Emulating QRPA Response Calculations

Surrogate models, or emulators, have been increasingly employed in the nuclear physics community to reduce the high cost associated with computing ground and excited state properties of many-body systems for large-scale systematic studies and uncertainty quantification applications. We have recently developed a physics-informed, machine learning-driven emulator for nuclear response functions, with applications to charge-exchange and like-particle modes. While not restricted to any given model, we have validated the approach with microscopic simulations from a covariant energy density functional (EDF) framework using the quasiparticle random phase approximation (QRPA) for the excited state properties. This emulator accurately reproduces collective response functions – including giant resonances, spin-isospin excitations, and other modes relevant to astrophysical processes – while operating at a fraction of the computational cost of full QRPA calculations. We benchmark the emulator by reproducing dipole polarizabilities and β-decay half-lives, demonstrating that its physics-informed structure yields stable and reliable predictions even in extrapolative regimes. By enabling rapid evaluation of response observables, this tool paves the way for efficient Bayesian calibration of effective nuclear interactions, offering a systematic means to constrain the nuclear equation of state, assess collective phenomena in excited states, and quantify theoretical uncertainties.