Efficient simulation of Grassmann tensor product states

Recently, the Grassmann-tensor-entanglement renormalization group (GTERG) algorithm has been proposed as a generic variational approach to study strongly correlated boson/fermion systems [Gu et al., arXiv: 1004.2563]. However, the weakness of such a simple variational approach is that generic Grassmann tensor product states (GTPS) with large inner dimension D will contain a large number of variational parameters which are hard to be determined through usual minimization procedures. In this paper, we first introduce a standard form of GTPS which significantly simplifies the representations. Then we describe a simple imaginary-time-evolution algorithm to efficiently update the GTPS based on the fermion coherent state representation and show that all the algorithms developed for usual tensor product states (TPS) can be implemented for GTPS in a similar way. Finally, we study the environment effect for the GTERG approach and propose a simple method to further improve its accuracy. We demonstrate our algorithms by studying some simple two-dimensional free and interacting fermion systems on honeycomb lattice, including both off-critical and critical cases.