[Pw_forum] ecutwfc convergence

Filipe Camargo Dalmatti Alves Lima flima at if.usp.br
Wed Sep 26 17:23:22 CEST 2012


Dear Yusuf,

I agree with everyone has said to you, but I just want to comment:

"When the cutoff is enough?" I think this is your question, right?

The best way to learn it is by testing your system with different
parameters.

For example: if you think that you need the best total energy description
you should create a series of test varying the
cutoff and testing the Total Energy.

So,

For a system you will run X simulations using different cutoffs 20 ry (200
ry), 25 ry (250 ry), 30 (300), 40 (400), 60 (600), ...
(Remember charge density description (ecutrho) should change as you
increase the kinect energy cutoff)

For instance, lets pretend you obtained the following total energies:

using 20 Ry = -20 eV
using 25 Ry = -22 eV
using 30 Ry = -22.501 eV
using 40 Ry = -22.623 eV
using 60 Ry = -22.625 eV

In a perfect world as much ecutwfc is higher better would be your result.
(if it does not happening, probably you are doing something wrong, like a
poor k point description, etc)

But do you need the best possible? In my example, there is a 2 eV
difference by between 20 Ry and 25 Ry, so, it might not be
a good idea use 20 Ry.

However, comparing 30 and 40 ry, you have ~0.1 eV difference. Is this
enough for the property that you are looking for?
In some cases it could be enough, in others it won't.  If you are looking
at conformational changes that are around 0.5 eV probably
you will need to increase your ecutwfc. You can use this procedure in all
properties that you are interested in.

In DOS case, you might compare if the DOS are changing as the cutoff
increases.

In summary:
- The parameters that you choose are not changing the physics of your
results.
- They have a good relationship between correctness/computational time.
- They are reasonable when you compare your work with the literature.
- Your tests should be (relatively) fast since you don't want waste your
machine time.

That's it,

I hope it will help you. Also, I hope I didn't make any mistake in my
comment! :)

Best regards,


-- 
_________________________________________
Filipe Camargo Dalmatti Alves Lima
PhD Student
University of São Paulo, Physics Institute, Materials Physics Department,
Nanomol Group, Brazil.
Phones:    (11) 3091-6881  (USP)
                (11) 8249-4628 (TIM)


On Wed, Sep 26, 2012 at 9:44 AM, Eduardo Ariel Menendez Proupin <
eariel99 at gmail.com> wrote:

> Dear Yusuf
> The cutoff depends on the pseudopotential that you use, not just on the
> atom. Different pseudopotentials for the same atoms have different cutoff.
> Also, some properties requiere higher cutoffs than others. For example,
> forces and relative energies need lower cutoff than total energy and
> stress. Using higher values does not harm, just take longer time and more
> computer resources.  Some programs include recommended cutoffs in the
> pseudopotential, but the user is still responsible of testing it.
>
> If you want a starting parameter for graphene, use
> C.pz-vbc.UPF with ecutwfc = 60.0 ,
> or
> C.pbe-rrkjus.UPF with ecutwfc = 30.0 and ecutrho = 180.0
>
>
> the above work reasonably well to relax the structures with fixed cell, and to get the bands and DOS. I fou want to calculate elastic properties or LDA+U or PBE+U, or calculate conductance, etc, etc, these cutoffs may be too low. Yo need to do tests increasing the values until you see no change in the property that you want to calculate.
>
> If you plan to adsorb other atoms (e.g. oxygen) on the graphene, then determine cutoffs that are valid for the full system and do all the calculations with the same parameters.
>
>
>
> Dear Manasaje
>
> The system is graphene with 32 carbon atoms. The cutoff is not specified in the pseudo and I need it before proceeding to other SCF and nscf calculations and believe, cutoff is a property of an atom in question. So can I still go ahead with the atoms in the unit cell. Thank you for your quick response
> Yusuf
>
>
> ______________________________
> __
>  From: Eduardo Ariel Menendez Proupin <eariel99 at gmail.com <http://www.democritos.it/mailman/listinfo/pw_forum>>
> To: pw_forum at pwscf.org <http://www.democritos.it/mailman/listinfo/pw_forum>
> Sent: Wednesday, September 26, 2012 7:18 PM
> Subject: Re: [Pw_forum] ecutwfc convergence
>
>
> Dear Yusufm
>
> ecutwfc and ecutrho depend only on the hardest of the pseudopotentials used for the atoms in the molecule or the crystal. If your molecule is C32 and have no other elements, you can make tests with a C2 (which does not exists, I guess), and the cutoffs for C32.
>
> Anyway, if your are going to calculate something that it is not the total energy or the DOS, let's say, the phonons, or the Raman spectrum, it is reasonable that you test the effect of the cutoff first on C2 and then on C32.
> ---------- Mensaje reenviado ----------
> From: Yusuf Zuntu <yzunt at yahoo.com <http://www.democritos.it/mailman/listinfo/pw_forum>>
> To: PWSCF Forum <pw_forum at pwscf.org <http://www.democritos.it/mailman/listinfo/pw_forum>>
> Cc:
> Date: Wed, 26 Sep 2012 03:15:32 -0700 (PDT)
> Subject: [Pw_forum] ecutwfc convergence
>
> Hi,
> I need clarification on kinetic energy cutoff. I have a system comprising 32 carbon atoms. Am I to consider total 32 carbon atoms in  convergence w.r.t kinetic energy cutoff scf calculation or the two carbon atoms in the unit cell.
> Thank you for anticipated reply.
> Yusuf
>
> --
>
>
> Eduardo Menendez Proupin
> Departamento de Química Fisica Aplicada
> Facultad de Ciencias
> Universidad Autónoma de Madrid
> 28049 Madrid, Spain
> Phone: +34 91 497 6706
>
> On leave from: Departamento de Fisica, Facultad de Ciencias, Universidad
> de Chile URL: http://fisica.ciencias.uchile.cl/~emenendez
>
>  "*Padece, espera y trabaja para gentes que nunca conocerá y que a su vez
> padecerán, esperarán y trabajarán para otros, que tampoco serán felices,
> pues el hombre ansía siempre una felicidad situada más allá de la porción
> que le es otorgada. Pero la grandeza del hombre está precisamente en querer
> mejorar lo que es. En imponerse Tareas. En el Reino de los Cielos no hay
> grandeza que conquistar, puesto que allá todo es jerarquía establecida,
> incógnita despejada, existir sin término, imposibilidad de sacrificio,
> reposo y deleite. Por ello, agobiado de penas y de Tareas, hermoso dentro
> de su miseria, capaz de amar en medio de las plagas, el hombre puede hallar
> su grandeza, su máxima medida en el Reino de este Mundo*".
> Alejo Carpentier, El reino de este mundo, (1949).
>
> "...man never knows for who he suffers and waits. He suffers and waits and
> works for peoples he will never meet, and who at the same time will suffer
> and wait and work for others who will also not be happy, as man always
> desires a happiness that is beyond the lot that he is given. But the
> greatness of man is precisely in wanting to improve what he is. In
> assigning himself Tasks. In the Kingdom of Heaven there is no greatness to
> conquer, for everything there is an established hierarchy, the unknown is
> unobscured, existence is without end, sacrifice, rest, and delight are
> impossibilities. Therefore, overwhelmed with hardships and Tasks, beautiful
> within his misery, capable of loving in the middle of calamity, man can
> only find his greatness, his full self-realization, in the Kingdom of this
> World." - Alejo Carpentier
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120926/880a0790/attachment.html>


More information about the users mailing list