Increasing the supercell's volume; after that, you optimize the structure first and then perform the desired characterization. This may work.

As Prakash Yadav wrote, change the cell volume. Did you check the k-mesh and ecutwfc? I see also that you use the Hubbard correction - in the first step, try to do calculations without the Hubbard flag, and if it works, you can improve the calculation by using DFT+U.

As seen from your screenshot, it's an issue related to your MPI procedures. It's very likely you need more memory for a larger cell. You'd better ask the administrators of your HPC for advice.

Przemyslaw Dziegielewski ,

K-mesh: in scf I used 2 2 2 0 0 0 and then in bands, I used R->gG->X->M->R->X->M->R (total 8) k-path (as FeS2 is Cubic structure).

ecutrho: 400 (scf) to 700 (bands)

ecutwfc: 50 (scf) to 100 (bands)

nbnd: Used 100 (in bands) instead of 544 (which is default) for faster calculation and for testing purpose

diag: davidson, not 'cg'

conv_thr: e-07 (scf) to e-11 (bands)

SCF ran good, but in bands calculations its showing the mentioned error.

With 12 atoms (which is one unit cell of FeS2) its showing good results, with 96 (2x2x2 supercell), its failing.

i'm attaching the output file. Please have a look Sir.

Dear Dipta Mahanta,

But you cannot do this:

"nbnd: Used 100 (in bands) instead of 544 (which is default) for faster calculation and for testing purpose"

nbnd is the number of bands to be calculated; you might want to increase it, not decrease it. Indeed, it corresponds to the number of states you make available to the electrons in your materials. QE will start allocating the electrons to each band and then probably crash because there are not enough.

Indeed, in your output, you can read

"number of electrons = 640.00

number of Kohn-Sham states= 100"

How can you fit 640 electron in 100 states? The code is going to crash somewhere.

Moreover,

"ecutrho: 400 (scf) to 700 (bands)"

"ecutwfc: 50 (scf) to 100 (bands)"

These two changes do not make much sense. Remember that the quality of the SCF calculation is in the SCF step and parameters. The calculations="bands" is a NON-SCF type of calculation -- it starts by reading the SCF output and builds from there interpolating the missing points. For example, the density, the central piece of information in DFT, in the non-SCF calculation is not changed.

My advise for a band calculation is to copy the SCF input file and then simply change calculation="bands" and the k-point mesh to what you need.

Finally, why do you want to run a band calculation with a supercell? The result ought to be identical to the case with a simpler cell, once you have unfolded the bands. In this respect, the band calculation offers no new insight into the physics of the supercell.

Best regards,

Roberto

Dear Sir Roberto D'Agosta ,

Thank you so much for your valuable reply. I will correct the input and put it to run this time and I will let you know the output.

The reason to run a band structure with a supercell is that after the pure FeS2 band structure, I will run again with Sulphur vacancy of

Dear Dipta Mahanta,

I was expecting your plan to do a vacancy or doping calculation -- please notice, however, that you don't need to do a band calculation for the pristine system in a supercell, as this would add nothing to what you already know. The band structure will come out identical once you unfold the bands.

I would therefore focus the computational time and effort in the supercell with vacancy/substitution, taking for granted the result for the single unit cell for the band calculation of the pristine system.

All the best,

Roberto

Roberto D'Agosta , Thank you Sir, I will consider that.

I have one more question.

As the supercell (a= 10.8076) is two times the dimension of the unit cell (a= 5.40381), does the K_POINTS coordinate becomes half of the original Unit cell (as its in reciprocal space)?

If yes, then the input in the bands.x file will be like the following right? --

....

For one cubic unit cell (with cell dimension, a= 5.40381), 12 atoms:

K_POINTS (tpiba_b)

5

R 0.5000 0.5000 0.5000

gG 0.0000 0.0000 0.0000

X 0.0000 0.0000 0.5000

M 0.0000 0.5000 0.5000

R 0.5000 0.5000 0.5000

to

For one cubic 2x2x2 supercell (with cell dimension, a= 10.8076), 96 atoms:

K_POINTS (tpiba_b)

5

R 0.2500 0.2500 0.2500

gG 0.0000 0.0000 0.0000

X 0.0000 0.0000 0.2500

M 0.0000 0.2500 0.2500

R 0.2500 0.2500 0.2500

Regards,

Dipta

Dear Dipta Mahanta,

As with other input parameters, QE offers many options to define them with respect to the parameters of the unit cell. For the k-points in the first Brillouin zone, these options are usually normalised with respect to the lattice constant. So, the high symmetry points defined with the tpiba_b format remain the same for the supercell and the fundamental unit cell.

On the other hand, if you wish to express the k-points in inverse length (Angstrom^-1 or Bohr^-1) then you would have to change them.

I hope this helps,

Roberto

Dear Sir Yue-Wen Fang ,

Thanks for the reply. The error is showing like that, but when I give the proper input, it shows results. From this I can say that with increasing the number of atoms, some input parameters should be changed too. The question is what parameters and by how much?

Because in 12 atoms, its giving good results but with 96, its showing the mentioned error, even after using MPI.