The ABINIT Users Mailing List ( CLOSED )

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

Hi,

I include my comments between your lines below :

On 10/28/06, john Bob <bobjohn.lzu@gmail.com> wrote:
> Dear Anglade:
>       I have not done the ecut converge test yet.I want to do it after
> optimization.
Nice idea but you must be carefull with that : You might still want to
have at least the kpt mesh converge as well as the ecut. For instance
starting with convergence to 10^-2 hartree for each and optimising the
forces down to 10^-3 then 10^-3 for kpt and ecut and 10^-4 for
forces...

A good idea if you want to start with rough low-cost calculations is
to had a large electronic temperatur wich will help k-point
convergency (tsmear+occopt).

> I also don't know how to create the kpoints for the
> orthorhombic v2o5,so i use the homogeneous sets.

I'm also not a master at reciprocal space 3D visualisation. That's why
I use the "kptrlen" + "prtkpt" (see the doc) to do an automatic
optimization. Try it ; your gonna love it.

> the number of bands
> is 42,
Looks good from your output. But if you use electronic temperature
don't forget to increase it.

> GGA pesuedopotential used comes from the abinit,org website.
>      The optimization ended after about 14 steps and converged.Below
> attachment is the .out file
A problem is that it is not really converged: you convergence
criterion on total energy is not strong enough. It might be better to
use toldff about 2 order of magnitude below tolmxf.


>      In SCF calculation ,there is no fix in y axes.I think  the fixy
> in the optimization step configure file is not so important because
> from the scf calculation I can see the froce in the y axes are all
> zero .Am i right?
NO
they are set to zero so that the atoms stay unmoved along this axis. I
think that this is the main source of your problem.

> Also
> I am going to do the optimization without the 'fixy'.
>    the attached file:  first is the result of optimization second is
> the configure file of the optimization ,third is the scf configure
> file.

A last advice: instead of diemac use "iprcel 4X" It is superior by a
wide margin.


> Best regards


--
Pierre-Matthieu Anglade