QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: deepwave on September 17, 2012, 14:26

Title: how to achieve self-consistent
Post by: deepwave on September 17, 2012, 14:26

it can be well self-consistent below 0.4 voltage, however it cann't be self-consistent  above 0.6 voltage in my system(tolerance=0.00008). i have change some parameters, such as:grid_mesh_cutoff, circle_points=50,real_axis_infinitesimal=0.0005*Hartree, integral_lower_bound=5.0*Hartree, Maybe it is wrong.so i need you help me!. thank you!
i provide the export file.log under 1.2 voltage,if you want the input file, i can provide by E-mail.

Title: Re: how to achieve self-consistent
Post by: nori on September 17, 2012, 15:07

Is it sure that your electrodes are 1-dimensional?
If not, you should increase k-point sampling both for SCF and physical quantities.

In addition, I strongly recommend for you to use the previous SCF result as an initial state for SCF.
(For instance 0.6V SCF result for 0.8V SCF calculation)

Title: Re: how to achieve self-consistent
Post by: deepwave on September 17, 2012, 15:47

i am sure that my electrodes are 1-dimensional.I have use the previous SCF result as an initial state for SCF(from 0,1.5,0.1).but it can't SCF when the voltage above 0.6.if you need the input file,i can post you by e-mail.

Title: Re: how to achieve self-consistent
Post by: nori on September 17, 2012, 16:50

So is it right that your system is 1-dimensional, spin-polarized, and has semi-conducting electrodes?
I can give you further advice although I'm not sure it works well...

First, you should add surface layers more in central region.
In general, 1-dimensional electrodes need more screening layers than 3d.
And it's also true for semi-conducting electrodes.

Second, the cause may be bound states in bias windows.
If so, you would improve convergence using SingleContour instead of DoubleContour.

Title: Re: how to achieve self-consistent
Post by: deepwave on September 18, 2012, 01:54

thank you for you answer!but This is a system work can't change the structure, and other cases it can be well self-consistent in the same parameters."SingleContour instead of DoubleContour"is meaning using the DoubleDZetaPolarized. but my computer memory is not enough when i using the  DoubleDZetaPolarizedinstead ofsingleDZetaPolarized? Are there any other way?

Title: Re: how to achieve self-consistent
Post by: mads.engelund on September 18, 2012, 13:36

No, nori is talking about Changing the Contour Integral parameters (under 'New Calculator/Calculator Settings'), not the Basis set.
Change the NEQ contour. We have changed the wording, 'DoubleContour' has become 'Weigthed' and 'SingleContour' has become 'Left' and 'Right'.
So nori's suggestion is to use either 'Left' or 'Right'.

Title: Re: how to achieve self-consistent
Post by: deepwave on September 19, 2012, 14:57

thanks for your explain! i have follow your advise,but it still can't SCF. What else can I do?

Title: Re: how to achieve self-consistent
Post by: Anders Blom on September 19, 2012, 15:29

We can't help unless we can see the system, complete script.

Title: Re: how to achieve self-consistent
Post by: Anders Blom on September 19, 2012, 15:31

General advise: don't modify the contour settings if it converges for lower bias. Use the latest version of ATK, your 11.2 is unsupported and old.