3.1.2: 'the signals are assumed/transformed to a normal distribution with mean zero'; required for the likelihood and the Csdelta calculation in Appendix B.
domain assumption Foregrounds are spectrally smooth, so their covariance is low-rank with m much less than nch and eigenvalues lambda_i much greater than 1 in the DCT basis.
Eq. (32) and Sec. 3.1.1: the separation into foreground and signal subspaces relies on the eigenvalue gap; the paper's own Fig. 2 shows degraded recovery when unresolved, non-smooth foregrounds are added.
domain assumption The reconstruction error covariance Rdelta is negligible compared with Rs (small GNILC error).
Sec. 3.1.2, Eq. (39): 'We assume to reach a small GNILC error delta such that the Rdelta term can be neglected.'
domain assumption The domain-averaged covariance estimate uses Np independent pixels and E(sq fq^T) = 0.
Sec. 3.1.1 and Appendix B, Eq. (46): the bias formula 2m/Np follows from this averaging model.
domain assumption The noise-free, Airy-disk-without-deconvolution simulation is representative of the foreground-removal-only scenario.
Sec. 5: the paper explicitly excludes instrumental noise and uses a truncated Airy beam, noting real beams (e.g., MeerKAT L-band, Asad et al. 2021) are more complex.
standard math The needlet window family satisfies the exact reconstruction condition sum_j [b_ell^{(j)}]^2 = 1.
Appendix A, Eqs. (A5)-(A9): standard needlet frame property from Marinucci et al. 2008 and Pietrobon et al. 2010.
standard math The type-II DCT with the scaling in Eqs. (1)-(2) is an orthonormal transform preserving the ILC inversion.
Sec. 2, Eqs. (1)-(2): the stated '2f' normalization makes the DCT orthonormal, which is required for a lossless change of basis along frequency.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Expanded Generalized Needlet Internal Linear Combination (eGNILC) Framework for the 21-cm Foreground Removal." pith.science (2026). https://pith.science/paper/G2BTKNHH

@misc{pith2026241116899,
  author       = {Pith},
  title        = {Pith review of: Expanded Generalized Needlet Internal Linear Combination (eGNILC) Framework for the 21-cm Foreground Removal},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G2BTKNHH}},
  note         = {Machine review of arXiv:2411.16899}
}

abstract

The Generalized Needlet Internal Linear Combination (GNILC) method is a non-parametric component separation algorithm to remove the foreground contamination of the 21-cm intensity mapping data. In this work, we perform the Discrete Cosine Transform (DCT) along the frequency axis in the expanded GNILC framework (denoted eGNILC) which helps reduce the power loss in low multipoles, and further demonstrate its performance. We also calculate the eGNILC bias to modify the criterion for determining the degrees of freedom of the foreground (dof), and embed the Robust Principal Component Analysis (RPCA) in mixing matrix computation to obtain a blind component separation method. We find that the eGNILC bias is related to the averaged domain size and the dof of the foreground but not the underlying 21-cm signal. In case of no beam effect, the eGNILC bias is negligible for simple power law foregrounds outside the Galactic plane. We also examine the eGNILC performance in the SKA-MID (SKA Phase-I in mid-frequency) and BINGO (Baryon Acoustic Oscillations from Integrated Neutral Gas Observations) simulations. We show that if the adjacent frequency channels are not highly correlated, eGNILC can recover the underlying 21-cm signal with good accuracy. With the varying Airy-disk beam applied to both SKA-MID and BINGO, the power spectra of 21-cm can be effectively recovered at the multipoles $\ell \in [20, 250]$ and $[20, 300]$ respectively. With no instrumental noise, the SKA-MID exhibits $\lesssim 20\%$ power loss and BINGO exhibits $\sim 10\%$ power loss. The varying Airy-disk beam only causes significant errors at large multipoles.

Figures

Figures reproduced from arXiv: 2411.16899 by the authors.

**Figure 1.** Sky maps of the synchrotron emission, free-free radiation, point sources, and 21-cm signal of Hi at the frequency 962.5 MHz. All maps are normalized with histogram equalized color mapping. The discrete cosine transforms of 60 channels (962.5-1257.5 MHz) for individual catalogues are plotted in the square panels, where only unmasked pixels are used. The horizontal ticks label the pixel index, the vertical ticks label… view at source ↗

**Figure 2.** Demonstrations of the RPCA and foreground effects on the GNILC foreground removal method with 20 channels. The left panel shows the 21-cm angular power spectrum at 1252.5 MHz, where the blue dashed line is the input power spectrum, the orange (green) solid line is the power spectrum recovered from “Extrapolation” power-law foregrounds without (with) RPCA. The red (purple) solid line is recovered from “Extrapolation”… view at source ↗

**Figure 3.** Same as [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Fluctuations of the recovered power spectrum by the plain GNILC in 1000 realizations of Hi 21-cm intensity maps, with the same foreground model and mask as in [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]

**Figure 5.** Figure 5: The sky coverage of the simulated SKA-MID and BINGO, projected in the Equatorial coordinate. The gray area is masked, which is derived from Haslam map. The valid area is reduced from the survey area of 20, 000 to 10, 000 deg2 and from the survey area of 5000 to 3000 de…

**Figure 6.** Figure 6: The recovered HI 21-cm angular power spectra of the simulated SKA-MID map with the Airy beam. The black thin line is the power spectrum of the orginal map (without beam), and the blue one is to convolve the black line with the beam (i.e. input spectrum). The orange thi…

**Figure 7.** Figure 7: Same as [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]

**Figure 8.** Figure 8: The window function b (j) ℓ as an example of B = 1.7. Mollweide view β (0), B=1.7 -8.38531e-05 8.38531e-05 Mollweide view β (3), B=1.7 -0.000357323 0.00110482 Mollweide view β (6), B=1.7 -0.00734435 0.0247761 Mollweide view β (8), B=1.7 -0.0508934 0.174024 [PITH_FULL_…

**Figure 9.** Figure 9: Mollweide-projected view of the needlet decomposition of a point source. harmonics domains (Guilloux et al. 2007). This requirement can be achieved through needlet analysis, which is an analogy to wavelet transform on the Euclid plane. The usual complex spherical harmo…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

69 extracted references · 11 canonical work pages

[1]

Abdalla, E., Ferreira, E. G. M., Landim, R. G., et al. 2022, A&A, 664, A14, doi: 10.1051/0004-6361/202140883

doi:10.1051/0004-6361/202140883 2022
[2]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022a, ApJ, 925, 221, doi: 10.3847/1538-4357/ac1c78 —. 2022b, ApJ, 924, 51, doi: 10.3847/1538-4357/ac2ffc

doi:10.3847/1538-4357/ac1c78
[3]

Abramowitz, M., & Stegun, I. A. 1972, Handbook of Mathematical Functions (United States Department of

work page 1972
[4]

2016, Astron

Adam, R., et al. 2016, Astron. Astrophys., 594, A10, doi: 10.1051/0004-6361/201525967

doi:10.1051/0004-6361/201525967 2016
[5]

Ade, P. A. R., et al. 2016, Astron. Astrophys., 594, A4, doi: 10.1051/0004-6361/201525809

doi:10.1051/0004-6361/201525809 2016
[6]

G., & Santos, M

Alonso, D., Bull, P., Ferreira, P. G., & Santos, M. G. 2015, Mon. Not. Roy. Astron. Soc., 447, 400, doi: 10.1093/mnras/stu2474

doi:10.1093/mnras/stu2474 2015
[7]

J., Luciw, N

Anderson, C. J., Luciw, N. J., Li, Y. C., et al. 2018, MNRAS, 476, 3382, doi: 10.1093/mnras/sty346

doi:10.1093/mnras/sty346 2018
[8]

Asad, K. M. B., Girard, J. N., de Villiers, M., et al. 2021, MNRAS, 502, 2970, doi: 10.1093/mnras/stab104

doi:10.1093/mnras/stab104 2021

Show all 69 references

[9]

2016, arXiv e-prints, arXiv:1610.06826, doi: 10.48550/arXiv.1610.06826

Battye, R., Browne, I., Chen, T., et al. 2016, arXiv e-prints, arXiv:1610.06826, doi: 10.48550/arXiv.1610.06826

2016 doi arXiv
[10]

A., Browne, I

Battye, R. A., Browne, I. W. A., Dickinson, C., et al. 2013, Mon. Not. Roy. Astron. Soc., 434, 1239, doi: 10.1093/mnras/stt1082

2013 doi
[12]

A., Brown, M

Battye, R. A., Brown, M. L., Browne, I. W. A., et al. 2012, arXiv e-prints, arXiv:1209.1041, doi: 10.48550/arXiv.1209.1041

2012 doi arXiv
[13]

A., Dickinson, C., Battye, R

Bigot-Sazy, M. A., Dickinson, C., Battye, R. A., et al. 2015, MNRAS, 454, 3240, doi: 10.1093/mnras/stv2153

2015 doi
[14]

A., Ma, Y

Bigot-Sazy, M. A., Ma, Y. Z., Battye, R. A., et al. 2016, in Astronomical Society of the Pacific Conference Series, Vol. 502, Frontiers in Radio Astronomy and F AST Early Sciences Symposium 2015, ed. L. Qain & D. Li, 41, doi: 10.48550/arXiv.1511.03006

2016 doi arXiv
[15]

G., Patel, P., & Santos, M

Bull, P., Ferreira, P. G., Patel, P., & Santos, M. G. 2015, Astrophys. J., 803, 21, doi: 10.1088/0004-637X/803/1/21

2015 doi
[16]

2003, Astron

Burigana, C., & Saez, D. 2003, Astron. Astrophys., 409, 423, doi: 10.1051/0004-6361:20031038

2003 doi
[17]

Chang, T.-C., Pen, U.-L., Bandura, K., & Peterson, J. B. 2010, Nature, 466, 463, doi: 10.1038/nature09187

2010 doi
[18]

J., & Ransom, S

Condon, J. J., & Ransom, S. M. 2016, Essential Radio Astronomy (Princeton University Press)

2016
[19]

2022, Mon

Cunnington, S., et al. 2022, Mon. Not. Roy. Astron. Soc., 518, 6262, doi: 10.1093/mnras/stac3060 de Oliveira-Costa, A., Tegmark, M., Gaensler, B. M., et al. 2008, Mon. Not. Roy. Astron. Soc., 388, 247, doi: 10.1111/j.1365-2966.2008.13376.x eGNILC foreground removal 23

2022
[20]

R., Parsons, A

DeBoer, D. R., Parsons, A. R., Aguirre, J. E., et al. 2017, PASP, 129, 045001, doi: 10.1088/1538-3873/129/974/045001

2017 doi
[21]

F., Jeune, M

Delabrouille, J., Cardoso, J. F., Jeune, M. L., et al. 2009, Astron. Astrophys., 493, 835, doi: 10.1051/0004-6361:200810514 Di Matteo, T., Perna, R., Abel, T., & Rees, M. J. 2002, Astrophys. J., 564, 576, doi: 10.1086/324293

2009 doi
[22]

2014, arXiv e-prints, arXiv:1405.7936, doi: 10.48550/arXiv.1405.7936

Dickinson, C. 2014, arXiv e-prints, arXiv:1405.7936, doi: 10.48550/arXiv.1405.7936

2014 doi arXiv
[23]

2020, Modern Cosmology (Elsevier Science), doi: 10.1016/C2017-0-01943-2

Dodelson, S., & Schmidt, F. 2020, Modern Cosmology (Elsevier Science), doi: 10.1016/C2017-0-01943-2

2020 doi
[24]

Draine, B. T. 2011, Physics of the Interstellar and Intergalactic Medium (Princeton University Press)

2011
[25]

2004, MNRAS, 348, 885, doi: 10.1111/j.1365-2966.2004.07409.x

Efstathiou, G. 2004, MNRAS, 348, 885, doi: 10.1111/j.1365-2966.2004.07409.x

2004
[26]

K., Mertens, F

Gehlot, B. K., Mertens, F. G., Koopmans, L. V. E., et al. 2019, MNRAS, 488, 4271, doi: 10.1093/mnras/stz1937

2019 doi
[27]

G., Ma, Y.-Z., Kittiwisit, P., et al

Gogo, T. G., Ma, Y.-Z., Kittiwisit, P., et al. 2022, MNRAS, 510, 1680, doi: 10.1093/mnras/stab3516 G´ orski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ, 622, 759, doi: 10.1086/427976

2022 doi
[28]

2007, arXiv e-prints, arXiv:0706.2598, doi: 10.48550/arXiv.0706.2598

Guilloux, F., Fay, G., & Cardoso, J.-F. 2007, arXiv e-prints, arXiv:0706.2598, doi: 10.48550/arXiv.0706.2598

2007 doi arXiv
[29]

2013, Phys

Hall, A., Bonvin, C., & Challinor, A. 2013, Phys. Rev. D, 87, 064026, doi: 10.1103/PhysRevD.87.064026

2013 doi
[30]

2009, PhRvD, 80, 063004, doi: 10.1103/PhysRevD.80.063004

Hanson, D., & Lewis, A. 2009, PhRvD, 80, 063004, doi: 10.1103/PhysRevD.80.063004

2009 doi
[31]

2018, Mon

Harper, S., Dickinson, C., Battye, R., et al. 2018, Mon. Not. Roy. Astron. Soc., 478, 2416, doi: 10.1093/mnras/sty1238

2018 doi
[32]

Haslam, C. G. T., Salter, C. J., Stoffel, H., & Wilson, W. E. 1982, A&AS, 47, 1 HERA Collaboration, Abdurashidova, Z., Adams, T., et al. 2023, ApJ, 945, 124, doi: 10.3847/1538-4357/acaf50

1982 doi
[33]

2002, Astrophys

Hivon, E., Gorski, K., Netterfield, C., et al. 2002, Astrophys. J., 567, 2, doi: 10.1086/338126

2002 doi
[34]

2012, ApJ, 753, 110, doi: 10.1088/0004-637X/753/2/110

Kogut, A. 2012, ApJ, 753, 110, doi: 10.1088/0004-637X/753/2/110

2012 doi
[35]

2017, PhRvD, 96, 063525, doi: 10.1103/PhysRevD.96.063525

Li, Y.-C., & Ma, Y.-Z. 2017, PhRvD, 96, 063525, doi: 10.1103/PhysRevD.96.063525

2017 doi
[36]

Madau, P., Meiksin, A., & Rees, M. J. 1997, Astrophys. J., 475, 429, doi: 10.1086/303549

1997 doi
[37]

2002, Mon

Maino, D., Farusi, A., Baccigalupi, C., et al. 2002, Mon. Not. Roy. Astron. Soc., 334, 53, doi: 10.1046/j.1365-8711.2002.05425.x

2002
[38]

2008, Mon

Marinucci, D., Pietrobon, D., Balbi, A., et al. 2008, Mon. Not. Roy. Astron. Soc., 383, 539, doi: 10.1111/j.1365-2966.2007.12550.x

2008
[40]

W., et al

Masui, K. W., et al. 2013, Astrophys. J. Lett., 763, L20, doi: 10.1088/2041-8205/763/1/L20

2013 doi
[41]

G., Mevius, M., Koopmans, L

Mertens, F. G., Mevius, M., Koopmans, L. V. E., et al. 2020, MNRAS, 493, 1662, doi: 10.1093/mnras/staa327

2020 doi
[42]

A., Ysard, N., Lavabre, A., et al

Miville-Deschenes, M. A., Ysard, N., Lavabre, A., et al. 2008, Astron. Astrophys., 490, 1093, doi: 10.1051/0004-6361:200809484

2008 doi
[43]

2011, International Journal of Modern Physics D, 20, 989, doi: 10.1142/S0218271811019335

Nan, R., Li, D., Jin, C., et al. 2011, International Journal of Modern Physics D, 20, 989, doi: 10.1142/S0218271811019335

2011 doi
[44]

C., Remazeilles, M., & Dickinson, C

Olivari, L. C., Remazeilles, M., & Dickinson, C. 2016, Mon. Not. Roy. Astron. Soc., 456, 2749, doi: 10.1093/mnras/stv2884

2016 doi
[45]

2020, Mon

Osato, K., Shirasaki, M., Miyatake, H., et al. 2020, Mon. Not. Roy. Astron. Soc., 492, 4780, doi: 10.1093/mnras/staa117

2020 doi
[46]

R., Backer, D

Parsons, A. R., Backer, D. C., Foster, G. S., et al. 2010, AJ, 139, 1468, doi: 10.1088/0004-6256/139/4/1468

2010 doi
[47]

2009, Mon

Pen, U.-L., Staveley-Smith, L., Peterson, J., & Chang, T.-C. 2009, Mon. Not. Roy. Astron. Soc., 394, 6, doi: 10.1111/j.1745-3933.2008.00581.x

2009
[48]

Pietrobon, D., Balbi, A., Cabella, P., & Gorski, K. M. 2010, Astrophys. J., 723, 1, doi: 10.1088/0004-637X/723/1/1

2010 doi
[49]

1998, Astrophys

Platania, P., Bensadoun, M., Bersanelli, M., et al. 1998, Astrophys. J., 505, 473, doi: 10.1086/306175

1998 doi
[50]

2011, Mon

Remazeilles, M., Delabrouille, J., & Cardoso, J.-F. 2011, Mon. Not. Roy. Astron. Soc., 418, 467, doi: 10.1111/j.1365-2966.2011.19497.x

2011
[51]

J., Bigot-Sazy, M

Remazeilles, M., Dickinson, C., Banday, A. J., Bigot-Sazy, M. A., & Ghosh, T. 2015, Mon. Not. Roy. Astron. Soc., 451, 4311, doi: 10.1093/mnras/stv1274

2015 doi
[52]

G., Cooray, A., & Knox, L

Santos, M. G., Cooray, A., & Knox, L. 2005, Astrophys. J., 625, 575, doi: 10.1086/429857

2005 doi
[53]

G., et al

Santos, M. G., et al. 2015, PoS, AASKA14, 019, doi: 10.22323/1.215.0019

2015 doi
[54]

R., & Lewis, A

Shaw, J. R., & Lewis, A. 2008, PhRvD, 78, 103512, doi: 10.1103/PhysRevD.78.103512

2008 doi
[55]

2014, Astrophys

Sitwell, M. 2014, Astrophys. J., 781, 57, doi: 10.1088/0004-637X/781/2/57

2014 doi
[56]

2015, PhRvD, 91, 083514, doi: 10.1103/PhysRevD.91.083514 Square Kilometre Array Cosmology Science Working

Pen, U.-L. 2015, PhRvD, 91, 083514, doi: 10.1103/PhysRevD.91.083514 Square Kilometre Array Cosmology Science Working

2015 doi
[57]

J., Battye, R

Group, Bacon, D. J., Battye, R. A., et al. 2020, Publ. Astron. Soc. Austral., 37, e007, doi: 10.1017/pasa.2019.51

2020 doi
[58]

1996, MNRAS, 281, 1297, doi: 10.1093/mnras/281.4.1297 24 Dai & Ma

Tegmark, M., & Efstathiou, G. 1996, MNRAS, 281, 1297, doi: 10.1093/mnras/281.4.1297 24 Dai & Ma

1996 doi
[59]

2017, Mon

Thorne, B., Dunkley, J., Alonso, D., & Naess, S. 2017, Mon. Not. Roy. Astron. Soc., 469, 2821, doi: 10.1093/mnras/stx949

2017 doi
[60]

2020, MNRAS, 498, 5916, doi: 10.1093/mnras/staa2727

Tramonte, D., & Ma, Y.-Z. 2020, MNRAS, 498, 5916, doi: 10.1093/mnras/staa2727

2020 doi
[61]

2019, MNRAS, 489, 385, doi: 10.1093/mnras/stz2146

Tramonte, D., Ma, Y.-Z., Li, Y.-C., & Staveley-Smith, L. 2019, MNRAS, 489, 385, doi: 10.1093/mnras/stz2146

2019 doi
[62]

M., Jordan, C

Trott, C. M., Jordan, C. H., Midgley, S., et al. 2020, MNRAS, 493, 4711, doi: 10.1093/mnras/staa414

2020 doi
[63]

2021, Mon

Wang, J., et al. 2021, Mon. Not. Roy. Astron. Soc., 505, 3698, doi: 10.1093/mnras/stab1365

2021 doi
[64]

B., Blake, C., et al

Wolz, L., Abdalla, F. B., Blake, C., et al. 2014, Mon. Not. Roy. Astron. Soc., 441, 3271, doi: 10.1093/mnras/stu792

2014 doi
[65]

Wuensche, C. 2019, J. Phys. Conf. Ser., 1269, 012002, doi: 10.1088/1742-6596/1269/1/012002

2019 doi
[66]

A., Villela, T., Abdalla, E., et al

Wuensche, C. A., Villela, T., Abdalla, E., et al. 2022, A&A, 664, A15, doi: 10.1051/0004-6361/202039962

2022 doi
[67]

2018, PhRvD, 97, 083504, doi: 10.1103/PhysRevD.97.083504

Xu, X., Ma, Y.-Z., & Weltman, A. 2018, PhRvD, 97, 083504, doi: 10.1103/PhysRevD.97.083504

2018 doi
[68]

2019, Research in Astronomy and Astrophysics, 19, 186, doi: 10.1088/1674-4527/19/12/186

Yohana, E., Li, Y.-C., & Ma, Y.-Z. 2019, Research in Astronomy and Astrophysics, 19, 186, doi: 10.1088/1674-4527/19/12/186

2019 doi
[69]

2021, Mon

Yohana, E., Ma, Y.-Z., Li, D., Chen, X., & Dai, W.-M. 2021, Mon. Not. Roy. Astron. Soc., 504, 5231, doi: 10.1093/mnras/stab1197

2021 doi
[70]

2019, Journal of Open Source Software, 4, 1298, doi: 10.21105/joss.01298

Zonca, A., Singer, L., Lenz, D., et al. 2019, Journal of Open Source Software, 4, 1298, doi: 10.21105/joss.01298

2019 doi
[71]

2018, Astron

Zuo, S., Chen, X., Ansari, R., & Lu, Y. 2018, Astron. J., 157, 4, doi: 10.3847/1538-3881/aaef3b

2018 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Abdalla, E., Ferreira, E. G. M., Landim, R. G., et al. 2022, A&A, 664, A14, doi: 10.1051/0004-6361/202140883

doi:10.1051/0004-6361/202140883 2022

[2] [2]

E., Alexander, P., et al

Abdurashidova, Z., Aguirre, J. E., Alexander, P., et al. 2022a, ApJ, 925, 221, doi: 10.3847/1538-4357/ac1c78 —. 2022b, ApJ, 924, 51, doi: 10.3847/1538-4357/ac2ffc

doi:10.3847/1538-4357/ac1c78

[3] [3]

Abramowitz, M., & Stegun, I. A. 1972, Handbook of Mathematical Functions (United States Department of

work page 1972

[4] [4]

2016, Astron

Adam, R., et al. 2016, Astron. Astrophys., 594, A10, doi: 10.1051/0004-6361/201525967

doi:10.1051/0004-6361/201525967 2016

[5] [5]

Ade, P. A. R., et al. 2016, Astron. Astrophys., 594, A4, doi: 10.1051/0004-6361/201525809

doi:10.1051/0004-6361/201525809 2016

[6] [6]

G., & Santos, M

Alonso, D., Bull, P., Ferreira, P. G., & Santos, M. G. 2015, Mon. Not. Roy. Astron. Soc., 447, 400, doi: 10.1093/mnras/stu2474

doi:10.1093/mnras/stu2474 2015

[7] [7]

J., Luciw, N

Anderson, C. J., Luciw, N. J., Li, Y. C., et al. 2018, MNRAS, 476, 3382, doi: 10.1093/mnras/sty346

doi:10.1093/mnras/sty346 2018

[8] [8]

Asad, K. M. B., Girard, J. N., de Villiers, M., et al. 2021, MNRAS, 502, 2970, doi: 10.1093/mnras/stab104

doi:10.1093/mnras/stab104 2021