The ABINIT Users Mailing List ( CLOSED )

[ธนูสิทธิ์ บุรินทร์ประโคน <thanusit@kku.ac.th>]

Dear Jigato

In your input file "gwaecl_2", the input parameter "ngkpt" is not specified. This seems (I may be wrong) to cause Abinit to generate all k-point in dataset1 (kpt1) to be the gamma (0 0 0), as can be seen in the output file. I'm afraid that your KSS file calculation may not be properly done. Just add in your input file

     kptopt1 1

     ngkpt1 8 8 8

would solve this.

Regarding  the k-points at which the GW corrections are calculated, this is specified by the parameters nkptgw and kptgw. In your case, nkptgw3 =1 and kptgw3 = 0 0 0 (the gamma), you will get only the GW calculation at gamma. To get the GW corrections at other k-points, you need to increase nkptgw3 and add more coresponding kptgw3. Note that the kptgw3 must be in the list of those used to gennerate the KSS file (kpt1).

Hope this helps.

 

Kind regards,

Thanusit Burinprakhon

On June 17, 2008 10:50:13 PM ICT, "Manuel P้rez Jigato" <Manuel.PerezJigato@fys.kuleuven.be> wrote:

> Dear all,
>
> when running GW for a cubic oxide, i seem to be getting different 
> behaviour for ABINIT in the screening/inverse dielectric matrix 
> calculation, depending on the
> way i set up the input (do it all in one run vs do SCF separately and 
> restart for the rest).
>
> Let me get more explicit.
>
> case 1
> If i follow the silicon example that appears in the GW1 tutorial, and 
> do the gamma point GW correction in one run (exactly as in input file 
> "tgw1_9.in"), ie,
> - SCF calculation
> - KSS file generation
> - screening matrix calculation
> - self-energy matrix elements and GW correction to the highest 
> occupied and lowest unoccupied bands
>
> in this case, the screening part seems to be doing what it should 
> (according to the silicon test case "tgw1_9") and it finds the whole 
> lot of q-points in the part printed as
>
> number of q-points found 85
> find q-points q = k - k1 and translate in first BZ
> q-points [reduced coordinates]
> -----then the list of q points follow.....
>
>
> case 2
> If instead, i do separately the SCF calculation from the rest 
> (according to the tutorial, you could save time doing SCF with full 
> symmetry, and the next calculation KSS-screening-selfenergy/GW using 
> symmorphi 0) in this case, the
> screening/inverse dielectric matrix part seems to do something 
> different to the above, ie, it computes chi0 and epsilon-1 only for 
> the gamma point:
>
> number of q-points found 1
> find q-points q = k - k1 and translate in first BZ
> q-points [reduced coordinates]
> 0.000010 0.000020 0.000030
>
> Let me enclose the input files corresponding to both cases, ans well 
> as the output for the "wrong" case (second case).
>
> Is there any good reason for the difference in behaviour? I guess when 
> you compute big systems, it is advisable to do the SCF part separately 
> with the full symmetry. In the convergence tests in GW1 for silicon, 
> it is convenient
> to do restarts as well.
>
> thanks
>
> Manuel Perez Jigato
> KU Leuven
>
>
>
>
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm