[ase-users] Forcing a (re)calculation

[Ask Hjorth Larsen]

Hi Melissa

2016-12-17 19:43 GMT+01:00 Melissa A. Hines via ase-users
<ase-users at listserv.fysik.dtu.dk>:
> ASE users:
>
> What is the correct way to force ASE to re-run a calculation (i.e., mark a
> calculation “dirty”)?
>
> Scenario: I run a calculation in which I set the maximum number of ionic
> steps to be N. On inspecting the calculation, I see that N ionic steps were
> taken, but the energy is not converged. I would therefore like to reload the
> calculation and take another N steps.
>
> One way to force a recalculation is to make a trivial change to the run
> parameters. For example with Vasp I could do the following:
>
>>>> calc = Vasp(restart=True) # Load old calculation
>>>> atoms = calc.get_atoms()
>>>> calc.set(nsw = N + 1) # Trivial change to run parameters
>>>> atoms.get_potential_energy()
>
> This change causes calculation_required to return True, thereby leading to a
> new calculation. Is there a better solution to this issue?

Different codes do this differently, so it depends on how ASE
interacts with that code.

In GPAW you can load a GPW file and set new convergence criteria, then
re-trigger the calculation.  In Octopus it will automatically do it
when running twice from the same directory.  Maybe someone else can
answer for Vasp?

(It is not so normal to restart single calculations though, because
most of them are simply not all that expensive.  So it is not
necessarily implemented)

Best regards
Ask

>
> Thanks,
>
> Melissa
> ____________________________________________________________
> Melissa A. ºHines Director, Cornell Center for Materials Research
> Department of Chemistry Professor of Chemistry
>    and Chemical Biology Voice/FAX: 607-255-3040
> Cornell University http://hines.chem.cornell.edu
>
>
> _______________________________________________
> ase-users mailing list
> ase-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/ase-users