The choice of Planck CMB likelihood in cosmological analyses

TL;DR

This work evaluates how cosmological inferences from Planck data depend on the choice of sky maps and likelihood pipelines (Plik-Legacy vs CamSpec-NPIPE, PR3 vs PR4). It introduces a foreground-marginalized CamSpec-NPIPE-lite likelihood to enable robust combinations with other CMB data, and it assesses LCDM and selected extensions using Planck TT/TE/EE+lowT+lowE, alone and in combination with ACT DR6. The main finding is that, for the Planck multipole range used with ground-based data, cosmological constraints are consistent across Planck products, with small shifts alleviated when adding ACT; polarization data and small-scale temperature data drive the extra constraining power, while full nuisance marginalization is crucial when truncating scales. These results support the reliability of Planck-plus-ground-based analyses for extended cosmologies and provide a practical lite likelihood for cross-dataset cosmology.

Abstract

We compare cosmological parameters from different Planck sky maps and likelihood pipelines, assessing robustness of cosmological results with respect to the choice of the latest Planck maps-likelihood combination. We show that, for the Planck multipole range retained in combination with ground-based observations, different products give very similar cosmological solutions; small remaining differences are reduced by the addition of other CMB datasets to Planck. In particular, constraints on extended cosmological models benefit from the addition of small-scale power from ground-based experiments and are completely insensitive to the choice of Planck maps and likelihood. For this work we derive and release a nuisance-marginalized dataset and CamSpec-NPIPE-lite likelihood for the Planck NPIPE data injected into the CamSpec likelihood - which are usually used to obtain the reference Planck PR4 cosmology. Using the extracted CMB spectra we show that the additional constraining power for cosmology is coming from polarization at all scales and from temperature at multipoles above 1500 when going from PR3 to PR4. We also show that full marginalization over the CamSpec foreground nuisance parameters can impact parameter inference and model selections when truncating some scales; our new likelihood enables correct combinations with other CMB datasets.

Figures (10)

• Figure 1: CMB foreground-marginalized CamSpec-NPIPE-lite spectra. The TE and EE spectra appear normal distributed around the best-fitting model, as does the TT spectrum below $\ell < 1500$. At $\ell > 1500$, the TT datapoints are highly correlated and the full behaviour of the data is captured in the covariance matrix.
• Figure 2: The TTTT subblock of the correlation matrix. The marginalization over foregrounds captures a lot of information in the $\ell > 1500$ part of the covariance, leading to high correlations between different bins.
• Figure 3: Posterior distributions of the foreground nuisance parameters from the multi-frequency likelihood (orange for the original CamSpec-NPIPE likelihood and green for our implementation in MFLike), and from the foreground marginalization procedure (blue). Due to the fact that each foreground appears in only one spectrum each, they are fully degenerate in the foreground marginalization procedure, and thus highly unconstrained. Orange and green posteriors are obtained with a fit to the full Planck TT/TE/EE+lowT+lowE dataset but receive contribution only from the high-$\ell$CamSpec-NPIPE likelihood. The vertical lines indicate the best-fitting values from the CamSpec-NPIPE chain (original likelihood, shown in orange).
• Figure 4: Posteriors of the $\Lambda$CDM cosmological parameters recovered from different implementations of the CamSpec-NPIPE likelihood. We find very good agreement in all five free parameters (we infer the optical depth $\tau_{\rm reio}$ through a prior), and see only a slight widening of $9\%$ on the constraint on $n_s$ from the lite likelihood due to the foreground marginalization. All posteriors are obtained with the full Planck TT/TE/EE+lowT+lowE dataset. The vertical lines indicate the best-fitting values from the CamSpec-NPIPE chain (original likelihood, shown in orange).
• Figure 5: Comparison of the constraints on select single-parameter extensions to $\Lambda$CDM as obtained from the full Planck TT/TE/EE+lowT+lowE dataset using the original CamSpec-NPIPE likelihood (orange), our implementation in MFLike (green), and our CamSpec-NPIPE-lite likelihood (blue). We show the constraints on the lensing peak smearing amplitude $A_L$ (top panel), the effective number of relativistic species $N_{\rm eff}$ (middle panel), and the running of the spectral index ${\rm d} n_s / {\rm d} \ln k$ (bottom panel). We observe that foreground marginalization has little effect on the constraints of extension models. The constraints from the lite likelihood are consistent and at most $5-10\%$ wider than those from the full likelihoods. The vertical lines show the expected values of these parameters in $\Lambda$CDM.