vasp.5.2.12 forces with hybrid functionals

Message

jacek · #1 Post by **jacek** » Mon Nov 14, 2011 8:17 pm

Hi all,

I have a problem with forces when using hybrid functionals.

Vasp either crashes (forrtl: severe (174): SIGSEGV, segmentation fault occurred), or hangs up, or gives forces which are wrong (do not correspond to energy derivatives over a series of fixed-geometry calculations).
The problem does not show in LDA/GGA runs.
Vasp behaves correctly once the call to fock_force in force is commented out, but the forces are useless since they miss the HF contribution.

I use ifort.11 with 11.1/038 mkl. I have been playing with different possible ways of compiling Vasp (scalar/parallel, different lapack, blas, fft, level of optimization, compiler options, and various unlimit settings) with no success. The most recent bugfix did not solve the problem neither.

Has anyone noticed a similar behaviour?
Best regards,
Jacek

#2 Post by **admin** » Fri Nov 25, 2011 8:21 pm

please tell us which vasp version do you use (release number and release date)?

jacek · #3 Post by **jacek** » Mon Nov 28, 2011 8:07 pm

I use vasp.5.2.12 11Nov11, but the problem existed also with the earlier vasp.5.2.12, the one before the most recent bugfix.

What happens looks like this (fixed atomic positions):
position(fractional) Energy(GGA) Force(GGA) Energy(hybrid) Force(hybrid)
-0.2 -10.766798 -1.006286 -14.773310 11.361726
-0.1 -10.842306 -0.504083 -14.858274 7.156272
+0.0 -10.867537 0.000000 -14.886666 0.000000
+0.1 -10.842301 0.504489 -14.858269 5.303473
+0.2 -10.766782 1.006334 -14.773293 5.801273

hybrid results obtained with:
PRECFOCK= N and HFSCREEN = 0.2