I'm looking to study time-dependent two-point correlation functions following the prescription laid out in
https://arxiv.org/abs/1207.0652
It seems like 'segment' boundary condition is what I'm looking for, but I'm confused on how to utilize this. What I want to do is use iDMRG to find the ground state of some model, and then use this to create a finite window to study the dynamics of some excitation. I'm confused on how to create a 'segment' from the results of iDMRG, and then how to properly use segment boundary conditions when time evolving?
My basic idea is
1) run iDMRG
2) enlarge the unit-cell of the ground state
3) convert the ground state and Hamiltonian to an MPS / MPO with segment boundary conditions
4) Apply a local operator to the new segment MPS
5) Time evolve the state
I'm confused on steps (3) and (5). I see the functions 'extract_segment' in the latest version of TenPy on GitHub, which seems fine for the MPS, but not for the MPO. First, the resulting MPO has 'infinite' boundary conditions in this function, is this an error? Secondly, If I just use these functions to get a new MPS / MPO, and then plug these into some time evolution algorithm, this will ignore the interactions between the segment and the environment. How would I incorporate the interactions with the environment in TenPy?
Best,
Nick