Siesta

Problems about using SISL to generate the device .nc file

Asked by Zifeng Wang

Dear Developers:

 I used the sisl Python module to generate a electrode .nc file and a two terminal device .nc file for a general graphene nanoribbon. The Hamiltonian is filled with empirical parameters, say, 0.0 for on-site energy and -2.7 for hopping energy between the nearest C atoms.

When I tried to calculate its transport properties under a certain bias, say 0.1 eV, I get the warning:

*** WARNING: transiesta and tbtrans bias not equivalent!
*** WARNING: Be sure to use an interpolation scheme!

Is that meaning the .nc files I have created only worked for non-bias calculation? If so, how should I get the corresponding bias-applied .nc file from my original files?

Here is my TBTrans input file:

#general information

SystemLabel 13AGNR

TBT.Voltage 0.1 eV

TBT.Directory ./13AGNR-0.1

#system information

TBT.HS WDEVICE.nc

TBT.T.All true
TBT.T.Bulk true
TBT.DOS.A true
TBT.DOS.A.All true

TBT.k [1 1 10]

%block TBT.ChemPots
  Left
  Right
%endblock TBT.ChemPots

%block TBT.ChemPot.Left
  mu V/2
%endblock TBT.ChemPot.Left

%block TBT.ChemPot.Right
  mu -V/2
%endblock TBT.ChemPot.Right

%block TBT.Elecs
  Left
  Right
%endblock TBT.Elecs

%block TBT.Elec.Left
  HS WELEC.nc
  chemical-potential Left
  semi-inf-direction -A3
  electrode-position 1
%endblock TBT.Elec.Left

%block TBT.Elec.Right
  HS WELEC.nc
  chemical-potential Right
  semi-inf-direction +A3
  electrode-position end -1
%endblock TBT.Elec.Right

Your reply will be highly appreciated!

Question information

Language:
English Edit question
Status:
Solved
For:
Siesta Edit question
Assignee:
No assignee Edit question
Solved by:
Nick Papior
Solved:
Last query:
Last reply:
Revision history for this message
Nick Papior (nickpapior) said : 		#1

You write:

Is that meaning the .nc files I have created only worked for non-bias calculation? If so, how should I get the corresponding bias-applied .nc file from my original files?

If the .nc files also *worked* for biased calculation we would never need NEGF ;)
To a first approximation the biased Hamiltonian depends on the bias through the Poisson equation, and from your description you haven't done this?

TBtrans allows one to use the non-bias calculation (in this case your TB) and then set boundaries as though there was a bias, but this does not mean that it *just works*.
To get a slightly better agreement you can specify:

Ef-fraction 0.5

in each of your electrodes to have some kind of potential drop from your electrodes. (undocumented feature)

Revision history for this message
Nick Papior (nickpapior) said : 		#2

Note that Ef-fraction 1 is equivalent to how GOLLUM does it.

Revision history for this message
Zifeng Wang (whumaple) said : 		#3

Yeah, but I am still confused how to perform a biased calculation.

I set the TBT.Voltage a non-zero value and different chemical-potential for two electrodes but get a warning which I do not know its actual meaning:

*** WARNING: transiesta and tbtrans bias not equivalent!

And you say:

To a first approximation the biased Hamiltonian depends on the bias through the Poisson equation, and from your description you haven't done this?

What fdf-flags should I use to add a a process for solving Poisson Equation?

Revision history for this message
Zifeng Wang (whumaple) said : 		#4

can I get some example input files or something ?

Revision history for this message
Best Nick Papior (nickpapior) said : 		#5

The warning will always show if it is a tight-binding calculation.

It is merely because transiesta and tbtrans are tightly integrated.

You can't process the Poisson equation in tbtrans. You have to empirically do this or use some other software.

PS. Doing biased tight-binding is not easy since it typically requires self-consistent tight-binding (this is currently not implemented in sisl or tbtrans)

Revision history for this message
Zifeng Wang (whumaple) said : 		#6

Thanks Nick Papior, that solved my question.