Dear developers,
I test the optical spectrum of graphene and surprisingly find that Abinit gives totally wrong spectrum with shiftk=(0,0,0). While with the default shiftk=(0.5,0.5,0.5), Abinit gives consistent result with Yambo and CASTEP. Increasing the ngkpt does not change this situation.
This is quite strange. Could you enlighten me? The test files and comparison graph are attached for your inspection.
Sincerely,
Guangfu Luo
P.S. Because Yambo gives correct spectrum with the KSS input file [shiftk=(0,0,0)] generated by Abinit, the problem must come from the DDK part and/or optic analysis of Abinit; I find Abinit also encontours some problem in the optical spectrum of 1D system, which maybe is related to the same cause and will be explained in a separate post.
[SOLVED]Wrong optical spectrum of graphene caused by shiftk
Moderators: mverstra, joaocarloscabreu
[SOLVED]Wrong optical spectrum of graphene caused by shiftk
- Attachments
-
Graphene_optics_shiftk_test.zip.in- remove suffix .in
- (87.11 KiB) Downloaded 348 times
Last edited by Robin on Fri Sep 02, 2011 4:29 am, edited 1 time in total.
Re: Wrong optical spectrum of graphene caused by shiftk
Dear All,
Similar to the case I discussed in another post with Matthieu (viewtopic.php?f=10&t=1261), this problem is also related to the NSCF WFK by iscf=-2, which gives incorrect occupations and causes incorrect electron transitions. When I use iscf=-3 to generate the NSCF WFK, both shiftks give correct results, except that they have different accuracy because of different nkpt.
Sincerely,
Guangfu Luo
Similar to the case I discussed in another post with Matthieu (viewtopic.php?f=10&t=1261), this problem is also related to the NSCF WFK by iscf=-2, which gives incorrect occupations and causes incorrect electron transitions. When I use iscf=-3 to generate the NSCF WFK, both shiftks give correct results, except that they have different accuracy because of different nkpt.
Sincerely,
Guangfu Luo