1021
General Questions and Answers / Re: The problem of device NEGF DFT convergence?
« on: June 28, 2017, 11:19 »
The structure looks OK. Which version of ATK are you using? In ATK-2017.b1, there is an option of using a more robust and faster Poisson solver: [Parallel] Conjugate gradient.
In the Device Algorithm settings, you may also consider using Neutral Atom instead of Equivalent Bulk for initial density.
I would not change the Contour Integral Settings for the moment, unless you are sure that there are deeply-lying states of Ni. In this case, you might need to increase the value of integral_lower_bound. This can be checked by band structure calculation for bulk Ni.
In the Device Algorithm settings, you may also consider using Neutral Atom instead of Equivalent Bulk for initial density.
I would not change the Contour Integral Settings for the moment, unless you are sure that there are deeply-lying states of Ni. In this case, you might need to increase the value of integral_lower_bound. This can be checked by band structure calculation for bulk Ni.