[CPMD-list] Troubles testing GaAs

[Ari.P.Seitsonen at iki.fi]

Hi Eyvaz,

> 1. Lanczos + Vanderbilt PsP combination gives an
> enormous error for
> B2MIN and B2MAX.

Sorry, but Vanderbilt pseudo potentials work neither with Lanczos NOR
k points!!

> 2. Vanderbilt type PsP is not compatible with
> Monkhorst-Pack k-points:

Please see above.

> 3. Test using ODIIS method and Vanderbilt type PsP
> seems never will be converged.
> 
>  ODIIS| Insufficient progress; reset!
> 5747  3.812E-01   7.778E-02    -725.679718   -3.155E+01      5.28
> 5748  1.264E+00   2.356E-01    -652.074943    7.360E+01      5.31
> 5749  6.561E-01   5.359E-02    -735.670382   -8.360E+01      5.40
> 5750  1.181E+00   8.692E-02    -714.601328    2.107E+01      5.49
> 
> In this case only \Gamma point is applied, isn't it?

I guess so... What was your input in this case?

> 4. I am not sure that results I found are good when
> Goedecker type PsP
> is applied.
> 
> a) Energy via lattice parameter is strange (it seems
> there are 2 minima).
> 
> 5.65 (B+E2+X-V) TOTAL ENERGY = -245.03314131 A.U.
> 5.66 (B+E2+X-V) TOTAL ENERGY = -244.97934288 A.U.
> 5.67 (B+E2+X-V) TOTAL ENERGY = -245.00667290 A.U.
> 5.68 (B+E2+X-V) TOTAL ENERGY = -245.00457030 A.U.
> 5.69 (B+E2+X-V) TOTAL ENERGY = -245.00533431 A.U.
> 5.70 (B+E2+X-V) TOTAL ENERGY = -244.94863815 A.U.
> 
> Of course, it may be connected with convergence
> problem (may be cutoff energy
> 36 Ry is not sufficient for 3d-electrons in Ga).

Indeed I would suspect the cutoff energy, especially the differences
are so small.

> b) Besides occupance numbers are strange: I have
> exactly 36 filled bands for
> 8 atoms in the cubic cell (13*4+ 5*4), but for k-point
> 3/4,3/4,3/4
> even 37-th band is filled.
> 
>  K POINT:       1    0.375000    0.375000    0.375000    0.125000
>     35      2.9491468     2.000       36      3.0419036     2.000
>     37      3.0419036     2.000       38      3.9066380     0.000
> 
>  K POINT:       2    0.375000    0.375000    0.125000    0.375000
>     35      2.7371278     2.000       36      3.3659589     2.000
>     37      3.8350205     0.000       38      4.6426493     0.000
> 
>  K POINT:       3    0.375000    0.125000    0.125000    0.375000
>     35      3.0925761     2.000       36      3.1181842     2.000
>     37      4.1882453     0.000       38      4.9720836     0.000
> 
>  K POINT:       4    0.125000    0.125000    0.125000    0.125000
>     35      3.5761476     2.000       36      3.6583125     0.000
>     37      3.6583125     0.000       38      5.4236853     0.000

Hmm, strange. On some k points the occupation must then be less than
36. Have you tested your pseudo potentials against ghost states?

    Greetings,

       apsi

-- 
-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-=*=-
 Ari Paavo Seitsonen / Ari.P.Seitsonen at iki.fi / http://www.iki.fi/~apsi/
 Tel +41 1 635 44 97 / Fax +41 1 635 68 38 / GSM +41 79 719 09 35      
 Anschrift: Physikalisch Chemisches Institut (PCI), Universität Zürich (UniZh)
 Indirizzo: Winterthurerstraße 190, CH-8057 Zürich
 Address:   Schweiz / Svizzera / Suisse / Svizra / Switzerland