[QE-users] Choosing convergence criteria for vc-relax

Nicola Marzari nicola.marzari at epfl.ch
Mon May 20 20:04:18 CEST 2019



Dear Michal,


it depends on how smooth is the dependences of the forces on the cutoff,
so one should make an informed decision depending on the 
pseudopotentials used. We e.g. look at phonon frequencies as a function 
of cutoff, and if you look at silicon, here below, even if we didn't 
test below 30Ry, I would bet that at 20 or even 15Ry forces would be 
reasonable:

https://www.materialscloud.org/discover/data/discover/sssp/convergences_efficiency/Si_8.0_conv_patt.png

If you try Ga, you can see that the phonons (the solid line) do not
particularly become worse in going down from 70Ry to 45Ry, but then 
progressively worsen:

https://www.materialscloud.org/discover/data/discover/sssp/convergences_efficiency/Ga_8.0_conv_patt.png

Still, for many pseudopotentials a decrease in cutoff doesn't worsen the 
forces particularly, but there is no 2/3 or 1/2 rule - one just need to
examine the tables above.

For ecutrho it might be a bit easier, and if you are not having 
magnetism often one can use a dual (the ration of ecutrho/ecutwfc)
equalt to 3, or even 2, for normconserving, and 6 or even 4 for ultrasoft.

All to be tested...

				nicola


On 20/05/2019 13:47, Michal Krompiec wrote:
> Dear Nicola,
> I am looking forward to reading the "short thing on this"!
> Can you comment on the widespread (?) practice of using much lower than 
> recommended (say, 2/3 or 1/2) ecutwfc for geometry optimization until a 
> minimum is found, and then increasing cutoffs until convergence is 
> achieved? Is it generally a good idea?
> 
> Best regards,
> 
> Michal Krompiec
> 
> Associate Director at Merck KGaA and Adjunct Prof at University of 
> Southampton
> 
> 
> On Sun, 19 May 2019 at 13:57, Nicola Marzari <nicola.marzari at epfl.ch 
> <mailto:nicola.marzari at epfl.ch>> wrote:
> 
> 
> 
>     Hi Robert,
> 
>     this issue of optimal parameters is a bit of a neverending story, we are
>     preparing a short thing on this. But in a nutshell - yes, the
>     electronic
>     convergence should be tighter, especially for a follow-up phonon
>     calculation. that number actually depends on the unit cell used, but
>     for
>     5 atoms, I would indeed try d-12 and d-14. For ecutwfc and ecutrho,
>     well, my advise is to use the SSSP suggestions (google it); and say if
>     you have 30 120, 40 160, 25 300, well, take the max of the ecutwfc
>     suggested, for the gloval ecutwfc, and same for ecutrho.
> 
>                              nic
> 
> 
>     On 17/05/2019 21:24, Appleton, Robert J wrote:
>      > Hello,
>      >
>      > I am trying to study the phonons of the material BaZrS3 to
>     determine if
>      > it is stable for a given structure. I first use phonopy and vasp and
>      > found negative frequency values for phoning band structure. My
>     advisor
>      > is concerned in the optimization of the structure and so we now
>     want to
>      > use quantum espresso. I have only used it a few times and mostly
>     used
>      > default or example inputs. The input I want to make sure is
>     sufficient
>      > is my criteria for convergence.
>      >
>      > Under the &control:
>      > force convergence criteria right now is
>      > forc_conv_thr = .0001 <— this value seems to big but how do I
>     choose a
>      > value that is best for optimization but does not take to much
>     computer time?
>      >
>      > Under the &electrons:
>      > conv_thr = 1.d-9 <— I am considering changing to 1.d-12. Is this
>     a good
>      > idea?
>      >
>      > Under &system:
>      > I am using ecutrho = 12 • ecutwfc
>      >
>      > If you can give me some advice on how best to choose convergence
>      > criteria for studying phonons that would be very good thanks.
>      >
>      > Robert Appleton
>      > Cal State LA - MS Physics Student
>      >
>      > Get Outlook for iOS <https://aka.ms/o0ukef>
>      >
>      > _______________________________________________
>      > Quantum ESPRESSO is supported by MaX
>     (www.max-centre.eu/quantum-espresso
>     <http://www.max-centre.eu/quantum-espresso>)
>      > users mailing list users at lists.quantum-espresso.org
>     <mailto:users at lists.quantum-espresso.org>
>      > https://lists.quantum-espresso.org/mailman/listinfo/users
>      >
> 
> 
>     -- 
>     ----------------------------------------------------------------------
>     Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
>     Director, National Centre for Competence in Research NCCR MARVEL, EPFL
>     http://theossrv1.epfl.ch/Main/Contact http://nccr-marvel.ch/en/project
>     _______________________________________________
>     Quantum ESPRESSO is supported by MaX
>     (www.max-centre.eu/quantum-espresso
>     <http://www.max-centre.eu/quantum-espresso>)
>     users mailing list users at lists.quantum-espresso.org
>     <mailto:users at lists.quantum-espresso.org>
>     https://lists.quantum-espresso.org/mailman/listinfo/users
> 


-- 
----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, EPFL
http://theossrv1.epfl.ch/Main/Contact http://nccr-marvel.ch/en/project


More information about the users mailing list