optimization

[Aadhityan]

How to get global minimum structure for device?

[zh]

"global minimum structure for device"
The global minimum search for a device structure is quite challenging because of two reasons: 1) the efficiency of optimization algorithm; 2) the computational cost if you use the DFT method for the local optimization of a structure.

At the present stage, no global minimum optimization method is implemented in the ATK for device.  Depending on the device problem you want to study, you may construct different device models according to your physics intuition  and then do the local optimization for each device configuration to find out the best one. This may be a feasible way.

[Jess Wellendorff]

We do have a tutorial on advanced device optimization: http://docs.quantumwise.com/tutorials/device_relaxation.html. The so-called 1DMIN method explicitly searched for the global minimium-energy device geometry (atom positions, central region length).

[Aadhityan]

Actually I am using two probe device.We use Cobalt for electrodes and some organic molecule as central region. Is it necessary to get global minimum? How to check my system is in global minimum or not?

[Jess Wellendorff]

In this context, global energy minimum means that your structure is at the geometry that minimizes the total energy. There are a few main ways to check this:
- If there are large forces on atoms, the structure needs geometry optimization (force minimization).
- If you chenge the length of the device central region just a little bit, and then relax the atoms, you will see if this leads to a lower total energy.

The 1DMIN method does the above automatically, but is not always needed. In particular, if the device was carefully made from relaxed "building blocks" you probably don't need to consider the central region length, so perhaps only relax forces on atoms.

[Aadhityan]

If i get no negative phonon vibration means, my device is in global minumum?

[Petr Khomyakov]

Getting no negative frequencies do not guarantee that the device structure is the most stable one, i.e., the total energy minimum found corresponds to the global minimum. In general, it only tells you that the structure is not at the saddle point, i.e., far from a phase transition.

Note that negative frequencies may also be an artifact of the calculation, e.g., using improper pseudopotentials can give rise to an inaccurate calculation of forces acting on atoms in the system.       

[Aadhityan]

 Thanks for your replies Petr Khomyakov, Jess Wellendorff and zh.