QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: Dipankar Saha on October 31, 2015, 12:06
-
biases = [0.000000, 1.000000, 2.000000, 3.00000 4.000000]*Volt
iv_curve = IVCurve(
configuration=device_configuration,
biases=biases,
........
........
........
........
selfconsistent_configurations_filename="ivcurve_selfconsistent_xyz.nc",
)
nlsave('xyz.nc', iv_curve)
nlprint(iv_curve)
___________________________________________
In case my simulation got stopped.....after convergence of the calculation considering 'case: 4_ bias voltage' (i.e. 3 Volt)....; then how can I utilize the "ivcurve_selfconsistent" objects to get the IV curve (let that be up to 3 Volt only, that's fine) ...??
Because, in the output file(xyz.nc)..... the "IV Curve" object (along with the details of the Tr. spectrum @ diff. bias voltages) gets generated.... only when the calculations for all the individual bias voltages are done!!
Regards_
Dipankar
-
See here http://quantumwise.com/publications/tutorials/item/820-ni-silicide-si-interfaces#h7-5-post-analysis-of-finite-bias-calculations
where an example script is also reported
-
Thank you Umberto..... :)
_______
Quite a few points_
1)
Some where it is written that..... "The calculation at 0.1 V will be done by restarting the one at 0.0 V. This method improves quite a lot the convergence of each SCF calculations" ? Why so? To save computational time? ? But, typically what I find that.... as bias voltage increases...time taken for the convergence, also increases.....!!
2)
I did not get, what this line is actually interpreting_
"In this case, since you are calculating the IV curve for an applied bias ranging from -0.5 V to 0.5 V you can in principle evaluate the TS only within -0.25 eV and 0.25 eV "
3)
Sometimes we find that the calculation ends up with an warning...which is something like,
"calculation did not converge.....as the computed multigrid residual is greater than the required accuracy"
However, The transmission object gets still generated....!!!
Do you suggest to discard such a transmission object, T(k, e).... ( or, it's okay as long as the T(k, e) is getting determined....) ??!!!
Regards_
Dipankar
-
1) Not exactly to save computational time, even though it may also be true.
This procedure is simply to get a better convergence of the finite bias state since you will have a well converged density to get started.
You can of course run a finite bias calculation from scratch (neutral atomic densities, for example) and get accurate results.
The suggestion indicated comes from our own experience.
2) this is because when in the IVCurve analysis you specify a bias of +0.5 V, the corresponding electrode voltages will be +0.25 V and -0.25 V for the left and right electrodes, respectively.
To be more precise, an energy window of -0.25 eV and 0.25 eV is really on the edge and you can also increase it a bit more to get a better description of the transmission spectrum.
3) yes, the TS as any other analysis wills still be calculated even if the SCF state did not converge properly. You should always check the log file to be sure your job is converged with no errors.
About this specific warning message, it has been posted many time on this forum.
-
1) Then, how the increase in bias voltage heavily increases the computational time..., memory requirement ??
2) Okay.....you say, it's unnecessary (for the IV calculation) to take a larger energy band (as described in the following )..../ Rather we should stick to something, that is as per the bias window....
biases = [0.000000, ............................. 3.000000]*Volt
.......................
.......................
biases=biases,
energies=numpy.linspace(-5,5,301)*eV,
................
...............
3) I understand this.... :)
Thanks & Regards_
Dipankar
-
1) more energy points to compute
2) I do not understand what you are asking.