system cannot be transformed to device configuration

[baizq]

Dear Colleagues,

I am following one tutorial of atk 11.2 : Spin transport in MTJ Fe-MgO-Fe, and encountered with an error as followed:

Error: the total width of the left and right electrodes may not exceed the width of the central region.

when I want to tranform the structure  from bulk configuration back to device configuration after geometry optimization.

Anyone knows what the problem is and how to solve it?

Thanks,
baizq

[baizq]

1. It seems that VNL cannot recognize the electrode length after geometrically optimizing the Fe-MgO-Fe structure as bulk configuration. As a result, it cannot convert the bulk configuration back to device and then exhibit the above error message. Anyone who knows how to deal with the problem?

2. Or other than relaxing the structure as bulk configuration as recommended by the tutorial, can we optimize it as device configuration? If we can, how to set the K-point grid and other main parameters?

3. It is said in the tutorial that geometry optimization as bulk configuration can speed up the computation. Why...?

[Anders Blom]

1. This is the be expected, given the concept of "electrode copy" in the central region. What you need to do is either constrain the "electrode copy" during the relaxation, or relax the system as a two-probe geometry.

2. You set the parameters of the method as usual, when you set up the device calculation (as if it was already relaxed).

3. It's faster because you are not running a two-probe calculation (no electrodes, no open boundary conditions). The results are very similar, but not identical.

However, you can do something different, there is a trick

If you set up the device calculation to be non-selfconsistent, ATK will automatically do the central region optimization as bulk, while imposing the relevant constraints. This is generally the best option.

[baizq]

Hi Anders,

Thanks very much for your reply, especially for the trick provided I am going to make a try.

baizq

[baizq]

Well....I have another question....

Do we only need to add iteration_control_parameters = NonSelfconsistent to the Device Calculator to make the device calculation a non-scf one, without specifying other optimization-related parameter, such as atomic constrants? The programme will automatically add constrain to the electrode copy region, right?

Is the script attached right for geometry relaxation?

Thanks! (*^__________^*)

baizq

[Anders Blom]

No, you still need to call Optimize function. However you don't need to set constraints.

[baizq]

Ok...I have modified it..Thanks..
So I do not have to add iteration_control_parameters = NonSelfconsistent to electrode calculator, right...?

Thanks!

[Anders Blom]

No, it's best to leave them self-consistent.

[nori]

Error: the total width of the left and right electrodes may not exceed the width of the central region.

Do you use VNL 11.2.0?
If so, I recommend that you use VNL 11.2.2 because Device From Bulk feature in Builder is improved.
I've confirmed that optimized bulk Fe/MgO/Fe structure can be converted to 2probe normally with VNL 11.2.2;)
 

[baizq]

Wow...That's pretty fine. Yes, I am using VNL 11.2.0 now and I will upgrade it to ver. 11.2.2.
Thanks, nori. 

[Anders Blom]

Actually, this algorithm will be further improved in 11.2.3, which will be released soon.