phonons with hexagonal symmetry and IBRION=6

[widom]

I am having trouble with the symmetry implementation of finite difference phonon calculations (IBRION=6). The problem is apparent in hexagonal structures (e.g. Mg) but not cubic structures (e.g. Al, Si).

Experimentally, the Gamma-point phonon frequencies for Mg are around 7.3 THz (nondegenerate) and 3.7 THz (2x degenerate). Using IBRION=5 (finite difference without symmetry), 7 (DFPT without symmetry) or 8 (DFPT with symmetry) I consistently find 7.4 THz and 3.8 THz in good agreement with experiment. These values are relatively insensitive to electronic k-point density, PREC settings, and the values of POTIM and NFREE. I’m using PAW_GGA potentials.

When I take IBRION=6 (finite difference with symmetry) I get 7.4 THz for the nondegenerate mode, but the 2x degenerate mode yields frequencies 2.1*i (NFREE=1), 4.8 THz (NFREE=2) and 3.4 THz (NFREE=4).

This may additionally be related to the difficulties being reported by users skaram, forsdan and jsfilhol in recent weeks.

[widom]

I have located and fixed the bug that gives improper results for IBRION=6. In the file finite_diff.F there is a variable initial_A that is used to save and later restore the initial undistorted lattice vectors. After A is re-initialized it is necessary to call LATOLD to reset the reciprocal lattice vectors B.

Insert the line “CALL LATOLD(OMEGA,A,B)� after the first “A = INITIAL_A� statement. This is line 537 in finite_diff.F as distributed with vasp5.2.2.

[jsfilhol]

Thank you very much for the patch, it seems to work fine : now I am getting the same results with IBRION 5 and 6

JS