Convergence problem again: What to do next after change the parameters ?

[fangyongxinxi]

Dear Sir,
There are lot'f of talk about the convergence in this forum, and I got a lot of help, but I still want to know: what we do next after we change the parameters ? Hope to get some guide on how to adjust the model.
Generally, we first do the scf, if it did not work, shown " over step",  we changed the parameters, such as electron T, K-point, and Cut-off energy. Many times, it worked, scf  accomplish. But sometimes, the scf fail, then what we do ? put the script to wastebin, or some other effort we can do ?

Take the attachments for example, this model stimulate the STM process, the central part is water layer. first, scf over_step, change the parameter, over_step again. So, there must be some problem on my model, how to deal with this situation?

Thank you.

[Anders Blom]

You are indeed correct, convergence is a very tricky point, and unfortunately there is no universal recipe.

An often overlooked parameter to obtain good convergence and quality results is the electrode depth (in the Z direction). Your system is a typical example; two Ag layers is way too few, you will need at least 4.

For more information on this point, see the important mini-tutorial http://quantumwise.com/publications/tutorials/mini-tutorials/99. Note that this tutorial, as most, refers to ATK 10.8, which in addition to new features also has improved convergence thanks to the more correct two-probe model implemented there. And, the upcoming version ATK 11.2 will contain even more advanced techniques to converge tricky systems.