Robust Non-Singular Bouncing Cosmology from Regularized Hyperbolic Field Space
Oleksandr Kravchenko
TL;DR
This work develops a complete, non-singular bouncing cosmology in a closed universe by introducing a regularized hyperbolic field space with the sigmoid metric $g^S_{\chi\chi} = \left(1 + e^{-2\alpha \phi / M_{\rm Pl}}\right)^{-1}$. The three physical boundary conditions—bounce, perturbative unitarity, and ghost-freedom—impose a unique, minimal-complexity solution solved by the logistic equation, linking hyperbolic geometry with $\alpha$-attractors. The resulting two-field model achieves a 60+ e-folds inflation after a non-singular bounce, preserves the NEC, and expands the basin of attraction by about $10^{21}$ relative to the pure exponential metric; curvature is efficiently diluted so that $|\Omega_k| < 0.001$ after inflation. Perturbations remain regular through the bounce and $\mathcal{R}$ is conserved on super-Hubble scales, with universal observational predictions $n_s \approx 0.967$, $r \approx 0.003$, independent of $\alpha$, consistent with Planck 2018 and testable by future CMB experiments.
Abstract
We present a complete and robust framework for non-singular bouncing cosmology in a closed universe, where the scalar field space is endowed with a regularized hyperbolic geometry. The field space metric $g^S_{χχ} = (1 + e^{-2αφ/M_{\mathrm{Pl}}})^{-1}$ is rigorously derived from fundamental physical principles: (i) exponential suppression during contraction enabling the bounce mechanism, (ii) asymptotic saturation to unity during inflation preserving perturbative unitarity, and (iii) positive-definiteness ensuring ghost-freedom. We prove that the sigmoid function emerges uniquely as the minimal-complexity solution satisfying these boundary conditions, with deep connections to hyperbolic geometry and cosmological $α$-attractors. Through comprehensive numerical validation, we demonstrate that the model achieves 60+ e-folds of inflation following a non-singular bounce without violating the Null Energy Condition. The sigmoid regularization expands the basin of attraction by $\sim 10^{21}$ compared to the pure exponential metric, achieving 100\% success rate across 16 orders of magnitude in initial velocities. The model is compatible with the BKL conjecture (anisotropic shear suppressed by spatial curvature) and solves the flatness problem (only $\sim 3.5$ e-folds needed for $|Ω_k| < 0.001$). Perturbation equations remain regular throughout the bounce, and the comoving curvature perturbation $\mathcal{R}$ is conserved on super-Hubble scales. Observable predictions ($n_s \approx 0.967$, $r \approx 0.003$) are independent of the regularization parameter $α$, in excellent agreement with Planck 2018 data and testable by next-generation CMB experiments.
