Abstract :
We construct an algorithm to simulate imaginary time evolution of translationally invariant spin systems with local interactions on an infinite, symmetric tree. We describe the state by symmetric infinite-Tree Tensor Network (iTTN) and use translation-invariant operators for the updates at each time step. The contraction of this tree tensor network can be computed efficiently by recursion without approximations and one can then truncate all the iTTN tensors at the same time. The translational symmetry is preserved at each time step that makes the algorithm very well conditioned and stable. The computational cost scales like image with the bond dimension image and coordination number image, much favorable than that of the iTEBD on trees [D. Nagaj, E. Farhi, J. Goldstone, P. Shor, I. Sylvester, Phys. Rev. B 77 (2008) 214431]. Studying the transverse-field Ising model on the Bethe lattice, the numerics indicate a ferromagnetic-paramagnetic phase transition, with a finite correlation length even at the transition point.
Keywords :
Infinite tree tensor network (iTTN) , Imaginary time evolution , Phase transitions , Ising model on trees