Power spectra based Planck constraints on compensated isocurvature, and forecasts for LiteBIRD and CORE space missions

Abstract

Compensated isocurvature perturbations (CIP), where the primordial baryon and cold dark matter density perturbations cancel, do not cause total matter isocurvature perturbation. Consequently, at the linear order in the baryon density contrast Δ, a mixture of CIP and the adiabatic mode leads to the same CMB spectra as the pure adiabatic mode. Only recently, Muñoz et al. showed that at the second order CIP leaves an imprint in the observable CMB by smoothing the power spectra in a similar manner as lensing. This causes a strong degeneracy between the CIP variance Δrms2 ≡ ⟨Δ2⟩ and the phenomenological lensing parameter AL. We study several combinations of the Planck 2015 data and show that the measured lensing potential power spectrum Cℓϕϕ breaks the degeneracy. Nested sampling of the ΛCDM+Δrms2(+AL) model using the Planck 2015 temperature, polarization, and lensing data gives Δrms2 = (6.9+3.0−3.1) × 10−3 at 68% CL. A non-zero value is favoured at 2.3σ (or without the polarization data at 2.8σ). CIP with Δrms2 ≈ 7 × 10−3 improves the bestfit χ2 by 3.6 compared to the adiabatic ΛCDM model. In contrast, although the temperature data favour AL ≃ 1.22, allowing AL ≠ 1 does not improve the joint fit at all, since the lensing data disfavour AL ≠ 1. Indeed, CIP provides a rare example of a simple model, which is capable of reducing the Planck lensing anomaly significantly and fitting well simultaneously the high (and low) multipole temperature and lensing data, as well as the polarization data. Finally, we derive forecasts for two future satellite missions (LiteBIRD proposal to JAXA/NASA and Exploring Cosmic Origins with CORE proposal to ESA's M5 call) and compare these to simulated Planck data. Due to its coarse angular resolution, LiteBIRD is not able to improve the constraints on Δrms2 or AL, but CORE-M5 (almost) reaches the cosmic variance limit and improves the CIP constraint to Δrms2 < 0.6 (1.4) × 10−3 at 68 (95)% CL, which is nine times better than the current trispectrum based upper bound and six times better than obtained from the simulated Planck data. In addition, CORE-M5 will exquisitely distinguish between Δrms2 and AL. No matter whether CIP is allowed for or not, the uncertainty of the lensing parameter will be σ(AL) ≈ 0.012, in the case where the simulated data are based on the adiabatic ΛCDM model with AL = 1.