QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: ph on May 28, 2013, 15:54
-
Dear Sir,
energies=numpy.linspace(-0.8,0.8,401)*eV,
I plotted transmission Spectra with (infinitesimal=1e-9) different number of energy points in given range in a particular direction but every time only particular peaks are varying. What is the role of increasing and decreasing number of energy points in a given transmission spectrum. As the number of points increasing the peaks also increasing. Where to stop number of points? Which plot is reasonably correct?. Kindly see attached.
regards
-
The role is very simple, for each point E you compute T(E). So if your transmission is a very rapidly varying function of E, the height of the peaks will strongly depend on exactly where you sample the peak.
How many energy points you need is mainly a question of what quantity you ultimately want to compute. If it's the current, then compare the current for each number of E - if it changes a lot, you need to increase E even further.
Without knowing anything about this system, I would imagine it has some 3D interfaces as electrodes. In that case it's not just the energy points, but also the number of k-points that has to be checked for convergence. Sometimes when you increase the k-point sampling (for the transmission) the curves become smoother.
-
Dear Sir,
My system is 2D RTD (1T.B-1W-1T.B.). I know that you are expert in that. a) I used (9,9,100)cal.settings for Elect.(P.B.cond.) and Central(PB+Dir.cond.). Are they OK? b) How to increase the k-point sampling (for the transmission funct.) to get smoother curves?. Since my transport is in particular longitudinal direction i used code as below. In this case I didn't use MonK.Grid. then how can I do this?. Kindly give some suggestion.
energies=numpy.linspace(-0.8,0.8,401)*eV,
kpoints=[[0.0,0.0,0.0],[0.0,0.5,0.0]],
kpoints_weights = [1.,2.],
Regards
-
The basic settings look reasonable, although it's always up to you to make sure.
You could use an asymmetric MonkhostPack grid like (9,100) to have more points in the longitudinal direction.
-
Thanks a lot for your kind advice.