Bug with PAW_GW Pseudopotentials
Posted: Mon Jan 06, 2014 4:15 pm
Hallo to everybody,
I signal here a problem I encoutered using new PAW_GW pseudopotentials, and I ask the community to help me in fixing it.
I'm trying to run HSE03 calculations on bulk wurtzite ZnO.
For Zn I used the "PAW Zn_sv_GW 01Dec2010" pseudopotential, with ZVAL=20. I played around with ALGO and TIME tags and I tried starting from scratch or from preconverged PBE calculations. In all cases some processors got stuck causing the others to wait. As a consequence the simulations had to be killed. The standard output then displayed an error like "EDWAV: internal error, the gradient is not orthogonal" for each stucked node.
Instead, if I use different pseudopotentials (e.g., "PAW Zn_pv_GW 09Oct2010" or "PAW Zn_GW 09Oct2010" with 18 and 12 electrons rispectively) I had no problem at all. Playing around with the tags became then just a matter of convergence speed.
If you have any hint of what's going on, please answer.
Thanks!
I signal here a problem I encoutered using new PAW_GW pseudopotentials, and I ask the community to help me in fixing it.
I'm trying to run HSE03 calculations on bulk wurtzite ZnO.
For Zn I used the "PAW Zn_sv_GW 01Dec2010" pseudopotential, with ZVAL=20. I played around with ALGO and TIME tags and I tried starting from scratch or from preconverged PBE calculations. In all cases some processors got stuck causing the others to wait. As a consequence the simulations had to be killed. The standard output then displayed an error like "EDWAV: internal error, the gradient is not orthogonal" for each stucked node.
Instead, if I use different pseudopotentials (e.g., "PAW Zn_pv_GW 09Oct2010" or "PAW Zn_GW 09Oct2010" with 18 and 12 electrons rispectively) I had no problem at all. Playing around with the tags became then just a matter of convergence speed.
If you have any hint of what's going on, please answer.
Thanks!