The ABINIT Users Mailing List ( CLOSED )

[Anglade Pierre-Matthieu <anglade@gmail.com>]

Hi,

Since a correct value for ecut is mostly conditionned by atomic
(pseudopotential) properties and not by atomic positions, it would be
possible and easier to perform preliminary ecut convergence test on
small (monoatomic) cells.

Once this is done, the "converged" value for ecut in a small cell is
likely a "converged" value in your large supercells. Yet it is always
better to check the accuracy of your planewave cut-off ont the actual
problem you are trying to solve. Indeed, each numerical approximations
(tsmear, kpt, ecut ...) used to solved the DFT problem may result in
varying degrees of inaccuracy on each problem.

Note that the "0.1 mHa convergence" is only a rule of thumb. For some
problems it is vastly oversetimated ; for others it results in a mere
underestimate of the correct cutoff energy.

regards,

PMA

On Tue, Jul 14, 2009 at 5:19 PM, Quan Phung Manh<manhquank9@gmail.com> wrote:
> Dear Abinit users
> I'm learning structures of metal oxide surfaces, the surfaces contain about
> 20-40 atoms with 10 Angstrom vacuum. As usual, I determine Ecut convergence,
> but I really don't know if Ecut is high enough. In some Abinit Tutorials, we
> increase Ecut until the change of total energy is about 0.1 mHa. But with
> large systems, Ecut need to be very huge to obtain such small change of
> Total energy and the computational cost and the memory requirements increase
> remarkably. With huge systems, I often choose Ecut until the Energy
> difference reaches 3mHa = 2kcal/mol. I wonder if this convergence condition
> is enough and there are rules to treat Ecut convergence? I only investigate
> structure, electron density and DOS of the systems.
>
> Sorry for such an old question.
> Best regards
>
>
> --
> Quan Phung Manh
> Khoa Hoá học - ĐH Khoa học Tự nhiên - 19 Lê Thánh Tông Hà Nội
>
>



-- 
Pierre-Matthieu Anglade