Mistake in modelling graphene nanoribbon with CONDUCATNCE VS. Vg

[donmbringer]

I have built the model of the device of a graphene nanoribbon with applied gate electric field, like shown in the .py file below. However, when I run the program in 1 and 2 to get the curve of conductance vs. gate voltage at temperature of 25K and 350 K, the results I obtained is shown in 4.jpg.

This is not correct, since theoretically it should be a "V" shape, with the minimum point at Vg=0. Could you please tell me the reason？

Thanks ahead!

[Anders Blom]

There is not enough information in the script, it only contains the geometry, not the calculation model. However, given the almost flat curves, I have a suspicion this might be caused by a recently discovered bug in 12.2.2 (fixed for 12.8 coming out soon) - check the log file of the calculation, if you see big WARNINGS then the calculation has not converged properly and the results are meaningless.

[donmbringer]

I choose calculator and transmission spectrum:
in calculator, I choose: n = 1, 1, 100, with SCF, and then change basic set to Cerda carbon.

In spectrum, I change energy range to -4 ~ 4, and calculation model to Krylov. Others are all in default.

there is a lot of repeated warning as follows:
################################################################################
# WARNING                                                                      #
#                                                                              #
# The computed multigrid residual is greater than the required accuracy.       #
#                                                                              #
# Computed residual :   7.91855e-012                                            #
# Required accuracy :   1.00000e-012                                            #
#                                                                              #
################################################################################

And then the final result was shown as before. Could you please tell me the reason?

[Anders Blom]

It's a bug. We have fixed it for 12.8, we may even make a 12.2.3. We will know better later this week when the release can be made. Apologies for the inconvenience.

[Anders Blom]

However, also note that you should adjust the boundary conditions. When we ask about the calculator settings, please always include the script, because there are many small details which may not be obvious, but help us determine the root cause of problems. In this case, it would be important to see if Neumann or Dirichlet boundary conditions apply in the Y direction.

[donmbringer]

Sorry. I forget to mentioned last time, I adjusted the first two (A ,B) boundary conditions to Neu, and the third one  (C) to Dir, which was the same as the one mentioned on Tutorial.  I just repeat the one on tutorial step by step, so I don't know why my result is different from the one on tutorial.

[Anders Blom]

Ok, that's correct then - but also the cause of the problem. As mentioned, there is a bug in 12.2.2 which gives wrong results for Neumann.

[donmbringer]

Hi Dr. Blom,

I found one possible problem in my method. To calculate the Conductance vs. gate voltage, I directly used the code on Page 19 of the tutorial "Graphene junction device".

The model the tutorial used is the one which has been built in software. However, my device is built by myself (as shown in previous attachment).

I am not quite familiar with the ATK language, but is this code also suitable for my model (such as defining the correct gate field at the correct region)?

If not, how to modify it?

Thanks ahead!
 

[Anders Blom]

Your structure is basically set up correctly. The gate is a bit close to the graphene, but ok if you want to simulate it as a substrate it might be a good idea - I would just make the dielectric a bit thicker.

I do note however that you electrodes (and the ribbon as a whole) is made from armchair graphene, which is not conductive, so you will not really see any current in this system. In general the electrodes should be metallic (or doped).

The code for varying the gate voltage and source-drain bias is rather general, so it will apply to any structure (of course with small modifications to import the correct structure etc).