The search for primordial gravitational waves via the curl-like (B-mode) polarization of the cosmic microwave background (CMB), parameterized by the tensor-to-scalar ratio rr, is limited by polarized Galactic foregrounds whose true complexity is not yet established observationally. Errard, Feeney, Peiris and Jaffe (2016) built the reference end-to-end forecasting framework — parametric component separation, delensing, and Fisher-matrix forecasts (parameter-error forecasts from the local curvature of the likelihood around a fiducial model) — for ground-based, balloon and satellite experiments, under two simplifications they explicitly deferred: only thermal dust and synchrotron are parametrically modeled, and the fitted foreground model is assumed to be correct. Subsequent work has sharpened both issues without resolving them. Planck data now demonstrate that the dust polarization spectral energy distribution varies across the sky; the moment-expansion formalism has been developed to model spectral distortions arising from line-of-sight and sky-scale parameter variations; and Simons Observatory and LiteBIRD studies in 2024–2026 show that complex foreground scenarios bias rr by more than 2σ2\sigma in default pipelines, and that the canonical foreground description will reach its limits at the σ(r)∼10−3\sigma(r)\sim10^{-3} sensitivity of next-generation missions. The problem remains open: which polarized components are present at the relevant level (including anomalous microwave emission, magnetic-dipole emission, and CO line emission, whose polarization is essentially unmeasured), what the statistics and angular scale of spectral-parameter variation are, how much bias foreground-model mis-specification injects into rr through the full component-separation-to-inference chain, and which experimental designs are robust to the observationally allowed complexity — with the decisive empirical characterization awaiting LiteBIRD, Simons Observatory and CMB-S4 data.

Background

The cosmic microwave background (CMB) carries linear polarization, conventionally decomposed into a gradient-like E-mode component and a curl-like B-mode component. A stochastic background of primordial gravitational waves from inflation would imprint degree-scale B modes whose amplitude is parameterized by the tensor-to-scalar ratio rr; current bounds place rr below a few times 10−210^{-2}. Two contaminants obscure this signal. First, polarized Galactic foregrounds: thermal emission from interstellar dust grains aligned with the Galactic magnetic field (dominant above roughly 100 GHz), and synchrotron radiation from cosmic-ray electrons spiraling in the same field (dominant below roughly 100 GHz). Possible additional polarized components include anomalous microwave emission (AME, usually attributed to rapidly spinning tiny dust grains), magnetic-dipole emission from thermal dust grains, and rotational line emission from carbon monoxide (CO). Second, gravitational lensing: deflections by intervening large-scale structure convert E modes into an additional B-mode contribution that must be statistically removed ('delensing'), using either the CMB's own four-point function or cross-correlation with tracers of the lensing potential such as the cosmic infrared background (CIB) or galaxy surveys.

Component separation exploits the different frequency dependence (spectral energy distribution, SED) of sky components. In the parametric maximum-likelihood approach, a mixing matrix A\mathbf{A} encodes each component's emission law — a power law with spectral index βs\beta_s for synchrotron, a modified blackbody with index βd\beta_d and temperature TdT_d for dust — whose parameters are fitted from multi-frequency maps before the CMB is recovered by linear inversion. Imperfect knowledge of A\mathbf{A} leaves foreground residuals in the cleaned map, conventionally summarized as an effective residual amplitude reffr_{\rm eff} in units of the primordial B-mode power at r=1r=1.

Errard, Feeney, Peiris and Jaffe (2016) combined these ingredients into a single forecasting pipeline for a suite of experiments and found that while sensitive post-2020 configurations could reach σ(r)∼10−4\sigma(r)\sim10^{-4}, their large-scale constraints could be biased by foreground residuals. Their analysis rested on simplifications the paper itself flags as open issues. Only dust and synchrotron were modeled; polarized AME, magnetic-dipole emission and CO lines were explicitly deferred. The foreground model was assumed correct — the bias from cleaning the sky with a wrong or incomplete model class was not computed (the paper cites a contemporaneous treatment of this systematic at large angular scales by Remazeilles et al. 2016). Spatial variation of the spectral indices was handled in two idealized ways: the 'npn_p approach', fitting independent spectral indices in patches of roughly 15 degrees across the observed sky, and the 'A\mathbf{A}-expansion', a first-order Taylor expansion of the mixing matrix in βd\beta_d and βs\beta_s; the paper states that data available at the time could not decide between them because the angular scale and amplitude of spectral-index variation were not yet measured. Residuals were treated as an additive Gaussian bias with parametrized amplitude and tilt, and instrumental systematics (gain, polarization-angle, beam and bandpass errors) were left for future work.

Since 2016, a substantial literature has grown around exactly these gaps. The moment expansion — extending the mixing matrix with moments of the joint distribution of dust parameters along each line of sight, generalizing the A-expansion — has become the standard parametrization of SED distortions caused by variations of dust temperature and emissivity index within the beam. Planck analyses have detected spatial variation of the dust polarization SED on large angular scales. End-to-end pipeline comparisons for the Simons Observatory (SO) and forecasts for the LiteBIRD satellite now treat foreground complexity, rather than instrumental noise alone, as the factor limiting credible constraints on rr at the 10−310^{-3} level.

Problem Statement

As posed by Errard et al. (2016) and sharpened by the subsequent literature, the open problem is to establish the complexity of the polarized Galactic foregrounds at CMB frequencies and the robustness of primordial B-mode constraints to how that complexity is modeled, at the sensitivity of experiments targeting σ(r)∼10−3\sigma(r)\sim10^{-3}. The experimental coverage required to resolve the three parts below is bounded to this next-generation suite — Simons Observatory, LiteBIRD and CMB-S4 — at their design sensitivities; coverage of further configurations is optional. Concretely:

(1) Empirical foreground complexity. Determine from multi-frequency polarization data which polarized components are present at observationally relevant levels — thermal dust, synchrotron, anomalous microwave emission, magnetic-dipole emission, CO line emission — and characterize their spectral energy distributions: the mean and dispersion of the dust and synchrotron spectral indices and dust temperature, how these vary across the sky, along the line of sight, and with angular scale (the characteristic coherence scale of the variation), so that the choice among foreground parametrizations — patchwise-constant spectral indices (the npn_p approach), mixing-matrix expansion, or moment expansion — can be grounded in measurement rather than assumption, for each experiment's chosen sky area and depth. A resolution is to be reached from already published multi-frequency polarization measurements (e.g. the Planck polarization analyses of refs. 11 and 16) and, where the required observation does not yet exist, from a stated forecast; for the components whose polarization is not yet measured — anomalous microwave emission, magnetic-dipole emission and CO line emission — an answer that reports them as undetermined together with their current upper limits, rather than as measured detections, is complete.

(2) Mis-specification bias. Quantify, by end-to-end propagation from multi-frequency maps through component separation, delensing and parameter inference, the bias and variance on rr produced when the sky is cleaned with an incomplete or wrong foreground model — including unmodeled components, spatially varying spectral parameters, and non-Gaussian or spatially correlated residuals — as a function of experimental configuration; and develop component-separation and inference methods whose mis-specification bias remains subdominant to the statistical error across the observationally allowed complexity envelope. An answer resolves this part either by providing such a method over the allowed envelope or by demonstrating the region of the envelope where subdominance is not attainable, together with the bias that remains there.

(3) Design robustness. From the empirically established complexity, determine the foreground-robust experimental choices — frequency allocations, sky splits, and the trade-off between deep small patches and wider shallower surveys — for the next-generation configurations named above, stating the assumed complexity envelope it is derived under. A one-time determination over these configurations resolves the sub-question; revision as new foreground data arrive is expected follow-up work, not a completion condition.

References

  1. Josquin Errard, Stephen M. Feeney, Hiranya V. Peiris, Andrew H. Jaffe (2016). 'Robust forecasts on fundamental physics from the foreground-obscured, gravitationally-lensed CMB polarization.' JCAP 03 (2016) 052, DOI 10.1088/1475-7516/2016/03/052, arXiv:1509.06770, https://arxiv.org/abs/1509.06770
  2. M. Remazeilles, C. Dickinson, H. K. K. Eriksen, I. K. Wehus (2016). 'Sensitivity and foreground modelling for large-scale cosmic microwave background B-mode polarization satellite missions.' MNRAS 458 (2016) 2032–2050, DOI 10.1093/mnras/stw441, arXiv:1509.04714, https://arxiv.org/abs/1509.04714
  3. F. Vansyngel, F. Boulanger, T. Ghosh, B. D. Wandelt, J. Aumont, A. Bracco, F. Levrier, P. G. Martin, L. Montier (2017). 'Statistical simulations of the dust foreground to cosmic microwave background polarization.' A&A 603 (2017) A62, DOI 10.1051/0004-6361/201629992, arXiv:1611.02577, https://arxiv.org/abs/1611.02577
  4. Byeonghee Yu, J. Colin Hill, Blake D. Sherwin (2017). 'Multitracer CMB delensing maps from Planck and WISE data.' Phys. Rev. D 96 (2017) 123511, DOI 10.1103/PhysRevD.96.123511, arXiv:1705.02332, https://arxiv.org/abs/1705.02332
  5. Alessandro Manzotti (2018). 'Future cosmic microwave background delensing with galaxy surveys.' Phys. Rev. D 97 (2018) 043527, DOI 10.1103/PhysRevD.97.043527, arXiv:1710.11038, https://arxiv.org/abs/1710.11038
  6. M. Remazeilles, A. J. Banday, C. Baccigalupi, et al. (CORE Collaboration) (2018). 'Exploring Cosmic Origins with CORE: B-mode Component Separation.' JCAP 04 (2018) 023, DOI 10.1088/1475-7516/2018/04/023, arXiv:1704.04501, https://arxiv.org/abs/1704.04501
  7. Maximilian H. Abitbol, David Alonso, Sara M. Simon, Jack Lashner, et al. (Simons Observatory Collaboration) (2021). 'The Simons Observatory: gain, bandpass and polarization-angle calibration requirements for B-mode searches.' JCAP 05 (2021) 032, DOI 10.1088/1475-7516/2021/05/032, arXiv:2011.02449, https://arxiv.org/abs/2011.02449
  8. L. Vacher, J. Aumont, L. Montier, S. Azzoni, F. Boulanger, M. Remazeilles (2022). 'Moment expansion of polarized dust SED: a new path towards capturing the CMB B-modes with LiteBIRD.' A&A 660 (2022) A111, DOI 10.1051/0004-6361/202142664, arXiv:2111.07742, https://arxiv.org/abs/2111.07742
  9. Danielle Sponseller, Alan Kogut (2022). 'Mitigating Bias in CMB B-modes from Foreground Cleaning Using a Moment Expansion.' The Astrophysical Journal 936 (2022) 8, DOI 10.3847/1538-4357/ac846f, arXiv:2207.13109, https://arxiv.org/abs/2207.13109
  10. Toshiya Namikawa, Anton Baleato Lizancos, Naomi Robertson, Blake D. Sherwin, et al. (Simons Observatory Collaboration) (2022). 'Simons Observatory: Constraining inflationary gravitational waves with multitracer B-mode delensing.' Phys. Rev. D 105 (2022) 023511, DOI 10.1103/PhysRevD.105.023511, arXiv:2110.09730, https://arxiv.org/abs/2110.09730
  11. Alessia Ritacco, François Boulanger, Vincent Guillet, Jean-Marc Delouis, Jean-Loup Puget, Jonathan Aumont, Léo Vacher (2023). 'Dust polarization spectral dependence from Planck HFI data. Turning point on CMB polarization foregrounds modelling.' A&A 670 (2023) A163, DOI 10.1051/0004-6361/202244269, arXiv:2206.07671, https://arxiv.org/abs/2206.07671
  12. T. Namikawa, A. I. Lonappan, C. Baccigalupi, N. Bartolo, et al. (LiteBIRD Collaboration) (2024). 'LiteBIRD science goals and forecasts: improving sensitivity to inflationary gravitational waves with multitracer delensing.' JCAP 06 (2024) 010, DOI 10.1088/1475-7516/2024/06/010, arXiv:2312.05194, https://arxiv.org/abs/2312.05194
  13. Kevin Wolz, Susanna Azzoni, Carlos Hervías-Caimapo, Josquin Errard, Nicoletta Krachmalnicoff, David Alonso, Benjamin Beringue, Emilie Hertig (2024). 'The Simons Observatory: component separation pipelines for B-modes.' arXiv:2407.06891, https://arxiv.org/abs/2407.06891
  14. Léo Vacher, Alessandro Carones, Jonathan Aumont, Jens Chluba, Nicoletta Krachmalnicoff, Claudio Ranucci, Mathieu Remazeilles, Arianna Rizzieri (2025). 'How bad could it be? Modelling the 3D complexity of the polarised dust signal using moment expansion.' A&A 697 (2025) A212, DOI 10.1051/0004-6361/202453066, arXiv:2411.11649, https://arxiv.org/abs/2411.11649
  15. Yiqi Liu, Susanna Azzoni, Susan E. Clark, Brandon S. Hensley, Léo Vacher, David Alonso, Carlo Baccigalupi, Michael L. Brown, Alessandro Carones, Jens Chluba, Jo Dunkley, Carlos Hervías-Caimapo, Bradley R. Johnson, Nicoletta Krachmalnicoff, Giuseppe Puglisi, Mathieu Remazeilles, Kevin Wolz (The Simons Observatory Collaboration) (2025). 'The Simons Observatory: assessing the impact of dust complexity on the recovery of primordial B-modes.' JCAP 11 (2025) 024, DOI 10.1088/1475-7516/2025/11/024, arXiv:2508.00073, https://arxiv.org/abs/2508.00073
  16. Vincent Guillet, Léo Vacher, Jonathan Aumont, François Boulanger, Alessia Ritacco, Jean-Marc Delouis, Andrea Bracco (2026). 'Variance of dust temperature and spectral index in Planck polarization data using spin-moment expansion.' A&A 705 (2026) A177, DOI 10.1051/0004-6361/202556491, arXiv:2510.18305, https://arxiv.org/abs/2510.18305
  17. S. Vinzl, J. Aumont, L. Vacher, R. T. Génova-Santos, D. Adak, A. Rizzieri, et al. (LiteBIRD Collaboration) (2026). 'Galactic Science with the LiteBIRD satellite: Spectral characterization of diffuse Galactic polarized emission at the angular power spectrum level.' arXiv:2607.20080, https://arxiv.org/abs/2607.20080
  18. S. Vinzl, J. Aumont, L. Vacher, R. T. Génova-Santos, D. Adak, A. Rizzieri, for the LiteBIRD Collaboration (2026). 'Contribution to the 2026 Cosmology session of the 60th Rencontres de Moriond: Galactic moments: Understanding polarized foregrounds complexity in the quest for CMB primordial B modes with LiteBIRD.' arXiv:2607.21247, https://arxiv.org/abs/2607.21247