[QE-users] Different SCF energy when # of MPI is changed

[Takahiro Chiba]

Dear Prof. Giannozzi,

Thank you for your explanation. Since then, I have tried many
combinations of {Fermi-Dirac, Gaussian} smearing and {0.0020, 0.0043,
0.0062, 0.0081, and 0.0570} degauss with respect to {24, 32, 40, 56}
MPI. By using Gaussian broadening, calculations were faster than
Fermi-Dirac, but I still obtained discrepancies in any combination of
smearing and degauss, unfortunately. I collected the results in the
PDF below. Differences in my input are shown in blue, and resulting
discrepancies are shown in red.

==Result table==
https://www.scl.kyoto-u.ac.jp/~tchiba/SCF-discrepancy.pdf
==

Can I conclude that this is unavoidable? Is there anything that I can try?


---Sender---
Takahiro Chiba
Graduate school of CSE, Hokkaido University, Japan
takahiro_chiba at eis.hokudai.ac.jp
-----



On Fri, Oct 7, 2022 at 5:42 AM Paolo Giannozzi <paolo.giannozzi at uniud.it> wrote:
>
> On 06/10/2022 09:18, Takahiro Chiba wrote:
> > It does not always happen. For example, no discrepancy was found with
> > the "ausurf" benchmark. Therefore, I expect there is something wrong
> > with my input rather than compile/install issues.
>
> it's not necessarily something wrong in your input. Apparently your run
> is quite hard to converge. Even the numerical noise that is unavoidable
> in a large calculation may occasionally lead the scf procedure towards a
> different path that converges badly (or better), for no apparent good
> reason. Notice that the run that exhibits the largest discrepancy in the
> energy (the one with 32 processors) is also the one that makes a
> disproportionate number of iterations (452 vs 63-67 for the three other
> runs).
>
> Something that might help is to use a larger broadening (degauss) and a
> gaussian broadening
>
> Paolo
>
> >
> >
> >
> > ---Sender---
> > Takahiro Chiba
> > Grad. school of chem. sci. and eng., Hokkaido Univ.
> > Group:   https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.cat.hokudai.ac.jp%2Fhasegawa%2FHOME.html&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C14ff559e8db1496a529308daa76b2e0b%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C638006376027518526%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=SFtIf53Uv8cu7y3HVzu607kQiM8OOFpK40TvpCz1DXY%3D&reserved=0
> > Mail:      takahiro_chiba at eis.hokudai.ac.jp
> > _______________________________________________
> > The Quantum ESPRESSO community stands by the Ukrainian
> > people and expresses its concerns about the devastating
> > effects that the Russian military offensive has on their
> > country and on the free and peaceful scientific, cultural,
> > and economic cooperation amongst peoples
> > _______________________________________________
> > Quantum ESPRESSO is supported by MaX (https://eur01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.max-centre.eu%2F&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C14ff559e8db1496a529308daa76b2e0b%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C638006376027518526%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=hkyGi1EsV7OWjpxPPxRrtNF3ewPOIWDu7dv0IXmXx2E%3D&reserved=0)
> > users mailing list users at lists.quantum-espresso.org
> > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.quantum-espresso.org%2Fmailman%2Flistinfo%2Fusers&data=05%7C01%7Cpaolo.giannozzi%40uniud.it%7C14ff559e8db1496a529308daa76b2e0b%7C6e6ade15296c4224ac581c8ec2fd53a8%7C0%7C0%7C638006376027518526%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=2oxldIe7U6fQ5Go173Oc3yzWRud4bzfMv6rpXDBV7v4%3D&reserved=0
>
> --
> 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