2
$\begingroup$

As the title says, I'm interested in performing Transiesta calculations with a device region that needs to be optimized. I read the slides reported here: https://departments.icmab.es/leem/SIESTA_MATERIAL/Docs/Tutorials/tlv14/slides12.pdf

However, I was wondering about the specific procedure regarding the connectivity atoms between the device region and the electrodes. In my case, I don't have a molecule but rather a bulk material presenting defects in the device region. This material is the same as used in the electrode, but without the defects. So should the procedure be:

  1. Define the device region and completely relax it
  2. Add one/two electrode layers, relax the structure but constraining the latter layers
  3. Add another one/two layers and relax the whole structure constraing the latter added layers
  4. Add the 'bulk' electrode with the converged amount of layers, relax the whole structure by constraining the bulk electrode and one/two connectivity layers.
  5. Proceed with the transport calculations.

I'm still unsure about step 4 and whether I should relax the connectivity layers.

Moreover, since in my case the structure is a simple graphene monolayer with an atomic defect in the device region, I was wondering if I should actually do this step by step or go directly to step 4 (and thus construct the whole scattering region and relax the structure by freezing the electrodes+connectivity atoms) would be sufficient. Thanks for your help!

$\endgroup$
4
  • $\begingroup$ What is it exactly in 4) you are not sure about? Is it the wording, or? The procedure you describe is a general recommendation. There are variations of it, sure, but effectively one should just be careful about the connectivity layers. $\endgroup$
    – nickpapior
    Dec 7, 2022 at 12:14
  • $\begingroup$ In your slides, you describe a case in which a molecule is the device region and the electrodes are a metal. In this case, I imagine adding layers and optimizing step-by-step considering some connectivity atoms is particularly important. However, let's say that the device and electrodes are the same material and the device region just presents some defects. In this case: 1. Is it important to have un-relax connectivity atoms? 2. Is it important to follow this step-by-step structure optimization or would it be enough to just relax the whole structure freezing the electrodes? $\endgroup$
    – Laura
    Dec 7, 2022 at 13:07
  • $\begingroup$ My slides are just an example of how to fast get a relaxed geometry. Relaxing a geometry with 1000 constraints is slower than relaxing two geometries with 500 constraints (conceptually). So it is not a requirement. Generally relaxing all in one go is the most generic and accurate procedure, and sometimes (if your initial configuration is a good guess) isn't that much longer than the above procedure. Typically the connectivity region should also be somewhat un-relaxed (same relative coordinates as the electrodes), but they can move a little. $\endgroup$
    – nickpapior
    Dec 8, 2022 at 11:44
  • $\begingroup$ Thank you very much Nick! You can add this as an answer so I can accept it. $\endgroup$
    – Laura
    Dec 8, 2022 at 11:57

1 Answer 1

2
$\begingroup$

The referred slides are just an example of how to fast get a relaxed geometry for complex geometries. Relaxing a geometry with 1000 constraints is slower than relaxing two geometries with 500 constraints (conceptually). So segmenting it for speeding up the process can in some cases be ideal.

It is definitely not a requirement to follow the procedure. Generally, relaxing all in one go is the most generic and accurate procedure, and sometimes (if your initial configuration is a good guess) isn't that much longer than the referenced procedure.

Typically the connectivity region should also be somewhat un-relaxed (same relative coordinates as the electrodes), but they can move a little. And it is thus recommended to have 1 or 2 layers with the same relative coordinates as the electrodes. But as everything, testing and knowing your electronic structure will let you bypass general recommendations. Testing is always a good idea.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .