SOC Band-structure convergence issue for a strained crystal

[sougat_purohit]

Dear all

I need to deform a Bi2FeCrO6 crystal with R3 symmetry along the x-direction by 1,3,5,-1,-3,-5% and get the SOC band structure of each deformed system with GGA = PE. I could perform the structure relaxation, self-consistent calculation and non-self-consistent without any issues. However, I am facing convergence issues with the SOC band structure calculation. I tried to converge the SOC band structure with ICHARG = 12, but still, it doesn’t seem like it’ll converge. Why do all other calculations converge but not the SOC band-structure calculation?
Any suggestions on how to overcome this issue?

I have attached the input files for the 1% deformed crystal in the x-direction. I have also attached the POSCAR file for the un-deformed crystal.

Regards
Sougat Purohit

[martin.schlipf]

Are these the exact inputs? If yes then the CHGCAR file is missing.

Otherwise could you tell me the exact steps to reproduce the issue.

[sougat_purohit]

Dear Martin Schlipf

The CHG and CHGCAR files made the folder too large so that I couldn't include them.
However, I am describing the exact steps I followed to reproduce this issue:

1. structure relaxation of the deformed crystal

1_str_rlx.zip

2. self-consistent run by taking the CONTCAR of structure relaxation as POSCAR

2_SC.zip

3. non-self-consistent run by taking the CONTCAR of self-consistent run as POSCAR

3_NSC.zip

4. SOC band structure calculation (BS.zip, previously attached)

I have attached my input files for every step.

Regards
Sougat

[martin.schlipf]

Thanks, I will have a look. It seems that the third step has the wrong KPOINTS file, but I will just use the one from your previous post.

One thing you should be aware however is that this calculation will be much more expensive than the 2nd step, so perhaps the issue is just that you did not allocate enough resources. In the second step you use 2 irreducible k points (after symmetry reduction). For the band structure you want 9 lines with 80 points each = 720 k points.

You could try to use fewer points per line or not to calculate all lines at the same time to see if that fixes your issue.

[sougat_purohit]

Dear Martin Schlipf

Thank you very much for the suggestion!
I'll redo the band-structure calculation with fewer points per line and check if it works.

Also, could you tell me what is wrong with the KPOINTS file in the third step? Do I need to use a gamma-centred grid?

Regards
Sougat

[martin.schlipf]

The attached KPOINTS file for the 3rd step is the same as for the 2nd one. Obviously you would want to run a KPOINTS file for the band structure there.

My test finished, I could produce a band structure with the KPOINTS file from the original post reducing the number of points per line to 50. To give you an idea with my setup the band structure calculation was about 150 times more expensive than the 2nd step, so make sure to scale your allocated time appropriately.

newplot.png

[sougat_purohit]

Understood. Thanks a lot for the help!