my calculation is not converged in SCF iteration.

[gkr5623]

Hi. i want your help. great guys.

because my calculation is not converged.

i use 2 electrode in two probe system. Au/Graphene

but. my calculation is can not converged.

when is use zero bias(0volt) it can converged.

but, when is use some bias for example 0,1 volts. it is not converged.

what is the reason?

here i uploaded my input file.

and second question is when i use zero bias, Fermi energy of transmission curve is only Fermi energy of Au.

i can not understand.  in Docs » ATK Manual » NanoLanguage Reference Manual » Classes » TransmissionSpectrum page said

"The Fermi levels of the left and right electrodes are not necessarily the same a priori, if for instance the electrodes consist of different materials or are doped differently. However, when the device calculation is set up, the right electrode Fermi level is shifted (and all energy eigenvalues along with it) such that the Fermi levels coincide at zero bias."

i can not understand. plz someone help me.

my english is very poor. If the question I made in English was rude, I would appreciate it if you let me know how to ask a question.
Thank you.

[Daniele Stradi]

Hi,

you are using the default SG15-Medium basis set, which is quite long-ranged. My initial suggestion would be to:

1) Increase the length of the Au electrode to 6 layers;
2) Increase the number of Au layers in the central region to at least 11 layers;
3) Increase the length of the graphene region in the central region to 10 graphene layers;
4) Increase the length of the graphene electrode to 4 graphene layers.

See the attached image.

Regards,
Daniele.

[gkr5623]

Thank you. your reply

i have one question.
you said. SG-15 is long range. so, when i use short range scattering region structure, what i use

and, what mean about long range ??

when i increase Au, Graphene region i need so many times for calculation. so i set up smallest structure(i think). Is it wrong???

[mlee]

Hi,

It depends on the element and basis set.
In the case of SG15 basis set, it effects longer range than other basis set. (Look at the basis plot)

You are right if it is converged without problem physically.
We also recommend to analyze the HatreeDifferencePotential to see the convergence.
Even in the convergence in SCF iteration loop, sometimes, we need longer length of screening area.

Kind regards,
Maeng Eun Lee.

[Daniele Stradi]

Hi,

the meaning of "long-range" is related to the basis-set used in the LCAO calculator, which is formed by pseudo-atomic orbitals which ampliude is strictly zero after a given distance from the nucleus, due to the use of a soft-confining potential. You can read more about the basis set in the section "LCAO basis set" in https://docs.quantumwise.com/manuals/ATKDFT.html.

When you deal with a short electrode, I would suggest you to switch to the FHI basis set, which has a shorter range than SG15.

Regards,
Daniele

[gkr5623]

thanks to Daniele Stradi

can i ask more ?

my new question is about increased electrode.

Up to now, i knew that bulk calculation only had to have at least layer.
in your previously answer, i do not understand why increase the gold and graphene electrode layers.

sorry my english is poor.

Thanks very much for your time and help!
Kyungyeol Gu.

[mlee]

In general, the minimum layered electrode with a periodicity works well. Especially, symmetry electrodes will be worked without convergence problem. But in your case, you have asymmetry electrodes and have a convergence problem with a long-ranged basis-set, SG15. Try the convergence using the increased layered electrodes. You can also check your electrode convergence using the electrode validator plugin.
It will help to look at the "Convergence of electrode parameters" in the linked tutorial, https://docs.quantumwise.com/tutorials/atk_transport_calculations/atk_transport_calculations.html

[Petr Khomyakov]

Up to now, i knew that bulk calculation only had to have at least layer.
in your previously answer, i do not understand why increase the gold and graphene electrode layers.

You are totally right, the bulk calculation of electrodes can be done for a single principle layer only. As a matter of fact, this is how it has already been implemented in the coming release of QuantumATK-2018.06.

However, the electrodes are effectively longer than just the length of a principle layer, and this is reflected in the length of the electrode extensions. The reason for that is that after the bulk calculations of the electrodes are done, one has to couple these electrodes to the central region to impose open boundary conditions in device calculations (instead of periodic boundary conditions in bulk calculations). The coupling takes place via self-energy matrices with the matrix dimensions that depend on the range (cutoff radii) of the atomic orbital basis set functions used in the DFT calculation. This is because the extension of the orbitals determine the effective interaction range (length) between the semi-infinite electrode and the central region. Larger the interaction range more principle layers should be used for the electrode extensions (and electrodes in the older versions of the QuantumATK).

To really understand how this works, you would have to read some literature on the electron transport calculations, e.g., a book by S. Datta, see https://www.cambridge.org/core/books/electronic-transport-in-mesoscopic-systems/1E55DEF5978AA7B843FF70337C220D8B.