Current and Future Constraints on the Primordial Power Spectrum

Abstract

The primordial scalar power spectrum provides a powerful window onto early-universe physics, including a broad class of inflationary scenarios. These models often predict a characteristic scalar spectral index $(n_s)$, nonzero running $(\alpha_s)$, or more general deviations from a pure power-law form. We find that freeing the number of light relic species $(N_{\rm eff})$ and the sum of the neutrino masses $(\sum m_\nu)$ broadens the $n_s$ error contours enough that key early-Universe models are no longer excluded. In particular, while both Starobinsky inflation and the Bi-thermal Big Bang model are ruled out at about the 95% confidence level in the six-parameter $\Lambda$CDM model, we find that in the nine-parameter $\Lambda{\rm CDM} + \alpha_s + N_{\rm eff} + \sum m_\nu$ model, CMB data from Planck, ACT, and SPT (CMB-PAS) combined with DESI DR2 give $n_s = 0.9758 \pm 0.0064$ and $\alpha_s = 0.0080 \pm 0.0064$, consistent with both models. Future CMB facilities will sharply improve constraints on $n_s$, $\alpha_s$, and, more generally, the binned primordial power spectrum $\mathcal{P}(k)$. For example, CMB-PAS constrains $e^{-2\tau}\mathcal{P}(k = 0.2\,{\rm Mpc}^{-1})$ to 0.5%, while SO-like and CMB-HD-like surveys would tighten this to 0.1%; CMB-HD would also extend this measurement to smaller, previously unprobed scales, constraining $e^{-2\tau}\mathcal{P}(k = 30\,{\rm Mpc}^{-1})$ to within a factor of ten. We release our forecast code and update the public CMB-HD likelihood and Fisher codes to support CLASS in addition to CAMB.