[Pw_forum] How to choose ecutwfc value?
Nicola Marzari
nicola.marzari at epfl.ch
Fri Oct 20 08:53:06 CEST 2017
Thanks Saif,
this is very useful. Sadly, it also shows the current limitations in the
field.
Basically, these days one can "converge" the computational parameters
quite well (increase the supercells, improve the k-point sampling, make
sure the smearing is correct, etc...), but often there remain a
significant uncertainty coming from the pseudopotentials used - i.e.
different pseudopotentials can give different results.
We almost never know which is the "right" result - it would require
all-electron calculations, and those in themselves require a great
deal of care (nowadays for something simple like the equation of state
of an elemental compound they seem to agree - at least looking at
the 2016 Science paper by Laejeghere - but for everything else noone knows).
Bottom line: we are stuck, and very few care about this. Hopefully
in the next few years we'll seem some more standardized all-electron
calculations, that will help refine further the pseudopotential
libraries. For the time being, we collect in the SSSP our best guesses.
nicola
On 18/10/2017 21:08, Saif Ullah wrote:
> Dear Nicola,
>
> First of all apology for such a delayed response and thank you very much
> for your reply. Actually, I wanted to cover as many possibilities as I can.
>
> I do not have VASP access right now and instead of Cl, I have used K as
> I have a very good K VDW-DF pseudo for SIESTA.
>
> I have used 30, 40 and 50 Ry ecutwfc and the adsorption energies were
> nearly the same. I have used different smearing method with various
> widths. I have used 3x3x1-6x6x1 kpoints. The values (40Ry) are here:
>
> graphene 5x5 SIESTA-VDW-DF QE-DFT-D2 .van QE-DFT-D2 sssp
> Li -1.11 -0.65 -1.5
> Na -0.67 -0.95 -0.94
> K -1 -1.36 -1.36
>
>
> The sssp refer to the link you mentioned. Moreover, our gaussian values
> are the same as VDW-DF-SIESTA values except for Na, for which we have
> -0.50 eV and is comparable to one of the VASP DFT-D2 studies. For Na,
> our QE-DFT-D2 values are also comparable to values reported by X.Fan
> using VASP.
>
> Now the question is for Li. The VASP DFT-D2 values are from -1.09 to
> -1.13 eV. The same range is also reported by Okamoto using QE-DFT-D2
> (DOI: 10.1021/acs.jpcc.6b05458). I have another paper with QE-DFT-D2
> (authored by Stefano Fabris) in which they have reported -1.4 eV using a
> 2x2 cell of graphene (DOI: 10.1103/PhysRevB.92.075403). So -1.4 eV for a
> 2x2 and -1.5 for 5x5 graphene is reasonable.
>
> One thing I'm still confused is that why in the above two references,
> they have reported two very different values ( -1.12 eV, and -1.4 eV)
> using same level DFT-D2 and code (QE). Additionally, VASP and QE both
> are PW codes, so i think the using PBE-D2 will lead to nearly the same
> adsorption energies.
>
> What to do now, please?
> If you need anything else, please inform.
>
> I wish to read from you and the community soon.
> Thanks
> Saif
> Department of Physics, Ufjf, Brazil
>
> On Sat, Oct 7, 2017 at 8:39 AM, Nicola Marzari <nicola.marzari at epfl.ch
> <mailto:nicola.marzari at epfl.ch>> wrote:
>
>
> Dear Saif,
>
> At the link below you find suggested pseudos and cutoffs - 30Ry for
> Li and Na, 50Ry for Cl try adsorption energies at 30, 40 and 50 Ry,
> first keeping the same geometry in all calculations, then relaxing
> everything. Please send to the forum a report with these 6 numbers,
> and those from
> SIESTA, Gaussian, and VASP.
>
> Thanks,
>
> nicola
>
> http://materialscloud.org/sssp/ <http://materialscloud.org/sssp/>
>
>
> On 07/10/2017 02:28, Saif Ullah wrote:
>
> Dear qe users,
>
> I'm a qe beginner, therefore, came up with a simple question. If
> i have to study the interaction of Li/Na with graphene, so in
> this case how will i choose the ecutwfc value? Should i do a
> convergence test w.r.t ecutwfc for pure graphene or I must check
> the convergence of Li adsorbed graphene and Na adsorbed graphene
> separately? For my pristine graphene, I found the 40Ry value to
> be enough to converge graphene but when I studied the Li
> interaction, my calculated adsorption value was very much
> underestimated from my SIESTA/Gaussian/VASP values. Finally, I
> checked the convergence of ecutwfc of Li/Na adsorbed graphene
> and found that 110Ry value is better for both the Na and Li.
> This being said that an enormous increase in computational time.
> I have seen a paper in which the authors used 30Ry and reported
> the accurate results. My question is, the 110Ry is ok for Li/Na
> graphene systems or I have done a mistake? Thanks in advance
>
> Saif
> Department of physics, ufjf, Brazil
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon>>
> Virus-free. www.avast.com <http://www.avast.com>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link>>
>
>
> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org <mailto:Pw_forum at pwscf.org>
> http://pwscf.org/mailman/listinfo/pw_forum
> <http://pwscf.org/mailman/listinfo/pw_forum>
>
>
>
> --
> ----------------------------------------------------------------------
> 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://theossrv1.epfl.ch/Main/Contact>
> http://nccr-marvel.ch/en/project <http://nccr-marvel.ch/en/project>
>
>
--
----------------------------------------------------------------------
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