UT theoretical chemistry code forum

Dear Prof. Greame,
I am running TS calculations for water dissociation on metal surfaces where the initial state is the water molecule adsorbed flat on the metal surface (optimized) and the final state is the water dissociating to H and OH to their corresponding stable states (optimized). If I use this optimized initial and final states for generating the POSCARs using nebmake.pl, I get a decent initial POSCAR files. Here the water molecule had to rotate a bit to get dissociated. This happens without any problem while generating the initial POSCARs (the input is fine). But during optimization both the hydrogen bonds break and go far away from the O atom. I am not able to get a optimized structure close to the initially generated POSCAR. interestingly, when I set the water molecule which looks similar to final state (direction of H and OH towards the final adsorption sites) but not the optimized structure, I get a very close estimate of the transition state geometry. The calculated barrier looks satisfactory. I would like to know whether transition state calculations are orientation dependent or we can get the close enough transition states with the actual optimized geometries.

Thanks and regards

The NEB will find a minimum energy path which is a local minimum from the initial path. Particularly with rotation, a linear initial path can result in initially high forces and a converged path which is not close to your initial state. In the case of water, for example, the hydrogen atoms might dissociate from the water molecule, and the NEB converge to a path involving hydrogen dissociation and recombination.

To deal with this, check your initial forces. If they are high, you can use a conservative optimizer (e.g. IBRION=3, POTIM=0.01) until the forces drop to 1 eV/Ang. Then check that your path is reasonable and switch to a more efficient optimizer if it is.

You can also use the nebavoid.pl script to avoid any short bond lengths which might be in your initial path.