Page 1 of 1

rfstrs RF problem for PAW pseudopotentials

Posted: Mon Sep 19, 2011 7:39 pm
by swan
Hi,

I'm trying to test new algorithm for RF calculations with PAW pseudopotentials. abinit 6.8.2 is compiled
with ifort 11.1.069 and openmpi 1.2.8 on Linux system. The rfphon perturbation part is OK, at the Gamma
point I get the frequencies close to those obtained with other pseudopotentials. However every time I add
the rfstrs command in the input file (stress perturbation), I get the segmentation fault in the mkcor3 subroutine.
Has anyone succeeded in these calculations?

Regards,
- Alexander

Re: rfstrs RF problem for PAW pseudopotentials

Posted: Fri Sep 23, 2011 2:48 am
by jzwanzig
rfstrs has not been coded for PAW, it is not supposed to work. However, the code should check and simply stop, not crash, when you try to use rfstrs with PAW.

Re: rfstrs RF problem for PAW pseudopotentials

Posted: Fri Sep 23, 2011 9:51 pm
by swan
Dear Josef,

Thank you for your reply. I hope the strain perturbation + PAW will be coded some day.

I'm afraid I was too optimistic saying that phonon RF + PAW pseudopotentials is OK. I have performed
much testing, and for TiO2 at the Gamma point I got this strange response:
Subroutine unknown:0:BUG
rfgd array must be allocated !
Action: contact ABINIT group.
Previously I saw this error one more time (I don't remember the substance I was testing).

Best regards,
- Alexander

P.S. With rfstrs enabled, the program does really core dump. The output of the system says:
Signal: Segmentation fault (11)
Signal code: Address not mapped (1)
Failing at address: 0x120
abinit682(mkcor3_+0x163c) [0x73e13c]
abinit682(eltfrxc3_+0xada) [0x715c2a]
abinit682(respfn_+0x9387) [0x478607]
abinit(driver_+0x380f) [0x453e6f]
abinit682(MAIN__+0x12f7) [0x44df57]
abinit682(main+0x3c) [0x44cb4c]