GW correction in an arbitraty point of the BZ

[claudio]

Dear Yambo developpers

I'm trying to calculate GW correction in a k-point that is not included in my original grid,
using an old procedure that was working in old yambo version, but in the new one I found two bugs
and this method does not work anymore.

I use ypp -k k to generate a shifted grid that contains my new point (ypp.in)
The first bug is that the new ypp grid seems not to contain the point I gave in input
probably it is an error of k-point units (see l_stderr file)

Anyway I calculate the bands in the new grid (file hbn.nscf_shifted.in) , I make yambo read the grid (yambo.in_setup_shifted)
but then when I try to perform the GW calculation it crashes (yambo.in_gw_shifted) :

Code: Select all

<---> EXS |                                        | [000%] --(E) --(X)[attacc-Latitude-E6430:30939] *** Process received signal ***
[attacc-Latitude-E6430:30939] Signal: Segmentation fault (11)
[attacc-Latitude-E6430:30939] Signal code: Address not mapped (1)
[attacc-Latitude-E6430:30939] Failing at address: 0x406ea9740

I put in attachment all input files, for DFT, yambo setup and GW.

The test in attachment takes less than 5 minutes

Best
Claudio

[Daniele Varsano]

Hi Claudio,
thanks a lot for reporting.

Daniele

[Davide Sangalli]

Ciao,
I confirm. Last working version is yambo-3.4

Changing line 232 of src/bz/bz_samp_index.F
(or nearby, depending on the version of the code)

Code: Select all

 !
 ! qindx_S(ik,iqbz,1)=okbz
 ! qindx_S(ik,iqbz,2)=iGo
 !
 ! qindx_C(ikbz,iqbz,1)=okbz
 ! qindx_C(ikbz,iqbz,2)=iGo
 !
- do ikbz=1,Xk%nbz
+ do ikbz=1,k%nbz

it works also in yambo-4.x.
Can you try that ?

Best,
D.

[claudio]

Ciao Davide

your correction works fine!

But there is still the bug in ypp.

I found the bug in yambo-4.2.0 GPL, the input k-points are transformed two time in iku units by mistake
and therefore ypp does not produce the correct grid.

Input k-points are transformed in iku in lines 174,175 of ypp/k-points/k_grid.F
and then again in lines 196,197 of the same file. (I referee to the line number of GPL yambo)

I removed the second loop from line 192 to line 199 and now ypp produces the right grid with the point I gave in input.

Best
Claudio

[Davide Sangalli]

Thanks.
I'll also include the fix to ypp.

Best,
D.

[claudio]

Hi again

here a small explanation how to make this kind of calculations:

GW correction on an arbitrary point of the Brillouin zone with Yambo
http://www.attaccalite.com/gw-correctio ... ouin-zone/

best
Claudio