I attempted calculations for hydrogen adsorption on transition metal-doped g-C3N4.
I want to know if it is possible to perform geometry optimization without enabling spin polarization, and then enable spin polarization for a second calculation with the optimized structure. Will the results obtained this way be correct?
Please be aware that DFT turns increasingly bad for any multiplicity going beyond a doublet. As a single-reference method, it will only include one of the microstates of a spin multiplet and therefore any more advanced spin properties will not be accurately accounted for.
So, if you ask if adding spin polarization in a second optimization is still "correct", the answer would be "yes, but only within the general spin accuracy limitation of DFT".
A priori I would say that you would need to optimise again the structure. While you might get something better than the initial configuration you might need to optimise again your structure. Indeed, this comes from the fact that the forces are evaluated starting from the electronic ground state.
The presence of spin-polarisation changes that ground state and the occupation of the atomic orbitals thus producing forces different from the un-polarised case. (This would be a problem with any optimisation step, irregardless if you are using a DFT or more sophisticated methods.)
If your calculation is particularly heavy, you might try with the unpolarised case and evaluate the forces again when you perform the polarised calculation. Clearly, the underlying assumption is that the relaxed structure you have obtained from the un-polarised case is close enough to the polarised case and that they belong somehow to similar structures, i.e., to the same energy basin.
Regards,
Roberto
But when I dope the transition metals that I set the spin-polarized to do the geometry optimization.
If I use the results(TM-g-C3N4) I got from the condition of spin-polarized to calculate hydrogen adsorption without spin-polarized.
(First geometry optimization )
And then the results I got from first optimization to do the geometry optimization again with spin-polarized.
Is it available?
I also want to ask if the original structure with spin-polarized that the following work all need to use spin-polarized?
Anything is possible, but clearly the point is how you interpret the results. Your structure was optimised with a spin polarised calculation. This means the forces have been calculated starting from spin up and down densities. This density is in general different from the one obtained without polarisation.
As I wrote, you might need to perform another optimisation with spin polarised.
Another potential issue is that if you expect to restart from each calculation using the restart mode of QE. In this case you are forced to maintain spin polarisation as the number of k points and densities will not agree and the restarting procedure will fail.
The package that I use is CASTEP.
I will do the optimization with spin-polarized again.
Will the results I get from this step(un-spin → spin) compare with the calculation in the condition of spin-polarized initially be similar approximately?
Well, that's the point of numeric optimization: you can end up in the global or a local minimum. So if you're lucky, you can end in the global minimum with both or you can end up in two local maxima.
Dear Alvin Chen ,
It depends mostly on how much the spin up and down densities differ from the unpolarised calculation. If you have a system that is strongly magnetic, that has a large spin imbalance, you might get that the two calculations are leading to different minima.
A check is provided by always evaluating the forces.
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Atmosphere