bug in noncollinear case optics calculation
Posted: Wed May 25, 2011 3:39 am
(1)
The noncollinear cases are assigned ISPIN=1, although spinor is actually used for the wavefunctions. This is fine until I need to do optics calculation. The NBVAL is incorrectly teated
NBVAL=(NINT(INFO%NELECT)+1)/2
(BTW, this might not be the only line that causes the trouble.) Therefore the band counting (NBVAL, NBCON) is messed up. For a noncollinear case, it ought to be
NBVAL = INFO%NELECT or NINT(INFO%NELECT)
A simple test, which I did, on a simple insulator rocksalt LiF, using both LDA and noncollinearLDA can demonstrate this point. LiF has a gap. NELECT = 10, with LDA LiF has NBVAL = 5, but with noncollinearLDA NBVAL = 10. If with noncollinear LDA NBVAL is taken to be 5, guess what happens to the dielectric function.
(2)
Using vasp.5.2.2, somtimes the OPTIC output file is printed, sometimes not (even using the same input files). Using vasp.5.2.11, I've never gotten the OPTIC
output, only the dielectric function in OUTCAR. I wonder why.
The noncollinear cases are assigned ISPIN=1, although spinor is actually used for the wavefunctions. This is fine until I need to do optics calculation. The NBVAL is incorrectly teated
NBVAL=(NINT(INFO%NELECT)+1)/2
(BTW, this might not be the only line that causes the trouble.) Therefore the band counting (NBVAL, NBCON) is messed up. For a noncollinear case, it ought to be
NBVAL = INFO%NELECT or NINT(INFO%NELECT)
A simple test, which I did, on a simple insulator rocksalt LiF, using both LDA and noncollinearLDA can demonstrate this point. LiF has a gap. NELECT = 10, with LDA LiF has NBVAL = 5, but with noncollinearLDA NBVAL = 10. If with noncollinear LDA NBVAL is taken to be 5, guess what happens to the dielectric function.
(2)
Using vasp.5.2.2, somtimes the OPTIC output file is printed, sometimes not (even using the same input files). Using vasp.5.2.11, I've never gotten the OPTIC
output, only the dielectric function in OUTCAR. I wonder why.