The convergence problem for transmission coefficient

[tigeryzz]

Hi，everyone，
i calculated transmissions for some 2D systems, like films. when i analyze the transmission coefficients, i test the brillouinZoneIntegrationParameters.
i use (k,1) to see if the coefficients can be converged. but the results are sad. It cannot converge even i use the kpoints of 1001*1. Any of you suggestions may beneficial to my work. Thank you~

the K points i used and the transmission coefficients for Spin.up and Spin.down are shown as follow:
101*1
1.97879210473e-009
2.08923627732e-009
201*1
0.00126760183962
4.66436520168e-005
301*1
2.06801372358e-009
2.13540912195e-009
401*1
1.97401526529e-008
1.24798756185e-008
501*1
0.000508560424445
1.87149713437e-005
601*1
6.24585146946e-009
5.30735471099e-009
701*1
7.91718763235e-007
2.2652668481e-008
801*1
0.000318090916723
1.17079069242e-005
901*1
1.75635227975e-008
8.55466170945e-009
1001*1
8.67660115626e-008
3.22559497865e-008
1101*1
0.000231421576928
8.52033531058e-006

[Anders Blom]

I assume these numbers are T(Ef) at zero bias? Do you expect the system to be conducting? Seems to me it's isolating, and you're trying to get an accurate estimate of ... zero

How does the whole transmission spectrum look around the Fermi energy? Is the system converged properly?

[tigeryzz]

er, the system is used to perform the magnetoresistance, so it is isolating.
but the transmission coefficients at fermi level under zero bias is important, like the case of Fe-Mgo-Fe.

[Anders Blom]

Well, only for one spin channel, right?

I think it's hard to draw any conclusions from single numbers. One had better look at T(E) in a range around the Fermi level, and also at T(Ef,k) as a function of k, to get a better understanding. Are there many narrow peaks (=need really many k) or just a flat but very low transmission (hard numerical problem to get accuracy).

[tigeryzz]

well, i'll try to know the T(E) as a function of k~~
thank you for suggestions.