Hi Forwordom,
If you have a sign problem then a nonzero EPSILON is needed just to make the updates ergodic. It will not help remove the sign problem. SOmetimes one can get rid of a sign problem by changing the basis - we can check that if you send us the model.
Matthias
On Dec 4, 2012, at 9:00 PM, "Forwordom" mljphy@qq.com wrote:
Hi ALPS users and developers I meet sign problem in simulating frustrated system using dirloop_sse quantum Monte Carlo. As for this problem, the tutorial give some suggestions taht one should adjust parameter EPSILON and check the results carefully. But there is not the details of how to adjust this parameters. I have tried to modify all kinds of parameters including EPSILON,WHICH_LOOP_TYPE and so on. However the warning "the hamiltonian has sign problem" always appears when running. Can I get reliable the results when sign problem exists? How to modify simulation parameters to avoid sign problem using dirloop_sse method for my simulating frustrated systems. Can you give help. Thanks advanced.