Effect of MAGMOM on energy barrier

[mukeshphy]

Dear VTST users and Prof. Henkelman,

I have been working on Single atom catalysts on a surface. We want to see the energy barrier of decomposition of K2S. So reaction will be

*K2S ---> *KS + *K

where * represent that system (K2S, KS and K) has attached on the surface.

Usually, I relaxed my system with spin-polarization, which is a bit more accurate than without spin because of relaxation of higher number of degree of freedom.

On relaxing both the initial and final structures show different magnetic moments of Fe atom.

So, Initial I started with same magmom as I have provided to relaxed structures.

But I am getting some weird energy barrier (attached image with_spin_and_six_images.png)

Also, I repeated the same calculation with eight images (attached image with_spin_and_eight_images.png)

Finally, I have did the energy barrier calculation without spin (attached without spin.png)

Note: all the calculation are conveged with EDIFFG = -0.05



I have read some of previous post, which suggest that we can tune the magnetic moment of images by performing one cycle single self consistent calculation. My questions are:

Q. What magnetic moments should I provide images as initial and final structures have different magnetic moments?

Q. Any suggestion of MAGMOM tag on energy barrier calculation will be highly appreciated.





thanks and regards
Mukesh Singh

[graeme]

Dealing with spin can be challenging, especially if you use DFT+U. You just can not expect that vasp with automatically reach a low-spin state, especially when it starts with a very high spin state by default. Instead, you need to help the calculations as much as possible using parameters such as MAGMOM.

In terms of your specific question, I can look at the electronic structure, but I'm pretty sure that images 1, 3-6 will have different spin states than 0 and 2. There is a similar issue for band 2, and likely for 3.

So then you have to decide if you take a consistent spin state across your band, or more typically assume that you want the lowest energy spin state. We have a few tricks for finding the lowest energy spin state, but I would start with looking at each image individually, using magmom to help get the spin state that you want, and then if you still have problems, doing single-point calculations on any problematic image and restarting the NEB using that charge state.

[mdhayesNU]

In my experience, magnetic systems can easily fall into different spin states across a NEB calculation. The first step is to ensure that NUPDOWN is set in your INCAR to the MAG value of your reactant and product states from the OSZICAR (these should be consistent). This will prevent the spin on individual electrons from changing during an optimization step.

Next, Occupation Matrix Control (OMC) can be used to set the correct occupation of orbitals in your system if they change in an unphysical way (see Allen J.P. and Watson G.W., Occupation matrix control of d- and f-electron localisations using DFT + U, Physical Chemistry Chemical Physics 16, 21016 (2014) https://github.com/WatsonGroupTCD/Occup ... ol-in-VASP). You can look at how the moments on each atom change easily using ASE's gui function & reading an OUTCAR from the calculation.

Its expected the magnetic moments will change somewhat on the involved atoms along the NEB path, but the total MAG should be consistent and individual moments should change in a way that is physically possible.