QuantumATK > General Questions and Answers

Number of K-point for transmission spectrum

(1/2) > >>

NW:
Hi,

I need to know how to choose the correct number of k-point (kA and kB) for calculation of transmission spectrum.

Thanks

mlee:
It is hard to suggest the correct number of K-point. It depends on the model. We guide the preset densities.

Anders Blom:
Or put differently: the only way to know is to keep increasing the number of k-points until the transmission spectrum (or better yet, the current) doesn't change.

Note that in general you need more k-points than for the SCF loop because the transmission coefficients as function of k (and, sometimes, energy too) can be quite spiky, whereas the density is more smooth and converges faster (excepts for some tricky systems, of course). Cf. the discussion here: https://docs.quantumatk.com/tutorials/nisi2-si/nisi2-si.html#transmission-spectrum-convergence

NW:
Dear Anders,

Thanks for your explanation. The transmission spectral of my system is so sensitive to the number of k-points and when I increased the number from 7 to 50 by choosing random numbers in between I got completely different results.

Anders Blom:
For the current? Well, yes, I guess it goes to show the importance of converging in this number, a point perhaps not made strongly enough in our tutorials or manuals.

The underlying reason is basically that the transmission is a resonance, so if the electrode band structure has several local minima/maxima in the energy range, then at any given bias the overlap between left/right states is almost random. Notably, you don't have to converge the whole transmission though, just the relevant range which is intergrated for the current. Often the band structure and hence transmission is more complex farther from the Fermi level, but this makes no difference for the current at low bias at least.

Navigation

[0] Message Index

[#] Next page