Improved Treatment of Frequency Sums in Propagator-Renormalized Perturbation Theories

We present a massively parallel algorithm for calculating the self-energy in self-consistent finite temperature perturbation theory for lattice models. The algorithm uses analytic functions with appropriate asymptotic high frequency behavior and fast Fourier transforms to accurately calculate the self-energy at low-frequency. Traditional methods that truncate the high frequency tails of the temperature Green's function lead to `contamination' of the low-frequency behavior of the self-energy. Our algorithm is both accurate and scalable. We compare results for the Hubbard model using various techniques for handling the high frequency tails of the temperature Green's function.