On 12 Jul 2007, at 13:09, Armin Sauter wrote:
Thank you for your answers. I can break down the problem to the case with no external field:
LATTICE="chain lattice" MODEL="spin" but with only this bondterm: -J*Sx(i)*Sx(j) h=Gamma=D=0 L=6,J=1
dmrg: SWEEPS=4,MAXSTATES=128 dirloop: THERMALIZATION=2000,SWEEPS=20000, EPSILON=0.2,T=0.01
Dirloop should not work on that model. I'll check this. Did you not get an error message?
The results for the ground state energy are: Fulldiag: -1.5 Dirloop: -1.0 Dmrg: -1.28
with CONSERVED_QUANTUMNUMBERS=Sz: Fulldiag: -1.0 Dirloop: -1.0 Dmrg: -1.28
and with -J*Sz(i)*Sz(j): Fulldiag: -1.5
I think there is no superposition of bases with Sz_total!=0 in the constraint case.
You should never set CONSERVED_QUANTUMNUMBERS=Sz for models that do not conserve Sz, such as your model.
To reproduce the problem and fix it, could you please send us your input files (parameter, models)
Matthias