[gpaw-users] Calc. hangs after Dipole Moment

[Ask Hjorth Larsen]

Hello

I think it could be a problem with MPI and the compilation.  Please
reconsider compilation options (compilers, MPI, etc.).

If there is no clear reason:  Can you reproduce the error within a
single node (presumably -N 1, -n 16)?  What about across two half
nodes using 8 cores on each?  Finally, in serial or on two cores (for
comparison with calculations on desktop)?

Best regards
Ask

2016-06-15 11:12 GMT+02:00 Ahmed Rizwan <rizwan.ahmed at aalto.fi>:
>
> ________________________________________
> From: gpaw-users-bounces at listserv.fysik.dtu.dk [gpaw-users-bounces at listserv.fysik.dtu.dk] on behalf of Ahmed Rizwan via gpaw-users [gpaw-users at listserv.fysik.dtu.dk]
> Sent: Tuesday, June 14, 2016 3:31 PM
> To: Ask Hjorth Larsen
> Cc: gpaw-users at listserv.fysik.dtu.dk
> Subject: Re: [gpaw-users] Calc. hangs after Dipole Moment
>
> ________________________________________
> From: Ask Hjorth Larsen [asklarsen at gmail.com]
> Sent: Tuesday, June 14, 2016 2:52 PM
> To: Ahmed Rizwan
> Cc: gpaw-users at listserv.fysik.dtu.dk
> Subject: Re: [gpaw-users] Calc. hangs after Dipole Moment
>
> Hello
>
> 2016-06-14 11:46 GMT+02:00 Ahmed Rizwan via gpaw-users
> <gpaw-users at listserv.fysik.dtu.dk>:
>> Hi,
>>
>> I have Ni(111)_3*3*3 slab. Calculations take only 20 mints (3 nodes) for
>> bare slab. But as soon as I have OH* on the surface, calculations hang
>> forever at the point of calculating forces ( Dipole Moment: [-1.25280403
>> -0.62580586 -0.19320845] ) just after few iterations. From here, it does not
>> output anything in the txt file. Please see the txt file. It does not print
>> any error message anywhere.
>>
>> I have tried different mixers, spinpol=True/False, and even gave initial
>> magnetic momentum to Ni atoms (magmom=0.6)... Nothing helped.
>>
>> Any idea? Thanks
>>
>> Regards,
>> Rizwan
>
> This looks like it could be tricky.  I notice that it ran several
> calculations before hanging.
>
> I tried running it with more coarse parameters but did not see any problem.
>
> To get closer to the issue, see if you can reproduce the problem with
> a smaller calculation: Less k-points, smaller grid, less cores.  That
> makes it easier to debug because it can be done on a single computer.
>
> It is not a convergence issue and therefore not related to mixers and spins.
>
> When you ran it multiple times, did(/does) it always hang at the same point?
>
> Best regards
> Ask
>
> Thanks for the suggestion.
>
> Yes, it does run few calculations before hanging. Yes, it does hang at the same point, but the number if ionic steps could be different.
>
> Let me test this with less aggressive parameters and then i will come back to you.
>
> Best regards,
> Rizwan
>
> Hi Ask,
>
> I have got same problem for fairly small system, with relatively less k and g points. I have changed the mode to 'lcao' but unable to get rid of the problem. Everything runs fine without OH on surface. You said you did not see any problem with coarse parameters, and I do not see any problem with python script. Could it be related to Batch script or processor architecture?
>
> slab = fcc111('Ni', size=(2,2,3), a=3.52245, vacuum=10.0)
> slab.append(Atom('O', (3.736, 2.157, 16.256),tag=0))
> slab.append(Atom('H', (2.736, 2.157, 16.370),tag=0))
> slab.center(axis=2)
> calc = GPAW(
>             mode='lcao',
>             basis='dzp',
>             gpts=(16,16,96), #h~0.27
>             kpts = (2,2,1),
>             xc='PBE',
>             poissonsolver=DipoleCorrection(PoissonSolver(), 2),
>             occupations=FermiDirac(0.1),
>             txt='Ni111-333_fixed-1.txt',
>             maxiter = 400)
>
> slab.set_calculator(calc)
> cons= FixAtoms(mask=[a.tag > 2 for a in slab])
> slab.set_constraint(cons)
>
> qn = QuasiNewton(slab, trajectory='Ni111-333_fixed-1.traj')
> qn.run(fmax=0.05)
>
> Best regards,
> Rizwan
>>
>>
>> _______________________________________________
>> gpaw-users mailing list
>> gpaw-users at listserv.fysik.dtu.dk
>> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users