QuantumATK Forum
QuantumATK => General Questions and Answers => Topic started by: deepwave on August 5, 2013, 02:06
-
Hi,professor:
we always calculate the charge transfer by ATK , in my paper, i calculate the scattering region charge transfer with the increase of bias voltage,but the reviewer refuse to accept my paper for the reason that "ATK uses the NEGF-DFT framework originally developed in McGill University, which assumes always keeping the charge neutrality of the whole central scattering region, at even non-equilibrium, i.e. there is no charge transfer at the boundary between lead and central region and the Hartree potential at the boundary should be continuous." which means ATK can not calculate the charge transfer and without any meaning?
thank you!
-
Note the following:
We use the NEGF-DFT approach described in Brandbyge et. al. PRB 65, 165401, which is formally equivalent to PRB 63, 245407. This is now a well established methodology cited by more than 1800 authors. In this method the central region is modelled as an open system where the electrodes are charge reservoirs and charge can move in and out freely.
For metallic electrodes, the electric field will be zero in the electrode region of the central region, thus from gauss theorem the total charge inside the central region must be zero. Thus, for a finite bias calculation, the charge will not change with the bias but stay constant if you have included a sufficient part of the electrodes in the central region.
If you use an electro-static gate potential, i.e. include metallic regions, then you can have a charge which depend on the gate, source and drain voltage. The total charge in the central region will then be the negative of the gate charge, thus, this is a way to determine the charge on the gate electrode.
-
Note the following:
We use the NEGF-DFT approach described in Brandbyge et. al. PRB 65, 165401, which is formally equivalent to PRB 63, 245407. This is now a well established methodology cited by more than 1800 authors. In this method the central region is modelled as an open system where the electrodes are charge reservoirs and charge can move in and out freely.
For metallic electrodes, the electric field will be zero in the electrode region of the central region, thus from gauss theorem the total charge inside the central region must be zero. Thus, for a finite bias calculation, the charge will not change with the bias but stay constant if you have included a sufficient part of the electrodes in the central region.
If you use an electro-static gate potential, i.e. include metallic regions, then you can have a charge which depend on the gate, source and drain voltage. The total charge in the central region will then be the negative of the gate charge, thus, this is a way to determine the charge on the gate electrode.
Hi there.
I've just accidentally found this conversation. I was wondering if there is no charge transfer between the central region and electrodes, how can we achieve the I-V curve of the system by applying DFT ? Here I found the example from Quantumwise. http://www.quantumwise.com/documents/tutorials/latest/ATKTutorialDevice/index.html/chap.iv.html
Thanks.
Jenny
-
The charge is transferred - but also leaves again, into the right electrode, and that's your current. What is meant above by "no charge transfer" is that there are no additional charges that localize in the central region.
-
The charge is transferred - but also leaves again, into the right electrode, and that's your current. What is meant above by "no charge transfer" is that there are no additional charges that localize in the central region.
Thank you Dr. Anders Blom.
But I do have another question. i did a calculation of the model as attached under zero bias. In the Mulliken Population analysis, I found sum of the central part atoms was 727.868. However the total out shell electrons should be 160*4+11*8=728. I used DFT with GGA_PBE and applied DZP for both carbon and copper atoms in the calculation. And the script for the result is Why the result didn't match the theoretical one, since there is no charge transfer between central part and electrode?
Best,
Jingyin
-
Because you are working with a numerical model, which has some small inaccuracies (approximations). You can reduce these approximations by having a longer electrode, a longer central region, more k-points (for the electrode), and above all probably more points on the complex and real parts of the contour integration.
In general, as long as the difference is just a fraction of an electron you are quite fine.