Dynamical Matrix of a Tiny Graphene Nanoribbon takes hours x hours

[Heinz]

Hi
I tried the inelastic spectrum tutorial of the silicon pn junction and the calculation of the dynamical matrix takes around  a min. However when I try to calculate the dynamical matrix of a tiny graphene nanoribbon as attached it takes ages. I'm sure I'm making an error in the parameters of the dynamical matrix calculation.

I checked the ATK manual and it says that the number of repetitions should be given as the repetitions of the structure.

1. As I understand if I build an 6-atoms wide GNR and repeat it 3 times in C direction, I have to set the repetitions in the calculation of the dynamical matrix as (1,1,3), is it right? But when I do this, it takes ages to calculate hours and hours. (attachment 1)

2. And another problem is that when I build a GNR device as attached (attachment 2), the settings for the z-repetitions in the dynamical matrix in VNL seems disabled however the structure is 1-D and I have to set it as the number of repetitions of the unit sell in z-direction? Is it right? When I set the repetitions as (1,1,4) in the Editor, this calculation (attachment 2) also takes hours and hours.

I'd be very glad if you could  tell me the trick about calcultaing the dynamical matrix in a min as in the Si pn junction tutorial, otherwise I cannot set it up and takes ages to calculate (and probably wrong).  Couldn't manage to work this for a few days. Thanks.

[Anders Blom]

The tutorial uses a classical potential whereas your setup treats the phonons in DFT. There is a huge difference in complexity there - you can parallelize your way out of it though, to some extent. But since it's just carbon and hydrogen, why not use a Tersoff potential?

[Heinz]

Hi
Thanks for your quick reply. So, also for the device calculation, can I use ATK-classical with Tersoff potential? And can I use just automatic repetitions?

[Anders Blom]

For  the phonon part, yes, that is exactly what the tutorial does.

[Heinz]

Thanks a lot.

[Jess Wellendorff]

You may be aware of this, but let me just add one note about configuration repetitions for the DynamicalMatrix when considering 2D or 1D structures:

The "automatic" determination of a suitable repetition scheme sometimes leads to erroneous repetition along non-periodic directions in 1D and 2D systems. For e.g. a nanotube, the repetition should be 1 along the directions perpendicular to the tube length (usually A and B). In case the "automatic" option comes up with something different, it is best to manually set the repetition (saves a lot of unnecessary calculations along non-periodic directions).

[Heinz]

Thanks for the repetition info. Then for example in my tiny sample, 4 repetitions of a GNR in z-direction should lead to a repetition of (1,1,4)?

And more importantly, what I also would like to ask is that when I build a device structure in VNL, in the dynamical matrix menu, editing of the repetition along z-direction is disabled and automatically set to 1, what's its reason? Thanks again.

[Jess Wellendorff]

For a GNR you should use a repetition of (1,1,x), where x = 3, 5, 7, etc. Even numbers along z is not allowed.

If you have a device, it does not make sense to repeat it along the transport direction. By construction, the device central region has the length it has, nothing more, nothing less. It is of course semi-periodic through the electrodes, but this is handled behind the scenes. Therefore, only repetitions of (x, y, 1) are allowed for devices.

[Heinz]

OK, thanks a lot for these valuable info.