Empty gw.unsorted.eig file

[Bruno]

Hello!

I've succefully finished a yambo G0W0 run which I intended to interface with Wannier90. So I followed the steps of the example23 of wannier90 and computed the QP corrections on a reduced grid, where the indexes were obtained with the k_mapper.py from a higher grid on the nscf, as instructed. The run seems to be ended fine and I have the o- and r- files as usual. On the same folder I have my main wannierization files: pb.amn, pb.mmn, pb.nnkp and pb.eig, this wannier run was also done on the same reduced grid as instructed. In order to get the gw.unsorted.eig I've ran ypp with the ypp.in input that you can find attached to this post, and the code seems to run fine and the QP were also correctly read:
/\ \ /\ \\ _ \ /"\_/`\/\ _`\ /\ __`\
\ `\`\\/"/ \ \L\ \/\ \ \ \L\ \ \ \/\ \
`\ `\ /" \ \ __ \ \ \__\ \ \ _ <" \ \ \ \
`\ \ \ \ \ \/\ \ \ \_/\ \ \ \L\ \ \ \_\ \
\ \_\ \ \_\ \_\ \_\\ \_\ \____/\ \_____\
\/_/ \/_/\/_/\/_/ \/_/\/___/ \/_____/


<---> [01] MPI/OPENMP structure, Files & I/O Directories
<---> [02] Y(ambo) P(ost)/(re) P(rocessor)
<---> [03] Core DB
<---> :: Electrons :: 82.00000
<---> :: Temperature [ev]:: 0.000000
<---> :: Lattice factors [a.u.]:: 11.10214 11.10214 11.10214
<---> :: K points :: 35
<---> :: Bands :: 1000
<---> :: Symmetries :: 48
<---> :: RL vectors :: 132307
<---> [04] K-point grid
<---> :: Q-points (IBZ):: 35
<---> :: X K-points (IBZ):: 35
<---> [05] CORE Variables Setup
<---> [05.01] Unit cells
<---> [05.02] Symmetries
<---> [05.03] Reciprocal space
<---> [05.04] K-grid lattice
<---> Grid dimensions : 8 8 8
<---> [05.05] Energies & Occupations
<---> [06] == Wannier 90 Interface ==
<---> QP db read, writing QP corrections in the .eig Wannier90 format
<---> gw.unsorted.eig written
<---> [07] Timing Overview
<---> [08] Game Over & Game summary

The problem is that the gw.unsorted.eig is empty and I have no idea why. The seed name(seed = "pb") seems to be right on the ypp.in since is the same name from the wannier files. The only thing that I did different from the tutorials is that the Wannier90 run was made in a different directory because I ran yambo on a different(more powerful) cluster. Then I manually transfered the files pb.amn, pb.mmn, pb.nnkp and pb.eig to the yambo folder. Is that a problem? I supposed the ypp just needed to read the nnkp file, but in any case I've also copied the other ones, am I missing some important file? If this is a problem, how can I work around this?

You'll find attached on the tar file the ypp input and output. Thank you!

ypp.in.txt

r-wannier.txt

Regards,

[AntimoMarrazzo]

Ciao Bruno,

The problem is that the gw.unsorted.eig is empty and I have no idea why. The seed name(seed = "pb") seems to be right on the ypp.in since is the same name from the wannier files. The only thing that I did different from the tutorials is that the Wannier90 run was made in a different directory because I ran yambo on a different(more powerful) cluster. Then I manually transfered the files pb.amn, pb.mmn, pb.nnkp and pb.eig to the yambo folder. Is that a problem?

No, it should not be a problem - amn,mmn, nnkp, eig, they are all text files so you can move them between different platforms (i.e. computers) with no problem

I supposed the ypp just needed to read the nnkp file, but in any case I've also copied the other ones, am I missing some important file? If this is a problem, how can I work around this?

A couple of questions for you:
-Did you run the tutorial (example 23) from start to end with the same yambo/wannier executables? Was it all working fine?
-Which k-grid did you use to run wannier90? If you used Quantum ESPRESSO, could you post also the nscf input files?
Thank you.

Best,
Antimo

[Bruno]

Hello!

Thank you for the reply. I did run the example 23 although not with this yambo, but the same wannier, I would run it with this yambo version but I don't have the wannier compiled on that machine because we got some restrictions on the moment. But everything seemed to work fine. I also used the ypp of this yambo version to interpolate the DFT bands and test ypp and they were also ok.

About the k-grid question. To generate the SAVE/ folder I used a 8x8x8 grid on the nscf calculation, but used a 6x6x6 reduced grid to to compute the QP corrections, which I obtained the indexes using the k_mapper.py utility so I could input it on yambo. To the wannier I also used a 6x6x6 k-grid, which I've generated using the kmesh.pl utility. Attached you will two nscf inputs: The nscf_gw.txt is the one used to generate the SAVE/ and the output of it was used to collect the indexes as discussed above. The nscf_wannier.txt is the one used to the wannier90 run.

nscf_wannier.txt

nscf_gw.txt

Regards,

[AntimoMarrazzo]

I think the problem is due to the fact the two k-grid are not commensurate, in other words the 6x6x6 is not a subset of the 8x8x8.
In practice, you do not have all the QP corrections that you would need for the Wannier grid.
If you use 8x8x8 for the yambo self-energy, the QP corrections can be computed only on the following grids i x j x k where i,j,k=1,2,4,8.
So if you redo your wannierisation with a 4x4x4 grid it should work.

[Bruno]

Hello,

Oh I forgot about this detail, my bad!! Thank you very much. I'll do the test and post an update here.

Regards,

[Bruno]

Hello!

In fact it did worked and now the gw.unsorted.eig has stuff inside. But now I'm having problems with the gw2wannier.py. It seems to me that I misunderstood something from the tutorial, here is what I did:

I calculated the QP corrections on the reduced 4x4x4 kgrid which lends to 10 indexes since I started from a 8x8x8 grid on the nscf. Since I wannierized just some bands around the gap(bands 82 e 83 are HOMO and LUMO and I used 26 WF), I asked to yambo to compute the QP from the band 60|110, in which this range contains my wannierized functions and I thought that I would maybe need some extra corrections because of the disentaglement proccess. But I think there is something wrong if this idea, because when I try to run the gw2wannier.py it writes a empty pb.gw.eig and then I get a traceback:

Traceback (most recent call last):
File "../wannier90-3.1.0/utility/gw2wannier90.py", line 216, in <module>
amn = AMN[BANDSORT[ik], :, ik]
IndexError: arrays used as indices must be of integer (or boolean) type

Sorry but can you point me out what I exactly I misunderstood? I also now tested the gw2wannier.py with both this yambo version and this wannier on the example23 and its working fine. You'll find attached my wannier input, yambo input and also the generated gw.unsorted.eig, which values seems kinda weird to me, because I suppose one column runs over bands and the other over k points, but the numbers don't seems to match and I can't find the reason.

wannier.win.txt

gw.unsorted.eig.txt

gw.in.txt

Regards,

[Bruno]

I thinking that maybe since I used 200 bands on the nscf that I used to do the wannierization, maybe I actually have to compute the QP from bands 1|200|, is that right?

Regards,

[AntimoMarrazzo]

I thinking that maybe since I used 200 bands on the nscf that I used to do the wannierization, maybe I actually have to compute the QP from bands 1|200|, is that right?

Regards,

Yes, the number of QP bands should match num_bands from Wannier90: this is because in practice we are replacing the entire .eig file with G0W0 corrections.
Hence, it is useful to use exclude_bands in Wannier90 to begin with, so to avoid computing QP corrections for very low bands that are not really necessary to wannierise around the Fermi level.

[Bruno]

Hello,

That is very interesting. So for exemple in my case I included num_bands=200, if I exclude_bands from 1 to 59 and from 111 to 200 from the wannierization I could actually compute the QP just from 60|110 is that right?

Thank you very much for you help!

best regards,

[AntimoMarrazzo]

That is very interesting. So for exemple in my case I included num_bands=200, if I exclude_bands from 1 to 59 and from 111 to 200 from the wannierization I could actually compute the QP just from 60|110 is that right?

Correct, you can check the .eig file from DFT: if you apply the exclude bands as you said then you will see that the .eig contains 50 bands and the band index goes from 1 to 50.