Using Parametric Matrix Models to Emulate Magnus IMSRG

The In-Medium Similarity Renormalization Group (IMSRG) in both its traditional and the so-called Magnus formulations is one among a variety of modern techniques that offer a path towards the exploration of medium-mass and heavy nuclei from first principles. To accelerate predictions and enable proper statistical sensitivity studies and uncertainty quantification, it is necessary to develop emulators for these approaches. Although techniques like Dynamic Mode Decomposition and Artificial Neural Networks have been applied to the IMSRG with some success, they sometimes struggle to provide sufficient accuracy or a substantial reduction in computational cost, especially in the Magnus IMSRG, where emulators need to be trained on data from late stages of the conventional flow. To overcome these issues, we employ Parametric Matrix Models (PMMs) to emulate Magnus IMSRG with relatively low computational complexity and high accuracy. We also test our new emulator's ability to calculate the IMSRG flows of different observables unseen by the original IMSRG flow. This new approach to emulating IMSRG offers a promising path forward to reduce computation time in future calculations, provide means of uncertainty quantification, and allow for the inclusion of higher order forces for relatively low computational cost.