Page 1 of 1

NEB path: general doubts

Posted: Wed Nov 05, 2008 1:25 pm
by gpkmohan
Dear All,

I work with a reaction path searching for the CH4 adsorption on Ni(111) surface, where CH4 is free to move and Ni slab is fixed. I have been used CI-NEB method from VTST tools with VASP and found a transition state[TS], then I like to perform a reaction path following by using a steepest descent [SD] algorithm from the TS and a normal mode analysis on the path in this system.

I would like to ask some doubts in general,

1. Equivalence of NEB image and SD image

As far as I know *all* the varieties of the NEB methods uses MASS UN-WEIGHTED forces and coordinates for its calculations. So I think - if I do a SD calculation in MASS UN-WEIGHTED coordinates[for xyz and forces] I can find/obtain (or very close geometries of) the converged NEB images during this SD search.

Is it possible ?
Somebody tried this before ie. retracing the NEB images by an SD from a TS?

2. relation between IRC and NEB

The ab-initio guys should familiar with IRC [Intrinsic Reaction Coordinate] calculations.
IRC is nothing but an SD in mass-weighted coordinates[in xyz and forces].
From a book: 'The reaction path in chemistry-current approaches and perspectives',edited
by Dr.D.Heidrich,[Leipzig] it is clear that steepest descent paths( in MASS UN-WEIGHTED *AND* MASS WEIGHTED )are equivalent in terms of the internal coordinate representation.

The main differences are
[1]. centre of mass is moves in mass un-weighted steepest descent.
[2]. centre of mass is not moves in mass weighted steepest descent [IRC]

Did someone checked out the equivalence/relations of IRC obtained points and NEB images ?
Does the movement of the centre of mass in the mass unweighted SD have some siginificat role
in the minimum energy path following ?

Hope someone can give your valuable comment on this matter.

With best regards and thanks in advance

Krishna Mohan
THEOR CHEM
University of Leiden

Re: NEB path: general doubts

Posted: Wed Nov 05, 2008 8:19 pm
by graeme
1) We have compared NEB images with a SD path. In the following paper:
http://theory.cm.utexas.edu/henkelman/p ... 134106.pdf
Section VIII and Figs. 9 and 10 show convergence to the minimum energy path (MEP) with a few different methods (NEB, string, and SD)

These paths are mass un-weighted; they follow the force from saddle to minimum.

2) The mass weighted MEP will be a different path and will not pass through the NEB images. Of course, it is simple to rescale the forces so that the NEB finds points along a mass-weighted MEP. Fortunately, the saddle point does not change with mass scaling, and it is the saddle point which is most important for reaction rates. Apart from tunneling, I can't think of any TST related calculation which uses the actual path away from the critical points. Well, maybe as a guideline for a free-energy calculation. Normal modes are along the mass-weighted directions, but these are evaluated only at the critical points.

For an isolated system, the NEB images can be shifted so that the center of mass does not move. This can also be done within the NEB calculation, by zeroing any net force on the images or any net displacement requested by the optimizer.

Ed Note: I see that Johannes Kastner and Paul Sherwood have implemented an option to use the NEB with mass weighted coordinates in the DL-FIND / ChemShell program.

Re: NEB path: general doubts

Posted: Thu Nov 06, 2008 9:01 am
by gpkmohan
Thank you verymuch Prof.Henkelman for your comments and new infos on it. I have looked at the figures on that paper, which is well decribed the SD and NEB relations.

As a newbies comment: sometime we need actual path [ie a lot of points on the path] in otherwords if we do vibrational anlaysis along the path, which can - in principle - give the idea on the relavance of the normal modes and thier importance on the reactivity. Ab-inito people are doing this for small systems and predict reactivity of the modes during the reaction[ie. first do an IRC then the vibrational analysis along that path]. Personally I found some difficulties for the same in planewave codes because: [in general ] lack of the good - point group analysis of the geometry and modes, un avialability of the internal coordinates and Hessians [all really smoothens the Hessian elements from one point to its nearest point].