Figure 7
From: Thermal bosons in 3d optical lattices via tensor networks
(Color online) (a) Graphical representation of the SU optimization, used in the TgPEPS algorithm. In contrast to the gPEPS algorithm61 where the optimization is performed for ground state tensors with one physical leg, in the TgPEPS the optimization is performed over thermal PEPO tensors with two physical degrees of freedom. (b) One-site and (c) two-site expectation values, as computed with a mean-field environment, in the TgPEPS scheme. All these diagrams are for a 2d honeycomb lattice, but they can be straifgtforwardly generalized to any 3d structure, such as the ones considered in the main text.
