Decision of Iteration control parameters

[sergio]

Hi,

I am running the VNL to calculate the transmission coefficients for some systems including around 500 atoms. For some systems we need to increase the max. step the under  "Iteration control parameters " otherwise we get:

"ATKError: Exceeded maximum number of self-consistent iterations.
Terminated Abnormally"

and of course it takes time!

1) Is there anyway to decide  the max. steps roughly before starting the calculation, so that we are able to save time?

2) Is there any other reason(s) of that error?

3) If we get this error and program terminates abnormally, using the produced nc file, is it possible to start the calculation at this point by increasing the step number?

Many Thanks,

[Anders Blom]

The short answers are 1. Unfortunately, no (and even if there were, it wouldn't really save any time...) 2. No 3. Yes (see below) To restart the calculation, you can use a very small script

Code

from ATK.TwoProbe import *
import ATK
old_scf = restoreSelfConsistentCalculation("checkpointfile.nc")
ATK.setVerbosityLevel(1)
ATK.setCheckpointFilename('new.nc')
new_scf = executeSelfConsistentCalculation(self_consistent_calculation=old_scf)

[sergio]

Thanks for your reply A.Blom, then how can I insert the increasing step number in this script?

[Anders Blom]

If you restart, using the script indicated in the previous post, it will take another 100 steps. In principle, if it still hasn't converged you can just repeat the procedure, but at that point you should probably rather investigate why it fails to converge. Thus, there is no need, in this case, to increase "max_steps", since this number applies "per run", not in total. Of course, if you run several of these simulations, and they generally require more than the default max steps (100), the relevant setting is something like

Code

iterationControlParameters(max_steps=200)

[sergio]

Many thanks for your quick reply,