Computation of ground-state properties of strongly correlated many-body systems by a two-subsystem ground-state approximation

We present a new approach to compute low lying eigenvalues and corresponding eigenvectors for strongly correlated many-body systems. The method was inspired by the so-called Automated Multilevel Sub-structuring Method (AMLS). Originally, it relies on subdividing the physical space into several regions. In these sub-systems the eigenproblem is solved, and the regions are combined in an adequate way. We developed a method to partition the state space of a many-particle system in order to apply similar operations on the partitions. The tensorial structure of the Hamiltonian of many-body systems make them even more suitable for this approach. The method allows to break down the complexity of large many-body systems to the complexity of two spatial sub-systems having half the geometric size. Considering the exponential size of the Hilbert space with respect to the geometric size this represents a huge advantage. In this work, we present some benchmark computations for the method applied to the one-band Hubbard model.