improved dimer method and ci-NEB

[askhetan]

1. The vasp website mentions about an improved dimer method to optimize the transition states by using frequencies. It asks for the use of the hardest mode values (i understand the dx,dy,dz values corresponding to the lowest imaginary frequency ??) to be used in the POSCAR file. In the example case of dissociative adsorption over a slab, will the modes values (dx,dy,dz) of the participating atoms only suffice? or do we need the mode values of all the atoms in the slab? Does this method really give us transition state from an initial and final state, in what form do we get the geometry of the intermediate transition state? How do we specify the final state? don't we need to specify the intermediate states ?

2. There is a supposedly more accurate variant of the NEB method called the ci-NEB method. Has it been incorporated into VASP? The UT Austin page on NEB mentions a few parameters like LCLIMB, etc... Does setting LCLIMB=.TRUE. ?? automatically turn on the ci-NEB method or should we first do a normal NEB and then use the ci-NEB ?
are there any other parameters that would be useful in implementing ci_NEB in a VASP run?

Very grateful for your help!

<span class='smallblacktext'>[ Edited ]</span>

[vasp_user]

You might consider using the UT Austin NEB forum for some of this. I just have a couple of comments about climbing image NEB from UT Austin package. You could try doing an even number of images and an odd number of images for the same initial and final config (both with LCLIMB off). Then for the one that has a lower activation energy start it again from scratch but with LCLIMB on and I believe the activation energy should climb at least as high as the highest one seen from the initial testing. You can double check with Austin forum. Another thing to be aware of is the LCLIMB just makes the highest image climb and does not move the others up (if I remember correctly). Of couse, this means that if your intial state and final state span multiple transitions states your cimbing image is only going to grab one of them at most.