**Charge convergence for non-equilibrium transport** |
- Date: 2020/08/12 02:33
- Name:
**Samuel Dechamps**
<samuel.dechamps@uclouvain.be>
- Dear all,
I'd like to run non-equilibrium Green's function (NEGF) calculations for some specific systems. I did beforehand some preliminary tests for the carbon chain in order to get a proper understanding of the convergence in the non-equilibrium case. I'm focusing my analysis on the Mulliken charges (MC) at the edges of the extended central region. Indeed, a recurrent issue in NEGF calculations is to obtained a flat electrostatic potential at the edges which would correspond to MC being identical to the ones in the lead. However, I'm always obtaining MC that are 5% to 10% off. For example (VDS=.5eV) : 1 C 2.030866540 2.030866540 4.061733080 0.000000000 2 C 2.011709486 2.011709486 4.023418972 0.000000000 3 C 2.121442689 2.121442689 4.242885378 0.000000000 4 C 1.915513226 1.915513226 3.831026451 0.000000000 I've played with many parameters without success. For example : -increase the mixing history, the Pulay start -decrease the mixing weight -set the Band iteration to 0 -switch the Poisson solver to finite differences -extend the central region and the leads However, the MC are not changing (meaning the excess MC would be intrinsic to VDS ?). How can I reduce the excess charges ?
Thanks for your consideration, S. Dechamps
| |