[Wien] Error in wien2wannier for spin orbit coupling case

[Niraj Aryal]

Dear wien2wannier developer

I am a novice wien2k user. I am running wien2k version 14.2 compiled with
intel compilers on machine with Centos operating system. I am trying to use
wien2wannier interface to get MLWF. I have been struggling with a strange
bug. I know this bug has been floating around for a while.I would really
appreciate your input.
So, here is my problem:

The crystal I am working is Zr2Te2P which has rhombohedral lattice
structure and has inversion symmetry. First I performed spin polarized with
spin orbit coupling DFT calculation and as per the direction in the manual
of w2w, I performed following steps:

->prepare_w2wdir dft_dir wannier_dir

->cd wannier_dir ; then init_w2w

-> x lapw1 -up -p ; x lapw1 -dn -p

-> x lapwso -up -p

-> x w2w -up -so -p

In this step, I get error like:

*forrtl: severe (39): error during read, unit 9, file wannier.vectorsoup_1*

and similar error for the down spin.

Strangely, I did not get this error for bcc iron when I was trying to
reproduce the example 18 result in the Wannier90 manual.

However I tried the similar thing for BaTiO3 crystal (by using the struct
file given in the lambda0 directory of BerryPi package) . And, I got the
same error. I wonder if this error occurs only for crystals with certain
symmetry.

The strange thing is w2w works fine without spin orbit coupling for all of
the above crystals.

I would really appreciate you feedback. Thank you!

[Elias Assmann]

Dear Niraj Aryal,

Sorry for the delayed reply. I will give my two Euro cents on your
report below. If you have solved the problem in the meantime, please
tell us what it was.

Post by Niraj Aryal
The crystal I am working is Zr2Te2P which has rhombohedral lattice
structure and has inversion symmetry. First I performed spin polarized
with spin orbit coupling DFT calculation and as per the direction in the
->prepare_w2wdir dft_dir wannier_dir
->cd wannier_dir ; then init_w2w
-> x lapw1 -up -p ; x lapw1 -dn -p
-> x lapwso -up -p
-> x w2w -up -so -p

This appears to be correct.

Post by Niraj Aryal
/forrtl: severe (39): error during read, unit 9, file wannier.vectorsoup_1/

This is not a lot of information to go by. Are there no other messages?

Does the same thing happen in a non-parallelized calculation? (If
necessary use fewer k-points, smaller RKmax etc.)


Elias

[Niraj Aryal]

Dear Elias
Thank you for your reply.
This is full error message I get when I run x w2w -up -so -p

forrtl: severe (39): error during read, unit 9, file
/home/aryal/work_wein2k/lapw/Zr2Te2P/wannier/./wannier.vectorsoup_1
Image PC Routine Line
Source
w2wc 000000000043CB33 Unknown Unknown Unknown
w2wc 0000000000465A44 Unknown Unknown Unknown
w2wc 0000000000429D82 read_vec_ 52
read_vec.f
w2wc 00000000004238FD MAIN__ 159 main.f
w2wc 0000000000403B9E Unknown Unknown Unknown
libc.so.6 00000037F501ED5D Unknown Unknown Unknown
w2wc 0000000000403AA9 Unknown Unknown Unknown
0.042u 0.037s 0:00.09 77.7% 0+0k 24+32io 0pf+0w


However, the program runs smoothly in serial version!

Also as I said here:
http://www.mail-archive.com/***@zeus.theochem.tuwien.ac.at/msg14497.html ,
I get no error at all when I consider bands starting from band index 1!

Post by Elias Assmann
Dear Niraj Aryal,
Sorry for the delayed reply. I will give my two Euro cents on your
report below. If you have solved the problem in the meantime, please
tell us what it was.

Post by Niraj Aryal
The crystal I am working is Zr2Te2P which has rhombohedral lattice
structure and has inversion symmetry. First I performed spin polarized
with spin orbit coupling DFT calculation and as per the direction in the
->prepare_w2wdir dft_dir wannier_dir
->cd wannier_dir ; then init_w2w
-> x lapw1 -up -p ; x lapw1 -dn -p
-> x lapwso -up -p
-> x w2w -up -so -p

This appears to be correct.

Post by Niraj Aryal
/forrtl: severe (39): error during read, unit 9, file

wannier.vectorsoup_1/
This is not a lot of information to go by. Are there no other messages?
Does the same thing happen in a non-parallelized calculation? (If
necessary use fewer k-points, smaller RKmax etc.)
Elias
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[wasim raja Mondal]

Dear Niraj,
Have you change symmetry operation 1 before running init_w2w?

Thanks
wasim

Post by Niraj Aryal
Dear Elias
Thank you for your reply.
This is full error message I get when I run x w2w -up -so -p
forrtl: severe (39): error during read, unit 9, file
/home/aryal/work_wein2k/lapw/Zr2Te2P/wannier/./wannier.vectorsoup_1
Image PC Routine Line
Source
w2wc 000000000043CB33 Unknown Unknown Unknown
w2wc 0000000000465A44 Unknown Unknown Unknown
w2wc 0000000000429D82 read_vec_ 52
read_vec.f
w2wc 00000000004238FD MAIN__ 159 main.f
w2wc 0000000000403B9E Unknown Unknown Unknown
libc.so.6 00000037F501ED5D Unknown Unknown Unknown
w2wc 0000000000403AA9 Unknown Unknown Unknown
0.042u 0.037s 0:00.09 77.7% 0+0k 24+32io 0pf+0w
However, the program runs smoothly in serial version!
I get no error at all when I consider bands starting from band index 1!

Post by Elias Assmann
Dear Niraj Aryal,
Sorry for the delayed reply. I will give my two Euro cents on your
report below. If you have solved the problem in the meantime, please
tell us what it was.

Post by Niraj Aryal
The crystal I am working is Zr2Te2P which has rhombohedral lattice
structure and has inversion symmetry. First I performed spin polarized
with spin orbit coupling DFT calculation and as per the direction in the
->prepare_w2wdir dft_dir wannier_dir
->cd wannier_dir ; then init_w2w
-> x lapw1 -up -p ; x lapw1 -dn -p
-> x lapwso -up -p
-> x w2w -up -so -p

This appears to be correct.

Post by Niraj Aryal
/forrtl: severe (39): error during read, unit 9, file

wannier.vectorsoup_1/
This is not a lot of information to go by. Are there no other messages?
Does the same thing happen in a non-parallelized calculation? (If
necessary use fewer k-points, smaller RKmax etc.)
Elias
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

--
With regards
Niraj Aryal
Grad. Student
Florida State University
Tallahassee, Florida-32304
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[Elias Assmann]

Post by wasim raja Mondal
Have you change symmetry operation 1 before running init_w2w?

Why would you do that?

[Niraj Aryal]

Thank you for taking part in this discussion Wasim.
No I have not changed symmetry operation to 1.

Post by wasim raja Mondal

Post by wasim raja Mondal
Have you change symmetry operation 1 before running

init_w2w?
Why would you do that?
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[Elias Assmann]

Dear Niraj,

Post by Niraj Aryal
Dear Elias
Thank you for your reply.
This is full error message I get when I run x w2w -up -so -p
forrtl: severe (39): error during read, unit 9, file
/home/aryal/work_wein2k/lapw/Zr2Te2P/wannier/./wannier.vectorsoup_1
Image PC Routine Line
Source
w2wc 000000000043CB33 Unknown Unknown Unknown
w2wc 0000000000465A44 Unknown Unknown Unknown
w2wc 0000000000429D82 read_vec_ 52
read_vec.f
w2wc 00000000004238FD MAIN__ 159 main.f
w2wc 0000000000403B9E Unknown Unknown Unknown
libc.so.6 00000037F501ED5D Unknown Unknown Unknown
w2wc 0000000000403AA9 Unknown Unknown Unknown
0.042u 0.037s 0:00.09 77.7% 0+0k 24+32io 0pf+0w

This looks a lot like the error reported recently by Wen Fong Goh on the
mailing list and here
<https://github.com/wien2wannier/wien2wannier/issues/10> (as well as by

Post by Niraj Aryal
However, the program runs smoothly in serial version!

does not fit.

So also in your case, I would advise you to try the latest development
version. If you send me your struct file and relevant parameters, I can
also try it out, if the calculation is not too large.

Post by Niraj Aryal
I get no error at all when I consider bands starting from band index 1!

Oops, I missed that message. (In the interest of “netiquette”, if you
find more information related to a previous e-mail, it is usually better
to reply to that e-mail rather than starting a new thread.)


Elias

[Niraj Aryal]

Thank you Elias. The newer version of w2w ran without giving any errors for
test case. I will do more tests and let you know.

However, the compilation of executables in SRC_trig directory gave
problems. This gives me error like:

write_win_backend.f(292): error #6460: This is not a field name that is
defined in the encompassing structure.

For this, I used the executables from the old SRC_trig directory and they
work fine with the new w2w executable.

Thank you again!

Post by wasim raja Mondal
Dear Niraj,

Post by Niraj Aryal
Dear Elias
Thank you for your reply.
This is full error message I get when I run x w2w -up -so -p
forrtl: severe (39): error during read, unit 9, file
/home/aryal/work_wein2k/lapw/Zr2Te2P/wannier/./wannier.vectorsoup_1
Image PC Routine Line
Source
w2wc 000000000043CB33 Unknown Unknown

Unknown

Post by Niraj Aryal
w2wc 0000000000465A44 Unknown Unknown

Unknown

Post by Niraj Aryal
w2wc 0000000000429D82 read_vec_ 52
read_vec.f
w2wc 00000000004238FD MAIN__ 159

main.f

Post by Niraj Aryal
w2wc 0000000000403B9E Unknown Unknown

Unknown

Post by Niraj Aryal
libc.so.6 00000037F501ED5D Unknown Unknown

Unknown

Post by Niraj Aryal
w2wc 0000000000403AA9 Unknown Unknown

Unknown

Post by Niraj Aryal
0.042u 0.037s 0:00.09 77.7% 0+0k 24+32io 0pf+0w

This looks a lot like the error reported recently by Wen Fong Goh on the
mailing list and here
<https://github.com/wien2wannier/wien2wannier/issues/10> (as well as by

Post by Niraj Aryal
However, the program runs smoothly in serial version!

does not fit.
So also in your case, I would advise you to try the latest development
version. If you send me your struct file and relevant parameters, I can
also try it out, if the calculation is not too large.

Post by Niraj Aryal
I get no error at all when I consider bands starting from band index 1!

Oops, I missed that message. (In the interest of “netiquette”, if you
find more information related to a previous e-mail, it is usually better
to reply to that e-mail rather than starting a new thread.)
Elias
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[Elias Assmann]

Dear Niraj,

Post by Niraj Aryal
Thank you Elias. The newer version of w2w ran without giving any
errors for test case. I will do more tests and let you know.

Good to hear, and thank you for reporting the compilation error.

Post by Niraj Aryal
For this, I used the executables from the old SRC_trig directory and
they work fine with the new w2w executable.

This should pose no problems, except that there have been some
bugfixes in write_win_backend.

Post by Niraj Aryal
write_win_backend.f(292): error #6460: This is not a field name that
is defined in the encompassing structure.

L. 292 is the empty line after

do i=1,inwf%Nproj
print fmt_centers, centers(i)
end do

I can only assume that the error refers to inwf%Nproj. This is a bit
odd because this and other fields of the same structure are used
throughout that file. Are there any other errors? What compiler and
version are you using?

Also, can you make sure that write_inwf_backend.f is compiled against
the new util_w2w.F? For instance, compile it in the wien2wannier
source directory rather than the Wien2k one (“cd
wien2wannier/SRC_trig; make write_win_backend”).


Elias

[Niraj Aryal]

Sorry, somehow I did not get the error again. I must have done something
stupid before. I am using intel compiler version 2016.1.150.

However, there is something serious thing I am encountering. When I do
wannier90 calculation after finding case.mmn, case.amn and case.eig (up and
down spins are added using w2waddsp), it gives me negative spread for Omega
I e.g:

Spreads (Ang^2) Omega I = -2549.235637370
================ Omega D = 1558.555122007
Omega OD = 3178.612948567
Final Spread (Ang^2) Omega Total = 2187.932433204

But we know that the spread has to be positive because it is expectation
value of (r- <r>)^2 . I found similar post here:
http://mailman.qe-forge.org/pipermail/wannier/2008-September/000199.html

Since Omega I is only calculated from the case.mmn file, I suspect that
there is some problem in doing so.

I am attaching input file. You are welcome to try this.

Post by wasim raja Mondal
Dear Niraj,

Post by Niraj Aryal
Thank you Elias. The newer version of w2w ran without giving any
errors for test case. I will do more tests and let you know.

Good to hear, and thank you for reporting the compilation error.

Post by Niraj Aryal
For this, I used the executables from the old SRC_trig directory and
they work fine with the new w2w executable.

This should pose no problems, except that there have been some
bugfixes in write_win_backend.

Post by Niraj Aryal
write_win_backend.f(292): error #6460: This is not a field name that
is defined in the encompassing structure.

L. 292 is the empty line after
do i=1,inwf%Nproj
print fmt_centers, centers(i)
end do
I can only assume that the error refers to inwf%Nproj. This is a bit
odd because this and other fields of the same structure are used
throughout that file. Are there any other errors? What compiler and
version are you using?
Also, can you make sure that write_inwf_backend.f is compiled against
the new util_w2w.F? For instance, compile it in the wien2wannier
source directory rather than the Wien2k one (“cd
wien2wannier/SRC_trig; make write_win_backend”).
Elias
--
Elias Assmann
Institute of Theoretical and Computational Physics
TU Graz <https://itp.tugraz.at/>
_______________________________________________
Wien mailing list
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien

[Elias Assmann]

Post by Niraj Aryal
However, there is something serious thing I am encountering. When I do
wannier90 calculation after finding case.mmn, case.amn and case.eig (up
and down spins are added using w2waddsp), it gives me negative spread
Spreads (Ang^2) Omega I = -2549.235637370
================ Omega D = 1558.555122007
Omega OD = 3178.612948567
Final Spread (Ang^2) Omega Total = 2187.932433204
But we know that the spread has to be positive because it is expectation
http://mailman.qe-forge.org/pipermail/wannier/2008-September/000199.html

Hm, I do not recall ever seeing negative values for the Ω components.
For completeness, note that the expectation value corresponds to the
*total* Ω; but nonetheless Ω_I must be nonnegative also on its own.
Also, your Ω_D and Ω_OD look very large. (Typically, they should be
small compared to Ω_I.)

Post by Niraj Aryal
Since Omega I is only calculated from the case.mmn file, I suspect that
there is some problem in doing so.

As Jonathan noted in the e-mail you linked, this is the case for an
isolated group of bands (no disentanglement). (This is not to say that
the error is not on the wien2wannier side, which it almost surely is.)

When you mention `w2waddsp`, does that mean you ran that program
manually? You should not need to do that, it is called by `x wannier90
-so`. Did you follow the procedure outlined in the wiki
<https://github.com/wien2wannier/wien2wannier/wiki/wien2wannier-with-SOC>?

If the problem persists, I can try it myself, but please give me any
parameters I need, i.e. anything that differs from

init -b -sp
runsp
initso
runsp -so

and the bands and initial projections you used.


Elias