bug in m-resolved spin-polarized DOS

[sipr]

There is a bug in probably all versions of abinit as concerns the
M-resolved spin-polarized DOS. I found in in versions 6.6.1,
6.10.2 and 6.12.1.

Namely, the M-resolved DOS is printed in such a way that that DOS from
spin-down states in the *DOS_AT???? files contains the sum of spin-up
and spin-down DOS'es and not the DOS for spin-down states only.

On the other hand, DOS for spin-up states is printed correctly. So to
get correct DOS for spin-down states, one has to subtract the spin-up
DOS from the spin-down DOS.

I summarized an example demonstrating this bug (a free-standing Co monolayer).
It can be downloaded as a tarball from here:
http://www.fzu.cz/~sipr/report-on-bug-in-dos.tar
There is also a readme.txt file in that tarball there spefying which file is which.
I tried to upload the tarball as an attachment to this report but the forum system
seems not to like me uploading files.

Ondrej

[mverstra]

Hello Ondrej

thanks for the bug alert. Your example file is quite large - do you have a smaller one, perhaps without any binary results?

The dos addition happens for the _m files only? The normal dos is correct for spin polarized systems afaik

Are you running in parallel?

M.