Page 1 of 1

problem related to the setup of images from nebmake.pl

Posted: Fri Dec 04, 2009 3:43 am
by whdxsn
Hi, NEB users

I have a problem related to the setup of images from nebmake.pl. I really need your help so I may write a bit more about the situation. thanks for your time :)

The system i study has a substrate and hydrogen atom bonded to the substrate surface atom.

i have the optimized geometries of the initial and final structures as POSCAR1 and POSCAR2.
i then run the nebmake.pl POSCAR1 POSCAR2 6 and get the image POSCAR in 01,02...

However, when I view the POSCAR in the 03 directory, I find that some of the atoms(atoms that bonded with H and deformed vertically with respect to rest of the atoms) on the substrate surface shift so much that they are separated from the rest of the substrate(say the normal atom bond length is 1A, the distances of the separated atoms are 1.5A away from the rest atoms on substrate, so those atoms look like a cluster sitting above the substrate surface)

My questions :
1. I am wondering that whether this unphysical intermediate images structures will be a problem for the following neb calculations.
2. I want to know whether it is possible or reasonable to fix some of the atoms when constructing the intermediate image structures. By reading the current nebmake.pl code, it looks like they take the pbc_difference of all corresponding atoms of the reactant and product.
3 . Furthermore, can i manually construct the guessed intermediate image structures in POSCAR and put them in the 01,02... and run the neb calculations and let the vasp optimize and find the transition states structures

Re: problem related to the setup of images from nebmake.pl

Posted: Fri Dec 04, 2009 3:59 pm
by dano
The nebmake.pl script linearly interpolates the images between the initial and finial POSCAR. Typically something unexpected like the cluster of atoms jetting up vertically from the surface, is do to the atoms not being in the same order for the two original POSCAR files.

1) unrealistic intermediates will be tough for the neb to converge most often do to the high forces predicted
2) the script itself does not allow to selectively fix atom positions, However if say atom 8 is in the same position in both the initial and finial POSCAR files then it will not move in any of the intermediates.
3) The nebmake.pl only uses a linear interpolation to guess the initial path. By no means does this always lead to the lowest transition state because the neb method only finds a local minimum energy path. The linear interpolation can be particularly bad for path which involve a rotation around a bond. You still need to make sure that atom 1 is the same atom for every image and so on.

Re: problem related to the setup of images from nebmake.pl

Posted: Fri Dec 04, 2009 8:00 pm
by whdxsn
thanks for help dano. I am wondering is there anyway to check or enforce the one-by-one correspondence? I has a file in *.car format from Material studio and load it into molden and convert it into POSCAR. Because of the amount of the atoms, checking by eye does not be a good solution though i am checking by eye now.:(
I know how to use openbabel to change formats, any secure way of format changing can preserve the atom order information or post-processing tool to ensure that ?

thanks again for your attention and help

Re: problem related to the setup of images from nebmake.pl

Posted: Sat Dec 05, 2009 6:42 pm
by dano
not that I am know of but what I do is convert one geometry and then use that geometry to make all my other states.