QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Yueyang on July 24, 2016, 09:49

Title: The calculation did not converge when the scattering region is long
Post by: Yueyang on July 24, 2016, 09:49

Hi,
I optimized a unit cell of graphene nanoribbon, and then repeat it to different lengths to calculate IV curves with the DFTB method.
However, I found that when the scattering region is long, 15 unit cells for example, the calculation is unable to converge to the requested tolerance (e-05) even when the maximum steps is set to be 1000.
But the problem is absent when the scattering region is short, 7 unit cells for example.
Is this normal?
Does that mean we can only study short structures with the software?

Thank you

Title: Re: The calculation did not converge when the scattering region is long
Post by: Jess Wellendorff on July 25, 2016, 09:54

No, that should not be the case. ATK works for large structures also. However, SCF convergence can sometimes become tricky for very large structures, especially at large bias. Can you attach the script and logging output? Thanks.

Title: Re: The calculation did not converge when the scattering region is long
Post by: Yueyang on July 26, 2016, 05:07

Here are two of the scripts that can't converge. The output files are too large to upload.

By the way, is it useful to tune the "Damping factor" or the "History steps" to get faster convergence?
Please give me more advices.
Thank you very much!

Title: Re: The calculation did not converge when the scattering region is long
Post by: Jess Wellendorff on July 26, 2016, 12:53

Perhaps it is the spin-polarization that give you trouble. I suggest non-polarized calculations. The damping factor is already quite low, but you could try to decrease the number of history steps to e.g. 10.

Title: Re: The calculation did not converge when the scattering region is long
Post by: Yueyang on July 27, 2016, 14:42

OK, I will try it.
Thank you!