Neb method with vasp

[ouarda]

Dear Vasp users,
can any one explain me how to do calculation with neb method to find transition state for migration energy. I'm intersting in the diffusion impurities in transition metals I use the vasp.4.6. Recently I have showed that the neb method work only in vasp parallel version and to do this, as was explained in http://cms.mpi.univie.ac.at/vasp/vasp/node134.html, we must creat subderectory wich contains POSCAR files. my question is how to create this POSCAR files when we have the initial and final positions.
which input parameter we need to specify in the incar file.
Thanks for your help and for your time

[forsdan]

I've never actually performed a NEB-calculation with VASP, but I'm about to start a simulation within two days so I'm quite eager to take part in the discussion as well.

As stated at the given adress the input parameters connected to the NEB-calculation will be IMAGES, SPRING and LSCALAPACK = .FALSE. , in addition it's recommended to not use CG-method for the atomic relaxation.

When it comes down to the generation of the POSCAR-files one often has some idea of how the diffusion process takes place. In my case I will perform calculation using a bcc-lattice and a reasonable starting guess in some cases for the interstitial octahedral-octahedral diffusion is that the tetragonal void will be the transition state. For substitutional diffusion via the vacancy mechanism the reaction coordinate will be quite straightforward. For these two cases I would choose to displace the impurity along the reaction coordinate in order to produce initial POSCAR-files for the images. The NEB-method will then keep the images from relaxtion to the structural minimum during atomic relaxation of each image.

Another situation is the direct exchange mechanism for substitutional diffusion. Here you will know the initial and final state, i.e. two atoms interchanged. By displacing the atoms towards each other along the line that interconnect them and maybe a bit perpendicular to this line, I've planned to generate the starting guesses for POSCAR-files for the images.

I hope I understood your question correct and that this might help you.

Best regards

Dan Fors

[ouarda]

thanks Dans,
In my case I calculate the diffusion barier fo impurities in fcc metals like silver or Ni. so Ihave to path way to explore i.e, Octa-Octa or Octa-Te-Octa. so to find exactly he transition state we must give intermidiates images between the initial and final state. So my problem is how to found the differentes POSCAR corresponding to the intermidiate images .
Thanks for all

[forsdan]

I'm still not sure that I understand you correctly. If we use the cartesian vectors a(1,0,0) , a(0,1,0) and a(0,0,1) we have for example a octahedral site at a(0.5,0.5,0.5) and a tetragonal void at a(0.25,0.25,0.75) if I'm not totally mistaken ? There will also be another octahedral site at a(0,0.5,1) right ?

If the diffusion process takes place via octa-tetra-octa diffusion, why can't you just displace the impurity along the a(-0.25,-0.25,0.25) direction thereby reaching the tetragonal site, and thereafter displacing the impurity along the a(-0.25,0.25,0.25) resulting in the other octahedral site ? Even if the POSCAR-files of the images don't lie at the MEP the NEB-method will relax them to it. I'm not sure the directions and coordinates have come out correctly now, but I hope you get the main idea or do you see any problems with this procedure ?

Best regards
Dan Fors
<span class='smallblacktext'>[ Edited Fri Jun 09 2006, 10:26PM ]</span>

[forsdan]

I thought of one more thing actually, because I might have misunderstood the question completly. Is it the spacing between the images that you're after in order to get images near the transition state ? Because to get an image exactly at the transition state may be hard in some cases if you don't have symmetry. The CI-NEB will as I understand force the images near the transition state to climb upwards and therefore produce a more dense mesh were it's needed, but I don't know how the regular NEB-method behaves. The solution otherwise is the choose to have more images around the transition state in order to get a more accurate result for the migration barrier.

Best regards

Dan Fors

[admin]

please
1) carefully relax the geometries of the initial and the final state
2) interpolate a certain number of geometries ('IMAGES') by linear interpolation between them (usually, 4 images is a good choice to start with). You can do this by writing a small program or script. Each of these geometries must be written into a POSCAR file
Please mind that cell size and cell shape have to be the same for all geometries.
The positions of all atoms should be interpolated linearly; please check that no atom leaves the cell at one side and enters on the other side (due to the periodic boundary conditions, this will give a pathologic path)
3) prepare (IMAGES+2) subdirectories, named 00, 01 ,..... 0(IMAGES+1). Each of these subdirectories must contain exactly
one POSCAR file
4) INCAR,KPOINTS and POTCAR have to be in thre parent directory (the directory which you start VASP from)
5) prepare INCAR (especially, SPRING=-5 and IMAGES have to be given)
6) start a parallel executable of VASP, the number of processors has to be an integer multiple of the number of IMAGES.

[ouarda]

Thanks for your all response. I have one question about INCAR file if I must also specify the input parameter related to the relaxation i.e, IBRION and ISIF in addition to the SPRING and IMAGES parameter.
Thanks

[admin]

ISIF need not be defined explicitely, the cell volume and shape have to be kept constant in an NEB calculation
concerning IBRION: usually it is best to choose it to 1 (or 3).
If the NEB fails to converge, please start the NEB with a few ionic steps
IBRION=1; NFREE=2
(or IBRION=2; SMASS=3)
and proceed with IBRION=1 after a certain pre-convergence is reached.