[Pw_forum] (no subject)

[eariel99 .]

Dear Shiva,


>I have 2 questions:
>1.I want to calculate bulk modules for the orthorhombic structure, but
ev.x just have sc,bcc,fcc and hex, so what shoul I do?
>2.Is the output of ev.x is trustable? I asked this, because it depends on
how many numbers are in the input files!!!
>please help me in this.
>thanks.



I think you cannot use ev.x as a black box. The fitted bulk moduli and
other parameters depend on the number of points, the accuracy of the
calculated energies, and the range of volumes. Fitting is an art.

Zero temperature calculations of bulk modulus can be compared with values
derived from measurements of ultrasound speed, where the volume variations
are tiny, much smaller than 1 %. However, for these volue variations the
energy changes are also tiny and you need very well converge calculations
(very large cutoffs, many k-points). Hence, a compromise must be achieved.
On the other hand, many e-v data improve statistics, check the errors given
by the fitting program. If ev.x doest not provide errors, then use gnuplot
or Origin.
Use volume variations between 1 and 10 %, and check the dependence of the
bulk modulus with the convergence parameters and with the number of
k-points and number of data. Check if the e-v fitting function is able to
reproduce the e-v data that are outside the fitting range. When the results
are carefully converged, everything matches.

I use the following methods to evaluate if a fitting is well done. Plot the
difference g(V)=E(V)-f(V), where E(V) are the calculated data and f(V) is
the fitting function. If g(V) looks random, the fitting proccess is fine,
and the statistical errors  given by the fitting program are meaningfull.
If g(V) is not random, then the functional form of the fitting function is
not adecuate, or the calculated values are biased (maybe due to poor
convergence).

Best regards
eduardo


Eduardo Menendez Proupin

*Grupo de Cálculos Cuánticos*
E.T.S.I.Telecomunicación

Universidad Politécnica de Madrid

Telefono: +34 91 549 57 00 Extensión 2310

URL: http://www.gdc.tat.upm.es


Direccion permanente:
Departamento de Fisica, Facultad de Ciencias, Universidad de Chile
URL: http://www.gnm.cl/emenendez
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20131002/8238e704/attachment.html>