This thread is locked.Only browsing is available.
Top Page > Browsing
OpenMx's Method of Avoiding Local Trap of Spin Configuration
Date: 2015/10/19 19:59
Name: Jong Hyun Jung

Dear OpenMx Developers,

I want to know the OpenMx's algorithm of avoiding local trap of spin configuration.

Because I want to calculate the spin configuration of metal hydride molecules during nudged elastic band method. I found that wrong spin changes total energy of molecule as much as 0.3 eV. I do not know the experimental spin state of the molecules.

As a test, my friend calculated O2 molecule with collinear magnetism. When he used openMx code and used set wrong initial configuration, during the convergence, openMx changes total spin abruptly ( |> to |> ) several times to avoid trapped in local minimums. Then it found correct ground state. I want to know the algorithm of global minimum search to determine how far i can rely on this algorithm. Then I would know whether I can use openMx to find global spin configuration of molecules. Because another DFT codes vasp does not seem to provide such feature, and vasp converges to nearby local minimum of spin configuration.

Thank you very much.
e
Page: [1]

Re: OpenMx's Method of Avoiding Local Trap of Spin Configuration ( No.1 )
Date: 2016/02/17 16:27
Name: Jong Hyun Jung

My own answer to my question:

Since openMX uses localized orbital basis, and vasp uses plane wave basis, changing a single electron state from spin up to down is more easier in openMX than in vasp. The number of basis concerned in this change is much small in openMX. So I think although openMX does not such global minimum of spin state, local minimum trap can be more easily avoided by using an algorithm avoiding a local minimum, though I am not sure whether this algorithm is implemented or not.
e

Page: [1]