[gpaw-users] RPA: RuntimeWarning: invalid value encountered in sqrt

[Jens Jørgen Mortensen]

On 11/14/2014 06:03 PM, Thomas Stenbæk Jauho wrote:
> Thanks for the answer, Jens Jørgen.
>
> I didn't use the newest version and getting the latest developer 
> version solved the issue.
>
> But I've noticed that some other changes have been to RPA within the 
> last month. I see that it is no longer possible to parallelize over 
> frequencies, but you parallelize over G vectors in stead. This 
> requires setting wstc=True. I can see that wstc is 
> WignerSeitzTruncatedCoulumb.
>
> What is the reason for this change?

Yes, wstc=False and parallelization unfortunately doesn't work right 
now, but it'll make it work as soon as possible.

We're currently testing wstc for RPA calculations and it seems to 
converge better with respect to number of k-points.

Jens Jørgen

>
> Best,
> Thomas
>
>
>
> Hej Jens Jørgen,
>
> Tak for svaret. Jeg brugte ikke den nyeste version, og det løste 
> problemet at opdatere til nyeste developer.
>
> Men jeg kan se at der er sket nogle andre ændringer i RPA inden for 
> den sidste måned.
> Jeg kan se at man ikke længere kan parallelisere over frekvenser, men 
> i stedet over G vektorer, hvilket så kræver at man sætter wstc=True. 
> Jeg kan se at wstc er WignerSeitzTruncatedCoulomb.
>
> Hvad er årsagen til denne ændring?
>
> On 2014-11-11 23:57, Jens Jørgen Mortensen wrote:
>> Den 11-11-2014 kl. 21:44 skrev Thomas Stenbæk Jauho:
>>
>>> Hi,
>>>
>>> I am doing RPA calculations for Na and 30 other systems. I have
>>> attached the scripts for Na, the RPA calculation is separated in
>>> three separate files; a PBE calculation, a diagonalization and the
>>> RPA.
>>>
>>> While 30 systems do work, Na does not. I receive the following
>>> error:
>>>
>>> /home/niflheim/s093006/gpaw/gpaw/response/chi0.py:302:
>>> RuntimeWarning: invalid value encountered in sqrt
>>> x_m = (-2 * df_m * deps_m / (omega.imag**2 + deps_m**2))**0.5
>>>
>>> I gather that it means that the square root (omega.imag**2 +
>>> deps_m**2))**0.5 receives a negative argument, but I don't know why.
>>>
>>>
>>> Previously I experienced the same problem with Li, but I resolved
>>> the issue by setting occupations=FermicDirac(0.01) i GPW(...) when
>>> making the DFT calculation. But this does not help for the Na
>>> system.
>>>
>>> Any help will be highly appreciated!
>>
>>  This could have been fixed in the latest development version:
>>
>>
>> https://trac.fysik.dtu.dk/projects/gpaw/changeset/12121/trunk/gpaw/response/chi0.py 
>>
>> [2]
>>
>>  Could you try with the latests version - if you are not already using
>> it?
>>
>>  Jens Jørgen
>>
>>> /Thomas
>>>
>>> _______________________________________________
>>> gpaw-users mailing list
>>> gpaw-users at listserv.fysik.dtu.dk
>>> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users [1]
>>
>>
>>
>> Links:
>> ------
>> [1] https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>> [2]
>> https://trac.fysik.dtu.dk/projects/gpaw/changeset/12121/trunk/gpaw/response/chi0.py 
>>