[Pw_forum] about the precision of energy and some extend problem.

[Hande Ustunel]

Dear Xu Yuehua,

You should fit your energies to a parabola and find the minimum of the
fit. For a second order fit, your minimum is at about 2.83. 2.77 is just a
local blip in your rather jumpy data. You cannot decide by looking at the
individual energies because they do not follow a smooth function. But even
for the "wrong" lattice constant, there's no physical reason why you should
get negative frequencies if you have converged your electronic and atomic
coordinates to high accuracy. I'm attaching the fit as an example.

I would also suggest increasing your cutoff a little because your data, as
I said, is a bit noisy near the minimum.

Best,
Hande

On Wed, 12 Dec 2007, xu yuehua wrote:

> hi ,uesers here :
> i relax one crystal lattice vs energy:
> i have compute several energy list :
> 
> c          E(RY)
> 2.795  -137.79266
> 2.78   -137.79248
> 2.77   -137.79233
> 2.8     -137.7927
> 2.81  -137.79275
> 2.82  -137.79275
> 2.83  -137.79278
> 2.84  -137.79276
> 2.85  -137.79279
> 2.86  -137.79272
> 2.87  -137.79265
> 2.88  -137.79258
> 2.9   -137.79229
> 2.95  -137.79122
> these energy points are smooth and  can be plotted  parabola.
> the system has 12 atoms.the energy is different most at the third digit
> after radix point. so i think the average energy/atom is less 0.0001.
> 
> i want to ask  the  within the precision of energy(my thinking is based on
> that :my above calculation is set ecutwfc:29, ecutrho:200,and the  criterion
> is 35,350, the convergence is 0.0049345 ry.)
> so this small difference of energy can skip,i.e we can choose 2.77 which is
> regarded as the best structure?
>  as why i ask this question,for the the structure of 2.85 has two  big
> negative frequencies  near -100cm-1, even if i add asr .
> but 2.77 has no negative frequency after i add asr.
> 
> 
> 

-- 
Hande Ustunel
Department of Physics
Office 439
Middle East Technical University
Ankara 06531, Turkey
Tel : +90 312 210 3264
http://www.physics.metu.edu.tr/~hande
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fit.eps
Type: application/postscript
Size: 14787 bytes
Desc: 
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20071212/c72f182f/attachment.eps>