[Solved] Electric field perturbation in RF with PAW psp

sponce · Post by **sponce** » Mon Jul 29, 2013 4:30 pm

Dear all,

I'm trying to compute the dielectric constant of a material by using DFPT. I did that calculation in 3 steps:
1) GS self-consistent calculation on the val. band
2) ddk calculation with

Code: Select all

rfelfd2   2       
   rfdir2   1 1 1         
    nqpt2   1
     qpt2   0.0 0.0 0.0   
  getwfk2   1     
  kptopt2   2             
    iscf2  -3        
  tolwfr2   1.0d-22

3) Electric field perturbation

Code: Select all

  rfphon3   1            
 rfatpol3   1 22          
  rfelfd3   3             
   rfdir3   1 1 1    
    nqpt3   1
     qpt3   0.0 0.0 0.0   
  getwfk3   1       
  getddk3   2        
  kptopt3   2 
  tolvrs3   1.0d-18
    iscf3  7

Everything went smoothly until dataset 3 were I get a Seg. Fault.

With gdb I've backtrace it to 66_paw/pawdijfr.F90 ligne 177 (version 7.4.1 of abinit)
The issue is that itypat=pawfgrtab(iatom)%itypat gives me the value "175626256" when prinited.
Obviously this is a random value because the itypat variable is not defined.

I have also found out that the routine that was calling this routine at that time was 72_response/dyfnl3.F90.
At the ligne 309 the variable "pawfgrtab_tmp" is allocated BUT never initialized or copy in any way and then passed to the routine "pawdijfr" 4 lignes later.

I do not see any ways for the variable itypat in the pawfgrtab_type data type to have any value and therefore it seems normal that it seg. fault.

Am I missing something? Is there a problem with my input variables? I have attached my full input file.

Thank you for your help,

Regards,

Samuel

sponce · Post by **sponce** » Tue Jul 30, 2013 11:36 am

The problem was that the pawfgrtab_tmp was not initialized in src/72_response/dyfnl3.F90.

I have made the changes and it work. I have commited my changes in sponce/7.4.1-private rev 456.

Have a good day!

suvas · Post by **suvas** » Fri Aug 30, 2013 4:12 am

Hi Sponse,

You said PAW had problem and you solved the issue. I have a problem with negative frequency. I got five negative frequency in both abinit 7.2.1 and 7.4.1 version. My convergence criteria and structural optimization are rahter strict. I just wanted to ask you, is there any problem in subroutine src/72_response/dyfnl3.F90. I would be so greatful if you could reply me soon. that would largely help spending time to calculate phonon frequency with PAW+LDA.

Thanks !

Suvas

sponce · Post by **sponce** » Fri Aug 30, 2013 11:08 am

Dear Suvas,

This is not the same issue. You might consider opening a topic about it.

Be sure to include your input files as well as any relevant data's.

Phonon computed in PAW+LDA should theoretically work but it might still be a bit unstable. Without more info I cannot help you any further.

Regards,

Samuel

Boris · Post by **Boris** » Fri Aug 30, 2013 12:31 pm

Hi Samuel

Did you check the results of your DFPT calculation related to response under an electric field?

It seemed to me that it was not available at the moment.

Boris

ABINIT Discussion Forums

[Solved] Electric field perturbation in RF with PAW psp

[Solved] Electric field perturbation in RF with PAW psp

Re: [Solved] Electric field perturbation in RF with PAW psp

Re: [Solved] Electric field perturbation in RF with PAW psp

Re: [Solved] Electric field perturbation in RF with PAW psp

Re: [Solved] Electric field perturbation in RF with PAW psp