Optimized reference ensembles for the in-medium similarity renormalization group

The in-medium similarity renormalization group (IMSRG) is an ab-initio method for computing the properties of medium mass and heavy nuclei. The IMSRG flow is a continuous unitary transformation of the Hamiltonian that decouples a chosen reference state from all excitations in a given many-body basis, thereby mapping it to the true ground state. In general, this transformation induces three-, four- and higher many-body interactions that cannot be tracked explicitly due to a prohibitive computational cost. We must truncate the induced operators at a certain point, which introduces error into the wavefunction solutions.

In principle, the IMSRG gives us control over this error through the choice of the basis operators that express the flow. Here, we explore a reference state ensemble as a means to equip this basis with information about the low-lying states of the many-body system. We show that a configurable mixture of basis states instead of a single state can robustly reduce error in the full eigenspectrum of the IMSRG evolved Hamiltonian, while preserving simultaneous observables of the system.