The computation failed to converge when the negative bias was applied

[iamddba]

Hi. everyone!
Hi, Blom! I have also tried to search the forum for the topics similar to the following problem but there are no findings. Sorry! 

I am calculating a two-probe system using 2008.10 version. The I-V characteristic under the positive bias has been computed successfully and effortlessly. However, the computation is very difficult to converge when the negative bias is applied on the system. Then I changed the electron temperature from 300K to 500K, and the problem was still existing. Can anyone tell me what is the reason and how can I do to solve the problem? Thanks in advance!

[Anders Blom]

k-point sampling is often crucial for convergence, to get an accurate Fermi level. Same for the number of layers in the electrode, which often is chosen too small. Otherwise we have improved convergence a lot in 10.8, so if possible you really should try it.

[iamddba]

Thanks a lot, Blom!
I have increased the K point sampling and the computation looks better in approaching the convergence.

[kaypu]

for my system, when i use k points 3*3*100 for convergence it is successful from -2.0V to 2.0v. but when i change the K points to 5*5*100 others are the same  ,it converges well in positive bias voltage but in negative, it doesn't converge at all.  http://quantumwise.com/forum/index.php?topic=888.0
what should i do?