Neural Network Construction of the Equation of State from Relativistic ab initio Calculations

Constraining the nuclear matter equation of state (EOS) beyond saturation density is crucial in nuclear physics and astrophysics. While the relativistic Brueckner–Hartree–Fock (RBHF) theory, an ab initio, nonperturbative nuclear many-body theory starting from realistic interactions, accurately describes nuclear matter properties near the saturation density ρ0 ≈ 0.16 fm−3, its applicability is limited to densities up to 3ρ0, necessitating a reliable extrapolation to higher densities. In this work, we employ supervised machine learning to train thousands of neural networks on low-density RBHF data. By enforcing thermodynamic consistency and smoothness, we select a subset of 264 optimal models employing the Swish activation function, which we identify as the most reliable choice for stable extrapolation after extensive testing and comparison. Using these models to extend the EOS to high density, we obtain the nuclear matter symmetry energy and then compute the neutron star mass–radius relation and tidal deformability, consistent with current astronomical observations. The corresponding extrapolation uncertainty originates from the combined contributions of both the 264 optimal models and the linear regression on nuclear matter EOS, yielding a symmetry energy of Esym(5ρ0) = 136.0 ± 52.8 MeV, a pressure of P(5ρ0) = 346.3 ± 97.4 MeV/fm3, a maximum neutron star mass of Mmax=2.18±0.18M⊙ , and a tidal deformability of Λ1.4M⊙=532±34 . This work establishes a general and data-driven framework to explore dense matter EOS by integrating ab initio calculations with modern machine learning techniques.