Cosmic-Enu: An emulator for the non-linear neutrino power spectrum

Cosmology is poised to measure the neutrino mass sum $M_\nu$ and has identified several smaller-scale observables sensitive to neutrinos, necessitating accurate predictions of neutrino clustering over a wide range of length scales. The FlowsForTheMasses non-linear perturbation theory for the massive neutrino power spectrum, $\Delta^2_\nu(k)$, agrees with its companion N-body simulation at the $10\%-15\%$ level for $k \leq 1~h/$Mpc. Building upon the Mira-Titan IV emulator for the cold matter, we use FlowsForTheMasses to construct an emulator for $\Delta^2_\nu(k)$ covering a large range of cosmological parameters and neutrino fractions $\Omega_{\nu,0} h^2 \leq 0.01$, which corresponds to $M_\nu \leq 0.93$~eV. Consistent with FlowsForTheMasses at the $3.5\%$ level, it returns a power spectrum in milliseconds. Ranking the neutrinos by initial momenta, we also emulate the power spectra of momentum deciles, providing information about their perturbed distribution function. Comparing a $M_\nu=0.15$~eV model to a wide range of N-body simulation methods, we find agreement to $3\%$ for $k \leq 3 k_\mathrm{FS} = 0.17~h/$Mpc and to $19\%$ for $k \leq 0.4~h/$Mpc. We find that the enhancement factor, the ratio of $\Delta^2_\nu(k)$ to its linear-response equivalent, is most strongly correlated with $\Omega_{\nu,0} h^2$, and also with the clustering amplitude $\sigma_8$. Furthermore, non-linearities enhance the free-streaming-limit scaling $\partial \log(\Delta^2_\nu / \Delta^2_{\rm m}) / \partial \log(M_\nu)$ beyond its linear value of 4, increasing the $M_\nu$-sensitivity of the small-scale neutrino density.