Dear Isobel,

the usual bottleneck for Tkwants numerical accuracy and speed is the 
calculation of the manybody integral.
In your script however, you are only evolving a onebody state, so this could 
not be the reason.
There are two other things which could go wrong when evolving a onebody state:

1. turn of the extrapolation of the perturbation as described here: 
https://tkwant.kwant-project.org/doc/dev/tutorial/onebody_advanced.html#time-dependent-perturbation

2. change the time-stepping accuracy (less likely however)

You can find both points in a new documentation section about common pitfalls:
 https://tkwant.kwant-project.org/doc/dev/tutorial/pitfalls.html

Still, it is strange that your calculation takes hours and not seconds or 
respectively minutes, as what I would expect for a typical onebody simulation.
Running the script with MPI on a cluster does not help either, as a onebody 
simulation is not parallelized over several cores.
So I wonder if there are no other problems involved.
One other possibility is that you don't have leads attached, so your will get 
reflections from the boundaries over time which will mess up your
smooth initial behavior. Second, I don't know what means "short" for the 
simulated time. Short and long time is always with respect to the relevant
energy scales of your system, which you should know in advance.

Best,
Thomas

Reply via email to