Hello Nick
Thank you, very much, for your insightful comments.
Indeed, we have started with a 0 V calculation in all our tests, and the
reported issue appears for V= 0 V. It also manifested when, after
performing a successful 0 V calculation on the same heterostructure
prior being relaxed in the slab geometry ( so only preoptimized cells in
the electrodes and barrier, we call it "ideal") and increased the bias
to 0.001 V or 0.5 V, independent calculations.
Also, the points:
- " you started a biased calculation (V) from a previous TranSiesta
calculation (V'), ..."
- "- your mixing weight is too high"
-"- your convergence parameters too loose during the Siesta SCF
can be discarded.
We shall check the remaining of your comments and further report to the
forum.
Kind regards from Bucharest,
Neculai
***
Neculai Plugaru
National Institute for R in Microtechnologies (IMT-Bucharest)
Simulation, Modelling and Computer-Aided Design - Laboratory L5
126A, Erou Iancu Nicolae Street, 077190, Ilfov
Bucharest, ROMANIA
E-mail: neculai.plug...@imt.ro
Tel: +4021 269 0777
https://www.imt.ro
***
On 2021-11-15 11:37, Nick Papior wrote:
Please use 4.1.5, it has checks for these kinds of things.
Probably what you are seeing is that the charge is leaking out of the
device region during your SCF. This is an indication of one or a
combination of the following:
- your initial Hamiltonian is far from the open-boundary one. I.e. your
electrode regions are too close to the scattering in the device
- your mixing weight is too high
- your convergence parameters too loose during the Siesta SCF
- your integration variables in TranSiesta could be better (more
points)
- you started a biased calculation (V) from a previous TranSiesta
calculation (V'), and the difference V - V' is too large for the system
to converge properly, try a smaller dV
There may be other things as well. Also, I hope you are doing a 0 V
calculation to start with.
Den fre. 12. nov. 2021 kl. 22.08 skrev Neculai PLUGARU
:
Dear SIESTA community
I attempt to perform conductance calculations on a heterostructure
with
two interfaces, (slab geometry) which was previously relaxed and then
the SCF calculation produced correct results.
I use SIESTA v4.1-b4, Compiler version: ifort (IFORT) 18.0.5 20180823,
PARALLEL version.
However, although the SIESTA cycle in the transport calculation
converges and the Harris, KS and Free energies are negative, TS starts
from (huge) positive energy values, then the calculation converges
according to the DM tolerance criterion, the energies are still
positive, the forces on atoms are huge and the populations on orbitals
are wrong. The leads are cut from the electrodes, so that the
scattering
region comprises the interfaces and the leads still are about one
nanometer long each.
I would greatly appreciate your comments to solve this issue.
Thank you,
Neculai
***
N. Plugaru
National Institute for R in Microtechnologies (IMT-Bucharest)
Simulation, Modelling and Computer-Aided Design - Laboratory L5
126A, Erou Iancu Nicolae Street, 077190, Ilfov
Bucharest, ROMANIA
E-mail: neculai.plug...@imt.ro
Tel: +4021 269 0777
https://www.imt.ro
***
--
SIESTA is supported by the Spanish Research Agency (AEI) and by the
European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
--
Kind regards Nick
--
SIESTA is supported by the Spanish Research Agency (AEI) and by the European
H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
