Author Topic: Gamma-point sampling result  (Read 1954 times)

0 Members and 1 Guest are viewing this topic.

Offline ao2207

  • New QuantumATK user
  • *
  • Posts: 3
  • Country: us
  • Reputation: 0
    • View Profile
Gamma-point sampling result
« on: January 23, 2018, 03:36 »
Is it true that the Gamma-point sampling result is more reliable for larger supercells?
I want to calculate transmission spectrum and I-V for a supper cell containing 350 atom. Can I use Gamma Point sampling to reduce computational cost?

Offline Petr Khomyakov

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 1290
  • Country: dk
  • Reputation: 25
    • View Profile
Re: Gamma-point sampling result
« Reply #1 on: January 23, 2018, 10:07 »
Is it true that the Gamma-point sampling result is more reliable for larger supercells?
No. You should always try using denser k-point grid to ensure that the results are converged with respect to the k-point sampling.

I want to calculate transmission spectrum and I-V for a supper cell containing 350 atom. Can I use Gamma Point sampling to reduce computational cost?
Nothing stops you doing it, but if you do self-consistent device calculations, you must use many k-points in the C (transport) direction for the electrodes. Regarding the two lateral (A and B) directions, you may in principle adopt a single k-point for self-consistent calculations, but this  really depends on the system and the supercell size.

For the transmission spectrum and I-V curve calculation, you only have to sample the 2D Brilloiun zone (BZ) corresponding to the lateral supercell mentioned in the previous paragraph. But you also have to make sure that these quantities are converged with respect to the k-point sampling. The G k-point is not that special for the electron transport, unless we are talking about normal tunneling where it gives the major contribution to the transmission. Otherwise, you may have transmission resonances all over the entire 2D BZ, and then you may need to use many k-points to converge the transmission spectrum and IV curve.

So, all these things are system-dependent, and have to be systematically investigated by doing convergence tests.