I am struggling with converging to the ground state of a BH model with unit filling with mps_optim with PBC.

For low hopping it produces the same result as in the OBC case ignoring the boundary condition.

I have the following input <pasted below>.  The state to which the mps_optim converges:

- has the same energy as the OBC hamiltonian for t<0.08 (the value of the constant varies it is larger for larger L) (see enclosed plot) and density artifacts at edges (similar to the OBC ground state)

- If I observe how the energy of a minimized state varies as a function of sweep number, then there are “plateaus” (for several iterations the energy seems to be converged, then it jumps down)..  It seems that for small t the GS computation of mps_optim is stuck at the first plateau. So for say t<0.07 after each sweep starting from 5 the energy is constant but for larger t, at say sweep number 10 there is a sharp drop.

- none of the above happens for the OBC.

- I include file containing the difference between the numerical result and the essentially exact result (strong coupling expansion up to 16th order - actually I copied the formula only until 10th order).  Clearly for all t between 0 and 0.06 the results are wrong, then they are fine again (and start to depart for larger t again but this is because I was to lazy to copy higher orders too)

- i have checked MAXSTATES=180 as well as MAXSTATES=320 (the results are exactly the same). For MAXSTATES=600 (and 120 sweeps and ietl_jcd_toll decreased to 1e-10)  the results are almost the same again but now problems are for t<0.08 not for t<0.07 so in some sense situation is worse.

*************** questions ******************

I have thus two questions: 

1. Is there a way to resolve it? 

2. It seems that propagation of the PBC across the sample length (L=40 sites) gets truncated somehow. Still increasing number of states to 600 did not help. Is it what is going on or is it a bug?

Best,

Mateusz Łącki

************* conf ******************

LATTICE="chain lattice"
MODEL="boson Hubbard"

L=40

MAXSTATES=180

measure_each=5
init_state="local_quantumnumbers"
initial_local_N='1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1'
nsweeps=60

t=0.06;
U=1.0;
N_total=40
Nmax=5;

CONSERVED_QUANTUMNUMBERS='N'
MEASURE[Entropy]=1
MEASURE_LOCAL[Local Density]='n'


{}