NEB calculation (2nd thread)

[ccccc]

The old thread has closed so i am writing again;

An example of a sentence from a paper "In the first levels, typically more than 100 iterations of NEB were carried out until the forces perpendicular to the reaction path were considerably small." or another says " The NEB method takes a set of images of the system between the initial and final states and optimizes them simultaneously in a subspace perpendicular to the imaginary line connecting the images."

Here i dont understand this. if the NEB calculation does a relaxation, the geometry goes its most stable structure. so what does those sentences above tell us? What does it means " the force perpendicular to the reaction path" and its relation with the NEB's relaxation?

thank you...
<span class='smallblacktext'>[ Edited ]</span>

[forsdan]

In short: The images in NEB calculations are only updated by using the force component, due to the potential, perpendicular to the band in combination with the parallell spring force. Therefore, relaxation of the images will not result in relaxation to a minima.

I recommend you to read some references about the NEB method. For example you can have a look at

http://theory.cm.utexas.edu/vtsttools/references/


Cheers,
/Dan
<span class='smallblacktext'>[ Edited Mon Mar 08 2010, 11:15AM ]</span>

[ccccc]

Thank you much for your interest.
One more question please: Will i specify this relaxation in INCAR with NSW parameter?

[forsdan]

The NSW paramter specifies the maximum number of ionic relaxations.

The required NEB parameters are given in the manual http://cms.mpi.univie.ac.at/vasp/vasp/node149.html .

If you want to run the Climbing-Image NEB instead, more information is available at http://theory.cm.utexas.edu/vtsttools/ .

Best regards,
/Dan

[ccccc]

I prepared an Incar file. Can you look it for a moment please?

SYSTEM = …..
ISTART = 0
IMAGES=8
ICHARGE= 1
ISYM = 2
SPRING = -5
NPACO = 256
NELM = 40
NELMIN = 8
PREC = HIGH
LREAL = .TRUE.
LELF = .TRUE.
NSW = 1
IBRION=1
NPAR=1
Does it look like true?

thanx

[admin]

NPACO and LELF are not needed for an NEB run,
NSW =1 (in standard vasp) would result in exactly one ionic step to be done. Is this what you want?
NELM = 40 may not be sufficient if the electronic convergence of the one of the images is bad (keep the default value instead)
8 Images may be rather much, please start with 4 and refine the calculation in the region around the barrier in a second NEB calculation.

[ccccc]

Firstly i want to thank you for answer.
I changed NSW parameter to 100. But at the end of the work, all images gave the same energy values in OSZICAR file.

Actually my question is simple. I have an initial and a final state geometry. I want to do a NEB calculation for to see the transition energy between them.

But i still dont understand the work of NSW parameter here. Doesn't it takes all middle geometries to the final state, if i give it a value different from 1?? And I can not find the parameter where i can specify the force which is perpendicular to the reaction path. Can i specify it with SPRING=0 ?

thank you again...

[forsdan]

The NSW parameters specifies the maximum number of ionic relaxations that are performed for each image. The relaxations will NOT take the middle geometries to the final state, since you only update the ionic coordinates with the NEB force (parallel spring force + perpendicular potential force). The spring part keeps the images equally spaced while the perpendicular force relax the images to the minimum energy path. The perpendicular potential force (and the parallel spring force) is obtain by projecting on a tangent between images (which is done internally in VASP). Please have a look at the references I linked to above to make sure you have understood the principle in the NEB method.

The SPRING parameter specifies the spring constant of the springs between the images.

Regarding that your statement that all energies become equal:

Please make sure you have followed the instructions at

http://cms.mpi.univie.ac.at/vasp/vasp/node149.html

In particular:

1) Make sure you have a parallel binary of VASP and that the number of nodes are dividable by the number of images.

2) Make sure you have put the POSCAR of initial geometry in 00/ and the POSCAR of the final geometry in XX/ (XX=number of images+1). Also make sure you have supplied POSCAR files in all intermediate folders where the geometries represent a reasonable interpolation between the initial and final state.

Cheers,
/Dan


<span class='smallblacktext'>[ Edited Mon Mar 15 2010, 09:56PM ]</span>

[admin]

which vasp version do you use? there was a small bug in all releases up to vasp.4.6.31, in which the total energies of all images were summed up before they were written to OSZICAR.
Anyway, if you want to obtain the total energies of the particular images, please have a look at the OUTCAR files in the respective subdirectories.

[ccccc]

I did those calculations with vasp 5.2.

[ccccc]

Dear Admin.

I read all references for NEB calculation which Forsdan gave before. I think i uderstood the principle. So i prepared an INCAR file like this:

SYSTEM = neb
ISTART = 0
SPRING = -5
ICHARGE= 1
ISYM = 2
NELM = 60
NELMIN = 8
PREC = HIGH
LREAL = .TRUE.
NSW = 150
IBRION = 2
NPAR = 1


Now, it is still not very clear for me that where i can read the parallel and perpendicular forces. I want to see their values for example the perpendicular component of Fneb is =.... ! Will i see it from OUTCAR file?