QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: rose on June 10, 2015, 18:19

Title: Convergence Problem in IV for 1V bias
Post by: rose on June 10, 2015, 18:19

Hi,
I have structure as attached and I check all the techniques that mentioned in the forum and tutorials to get convergence but i did not success. For short I check following options: increasing  k points in x y and z, mesh cut off, electron temperature, history steps,circle points, Integral lower boundary and reduce damping factor. I use both DZP, SZP, LSDA, sGGA but i not get any divergence while bias is 1V. I got divergence with zero and 0.5V bias.
any idea how can i fix it?
thanks
rose

Title: Re: Convergence Problem in IV for 1V bias
Post by: Anders Blom on June 10, 2015, 23:39

The system seems to be too short in the Z direction. The electrode should probably be at least 8 Å, and that will also make the central region longer, which helps convergence.

Title: Re: Convergence Problem in IV for 1V bias
Post by: rose on June 11, 2015, 12:47

Hi
Thanks. I check it with 15unitcell in z direction and larger supercell as electrodes, but still i have problem in bias voltage 1V.
Should I do it with larger channel?
yours
Rose

Title: Re: Convergence Problem in IV for 1V bias
Post by: Anders Blom on June 11, 2015, 14:17

Check your points, 2x2 is wrong, one direction is not periodic. 1x9 is more correct.

Title: Re: Convergence Problem in IV for 1V bias
Post by: rose on June 21, 2015, 18:01

Hi
thanks.
I used 1*10*100 points for k sampling but it did not converged for 1V
any other ideas?
rose

Title: Re: Convergence Problem in IV for 1V bias
Post by: Umberto Martinez on June 22, 2015, 09:06

- now that you are using the correct k-point sampling try to use the default parameter for DoubleContourIntegralParameters and IterationControlParameters.
- A usual, it is a good idea to converge your system at zero bias, and converge your finite bias calcualtions starting from this state and increasing bias by small steps.