Converging k-points in Transiesta/TBTrans

Asked by Tristan Zaborniak on 2019-12-03

I believe that there are some gaps in my understanding of how to converge k-points in Transiesta/TBTrans (version 4.0). Below I've compiled what I've read/know, asking questions where unsure.

Converging k-points in Transiesta:

1. For the electrodes, I've read that one should use a high number of k-points in the transport direction (~100), and then converge the transverse direction k-points. Is this so? I would've thought that because the electrode calculation is a Siesta calculation, one would have to converge k-points in the transport direction as well.

2. After the k-points of the electrode are converged, the transverse-direction k-points of the device should be the same as for those of the converged electrode. Is this true? And, because Transiesta truncates the transport k-points to 1, do we need to worry about converging k-points for the device region at all? Or do we converge for the device region first, and then use the transverse-direction k-points in defining those of the electrode?

Converging k-points in TBTrans:

3. Here I've read that one must converge transverse-direction k-points by looking at the averaged transmission. This is done by overlaying transmissions calculated using different k-point samplings, and finding the density at which the superimposed curves fall on top of one another, yes?

And, finally, regarding the last column of the k-points block: is it a good rule of thumb to set each entry (offset) to 0.5? Are there cases where setting each entry to 0, or some other number, is preferable?

Thank-you for your help.

Question information

English Edit question
Siesta Edit question
No assignee Edit question
Solved by:
Nick Papior
Last query:
Last reply:
Nick Papior (nickpapior) said : #1

1) Yes, this is so. Essentially you could also converge k-points along semi-infinite direction. However it is much simpler to just use a very high number (like 100)

2) Yes, they should be the same.
Sometimes it can work better to have k-points along the semi-infinite direction for the device. Other times not so much. In any case 1 along the semi-infinite direction in the device is fine.

3) This is done by analysing the k-averaged transmission for various k-samples. If this was what you meant, then yes.

Your final question is not specific to transiesta. This relates equally well to regular siesta calculations. So your choice.

Tristan Zaborniak (tristanz) said : #2

Hi Nick. Thank-you for your response.

I have one further question: because TBTrans uses the output of a Transiesta calculation run at a particular bias, must we converge k-points in TBTrans for each bias we run a structure under?

Best Nick Papior (nickpapior) said : #3

For consistency you should use the same TBtrans k-grid for all TBtrans calculations (otherwise comparing may be difficult).

Tristan Zaborniak (tristanz) said : #4

Thanks Nick Papior, that solved my question.