Dear Joe, Many thanks for your reply. After replacing all the manually given hoppings by lat.neighbors() method, I still obtain the same result. Also, the system is plotted for the length L of even and odd numbers, which gives corrected tight-binding system. So I think as Adel said in his reply, the result is the expected physical one. After checking some materials, I found it is a known effect in clean 1D quantum wires which is called even-odd oscillation, for example, discussed in PRB 78, 045428 and PRB 41, 8553. Best regards, feng Joseph Weston <joseph.weston08@gmail.com>于2017年9月6日周三 上午1:08写道：

Dear Liu,

...

I have met a strange problem in computing conductance of 1D quantum wire. The setup is as follows: The hopping of the quantum wire is given by the first kind of Bessel function, for example, J0(A). The coupling of the lead is set as 1. As we know in this case the transmission should be proportional to J0(A)^2. However, I get very different results for wire of even and odd sites numbers as shown in the attachment... Is this a bug or do I miss something else?

This is probably not a bug in Kwant. I see in your script that you do not construct your system in a very idiomic manner (you manually add nearest-neighbor hoppings etc). Are you sure you did not make a mistake during system construction? Did you plot the system in the two cases (even/odd) and see that everything is as you expect?

Happy Kwanting,

Joe