[Wien] Large oscillations in total energy

[Peter Blaha]

As the charge distance is not going bad, I don't expect it has to do 
with mixing, nor with any scf problem.

I'd guess, only E-tot is calculated wrongly in mixer or the run script 
has a problem .
It could be that you have 2 times the same iteration number in the scf 
file (:ITE) or that for some reasons once contribution to :ENE is not 
listed/listed twice, ...

Please check the scf file for :ITE, but also manually for the different 
E-tot contributions like (grep :ENE case.scf)

:ENE
:DEN
:SUM
and the that the core eigenvalues are listed only once.

ALSO: save the calculation, and continue with another run_lapw
Eventually it should stop after 3 cycles, because everything is converged.


On 12/07/2017 11:20 AM, Pavel Ondračka wrote:
> Dear Wien2k mailing list,
>
> I'm seeing some strange energy oscillations in one calculation near the
> end when everything seems almost converged.
>
> :ENERGY convergence:  0 0.0001 .0039009700000000
> :ENERGY convergence:  0 0.0001 .0168906450000000
> :ENERGY convergence:  0 0.0001 .0096598000000000
> :ENERGY convergence:  0 0.0001 .0162053350000000
> :ENERGY convergence:  0 0.0001 .0089683200000000
> :ENERGY convergence:  0 0.0001 55.6398533450000000
> :ENERGY convergence:  0 0.0001 27.8210445050000000
> :ENERGY convergence:  0 0.0001 27.8235252600000000
>
> the charge is converged already:
>
> :CHARGE convergence:  1 0.001 -.000920
> :CHARGE convergence:  1 0.001 -.000977
> :CHARGE convergence:  1 0.001 -.000933
> :CHARGE convergence:  1 0.001 -.000972
> :CHARGE convergence:  1 0.001 -.000984
> :CHARGE convergence:  1 0.001 -.000988
> :CHARGE convergence:  1 0.001 -.000997
> :CHARGE convergence:  1 0.001 -.000998
>
> This is quite large calculation of TiSiO amorphous-like cell (96 atoms,
>  struct file attached, compressed in order to pass the moderation
> queue). The numerical parameters are quite standard,
> 2x2x2 shifted k-grid (can also reproduce this with 3x3x3), Rkmax 7.5,
> PBE, rest of stuff at default. I have done a lot of calculation with
> similar cells and such parameters without any problems (albeit mostly
> some time ago with older Wien2k versions)...
>
> I'm trying to calculate the binding energy of the O 1s for the first
> oxygen using the Slaters transition state approach. Surprisingly when I
> use almost touching spheres it converges (O:1.48, Ti: 1.72, Si:1.48),
> however when I reduce them by 3% (as in the attached struct file) this
> problems surfaces. However I need this reduced spheres for comparison
> with another calculation. BTW the spheres around 1.5 for oxygen
> should be reasonable anyway?
>
> No errors or warnings anywhere, except for the expected
> WARN  :        CHARGED CELL with  -0.500
>
> This is with Wien2k 17.1, compiled with MKL 2017.0.098 and ifort.
>
> Any ideas what might be causing this problem?
> Best regards
> Pavel
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:  http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at    WIEN2k: http://www.wien2k.at
WWW:   http://www.imc.tuwien.ac.at/TC_Blaha
--------------------------------------------------------------------------