Two imaginary vib modes for TS

[kra268]

Hello Graeme and Andri,

Hope you are doing well!

I am trying to find the vib modes for a possible TS (CH3CHCOOH to CH3CCOOH) of dehydrogenation. The surface is a Ni[111] 5 layers and bottom two constrained. I obtained the saddle point from NEB followed by dimer calculations. The forces on the dimer was converged to EDIFFG = -0.001. I initially started the dynamical matrix calculation with displacements only for the molecule on the surface (see attached files). I was testing for convergence w.r.t finite diff displacements of 0.005 and 0.0025. But I stopped because in both cases, there are two imaginary frequencies. I see this in other saddle points pertaining to other elementary steps too. Either I am consistently making a mistake or missing something. More info: ISYM=0; used dymseld.pl with initial and saddle point POSCAR to set up DISPLACECAR. POSCAR contains the saddle point. Kindly let me know if there's a way to share these files. Your opinions and suggestions will be highly important. Thank you.

Regards,
Keerthan.

[kra268]

My apologies, I forgot to post the freq.dat: For displacement of 0.005:
857.231119 cm^{-1} ... 1
171.311052 cm^{-1} ... 1
85.785600 cm^{-1} ... 0
103.662689 cm^{-1} ... 0
136.663746 cm^{-1} ... 0
209.776436 cm^{-1} ... 0
238.833204 cm^{-1} ... 0
284.304291 cm^{-1} ... 0
318.386608 cm^{-1} ... 0
360.947309 cm^{-1} ... 0
434.178180 cm^{-1} ... 0
551.598022 cm^{-1} ... 0
563.117506 cm^{-1} ... 0
664.155613 cm^{-1} ... 0
695.014763 cm^{-1} ... 0
854.494415 cm^{-1} ... 0
964.265884 cm^{-1} ... 0
1048.986775 cm^{-1} ... 0
1083.675972 cm^{-1} ... 0
1181.307014 cm^{-1} ... 0
1338.368489 cm^{-1} ... 0
1351.249002 cm^{-1} ... 0
1407.138199 cm^{-1} ... 0
1434.281567 cm^{-1} ... 0
1497.564559 cm^{-1} ... 0
1907.991695 cm^{-1} ... 0
2739.599517 cm^{-1} ... 0
2964.967621 cm^{-1} ... 0
3050.784764 cm^{-1} ... 0
3646.404812 cm^{-1} ... 0

[graeme]

I see two possible issues. The first is that the initial electronic structure calculation is not converged; it reached the limit of 60 electronic steps and then moved on. This is normally not a problem in a calculation, but for the dynamical matrix, you need to have every iteration converged. You can increase NELEM to solve this.

Second, make sure that you are using the CENTCAR from the dimer calculation. The initial force is a little high. This could be related to the first point, but it could also be that you used the CONTCAR from the dimer calculation, which is slightly displaced from the center of the dimer, which is the CENTCAR geometry.

[kra268]

Thank you Graeme! I finally got one imaginary freq and I believe it was both those issues you pointed to. Now I only need to converge it w.r.t displacements and include more surface atoms!