UT theoretical chemistry code forum

Hey Dr. Graeme

I have been running dimer method for CO oxidation to CO2 on Pd adatom over ceria 100. However, I am getting significant double negative frequencies on vibration calculation for converged dimer TS structure. Is there anything you would suggest to further refine the dimer run and remove the second -ve frequency not on rxn coordinate ? Technically, the TS structure from dimer run do look like a reasonable transition state structure for this reaction.I already tried using the max negative frequency as the input MODECAR for the next dimer run but still could not remove the second negative frequency which is not on rxn coordinate. I have used 111 kpoints and smaller ENCUT here since I was so close to convergence.

Attached is the dimer run with VASPVIB.MOLF file containing all the vibration frequencies.

Thanks

It's not clear to me how the vibrational modes were calculated. The modes in the VASPVIB.MOLF look a little strange in that they contain many zeros - were these calculated on a subset of coordinates?

Anyway, frequency calculations at saddles do take some care. I would start by reducing the force in your dimer calculation to 0.01 eV/Ang, and then make sure that you have sufficient accuracy in your dynamical matrix calculation (a low ediff and a small finite difference displacement or use the linear response algorithm). Numerical issues could be exacerbated by low frequency 'wag' modes of the CO molecule.

And make sure to use the CENTCAR, rather than the CONTCAR for your frequency calculation. They are very similar, but again, you have to be careful for frequency calculations.

Hey graeme

Thanks for your comments. The vibrational modes are calculated by freezing the surface atoms to save time and just check if the obtained TA state has a single imaginary frequency. I will rerun the dimer calculation as advised and keep you posted.

I will check with the linear response method. I assume you are talking about using IBRION 7 or 8 depending on symmetry.


Thanks