QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: yang su on September 6, 2012, 01:40

Title: how can I make a device periodical in one direction
Post by: yang su on September 6, 2012, 01:40

Hi all,

I am making a 2d FET device and what I care is just the transporting direction and the vertical direction. How can I make the other direction infinite using periodic boundary conditions and k-point sampling? Should I specify that in the builder or calculator? I wanna save some simulation time by doing that. By the way, If I wanna specify my contacts as gold. Should I just use spacial region in the builder of I need to actually add gold atoms at the contact area? Thank you so much.

Title: Re: how can I make a device periodical in one direction
Post by: esp on September 7, 2012, 02:04

Yes I am wondering this as well, and I suppose the other thing is, what will ATK assume about periodicity of the system, I saw other posts that said it is assumed to be periodic in X and Y, but that cannot be the case, for example with Graphene, since it is only periodic in Z ... so then how does it know which directions are to be periodic?  since I use the graphene builder tools, I am not sure since you do not directly set the unit cells, etc .. but I would guess it has to do with how you setup the unit cell is that correct? So then maybe to be periodic in Y but not X, you have to leave some space in the X directions maybe? Also Yang, when you run a TransmissionSpectrum, you can specify the kpoints parameter to use .. for example:

Code

			transmission_spectrum = TransmissionSpectrum(
				configuration = device_configuration,
				energies = transSpecEnergies,
				kpoints = MonkhorstPackGrid(1,1),
				energy_zero_parameter = AverageFermiLevel,
				infinitesimal = 1e-06*eV,
				self_energy_calculator = KrylovSelfEnergy(),
			)

Title: Re: how can I make a device periodical in one direction
Post by: mads.engelund on September 10, 2012, 10:11

You are correct esp,
Basically, if you set some unit cell ATK will assume that all directions are periodic (or A and B for device). Then if you don't want to actually model a periodic system you makes sure that the periodicity is large enough so that they don't actually interact.

If you are not actually modelling a periodic system then there is no reason to have more than 1 k-point in that direction.