FAQ#

Problems & Bugs#

If you run into problems (especially if you encounter nonphysical results), or if you are pleased with our platform, please contact us.

If the interacting cluster combined with the bath is degenerate, one should, at a minimum, keep as many low-lying states as the ground state degeneracy. This is because the measurements are partially done on the fly during the DMRG sweeping, and the ground state order may change in the case of degeneracies.
We have limited the accuracy of the underlying DMRG solver for the free beta. Please check the quality measure in the output. If that number is high, then more resources are needed for the DMRG.

Why do simulations for a single site cluster sometimes produce strange results, e.g. the filling on the cluster is not the desired one?

The number of degrees of freedom used to optimize the mean-field description might be too low.

Do single cluster systems make sense?

Yes. In this case the system is completely handled within DMRG and there is no embedding.

Does iterating make sense?

Since there is no embedding, there is no need to iterate, and the code will stop after the first iteration.

Which DMRG code is used?

PRG++, an open source DMRG implementation, which is available at https://projects.tuxfamily.org/?do=group;name=prg

Why doesn't the DMRG code achieve the same accuracy as given in certain publications?

It is usually a matter of fine tuning of the parameter and the amount of resources available. E.g., for the results of the full counting statistics of the interacting resonant level model millions of CPUh were spend to obtain a single plot.