uniform_mps

• full name: tenpy.networks.uniform_mps

• parent module: tenpy.networks

• type: module

Classes

UniformMPS(sites, ALs, ARs, ACs, Cs[, norm])

A Uniform Matrix Product State, only defined in the thermodynamic limit.

Module description

This module contains a base class for a Uniform Matrix Product State.

This is an extension of the MPS class for tangent space algorithms like VUMPS and TDVP (even though the current TDVP algorithm does not use this).

A uniform MPS differs from a canonical MPS in the tensors that are stored on each site. In a canonical MPS, we store a single tensor on each site and diagonal Schmidt coefficients on each bond. From these, we can construct any desired form of a tensor on each site; e.g. given B_i, we can construct A_i = S_i B_i S_{i+1}^{-1}. On every site, we assume that $AS = SB$, which is guaranteed (up to numerical noise) after calling canonical form. In a uniform MPS, however, we are not guaranteed that this condition holds. Instead, we store an AL tensor (left canonical, A in MPS notation), AR tensor (right canonical, B), and an AC tensor (one-site orthogonality center, Theta) on each site. On each bond we store a C tensor that is not guaranteed to be diagonal.

A uniform MPS is only defined in the thermodynamic limit.

The functions in the class are mostly trivial copies of the functions from MPS that account for the additional type of tensor structure.