Matthias,

Yes, I'm trying to run the 4x4 square lattice with periodic boundary conditions (although I explicitly set the latter, perhaps that is an error with my input file).

I recognize that DMRG does not do well for 2D systems but I figured that it would just give me an upper bound, if just a very loose one. The naive DMRG for 2D lattices just converges very poorly in the number of states, m, and thus isn't very useful, but this does not explain how I get below the exact answer.

Thanks,

Jeff

Matthias Troyer wrote:

Hi Jeff,

I'll ask Adrian to look into this, but I wanted to warn you that DMRG does not work well for 2D lattices and also has problem with periodic boundary conditions. Are you trying to do 2D calculations with DMRG?

Matthias

On 26 Apr 2007, at 09:38, Jeff Hammond wrote:

...

I am trying to run DMRG for the half-filled 4x4 Hubbard using Alps 1.3b and am getting energies which are substantially below the exact value. Granted, the jobs haven't finished, but all my (limited) knowledge of DMRG suggests that the energy will never drop below the exact value during any intermediate point in the calculation.

Hopefully someone can tell me what I am doing wrong with my input file or how my thinking is incorrect. I am confused that it got below E=-16 when the exact result is around E=-13.

Thanks,

Jeff

hubbard.parm:

SWEEPS=10 MAXSTATES=1000 MODEL="fermion Hubbard" LATTICE_LIBRARY="/home/jeff/ALPS/lib/xml/lattices.xml" MODEL_LIBRARY="/home/jeff/ALPS/lib/xml/models.xml" CONSERVED_QUANTUMNUMBERS="Nup,Ndown" LATTICE="square lattice" {L=4;Nup_total=8; Ndown_total=8; t=1; U=4} =========================================================