UT theoretical chemistry code forum

Hi,

I am a frequent user of NEB and now I am trying to learn to use the dimer method. I find NEB somewhat intuitive, I will give an example with a simple molecule. If I want to calculate the barrier for hydrogen abstraction from CH4 on a surface, I will relax two endpoints CH3 + H* and do an interpolation and run NEB as usual. The same for the subsequent step (CH2+H*, CH+H* and C+H*).

How does that compare to the dimer method? If I put a CH4 molecule on the surface, how will the code know that I am particularly looking for the barrier for the first hydrogen abstraction?

Thanks,
Alexandra

If you know the reaction mechanism that you want to model, the NEB is a good way to find that specific minimum energy path. The dimer method can also help with this. If you have a geometry in the vicinity of a saddle point, you can use the dimer method to converge to the saddle. The neb2dim script makes this pretty easy. Since the dimer method uses a single image, the cost of convergence is generally lower than relaxing an entire band. The dimer method is also useful if you want to reconverge on a saddle point with different calculation settings, such as when you do convergence tests.

The dimer method can also be used to find saddles that may correspond to unknown or complex reaction mechanisms. For that, we usually start by displacing the atoms in some active region of configuration space. In your example, this could be the CH4 molecule, or decomposition fragments on the surface. Then, the dimer method will calculate and follow the lowest curvature mode in the system and try to converge upon a nearby saddle. As you point out, this is not ideal if you want to find a particular saddle, but it is very good if you want to explore the potential surface and find saddles for reaction mechanisms that you don't nessecarily know about or are not able to anticipate.