Issue with Improved Dimer Method on Transition State at More Negative Potentials (VASPsol++)

[shweta_mehta1]

We are modeling the transition state (TS) for CO₂ hydrogenation on a solid-state catalyst using a cluster with explicit solvent including an H₃O⁺ ion, under different applied potentials via VASPsol++. Our workflow was as follows:

1. Optimized initial and final states at −0.5 V vs SHE.
2. Performed a CI-NEB calculation at −0.5 V vs SHE.
3. Carried out a frequency calculation on the highest energy image, which showed an imaginary frequency corresponding to the reaction coordinate.
4. Used the Improved Dimer Method (IDM), which converged to a TS with one imaginary frequency along the reaction path.

Next, we attempted to take this optimized TS to a more negative potential. Frequency calculations still showed an imaginary frequency along the reaction path, but when we used this as a starting point for IDM, the structure “exploded” during the optimization.

Here is the INCAR we used for IDM:

ISTART = 0
ICHARG = 1
ISMEAR = 0
SIGMA = 0.2
EDIFF = 1E-06 # Tried 1E-08, same result
EDIFFG = -0.05
ALGO = Fast
NSW = 1000
IBRION = 44
POTIM = 0.05 # Tried 0.01, still exploded
ISYM = 0
LREAL = Auto # Tried LREAL = FALSE, still exploded
IVDW = 11
LSOL = .TRUE.
ISOL = 2
C_MOLAR = 1.0
R_ION = 4
EFERMI_ref = -3.57

Any advice or insights would be greatly appreciated. Thank you.

[andreas.singraber]

Hello!

I will move your post to the "From users for users" board because your questions target at a very specialized workflow which also involves third-party software and modifications to VASP. I have no experience in the procedures you are describing, therefore I am not able to contribute here. However, maybe other users could help if you provide them with some more information, in particular the input files you used when running the improved dimer method and observed the structure "exploding".

All the best,
Andreas Singraber

[longfei_guo1]

Hi，

Could you upload the structures? I am calculating the same models of CO2RR.

Best
Longfei

[rwindiks]

Hi,

to exclude that the reported issue due to VASPsol++, you might want to do the following:
1. From the failed calculation with VASPsol++ and the more negative potential extract the number of electrons used (defined by the directive NELECT)
2. Repeat the failed calculation but with setting LSOL = .FALSE. and using NELECT to the value used in the previous step

You might want to freeze all atoms that are not involved in the reaction and rearrangement.

Make sure that in all steps of the Dimer method the energy is converged to a negative value. Otherwise the calculated forces used in the Dimer method are rubbish. Use ALGO = All with a timestep of 0.2 and increase the number of SCF iterations to 300 or even 400 steps. Accurate forces are essential for the Dimer method!

Thanks.

René