Lost electron in device region during transiesta run

I’m trying to run a transiesta simulation with a Ag wire system( 4 layer Ag buffer | 4 layer Ag electrode | 4 layer Ag screening | 4 layer Ag with some defect as device | 4 layer Ag screening | 4 layer Ag electrode ), the initial charge looks weired as

Total charge [Q] : 874.00000
Device [D] : 417.12568
Left [E1] : 151.12545
Left / device [C1] : 0.88087
Right [E2] : 151.56733
Right / device [C2] : 0.86775
Buffer [B] : 151.56649
Other [O] : 0.86642
Excess charge [dQ] : 3.63798E-12

ts-q: D E1 C1 E2 C2 B dQ
ts-q: 372.849 151.125 0.887 151.567 0.891 151.566 -.442E+02
ts-Vha: 0.21024793E+03 eV

During the run, electron in device region goes to zero as:

ts-scf: 45 195316.366504 195316.366503 195316.366503 0.000000 -3.128386 0.000424
ts-q: D E1 C1 E2 C2 B dQ
ts-q: 0.014 151.125 -0.016 151.567 -0.016 151.566 -.419E+03

there must be some major problem, then what is that?
Thank you

In the manual, it will recommand us to extend the eletrode, but I think add buffer region can solve this problem quickly.

Well, actually the problem is the lowest eigen value of Ag is very low ( due to the 4s & 4p semi-core states I guess) at about -100 eV, and the default value generated by the tselect.sh:

%block TS.Contour.C-Left
part circle
from -40. eV + V/2 to -10 kT + V/2
points 25
method g-legendre
opt right

only covers -40 ev and above. Decrease that -40 eV to like -120 eV solves the problem.

Simply adding buffer or extend the electrode won’t help, and if you use “TS.dQ buffer”, even though it converges, the result will be unphysical.

that is useful for me too!

I find that decrease the value from -40eV to -120eV can’t solves my problem.
my system is ( 4 layer Mo buffer | 4 layer Mo electrode |8 layer Mo as device | 4 layer Mo electrode| 4 layer Mo buffer ) just a simple setting for test Mo.

but when i change the system in ( 4 layer Mo electrode | 4 layer Mo buffer|8 layer Mo as device | 4 layer Mo buffer | 4 layer Mo electrode ) the dQ converges

so I don’t know is the second setting is right, perhaps you know the reason why the first setting can’t converges?

Thank you!

Actually I dont have a very clear idea about buffer either