[Pw_forum] problem with final scf calculation with vc-relax

[Leonid Kahle]

Thanks a lot for your comment, Ilya.
It was precisely my point that it is "typical for the variable-cell 
optimization".
But it is great to have confirmation that the final structure can be 
reused without any problems.

Best,

     Leonid


On 10/15/2017 04:41 PM, Ilya Ryabinkin wrote:
> Hi Leonid:
> you error is typical for the variable-cell optimization: the problem
> is in the size of PW basis for the fixed energy cutoff. If the size of
> a cell is changing, there are more/fewer basis plane waves (G
> vectors)for the same energy cutoff. Just ignore this error and re-run
> SCF calculations with new cell parameters. Alternatively, to make sure
> that everything is fine, restart vc-relaxation with new cell
> parameters, it should converge very quickly with no errors.
>
> --
> Ilya
>
> On Sat, Oct 14, 2017 at 3:04 AM, Leonid Kahle <leonid.kahle at epfl.ch> wrote:
>> Dear developers,
>>
>> There are frequent problems with the "final scf" after the convergence of a
>> vc-relaxation with the recalculated G-vectors.
>> In this case, the volume grows during the vc-relax by 3.5%, and G-vectors
>> are assigned from
>>
>> sticks:   dense  smooth     PW     G-vecs:    dense   smooth      PW
>>       Min          77      38     11                 1889      668     114
>>       Max          78      39     12                 1891      671     118
>>       Sum        1393     685    211                34015    12039    2085
>>
>> to:
>>
>>       sticks:   dense  smooth     PW     G-vecs:    dense   smooth      PW
>>       Min          78      38     11                 1973      689     115
>>       Max          79      39     12                 1977      693     120
>>       Sum        1417     701    211                35557    12429    2109
>>
>>
>> for the final scf. The final scf-calculation crashes with the following
>> error:
>>
>>       Error in routine ggen (78):
>>       smooth g-vectors missing !
>>
>> Before I get told about strange input:
>>
>>   - The input is 2 atoms (LiI) in a reasonable cell (and the vc-relax goes
>> fine)
>>   - 10x10x10 kpoints, dense enough for an insulator
>>
>> Furthermore, this is QE6.0 compiled with Intel. I run on 18 cores on a Cray
>> XC40, with no command line switches.
>>
>> If I'm doing something wrong, please let me know.
>> But regardless, I would be happy to see the final scf in a vc-relax being
>> optional. As said,  this is happening frequently for all kinds of systems.
>> I've attached input, submission script,  and a (shortened for size
>> limitations) output file for this mentioned case.
>>
>> Thanks a lot and best regards,
>>
>>      Leonid Kahle
>>
>> --
>>
>> Doctoral Assistant
>> EPFL STI IMX THEOS
>> ME-D2 1019 (Bâtiment ME)
>> Station 9
>> CH-1015 Lausanne (Switzerland)
>>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://pwscf.org/mailman/listinfo/pw_forum
>
>