📝 SummarXiv

Abstract

Cluster Dynamical Mean Field Theory (CDMFT) with an exact diagonalization (ED) impurity solver faces exponential scaling limitations from the Hilbert space dimension. We introduce subbath CDMFT (SB-CDMFT), an alternative to the conventional CDMFT-ED method in which the discrete bath is subdivided into separate subbaths, each coupled to the cluster with distinct hybridization functions. In this approach, only one subbath at a time is actively involved in ED, dramatically reducing the computational cost by replacing a single large impurity problem with multiple smaller separate ones. Our method successfully reproduces key physical properties including particle-hole symmetry and Mott physics, while allowing for an extended bath representation at a fraction of the typical computational cost.