Author Topic: some questions about ivcurve converge  (Read 3558 times)

0 Members and 1 Guest are viewing this topic.

Offline 395235863

  • Heavy QuantumATK user
  • ***
  • Posts: 81
  • Country: cn
  • Reputation: 0
    • View Profile
some questions about ivcurve converge
« on: May 7, 2015, 11:03 »
1.There are two device models for semiconductor:

first,use Au(1,1,1)-semiconductor interface''http://quantumwise.com/documents/tutorials/latest/Au_Pentacene/index.html/index.html'';

second,use doping inATK''http://www.quantumwise.com/publications/tutorials/item/820-ni-silicide-si-interfaces''.

what the differences between the two-probe models?and which one is better to be converged?

2.There are two ways to calculate the iv curve:

first,use IVCurve and use biases = [0.0, 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4, 1.6, 1.8, 2.0]*Volt;

second,use code in the web ''http://quantumwise.com/publications/tutorials/item/114-i-v-curve-and-voltage-drop'';

what the differences between the two ways?and which one is better to be converged?

and what the differences between biases = [0.0, 0.2, 0.4, 0.6, 0.8, 1.0] and biases = [0.0, 0.1, 0.2, 0.3, 0.4, 0.5,0.6,0.7,0.8,0.9,1.0]?is the latter one is converged better than the former one?

3.now i have studyed in 2D materials,i what to know what the range of  center region's length should be use so that the model can be converged better?

Thank you!

Offline Umberto Martinez

  • Supreme QuantumATK Wizard
  • *****
  • Posts: 479
  • Country: dk
  • Reputation: 26
    • View Profile
Re: some questions about ivcurve converge
« Reply #1 on: May 8, 2015, 10:19 »
1) ??

2. The IVCurve analysis will
- run the finite bias SCF calculation
- store it an nc file IF you set selfconsistent_configurations_filename parameter
- calculate the TS
- loop over the bias voltages
- from the TS calculate and store the IV curve object that you can easily plot from the LabFloor.
Easy to set up but the drawback of this method is that if your job is interrupted you will loose the transmission spectra which you will have to recalculate from the saved SCF states.

The second method you refer does exactly the same the steps above are done "manually" with python scripts.

3. Check this tutorials:
http://quantumwise.com/documents/tutorials/latest/ATKTutorialDevice/index.html/
http://quantumwise.com/documents/tutorials/latest/VNLTutorial/index.html/chap.gnr.html




Offline 395235863

  • Heavy QuantumATK user
  • ***
  • Posts: 81
  • Country: cn
  • Reputation: 0
    • View Profile
Re: some questions about ivcurve converge
« Reply #2 on: May 8, 2015, 17:05 »
1.i mean there are two two-probe models we can use.
one is like 1.png , that uses Au as electrode.
the other uses doping code.the code i have mentioned in http://quantumwise.com/forum/index.php?topic=3375.msg15503#new

i what to know the differences between them.

Thank you very much!

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5394
  • Country: dk
  • Reputation: 89
    • View Profile
    • QuantumATK at Synopsys
Re: some questions about ivcurve converge
« Reply #3 on: May 8, 2015, 22:54 »
It's not really two different two-probe "models" - the calculations are performed exactly the same way. It's simply different physical systems to model: in the first case, it's a heterojunction (single interface) between a metal and a molecular crystal (not quite a semiconductor), in the second it's also a heterojunction but now between two doped semiconductors. And, in your picture it's a metal-semiconductor-metal structure, although the semiconductor in this case is more like a large molecule, but that's also fine.

This is, you can say, one of the strengths of ATK - you model what you want to model. You can study different physical setups and configurations, which are relatively close to the experimental "reality", without making too many of the traditional assumption (slab model, no doping, etc). If you want to model a doped system like a p-n junction you do that, if you want to know the properties of a heterojunction you set up such a system, if it's a M-S-M system that you are trying to understand, as in your picture, then you choose that.

Offline 395235863

  • Heavy QuantumATK user
  • ***
  • Posts: 81
  • Country: cn
  • Reputation: 0
    • View Profile
Re: some questions about ivcurve converge
« Reply #4 on: May 9, 2015, 14:40 »
i just want to simulate 2D material field effected transistor,and i don't know which model can tell the intrinsic properties ,like ivcurve.

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5394
  • Country: dk
  • Reputation: 89
    • View Profile
    • QuantumATK at Synopsys
Re: some questions about ivcurve converge
« Reply #5 on: May 10, 2015, 20:27 »
As mentioned, there is only one model, both examples you mentioned use the same method to compute the I-V curve.
See http://quantumwise.com/documents/tutorials/latest/GrapheneDevice/index.html/ for something similar to what you want to study.

Offline 395235863

  • Heavy QuantumATK user
  • ***
  • Posts: 81
  • Country: cn
  • Reputation: 0
    • View Profile
Re: some questions about ivcurve converge
« Reply #6 on: May 11, 2015, 04:14 »
As mentioned, there is only one model, both examples you mentioned use the same method to compute the I-V curve.
See http://quantumwise.com/documents/tutorials/latest/GrapheneDevice/index.html/ for something similar to what you want to study.
Thank you!
the question is that how to model the electrode.
Must electrode be conductive?
the graphene use itself as the electrode,so the iv curve is intrinsic properties?for my study in MoS2,can i use itself for electrode?or use some tricks?

Offline Anders Blom

  • QuantumATK Staff
  • Supreme QuantumATK Wizard
  • *****
  • Posts: 5394
  • Country: dk
  • Reputation: 89
    • View Profile
    • QuantumATK at Synopsys
Re: some questions about ivcurve converge
« Reply #7 on: May 11, 2015, 13:00 »
The electrodes should generally be conductive if you want an I-V curve, but again you should attempt to compute the device as close to the "real" situation as possible. If you are looking for purely intrinsic properties of a material, then you just need to model it as such, using periodic boundary conditions - you don't need a 2-probe set up at all.

Of course, then you can't apply a bias and get an I-V curve, but you can do that (in the coherent, ballistic picture) for any material - you can't apply a finite bias across a perfectly periodic, infinite structure because then your electrodes are at infinite distance and the voltage drop is undefined across any finite segment (the central region in the device simulation).

A device needs to be "finite" in some respect, due to a gate, an interface, etc, so there is scattering such that there is a well-defined voltage drop. But in any case, your "electrodes" should be realistic - if they are made from updoped semiconductors your current will be very small, in the experiment as well as in the simulation.