Band structure calculation using SOC

Running Abinit Ground state
1

Dear all,

I am new to ABINIT and trying to run the band structure calculation of Bi2Te3 using SOC in abinit and unfortunately the calculation stops at a particular point, which I understand is because of memory reasons (most probably) but also I can not run abinit on multiple nodes also due to some parallelization contrainsts. Here I am putting my input file and submit script. Kindly let me know why the calculation does not run successfully.

Submit script:
#!/bin/bash
#SBATCH -t 90:00:00
#SBATCH --ntasks-per-node=8
#SBATCH -c 16
#SBATCH -J “abinit-test”
#SBATCH --account=xxxx

soc_input.txt (2.6 KB)

#SBATCH -p normal

export OMP_NUM_THREADS=1

Run with full MPI task count

mpirun -np 64 abinit bi2te3.abi > abinit.out

Thank you very much,
Bibek Samal

2

Hi Bibek,

I can suggest a couple of ways to improve your experience with this set of calculations:

  1. perform a convergence test with respect to the vacuum distance. I notice in your input you use almost 122 angstrom. If your system is pristine (as in not doped or having defects etc.), I would suggest starting with 12 Angstrom vacuum distance and increment until convergence of total energy.
  2. 24x24x1 k-point grid is also an enormous grid of k-points. Againt what property did you converge this grid?
  3. is your rprim defined correctly for your 2D hexagonal cell?

Let me know!
Bogdan

3

Hi Bogdan,

Thank you very much for you reply.

To compare the band structure with that of Quantum espresso I took 24241 k-point grid because there in the literature I found someone using the same gird. I guess the rprim definition might be wrong because I am not getting the correct band structure even in the non-relativistic limit. I am playing with various parameters. I will let you know soon.

Thanks,
Bibek

4

Hi Bogdan,

The primitive lattice vectors (rprim) used in my calculation is “rprim
1.0 0.0 0.0
0.5 0.866 0.0
0.0 0.0 1.0”

Bibek