optomization of monolayer silicene device

[beauyy]

Dear sir,
     I have built the bulk of monolayer silicene and optimized it according to http://www.quantumwise.com/publications/tutorials/item/512-opening-a-band-gap-in-silicene-and-bilayer-graphene-with-an-electric-field. But the devices built from the optimized bulk is difficult to SCF. What is wrong with my device structure and the device optimization?
The attached file is the script for device optimization.

[beauyy]

Can anybody answer this question?

[Umberto Martinez]

you should first try to converge at zero bias and only afterwards increase bias by restarting from this state.
This has been discussed various time in the forum, you can search for similar topics.

[beauyy]

Thanks for your reply.  I find that the structure of monolayer silicene can easily to achieve SCF under lower bias,  but the monolayer silicene is still difficult to achieve SCF under 1.6 V when I use the optimized structure under 1.4 V. What can I do for this problem?

[Umberto Martinez]

Does a full SCF state converge under high bias?
note that the effect of the finite bias on the geometry optimization is most probably very small and you can use the zero bias geometry for your finite bias calculations.
Since you managed to converge the forces at 1.4 V you can see if the geometry is very much different from the zero bias one.
If indeed they are very similar you can safely use the zero bias geometry.

[Anders Blom]

There is very, very rarely any significant geometry changes when you apply the bias, so I would advise against trying to optimize the structure for each bias point.

[beauyy]

The geometric optimization under 1.4 V is easy to achieve full SCF. The geometry is the same to that under zero bias. But when I  do the geometeic optimization  under 1.6 V, it is hard to reach full SCF whatever the original geometry is under zero or 1.4 V. It is a big trouble to me.

[beauyy]

Is the geometry of monolayer silicence unavailable under the higher bias than 1.4 V ?

[Umberto Martinez]

As we say, you can safely use the geometry obtained at zero bias.
In order to achieve convergence you can try to increase your k-point sampling.