[Pw_forum] Difference in total energies (serial vs parallel) for calculations with hybrid functional using norm conserving pseudopotential: Possible bug

[Prasenjit Ghosh]

Thanks a lot Nicola and Paolo.

Paolo, is there some quick fix for the bug?

With regards,

Prasenjit

On 28 August 2017 at 21:17, Paolo Giannozzi <p.giannozzi at gmail.com> wrote:

> There is a problem with reduced-cutoff calculations (using ecutfock <
> 4*ecutwfc) and k-points. Thank you for reporting this
>
> Paolo
>
> On Mon, Aug 28, 2017 at 8:33 AM, Prasenjit Ghosh <prasenjit.jnc at gmail.com>
> wrote:
>
>> Dear all,
>>
>> We are doing calculations for my system (PbI2) with hybrid functionals
>> (using the ACE algorithm) and norm conserving pseudopotentials. Presently
>> we are performing this calculations using scalar relativistic
>> non-conserving pseudopotentials for Pb and I. The pseudopotentials have
>> been generated by me using PBE functionals. I have tested these (at the
>> level of GGA) and it can reproduce well the lattice parameters and
>> electronic bandstructure of a range of different compounds (from metallic
>> Pb to ionic PbI2 and CsPbBr3).
>>
>> When we perform the hybrid calculations using GAU-PBE, we find that the
>> total energy differs when we use 1 core and 4 cores. The energy difference
>> is about 0.23 Ry. However, for both the cases, the standard GGA-PBE
>> calculations give same results (total energy). Even the band gap with
>> GAU-PBE (using 1 and 4 cores) are same. A look into the output file
>> suggests that the difference comes from
>> -averaged Fock potential + Fock energy.
>>
>> Since I have generated the pseudopotential, I thought that there might be
>> some issue with this. Hence to cross check, I downloaded the normconserving
>> pseudopotentials for Pb and I from http://theossrv1.epfl.ch/Main/
>> Pseudopotentials (pbe.0.3.1.tgz). For these two pseudopotentials also
>> the total energies (1 core vs 4 cores) differ. The input files,
>> pseudopotential files and the output files can be downloaded from the link
>> given below:
>>
>> http://www.iiserpune.ac.in/~pghosh/QE/
>>
>> These results are reproducible on different machines, using different
>> compilers (eg. gfortran vs intel).
>>
>> I think that there is a bug in the code, particularly the part where the
>> contributions to total energy from EXX is collected. If the developers can
>> suggest where exactly to look for in the code, I can give it a try to sort
>> it out.
>>
>> With regards,
>>
>> Prasenjit.
>>
>> --
>> PRASENJIT GHOSH,
>> IISER Pune,
>> Dr. Homi Bhabha Road, Pashan
>> Pune, Maharashtra 411008, India
>>
>> Phone: +91 (20) 2590 8203 <+91%2020%202590%208203>
>> Fax: +91 (20) 2589 9790 <+91%2020%202589%209790>
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://pwscf.org/mailman/listinfo/pw_forum
>>
>
>
>
> --
> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> Phone +39-0432-558216, fax +39-0432-558222
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>



-- 
PRASENJIT GHOSH,
IISER Pune,
Dr. Homi Bhabha Road, Pashan
Pune, Maharashtra 411008, India

Phone: +91 (20) 2590 8203
Fax: +91 (20) 2589 9790
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20170829/93c5ee4b/attachment.html>