Inverse-mapped density-dependent relativistic mean-field inference of the neutron-star equation of state with terrestrial nuclear, NICER, and gravitational-wave constraints

Abstract We perform a Bayesian inference of the cold dense-matter equation of state within a density-dependent relativistic mean-field (DD-RMF) model. The density-dependent couplings are reconstructed through an inverse mapping from a 12-dimensional set of nuclear-matter and high-density shape parameters. Two independent offsets, $\Delta d_\sigma$ and $\Delta d_\omega$, decouple the scalar and vector Typel--Wolter shift parameters from the isovector channel and provide a flexible isoscalar evolution at suprasaturation density. Thermodynamic consistency, beta stability, nonnegative core pressure, and the causal sound-speed condition are imposed on every accepted EOS. We analyze four likelihood combinations built from NICER mass--radius data, GW170817 tidal information, and heavy-ion collision flow constraints on the pressure of symmetric nuclear matter: NS-only, NS+HIC, NS+GW170817, and NS+HIC+GW170817. In the GW-conditioned comparison, the NS+HIC+GW170817 posterior gives $R_{1.4}=11.91^{+0.35}_{-0.35}~\mathrm{km}$, $R_{2.0}=11.40^{+0.42}_{-0.49}~\mathrm{km}$, and $M_{\max}=2.131^{+0.079}_{-0.068}\,M_\odot$ at 68\% credibility. Relative to the corresponding neutron-star baseline, HIC information narrows the high-density isoscalar sector, favors correlated scalar and vector density dependences, and reduces the high-mass stellar radius. The central microscopic response is therefore not only a narrowing of the pressure band, but a selection of the scalar-attraction and vector-repulsion balance that generates the pressure of symmetric matter. Evidence and sensitivity tests are used as diagnostics within the adopted likelihood construction: the results show a moderate dependence on the HIC uncertainty width, while using a fixed reference saturation density in the HIC likelihood leaves the conclusions essentially unchanged. These intervals should be read as conditional results within the adopted 12-parameter inverse-mapped DD-RMF family rather than as model-independent measurements.