VASP and Wannier90 with SOC

Problems running VASP: crashes, internal errors, "wrong" results.

Moderators: Global Moderator, Moderator

Locked
Message
Author
eggerd
Newbie
Newbie
Posts: 1
Joined: Thu Nov 21, 2019 6:23 am

VASP and Wannier90 with SOC

#1 Post by eggerd » Thu Jul 02, 2020 8:03 am

Dear VASP developers,

I would like to report an issue regarding usage of VASP 5.4.4 and wannier90 2.1.0:

First, VASP has been compiled with the relevant -DVASP2WANNIER90v2 compiler option. Also, W90 is run in library mode with the lib file being properly linked to VASP as far as I can see.

In the attached files, you will see that the code runs as it should once "LSORBIT=FALSE", so that at the moment I doubt this is an installation issue.

However, once I use "LSORBIT = T" and the interface to Wannier90, then the VASP run goes through normally, however, the interface (i.e., the actual W90 calculation) crashes.

I'm now providing additional information from my debugging and tests that may be of help to you:

the most common thing to happen is that the number of projections simply runs too far. Please see attached tests, where I have saved the value of variable "nproj" into the sdout file of VASP. E.g., I expect to see an output related to ten projections, but the index runs further.

Further information, potentially not reliable: I think it is related to the file "mlwf.F" and the following cycle:

IF (proj_l(ICNTR)==0.AND.proj_m(ICNTR)==0.AND.proj_radial(ICNTR)==0) CYCLE sites

I'm speculating that something between wannier1.2 and wannier2.0 in terms of how spinors are handled must have changed, so that this condition (or another one close to the place of this one in the file mlwf.F) may require an update.

I'm happy to run further tests or provide further information.

Your help is greatly appreciated, thanks a lot!

Best wishes,
David

Attachment doesn't work at the moment, so here is a weblink to the test files I produced:

https://www.dropbox.com/s/9g2kuvhzdeucf ... s.zip?dl=0

User avatar
juergen.furthmueller
Newbie
Newbie
Posts: 5
Joined: Fri Nov 08, 2019 9:46 am

Re: VASP and Wannier90 with SOC

#2 Post by juergen.furthmueller » Mon Jul 06, 2020 12:45 pm

There are actually several problems and bugs (incomplete argument lists,
wrong countings, ...) as far as I understood. For working properly with spinors
various changes are needed in mlwf.F (and VASP needs to be interfaced with
version 2.x of wannier90). Chengcheng-Xiao has published a patch under

https://github.com/Chengcheng-Xiao/VASP2WAN90_v2_fix

which fixes all bugs and problems. With this patch there should be a good
chance that everything works properly as it should. Would be nice if one of
the vaspmasters in Vienna could incorporate this patch / corresponding
changes in the official VASP version(s).

martin.schlipf
Global Moderator
Global Moderator
Posts: 164
Joined: Fri Nov 08, 2019 7:18 am

Re: VASP and Wannier90 with SOC

#3 Post by martin.schlipf » Tue Jul 07, 2020 7:21 am

We are working on upgrading the interface with Wannier90 to v3.1, which will be released in Vasp 6.2.

User avatar
chengcheng_xiao1
Newbie
Newbie
Posts: 18
Joined: Sun Nov 17, 2019 6:23 pm
Location: London, UK
Contact:

Re: VASP and Wannier90 with SOC

#4 Post by chengcheng_xiao1 » Wed Jul 08, 2020 10:22 am

juergen.furthmueller wrote:
Mon Jul 06, 2020 12:45 pm
There are actually several problems and bugs (incomplete argument lists,
wrong countings, ...) as far as I understood. For working properly with spinors
various changes are needed in mlwf.F (and VASP needs to be interfaced with
version 2.x of wannier90). Chengcheng-Xiao has published a patch under

https://github.com/Chengcheng-Xiao/VASP2WAN90_v2_fix

which fixes all bugs and problems. With this patch there should be a good
chance that everything works properly as it should. Would be nice if one of
the vaspmasters in Vienna could incorporate this patch / corresponding
changes in the official VASP version(s).
Thanks Juergen for pointing out my repo!
martin.schlipf wrote:
Tue Jul 07, 2020 7:21 am
We are working on upgrading the interface with Wannier90 to v3.1, which will be released in Vasp 6.2.
Glad to know that! (except I don't have a valid license for VASP6...) Hopefully my patch is useful to some degrees for you guys! ;)

Just a small question, will the new interface in VASP v6.2 have the full capabilities as pw2wannier90? (spn, uhu, uiu, dng, sym, etc)

henrique_miranda
Global Moderator
Global Moderator
Posts: 234
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: VASP and Wannier90 with SOC

#5 Post by henrique_miranda » Fri Jul 10, 2020 9:06 am

Just a small question, will the new interface in VASP v6.2 have the full capabilities as pw2wannier90? (spn, uhu, uiu, dng, sym, etc)
We currently not planning to add support for symmetry adapted Wannier functions.
We might still include the writing of the spn, uhu, uiu files though.
About 'dng', I don't know what you referring to.

User avatar
chengcheng_xiao1
Newbie
Newbie
Posts: 18
Joined: Sun Nov 17, 2019 6:23 pm
Location: London, UK
Contact:

Re: VASP and Wannier90 with SOC

#6 Post by chengcheng_xiao1 » Tue Jul 14, 2020 3:12 pm

Ah, I mean "dmn" matrix for the symmetry adapted mode.

So I guess SCDM is also not going to happen in the near future. Anyhow, thanks for the update! Looking forward to it! ;D

Locked