geometric Reoptimise Transition state with original KPOINTS after dimer/NEB

Vasp transition state theory tools

Moderator: moderators

Post Reply
sxd375
Posts: 28
Joined: Tue Feb 13, 2018 3:30 pm

geometric Reoptimise Transition state with original KPOINTS after dimer/NEB

Post by sxd375 »

Dr. Graeme

As suggested by you in this forum throughout, during a dimer or NEB run approaching forces lesser than 0.1 ev, I lower the constraints on KPOINTS from original to 1 1 1 which lets the run computationally faster and I get the Transition state with this lowered KPOINTS. However, when I try to reoptimise this TS state with original KPOINTS and IBRION 1, low POTIM like 0.05/0.1, then in some cases the forces donot lower till EDIFFG -0.05(I keep getting forces like 0.1-0.2), and I donot increase the POTIM fearing that it will fall to either the initial or final state.
In such a case, can I just take a single point calculation with the original KPOINTS for the TS state, or do you suggest anything for the geometric reoptimisation ?

Thanks
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: geometric Reoptimise Transition state with original KPOINTS after dimer/NEB

Post by graeme »

No, don't do a single point calculation. If there is a problem converging with more accurate settings, you need to get to the bottom of it.
sxd375
Posts: 28
Joined: Tue Feb 13, 2018 3:30 pm

Re: geometric Reoptimise Transition state with original KPOINTS after dimer/NEB

Post by sxd375 »

OK
As then what do you suggest for re-optimisation if lower POTIM values do not work and higher POTIM takes the strucutre away from the obtained TS state ?

thanks
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: geometric Reoptimise Transition state with original KPOINTS after dimer/NEB

Post by graeme »

If this is a dimer calculation, make sure that you get a negative mode for the initial structure. You can increase DRotMax parameter to allow for more rotations to accurately find the negative mode.

It would be very surprising to me if you really did converge to a saddle point with a low k-point mesh and then were not able to reconverge with a finer mesh. If you believe that is really the case, I would be interested to see the calculations.
Post Reply