Pith. sign in

REVIEW 2 major objections 5 minor 41 references

On likelihood-based analysis of the gravitationally (de)lensed CMB

T0 review · 2 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read The exact CMB lensing-spectrum score reduces to a quadratic-estimator-like gradient spectrum on delensed maps, with mean-field subtraction and realization-dependent debiasing.

desk verdict A rigorous, clearly-written theory paper that unifies beyond-QE CMB lensing statistics with the QE framework; its main derivations hold up, with one honest but unquantified approximation about the discarded 2-point term. read the letter →

arxiv 2502.02399 v2 pith:2FZ5OKLY submitted 2025-02-04 astro-ph.CO

Julien Carron This is my paper · ORCID
classification astro-ph.CO
keywords cosmicmicrowavebackgroundgravitationallensingdelensingquadraticestimatorslikelihoodscorerealization-dependentdebiasingCMBspectrum
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 derives the score of the gravitationally lensed CMB likelihood and shows that the optimal lensing-spectrum statistic has a direct quadratic-estimator structure. The exact lensing-spectrum score is the posterior average, over plausible lensing maps, of the power spectrum of a quadratic-estimator-like gradient evaluated on maps delensed (remapped by the inverse deflection) by that map, complete with mean-field subtraction and a realization-dependent debiasing term ($\mathrm{RD}\text{-}\hat n^{(0)}$); the one additional term, carrying delensed two-point information, is argued to be degenerate with ordinary CMB power-spectrum analysis and safely discardable. The paper also proves that the log-likelihood gradient at a non-zero delensing map is exactly a generalized minimum-variance quadratic estimator applied to delensed maps. This matters because deep, high-resolution CMB experiments can then reach near-optimal delensed information while keeping the stability, redundancy, and computational practicality of standard quadratic-estimator analyses.

What carries the argument

The load-bearing object is the score function of the lensed CMB likelihood, $s_\theta=\partial_\theta\ln p(X|\theta)$, and for spectrum estimation the posterior-averaged gradient spectrum of Eq. (3.6). The key mechanism is the curvature of the lensing log-likelihood: the realization-dependent debiaser $\mathrm{RD}\text{-}\hat n^{(0)}_L$ is defined as the negative diagonal of that curvature restricted to the linear lensing response, and it subtracts the Gaussian four-point contractions that are not residual-lensing signal. At non-zero delenser, the chain-rule identity $\hat g = |A_\kappa|\,\hat q$ shows that the gradient is the magnification determinant times a generalized minimum-variance quadratic estimator applied to delensed maps, making the quadratic-estimator analogy exact rather than heuristic. This machinery turns a formally intractable posterior average into a debiased gradient spectrum that can be computed with maximum-a-posteriori delensers and simulation-based calibration.

What would settle it

In a simulated deep, high-resolution lensed CMB map, compute the exact lensing-spectrum score of Eq. (3.6) as a posterior average over lensing maps and compare the Fisher information it provides to that of the simplified score with the delensed two-point term discarded; if the simplified score loses more than the statistical uncertainty on multipoles where projection effects are strongest, the degeneracy claim is falsified.

Watch

Extended reading notes

Core claim

The central discovery is a first-principles decomposition of the CMB lensing-spectrum score. Starting from the marginalized likelihood $p(X|C)=\int D\kappa\,p(X|\kappa)p(\kappa|C)$, the paper rewrites the derivative with respect to $C_L$ as the posterior average over the lensing map of a statistic $\hat C_L[X,\kappa]$ built from the likelihood gradient evaluated at delenser $\kappa$. That statistic is the spectrum of the mean-field-subtracted quadratic gradient $\hat g^{\rm QD}_{LM}[\bar X^\kappa,\bar X^\kappa]$, minus the realization-dependent debiaser $\mathrm{RD}\text{-}\hat n^{(0)}_L$, plus a second-order term. The paper argues that the second-order term is equivalent to delensed two-point information and therefore adds nothing beyond a CMB spectra analysis, leaving the gradient-spectrum term as the informative statistic. A separate identity (Eq. 2.7) shows the likelihood gradient at non-zero delenser is, up to the magnification determinant, exactly a generalized minimum-variance quadratic estimator on the delensed maps. Consequently, the practical beyond-QE lensing analysis is a quadratic-estimator-like spectrum reconstruction on maps delensed by the most probable lensing map.

Load-bearing premise

The simplified score drops the delensed two-point information term on the assumption that it duplicates information already contained in CMB power-spectrum measurements, but that equivalence is proven only to leading order and is broken by projection from delensed to observed positions, so the simplification may be incomplete in some regimes.

Editorial extensions

If this is right

  • Near-optimal lensing constraints can be obtained by running a quadratic-estimator-like spectrum analysis on maps delensed by the most probable lensing map, with mean-field subtraction and realization-dependent debiasing, without sampling the full lensing posterior.
  • The MUSE lensing spectrum score is precisely such a delensed-map quadratic-estimator spectrum, and adding mean-field and realization-dependent debiasing to it should reduce its sensitivity to mismodeled noise and simulations.
  • The analytically modeled score of Eq. (4.5) requires a normalization $A_L$ whose value is not fully under analytical control, but which is independent of the data noise and can be calibrated from simulations to within a few percent.
  • Standard quadratic-estimator stability tools, including shear-only, EB-only, and bias-hardened gradients, can be carried over to beyond-QE analysis by swapping the gradient while holding the delenser fixed, preserving the variance reduction from delensing.
  • In this score-based formulation the lensing map serves mainly as a delenser; what must be modeled is the likelihood gradient spectrum, which reframes stability questions around the gradient rather than the reconstructed map.

Reading between the lines

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

  • If the claimed degeneracy of the delensed two-point term holds beyond leading order, then lensing-spectrum inference and delensed CMB spectrum inference factor exactly: the gradient spectrum plus the delensed spectra would exhaust the information, and end-to-end pipelines could be assembled modularly.
  • The paper's separation of delenser from gradient suggests a testable family of hybrid estimators: keep a fixed maximum-a-posteriori delenser but use a shear-only or bias-hardened gradient, then measure residual lensing signal and systematic bias in simulations with injected foregrounds; this would show whether stability is gained at small variance cost.
  • Because $\mathrm{RD}\text{-}\hat n^{(0)}$ is identified with the diagonal of the likelihood curvature, it may be computable from the data map and only a handful of simulations, instead of the large simulation ensembles currently used, which would lower the cost of moving beyond-QE analyses to wide-area surveys.
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.

Referee Report

2 major / 5 minor

Summary. This paper derives the likelihood score (gradient of the log-likelihood) for CMB lensing and lensing-spectrum estimation, and connects each term to quadratic-estimator (QE) constructions. The main results are: (i) the log-likelihood gradient evaluated at a nonzero delensing map is formally equivalent to a GMV quadratic estimator applied to delensed maps, up to the Jacobian between observed and delensed positions (Eq. 2.7, Appendix C); (ii) the exact lensing-spectrum score can be written as a posterior average over lensing maps of a QE-like gradient spectrum that includes mean-field subtraction and a realization-dependent debiaser RD-n(0), plus a second-order term that the paper argues is degenerate with delensed CMB two-point information (Eq. 3.6); (iii) practical implementations, namely the Geneva score and the MUSE score, are discussed as approximations to this exact score, with explicit attention to robustness, mean-field subtraction, and debiasing. The paper also clarifies the relation between the joint and marginalized MAP formulations and suggests concrete robustness modifications (e.g., bias-hardened or shear-only gradients) for beyond-QE analyses.

Significance. If the central claims hold, this paper provides a clean theoretical foundation for beyond-QE CMB lensing analysis: it shows that the exact score has a direct QE-like interpretation and that practical methods such as MUSE can be understood as specific approximations with identified robustness trade-offs. The derivation is self-contained, does not involve fitting parameters to data, and is supported by detailed appendices (especially Appendix C and D). The paper also makes actionable connections to current analyses, including the SPT-3G MUSE implementation, and proposes robustness improvements that could be tested with existing simulation pipelines. The main caveats, both acknowledged in the text, are the unquantified projection correction in the degeneracy of the second-order term and the analytically uncontrolled normalization in the Geneva score; these do not invalidate the theoretical framework but need sharper treatment before the practical scores can be claimed to be optimal in a fully quantitative sense.

major comments (2)
  1. [III.A, Eq. (3.6); Appendix D.2] The decision to discard the second-order term in Eq. (3.6) rests on a leading-order degeneracy proof stated for the delensed-space curvature tilde-H (Eq. D4), while the score itself involves the observed-space H. The text asserts that the degeneracy is 'only slightly broken by projection effects' and that the term carries 'at most little independent value,' but no quantitative estimate is supplied. This is load-bearing because both the Geneva score (Eq. 4.5) and the MUSE score (Eq. 4.10) omit this term; if the projection correction is not small, these statistics are not the exact lensing-spectrum score and may miss information not contained in a separate CMB power-spectrum analysis. I ask for an explicit estimate of the difference between the M-averaged diagonals of H and tilde-H (e.g., in a flat-sky or simulated setting, as a function of L and noise level), or for the text to be restricted to a leading-order statement with this caveat stated prominently.
  2. [IV.A, Eq. (4.5)] The Geneva score requires a normalization A_L that is 'not fully under analytical control' and is described as 'probably the major defect of this approach.' Because this normalization sets the relative weight of the squared gradient and RD-n(0), any error in A_L directly biases the estimated lensing spectrum. The paper cites successful calibration in prior works; please include the relevant numerical evidence (or a concise summary of it) and state precisely what accuracy is required for the science targets discussed, so that the claim that this approximation matches predictions is reproducible from the present manuscript.
minor comments (5)
  1. [II.A] The phrase '(sUSABLE)' appears to contain a formatting artifact; it should presumably read 'usable' or be set as a named acronym. Please fix.
  2. [III.A, Eq. (3.6)] The simulation maps S and S' in the RD-n(0) term are not defined explicitly; please state that they are independent realizations of the fiducial model conditioned on kappa, so that the cross-spectrum averaging is unambiguous.
  3. [III.B, Eq. (3.13)] The notation '<3.12>' is confusing; please write it as the average of Eq. (3.12) or define the averaging explicitly (fixed delenser, average over CMB realizations).
  4. [Appendix C, Eq. (C7)] The inverse-noise matrix N_kappa is introduced as the inverse of N_kappa^{-1} although the latter is not invertible in practice; the 'notational trick' is clear, but please state explicitly that all final expressions depend only on combinations like D_kappa^dagger B^dagger Cov^{-1}_kappa, so that no practical invertibility assumption is used.
  5. [VI] There is a grammatical error in 'The latter begin a significantly more complicated process'; it should be 'The latter being a significantly more complicated process.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's formal derivation of the lensing-spectrum score is self-contained, and its self-citations and approximations are not load-bearing circular inputs.

full rationale

The central derivation does not reduce to its inputs. Eq. (3.6) is obtained from exact score-function identities: the lensing-spectrum score is rewritten using Eq. (3.5), which replaces prior spectrum derivatives with curvature derivatives, followed by integration by parts and the Gaussian-prior score. No parameter is fitted to data, and no posterior average is replaced by a predicted quantity. The identification of likelihood gradients with GMV quadratic estimators on delensed maps is proven in Appendix C (Eqs. C5–C11), starting from the likelihood gradient expression and manipulating operators; the cited [8] formula is an approximate position-space form, and Appendix C itself establishes the QE structure rather than importing it as an unproved premise. The claim that the f^(2) term in Eq. (3.6) is degenerate with delensed 2-pt information is explicitly analyzed in Appendix D.2: the degeneracy is demonstrated for the delensed-space curvature eH (Eq. D10), and the paper openly states that for the observed-space H the degeneracy is 'not expected to be perfect' and 'still fully there to leading order.' That is an admitted, leading-order approximation—a correctness/robustness caveat, not a circular reduction. Self-citations to Legrand & Carron [22,23] and Millea & Seljak [20] are used as background for practical implementations and for calibration statements that the paper itself describes as 'not fully under analytical control'; the formal claims do not depend on these citations being true. The MUSE and Geneva scores are presented as reinterpretations and comparisons, not as newly derived results whose content is identical to their premises. No fitted parameter is renamed as a prediction, and no self-citation is invoked to forbid alternatives. Accordingly, the paper exhibits no circular reasoning.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The paper introduces no new physical entities or fitted parameters. It relies on standard statistical assumptions (Gaussianity of the CMB given lensing, score-Fisher equivalence) and domain assumptions (pure gradient deflection, invertibility of the deflection, MAP approximation, degeneracy of 2-point term). No free parameters are fitted to data.

assumptions (6)
  • domain assumption The observed CMB is a zero-mean Gaussian random field conditioned on the lensing map, with covariance Cov_kappa = B D_kappa C_X D_kappa^dagger B^dagger + N.
    Used throughout the derivation of the score and gradients (Eq. A1, A2).
  • domain assumption The lensing deflection is a pure gradient, neglecting the curl mode.
    Stated in Sec. II.B as of no relevance to what follows.
  • ad hoc to paper The deflection field is invertible.
    Used in Appendix C to show the log-likelihood gradient is a GMV quadratic estimator on delensed maps; the paper states this is harmless in practice but not needed for reconstruction.
  • standard math The score function provides lossless compression, with score covariance equal to Fisher information, when fiducial parameters are close to truth.
    Standard statistical result invoked in Sec. II.A to justify optimality.
  • ad hoc to paper The posterior average over lensing maps can be approximated by the MAP point.
    Used to construct practical scores (Geneva, MUSE) from Eq. (4.1) and Sec. IV.
  • ad hoc to paper The second-order response term in the score is degenerate with delensed CMB power spectrum to leading order, so it can be discarded.
    Argued in Sec. III.A and Appendix D.2; the degeneracy is broken by projection effects, but this is neglected.

how reviews work

0 comments
Cite this review

Pith. "Pith review of On likelihood-based analysis of the gravitationally (de)lensed CMB." pith.science (2026). https://pith.science/paper/2FZ5OKLY

@misc{pith2026250202399,
  author       = {Pith},
  title        = {Pith review of: On likelihood-based analysis of the gravitationally (de)lensed CMB},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2FZ5OKLY}},
  note         = {Machine review of arXiv:2502.02399}
}
read the original abstract

By reducing variance induced by gravitational lensing, likelihood-based de-lensing techniques have true potential to extract significantly more information from deep and high-resolution Cosmic Microwave Background (CMB) data than traditional methods. We derive here optimal data compression statistics for the lensed CMB, and clarify the role of each term, demonstrating their direct analogs in the quadratic estimator (QE) framework. We discuss in this light pros and cons of practical implementations, including the MUSE approach, as used in the latest SPT-3G cosmological analysis, and give improvements. We discuss pathways for porting the large robustness and redundancy toolbox of the QE approach to beyond-QE with simple means.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 14 canonical work pages

  1. [1]

    Gradient inversion 13

    work page

  2. [2]

    On likelihood-based analysis of the gravitationally (de)lensed CMB

    Delensed 2-pt information 14 I. INTRODUCTION The Cosmic Microwave Background (CMB) provides one of the most powerful observational windows into the early ∗ julien.carron@unige.ch universe and fundamental physics. The study of gravitational lensing of the CMB, caused by the large-scale distribution of matter along the line of sight, has become an important...

    work page Pith review arXiv 2025

  3. [3]

    If the approximation captures the relevant elements, one will then make a good job at inference, by computing this observable from the data and building a likelihood for it

    Obviously the score is rarely as simple or practical than this example, but may still be used to try and design efficient observables, by approximating it with some- thing usable (‘sUSABLE’). If the approximation captures the relevant elements, one will then make a good job at inference, by computing this observable from the data and building a likelihood for it

    work page

  4. [4]

    (2.2), where ⟨sCL ⟩ = 0 in for consistent fidu- cial model))

    The score always vanishes in the mean, provided the fiducial model matches the truth (as can be seen clearly in Eq. (2.2), where ⟨sCL ⟩ = 0 in for consistent fidu- cial model)). This suggests another approach, poten- tially useful for example when the construction of the score statistic or it approximation heavily depends on a fiducial value for the param...

    work page

  5. [5]

    Finally, the prefactor (2L + 1)/2C 2 L in Eq. (2.2) is the inverse well-known variance of the spectrum statis- tic ˆCL: whenever the score can be treated as lin- ear in the parameter of interest, the latter is always inverse-variance weighted. A less trivial example is that of CMB lensing (and other) quadratic estimators, for which the normalization (the ...

    work page

  6. [6]

    Lewis and A

    A. Lewis and A. Challinor, Weak gravitational lensing of the cmb, Phys. Rept. 429, 1 (2006), arXiv:astro-ph/0601594 [astro- ph]

    arXiv 2006

  7. [7]

    Carron, M

    J. Carron, M. Mirmelstein, and A. Lewis, CMB lensing from Planck PR4 maps, JCAP 09, 039, arXiv:2206.07773 [astro- ph.CO]

    arXiv

  8. [8]

    M. S. Madhavacheril et al. (ACT), The Atacama Cosmology Telescope: DR6 Gravitational Lensing Map and Cosmological Parameters, (2023), arXiv:2304.05203 [astro-ph.CO]

    arXiv 2023

Show all 41 references
  1. [9]

    Ge et al

    F. Ge et al. (SPT-3G), Cosmology From CMB Lensing and De- lensed EE Power Spectra Using 2019-2020 SPT-3G Polariza- tion Data, (2024), arXiv:2411.06000 [astro-ph.CO]

    2024 arXiv

  2. [10]

    Ade et al

    P. Ade et al. (Simons Observatory), The Simons Observatory: Science goals and forecasts, JCAP 02, 056, arXiv:1808.07445 [astro-ph.CO]

    arXiv

  3. [11]

    P. A. R. Ade et al. (BICEP, Keck), Improved Constraints on Primordial Gravitational Waves using Planck, WMAP, and BI- CEP/Keck Observations through the 2018 Observing Season, Phys. Rev. Lett. 127, 151301 (2021), arXiv:2110.00483 [astro- ph.CO]

    2021

  4. [12]

    Abazajian et al

    K. Abazajian et al. (CMB-S4), CMB-S4: Forecasting Con- straints on Primordial Gravitational Waves, Astrophys. J. 926, 54 (2022), arXiv:2008.12619 [astro-ph.CO]

    2022 arXiv

  5. [13]

    Belkner, J

    S. Belkner, J. Carron, L. Legrand, C. Umilt `a, C. Pryke, and C. Bischoff (CMB-S4), CMB-S4: Iterative internal delensing and r constraints, (2023), arXiv:2310.06729 [astro-ph.CO]

    2023 arXiv

  6. [14]

    Green, J

    D. Green, J. Meyers, and A. van Engelen, CMB Delensing Be- yond the B Modes, (2016), arXiv:1609.08143 [astro-ph.CO]

    2016 arXiv

  7. [15]

    W. R. Coulton, P. D. Meerburg, D. G. Baker, S. Hotinli, A. J. Duivenvoorden, and A. van Engelen, Minimizing gravitational lensing contributions to the primordial bispectrum covariance, Phys. Rev. D 101, 123504 (2020), arXiv:1912.07619 [astro- ph.CO]

    2020 arXiv

  8. [16]

    S. C. Hotinli, J. Meyers, C. Trendafilova, D. Green, and A. van Engelen, The benefits of CMB delensing, JCAP 04 (04), 020, arXiv:2111.15036 [astro-ph.CO]

    arXiv

  9. [17]

    K. N. Abazajian et al. (CMB-S4), CMB-S4 Science Book, First Edition, (2016), arXiv:1610.02743 [astro-ph.CO]

    2016 arXiv

  10. [18]

    Okamoto and W

    T. Okamoto and W. Hu, CMB lensing reconstruction on the full sky, Phys. Rev. D67, 083002 (2003), arXiv:astro-ph/0301031 [astro-ph]

    2003 arXiv

  11. [19]

    Hu and T

    W. Hu and T. Okamoto, Mass reconstruction with cmb polariza- tion, Astrophys. J. 574, 566 (2002), arXiv:astro-ph/0111606

    2002 arXiv

  12. [20]

    A. S. Maniyar, Y . Ali-Ha¨ımoud, J. Carron, A. Lewis, and M. S. Madhavacheril, Quadratic estimators for CMB weak lensing, Phys. Rev. D 103, 083524 (2021), arXiv:2101.12193 [astro- ph.CO]

    2021 arXiv

  13. [21]

    C. M. Hirata and U. Seljak, Analyzing weak lensing of the cos- 10 mic microwave background using the likelihood function, Phys. Rev. D67, 043001 (2003), arXiv:astro-ph/0209489 [astro-ph]

    2003 arXiv

  14. [22]

    C. M. Hirata and U. Seljak, Reconstruction of lensing from the cosmic microwave background polarization, Phys. Rev. D68, 083002 (2003), arXiv:astro-ph/0306354 [astro-ph]

    2003 arXiv

  15. [23]

    Carron and I

    J. Carron and I. Szapudi, Optimal non-linear transformations for large scale structure statistics, Mon. Not. Roy. Astron. Soc. 434, 2961 (2013), arXiv:1306.1230 [astro-ph.CO]

    2013 arXiv

  16. [24]

    Alsing and B

    J. Alsing and B. Wandelt, Generalized massive optimal data compression, Mon. Not. Roy. Astron. Soc. 476, L60 (2018), arXiv:1712.00012 [astro-ph.CO]

    2018 arXiv

  17. [25]

    Millea and U

    M. Millea and U. Seljak, Marginal unbiased score expansion and application to CMB lensing, Phys. Rev. D 105, 103531 (2022), arXiv:2112.09354 [astro-ph.CO]

    2022 arXiv

  18. [26]

    Millea, E

    M. Millea, E. Anderes, and B. D. Wandelt, Sampling-based in- ference of the primordial CMB and gravitational lensing, Phys. Rev. D 102, 123542 (2020), arXiv:2002.00965 [astro-ph.CO]

    2020 arXiv

  19. [27]

    Legrand and J

    L. Legrand and J. Carron, Lensing power spectrum of the cosmic microwave background with deep polarization exper- iments, Phys. Rev. D 105, 123519 (2022), arXiv:2112.05764 [astro-ph.CO]

    2022 arXiv

  20. [28]

    Legrand and J

    L. Legrand and J. Carron, Robust and efficient CMB lens- ing power spectrum from polarization surveys, (2023), arXiv:2304.02584 [astro-ph.CO]

    2023 arXiv

  21. [29]

    Horowitz, S

    B. Horowitz, S. Ferraro, and B. D. Sherwin, Reconstructing Small Scale Lenses from the Cosmic Microwave Background Temperature Fluctuations, Mon. Not. Roy. Astron. Soc. 485, 3919 (2019), arXiv:1710.10236 [astro-ph.CO]

    2019 arXiv

  22. [30]

    Hadzhiyska, B

    B. Hadzhiyska, B. D. Sherwin, M. Madhavacheril, and S. Fer- raro, Improving Small-Scale CMB Lensing Reconstruction, Phys. Rev. D 100, 023547 (2019), arXiv:1905.04217 [astro- ph.CO]

    2019 arXiv

  23. [31]

    M. M. Schmittfull, A. Challinor, D. Hanson, and A. Lewis, Joint analysis of CMB temperature and lensing-reconstruction power spectra, Phys. Rev. D 88, 063012 (2013), arXiv:1308.0286 [astro-ph.CO]

    2013 arXiv

  24. [32]

    Peloton, M

    J. Peloton, M. Schmittfull, A. Lewis, J. Carron, and O. Zahn, Full covariance of CMB and lensing reconstruction power spec- tra, Phys. Rev. D95, 043508 (2017), arXiv:1611.01446 [astro- ph.CO]

    2017 arXiv

  25. [33]

    Schaan and S

    E. Schaan and S. Ferraro, Foreground-Immune Cosmic Mi- crowave Background Lensing with Shear-Only Reconstruction, Phys. Rev. Lett. 122, 181301 (2019), arXiv:1804.06403 [astro- ph.CO]

    2019 arXiv

  26. [34]

    F. J. Qu, A. Challinor, and B. D. Sherwin, CMB lensing with shear-only reconstruction on the full sky, Phys. Rev. D 108, 063518 (2023), arXiv:2208.14988 [astro-ph.CO]

    2023 arXiv

  27. [35]

    Carron and A

    J. Carron and A. Lewis, Spherical bispectrum expansion and quadratic estimators, (2024), arXiv:2404.16797 [astro-ph.CO]

    2024 arXiv

  28. [36]

    Namikawa, D

    T. Namikawa, D. Hanson, and R. Takahashi, Bias-Hardened CMB Lensing, Mon. Not. Roy. Astron. Soc. 431, 609 (2013), arXiv:1209.0091 [astro-ph.CO]

    2013 arXiv

  29. [37]

    Millea, E

    M. Millea, E. Anderes, and B. D. Wandelt, Bayesian delens- ing of CMB temperature and polarization, Phys. Rev. D 100, 023509 (2019), arXiv:1708.06753 [astro-ph.CO]

    2019 arXiv

  30. [38]

    Carron, Optimal constraints on primordial gravitational waves from the lensed CMB, Phys

    J. Carron, Optimal constraints on primordial gravitational waves from the lensed CMB, Phys. Rev. D 99, 043518 (2019), arXiv:1808.10349 [astro-ph.CO]. Appendix A: Marginalized vs joint posterior The log-posterior density for the lensing map is given by the combination of the an...

    2019 arXiv

  31. [39]

    One operational consequence is to remove in the curvature matrix all terms without explicit dependence on the data

    Gradient inversion Consider ignoring mean-fields, which is often adequate on small scales. One operational consequence is to remove in the curvature matrix all terms without explicit dependence on the data. Another consequence is that ¯X is driven to zero close to a maximum li...

  32. [40]

    (D5) As usual, ˆn′ is the undeflected position corresponding toˆn according to the delenser, ˆn′ ∼ ˆn+∇ϕ(ˆn)

    where K ≡ B†Cov−1 κ B. (D5) As usual, ˆn′ is the undeflected position corresponding toˆn according to the delenser, ˆn′ ∼ ˆn+∇ϕ(ˆn). Viewed as a matrix acting on arbitrary lensing deflection vector fields,H sends to zero any vector field orthogonal to the undeflected, Wiener-f...

  33. [41]

    We justify the intuitive claim that this is (close to) equivalent to the 2-pt information of the delensed CMB, and hence carries little independent information

    Delensed 2-pt information Part of the lensing spectrum score function is the M-average of the diagonal of the non-perturbative second order response terms. We justify the intuitive claim that this is (close to) equivalent to the 2-pt information of the delensed CMB, and hence ...

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.