2.17. Frequently asked questions¶
Make sure that Tkwant is correctly build, see Installation. If Tkwant can be imported inside the Tkwant repository folder but not outside, the package is very likely not in the Python search path. If a symbolic link has been created, make sure that it points to the correct destination and that its origin is in the Python search path.
Enable Logging to see possible warning messages.
No open modes are below the Fermi energy (chemical potential \(\mu\) at \(T = 0\)). By default, the Fermi energy is assumed to be zero. Plot the lead spectrum as in the first example in Getting started: a simple example with a one-dimensional chain to check that open modes are below the Fermi energy.
Only the MPI root rank carries the simulation result, all other ranks have
None as result.
am_master() strategy from Parallelization with MPI to save the result
from the root rank only.
On all non-root MPI rank the result is
None, preventing arithmitic
operations like \(+, -\) and \(*\) on these ranks.
Perform arithmetic operations only on the output from the MPI root rank
by using the
am_master() strategy from Parallelization with MPI.
The number of OMP threads must be one. Set
When the manybody integrand has a complicated structure (for systems having resonances, strong perturbations etc.) the adaptive refinement does not make progress. One should first enable Logging to check whether the simulation is indeed traped in the refinement cycle. If this is really the case, one can try to change the accuracy of the refinement process and also limit the total number of cycles. Alternatively, one might run the manybody simulation with the “low-level” approach using a fixed number of onebody states, see Solving the many-body problem.
Very likely, the manybody integral is poorly approximated. For the adaptive solver, check the estimated error and apply adaptive refinement. For the “low-level” approach, increase the number of integral splits, see Solving the many-body problem.
Update Tkwant and the required packages. Almost all tutorial examples execute real code to generate the output.