Hi, when I did a 10 images neb calculation about the H2 dissorciation on the surface. When the NEB converged, I found two local maximum during the pathway. It looks one of them related the break H-H bond and the other related to break H-O(surface atom) bond. Is this reasonable?
Also, when I check the frequencies of these, I found two negative ones for breaking H-H bond maximum. So, what does this mean? and how can find a real transition state?
Thanks
two local maximum founded
Moderator: moderators
It is reasonable. One of the strengths of the NEB method is that it can show you a reaction pathway that you were not expecting. If you use the climbing image, the higher of the two saddles should converge to the saddle point. If you want to find the lower saddle, you can break up your calculation into two NEBs, or start a dimer/lanczos/bfgs(ibrion=1) calculation from your best guess of the lower transition state.
As Andri mentioned in another recent post, finding accurate frequencies with vasp needs to be done carefully - especially for transition states. You must be starting with an image that is well converged to a saddle point (EDIFFG<0.005 eV/Ang). Then you need to use small finite difference displacements, on the order of 0.01 Ang, and use a fine force criterion, (EDIFF=1e-8 is safe).
There are some rare cases in which the NEB can settle on second order saddles, but this is highly unusual, and generally a consequence of having a highly symmetric system. Definitely, start by checking the accuracy of your frequency calculation. You can be confident in the frequencies that you have calculated when they are insensitive to your choice of finite difference displacement.
As Andri mentioned in another recent post, finding accurate frequencies with vasp needs to be done carefully - especially for transition states. You must be starting with an image that is well converged to a saddle point (EDIFFG<0.005 eV/Ang). Then you need to use small finite difference displacements, on the order of 0.01 Ang, and use a fine force criterion, (EDIFF=1e-8 is safe).
There are some rare cases in which the NEB can settle on second order saddles, but this is highly unusual, and generally a consequence of having a highly symmetric system. Definitely, start by checking the accuracy of your frequency calculation. You can be confident in the frequencies that you have calculated when they are insensitive to your choice of finite difference displacement.