nupdown

Message

cpopa · #1 Post by **cpopa** » Thu Jun 03, 2010 9:35 am

Dear all,

I'm running calculations using nupdown, to see how is my system in a specific magnetic configuration. Then I rerun, using the previous CONTCAR as POSCAR. In INCAR the only difference is that I comment out the NUPDOWN flag.
Usually, releasing the spin, I obtain the same magnetic localization but higher energy (few meV - 0.2 eV)! This is the problem. Why? I thought that taking away the spin constrain should lead to lower energy. Which is the correct energy then?
I also need dos, and correct is described without nupdown. What is the solution?

Thank you for your comments
cristina

forsdan · #2 Post by **forsdan** » Thu Jun 03, 2010 5:10 pm

Can you please elaborate on what you are investigating (molecules or extended systems)? Without more information I spontaneously only think about two things at the moment:

1. If you perform relaxations in extended systems where the volume and shape changes you have to consider that the change of the basis set. Since when you restart VASP adopts a new spherical cutoff and the energy can change discontinuously. Please have a look at

http://cms.mpi.univie.ac.at/vasp/vasp/V ... tress.html

2. Note that the electronic relaxation can end up in a local minimum corresponding to a magnetic configuration with higher energy, provided that there are several possible minima. Convergence to the global minimum is never guaranteed. It depends on your initial guess. Your constrained solution may be closer to a local minimum than the global one.

As a side note: I would not recommend to use NUPDOWN for extended systems without caution. I would rather try different initial magnetic moments att explore the possible configurations.

Finally, I don't understand your question about the DOS so you will have to clarify it.

Best regards,
/Dan

<span class='smallblacktext'>[ Edited Sat Jun 05 2010, 10:56AM ]</span>

cpopa · #3 Post by **cpopa** » Fri Jun 04, 2010 11:28 am

Thank you, Dan, for your comments.
I investigate O defective ceria surface suing PBE+U approach. This system is well known to be difficult...
1. I don't relax the unit cell, just the atoms inside it.
2. Yes, I'm aware of the local minima and I'm looking for them in my calculations. NUPDOWN is the way to try and see where the spin is localized (on 2 Ce atoms, but which ones?). This is an old debate in literature, and I'm doing these calculations just for using THE hopefully global minimum for further calculations, putting adsorbates on ceria.
It's very awquard to obtain a higher energy when this constrain is taken away. Or is it a restrain for NUPDOWN for calculation of the energy?
I just repeat one calculation and up to now, it seems to go in the same way. I checked all the things which I could think about.
About DOS, the Fermi energy is not correct when using NUPDOWN, it's just for the spin down. Please, see http://cms.mpi.univie.ac.at/vasp-forum/ ... .php?3.709

best wishes,
cristina