question on The number of k-points in the equivalent nulk calculation

[jjhskang]

Hi!the

I wonder how ATK sets the number of k-points in the equivalent bulk calculation for a two-probe system. Does it use Gamma-point only?

In relation to this question, it is strange that I do not get the same Fermi level for an electrode and the equivalent bulk calculation in the two-probe configuration for a perfect metal wire. We should expect the same value of Fermi energy for the two calculation!

I guess that the observed difference comes from the difference in the number of k-points, i.e., the difference in the accuracy of two calculations. As everbody does, we can specify the number of k-points for the elctrode in the two-probe calculation, say, 40. Then, how is the number of k-points for the equivalen bulk set ? 

The reason why I am asking this is if we can make an estimation of charge transfer between the electrode and the scattering region by comparing values of the Fermi level from the two calculations.

[jjhskang]

Hi!

Let me make some additional description for the probem of my intesest.

If LUMO of an isolated molecule calculated from calcuteMolecularEnergySpectrum is lower than the Fermi energy Ef of an electrode of concern, we will expect a charge transfer from the reservoire of the electrode to the molecule, when it is immsered in the two-probe geometry.
 
How can we find if there would be a charge transfer in the two-probe calculation if the comparison of Ef from the electrode and that of the simple equivalen bulk calculation does not show it? is there clearer way of finding it?

[Nordland]

If you use the kpoints (ni, nj, nk), then it uses (ni, nj, 1) for the equivalent bulk run.

[jjhskang]

So you mean that ATK uses 1 k-point along Z direction for the equivalent bulk calculation for the scattering region ?

[Nordland]

Yes.

[jjhskang]

Thanks!

That's what I guessed.
 
By the way, can you answer my second question? is there any simple method how we can see charge transfer from/to the scattering region? it will be good if we can adjust the number of k-points along Z axis for that purpose so that the Fermi level of the equivalent bulk can be accurately located if desired, though not always.

[Nordland]

The charge transfer is quite straight forward to see once, you have a converged twoprobe calculation.
For step 0, there is a  charge value printed, something like:

# sc  0 : q = 1309.00000

And once it is converged, it will report the a new value for the charge

# sc  14 : q = 1308.985

The charge transfer is then the difference between these two numbers.

[ipsecog]

Nordland, I have noticed in some calculations that the value of "q" in the SCF loop is not necessarily the same as the sum of all Mulliken charges. Which value is more correct? Isn't it better to sum up the Mulliken charges in the central region, for the charge transfer?