is this possible?

[fanjiaping]

Dears:     
 I get the I-V curve which contained a strange phenomenon. I doublt that this case can't be occurred. So I want to get authority reply from you. the picture was in additional information. thank you!

[nori]

I think your calculation is something wrong because current shouldn't be positive in negative bias voltage.
First of all, you need to check the transmission spectrum carefully at about -1.8V.
And then it is expected that you can see the large negative transmission coefficients within bias window.

In most of my experience, large negative value of transmission coefficients comes from inaccuracy of KrylovSelfEnergy.
So if my speculation is right, you would be able to solve this issue by using RecursionSelfEnergy or DirectSelfEnergy.

[yongjunwinwin]

What is the difference of KrylovSelfEnergy and the two you mentioned? Generally KrylovSelfEnergy is the default parameter in the device calculation, what is the difference of the result using those algorithm parameters?

[nori]

What is the difference of KrylovSelfEnergy and the two you mentioned?

Briefly some (fast) decaying modes are ignored in KrylovSelfEnergy and in some cases this causes inaccuracy.
Please see the manual and the following literature for more information:

• KrylovSelfEnergy: Phys. Rev. B, 77, 155301, 2008
• RecursionSelfEnerghy: J. Phys. F., 15, 851, 1985
• DirectSelfEnerghy: Phys. Rev. B, 59, 11936, 1999

[fanjiaping]

the problem wasn't solved after I use DirectSelfEnergy instead of KrylovSelfEnergy. So I confused by this case. Could you check the file for me again.thanks!

[Anders Blom]

On a general note, the unit cell is somewhat small in the XY directions, it will not be very expensive to add a bit of vacuum.

However, the main problem with the script is probably the use of Neumann boundary conditions in XY, which seems inappropriate.

It is basically impossible for the current to change sign unless the transmission spectrum is negative, so without seeing T(E) for each bias one cannot really advise further. My suspicion is that T(E) is basically zero at all bias values, but sometimes infinitesimally negative (which is never an issue as long as the real transmission is finite), causing the apparent negative current. However, with such a large value of the current in nA, that doesn't seem to be the case either...

Which ATK is this?

[fanjiaping]

the attachment is the transmission spectrum. Did the size of unit cell is the factor to cause the problem ? How can i choose the periodic boundry?

[Anders Blom]

Well, that certainly explains the I-V curve, doesn't it

So, why negative transmission? I can't say for sure, but definitely using Neumann boundary conditions makes no sense, so you should go back to the Script Generator and keep the default FFT2D, and not change it as you did to MultiGrid with Neumann in X/Y.