# [DFTB-Plus-User] DFTB+NEGF: Net accumulated charge and supercell calculations

Gabriele Penazzi gabriele.penazzi at bccms.uni-bremen.de
Wed Mar 20 10:44:52 CET 2013

```On 03/19/2013 07:37 PM, Yuranan Hanlumyuang wrote:
> Hi,
>    I have a few more questions about nanowire calculations. I would
> like to determine the net accumulated charge
> on a nanowire under some applied gate bias.
>
>    (1) I obtained the accumulated charge by summing the net atomic
> charges reported in the file 'detailed.out'
> With zero bias the accumulated charge is however not zero.   In a
> normal SCC calculation,  the accumulated
> charge is in the order of 10^-8 - 10 ^-7.  In the GreenFunction case,
> I got the net charge of about  -0.001. Is this number
> considered insignificant? Is this non-zero charge a result of my
> choice of parameters in the Eigensolver (GreenFunction)?

Hi Yuranan,

generally speaking in the NEGF calculation the number of particles in
the system is not a constant and there could a small accumulated charge
at equilibrium due to some numerical error. This introduces a small
artificious scattering barrier at the device-contact interface. I would
say that the amount of precision you need depends on what you're doing.
For example, if I run an homogenous 1d system like a nanowire, I always
check that the transmission is step-like at equilibrium: a charge
accumulation in the device region will in fact cause a small fictitious
scattering.

If you want to reach a good precision on the charge in the system, it is
important that you're Fermi level at the electrodes is defined
precisely, so that a small charge transfer is not induced by numerical
errors. If you run a semiconducting wire at 0 temperature, you can put
the Fermi level in the middle of the gap by hands. If your system is
metallic, use a very fine k-point grid when you calculate the electrode
and check carefully the convergence of the Fermi level.

A finer Poisson grid can also help. If you're at equilibrium the
eigensolver usually works good, you can try to increase the integration
points but for my experience, more likely it's one of the aforementioned
points

>
>    (2) A cluster geometry is used in the above calculations.  I
> haven't had any success with the supercell geometry. I kept getting
> message like
>    scfsum: 6 NaN NaN NaN

Are you sure that your lattice vectors are correct and prevent periodic
copies from overlapping? If yes, can you send the dftb_pin.hsd file?

>
>     These messages only appear in supercell calculations. Using the
> same input parameters with cluster geometry gives me
> no error messages. Is using cluster geometry sufficient for doing a
> nanowire calculation?

Cluster geometry is perfectly fine for a nanowire calculation

Best
Gabriele

>
>    Thank you. I am gladful for all guidances I have received  from
> this mailing list.
>
> Best,
> Yuranan Hanlumyuang
> University of Houston
>
>
>
>
>
>
>
>
>
>
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at dftb-plus.info
> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user

--
Dr. Gabriele Penazzi
BCCMS - University of Bremen

http://www.bccms.uni-bremen.de/