Numerical calculation of correlation functions by boundary condition averaging: test on the one-dimensional Hubbard model Original Research Article

We diagonalize exactly the Hamiltonian of the one-dimensional half filled Hubbard model on small clusters of sites with generalized boundary conditions (BCs), and we calculate various correlation functions. We choose the ratio of the onsite repulsion to the hopping matrix element Ut = 1, i.e. an intermediate regime. By comparing different system sizes, we find that when such correlations are averaged over different BCs, a quick convergence of the correlation functions in their dependence on the cluster size is obtained. In particular, the use of spin-dependent BCs optimizes the results for spin correlation functions.