Nonlocal vdW functionals with non-collinear (LSORB)
Posted: Thu Jul 28, 2016 10:50 am
Dear all,
I have noticed that the nonlocal vdW-DF functionals does not work (have no effect) with the LSORB = .TRUE. tag.
I think it is well known problem in the vdW-DF non-local approach where it seems not be implemented for SOC.
(See link here for the discusstion given by Torbjörn Björkman: https://www.researchgate.net/post/Why_i ... tly_larger)
By inspecting vdw_nl.F and xcspin.F, we can easily figure out that the problem is originated from calling
the routine VDW_NONLOC_spin, which is called in xcspin.F and located in vow_nl.F.
There is two problems in calling VDW_NONLOC_spin.
One is that VDW_NONLOC_spin does not properly work when NGZhalf and NGXhalf tag is deactivated (this is default in non-collinear source code, vasp-ncl).
Second one is that there is no lines for activating vdW-DF calculations when non-collinear calculation is set, on the other hand,
when vdW-DF tag is not set in INCAR, vdW-DF calculation is activated by "FEXCGS_VDW_" routine in xcspin.F.
To solve this problem, I have added corresponding subroutines and calling routines for vdW-DF calculations within non-collinear calculations, so called "FEXCGS_NONCOL_VDW_" in xcspin.F.
This works find for my test cases, including TMD layered materials.
I hope this helps and hope this issue to be improved in the next vasp release.
Thank you.
SIncerely,
Hyun-Jung Kim.
(Infant@kias.re.kr)
Note: Since the modified source code is not allowed to attach in this message, please contact me if you have any request.
And also note that Prof. I. Hamada (NIMS, Japan) also participated in modifying the source code.
I have noticed that the nonlocal vdW-DF functionals does not work (have no effect) with the LSORB = .TRUE. tag.
I think it is well known problem in the vdW-DF non-local approach where it seems not be implemented for SOC.
(See link here for the discusstion given by Torbjörn Björkman: https://www.researchgate.net/post/Why_i ... tly_larger)
By inspecting vdw_nl.F and xcspin.F, we can easily figure out that the problem is originated from calling
the routine VDW_NONLOC_spin, which is called in xcspin.F and located in vow_nl.F.
There is two problems in calling VDW_NONLOC_spin.
One is that VDW_NONLOC_spin does not properly work when NGZhalf and NGXhalf tag is deactivated (this is default in non-collinear source code, vasp-ncl).
Second one is that there is no lines for activating vdW-DF calculations when non-collinear calculation is set, on the other hand,
when vdW-DF tag is not set in INCAR, vdW-DF calculation is activated by "FEXCGS_VDW_" routine in xcspin.F.
To solve this problem, I have added corresponding subroutines and calling routines for vdW-DF calculations within non-collinear calculations, so called "FEXCGS_NONCOL_VDW_" in xcspin.F.
This works find for my test cases, including TMD layered materials.
I hope this helps and hope this issue to be improved in the next vasp release.
Thank you.
SIncerely,
Hyun-Jung Kim.
(Infant@kias.re.kr)
Note: Since the modified source code is not allowed to attach in this message, please contact me if you have any request.
And also note that Prof. I. Hamada (NIMS, Japan) also participated in modifying the source code.