Transition State Search for Spin polarization Calculations

[zongtang]

Hi,

I am having an convergence issue for the transition state search for the spin polarization calculations.
For example, the reactant and product have different spins; along the pathway, there exists a point of where the spin has to change.

I could get the forces of most images less than 0.05 by specifying spin using an initial guess of MAGMOM, however, I could not get all converged.

One example is like this if I do nebef.pl

0 0.000000 -1765.296900 0.000000
1 0.000003 -1765.281100 0.015800
2 0.000006 -1765.199700 0.097200
3 0.000003 -1764.502300 0.794600
4 0.000004 -1764.269300 1.027600
5 0.633889 -1763.422700 1.874200
6 0.000000 -1764.065000 1.231900

most likely the fifth image was optimized to a wrong spin state.
is there a way to get it fixed ?

Thanks,

Best,

Zongtang

[graeme]

This is a tricky problem. Using MAGMOM is a good strategy. You can also do a single-point calculation for the problematic image, aiming to get the right electronic structure. Then, if you load that WAVECAR and CHGCAR in the NEB calculation, it will be more likely to be on the correct energy surface.

[zongtang]

Thanks a lot. Will give it a try to see how it goes.

[zongtang]

Yes, it worked for some calculations. Sometimes a different optimizer is also helpful for some challenging jobs.

By the way, I found out only one WAVECAR can be read in a TS calculation. Is there a way to read multiple WAVECARs for multiple images ?

Thanks a lot.

Zongtang

[graeme]

If you are doing an NEB calculation, each image should read the wavecar (if present) in its own directory.

[zongtang]

Hi Graeme,

Thanks, yes, multiple WAVECARs can be read if present in the folders for a NEB calculation; I didn't check the OUTCARs carefully.

Best,

Zongtang