Thermal properties of the $ρ$ meson from a machine-learning-improved holographic QCD model

Abstract

The thermal properties of the \(ρ\) meson are revisited within the soft-wall AdS/QCD framework. The dilaton background is calibrated against lattice QCD screening-mass data through a neural-network-assisted inverse procedure. Using this calibrated background, we compute the temporal vector spectral function and the lowest vector quasi-normal modes. The pole mass \(m_{\mathrm{pole}}(T)\) and thermal width \(Γ(T)\) are then extracted from the corresponding complex quasi-normal mode frequencies. Adopting the vanishing real part of the quasi-normal mode frequency, \(ω_{\mathrm{Re}}(T_d)=0\), as a criterion for meson dissociation, we estimate \(T_d\approx168\,\mathrm{MeV}\). As the temperature increases, the thermal width grows, and the spectral function evolves from a narrow Breit-Wigner-like resonance peak into a transport peak, confirming the QNM prediction that the vanishing of the pole mass is accompanied by the emergence of a transport peak and making it difficult to identify a quasiparticle peak. A numerical extraction of the dispersion relation shows that the propagator can be described by a finite-temperature simple-pole form with momentum-dependent thermal width, characterized by a diffusion coefficient \(D(T)\) and an effective propagation speed \(u(T)\), whose temperature dependences are also extracted. This form modifies the usual relation between the screening and pole masses. These results may provide a useful reference for effective studies of the hot hadronic medium.