[Pw_forum] Intermediate images in NEB

[Guido Fratesi]

Dear Daniele,

 | However, there are case when it would be desirable to use EXACTLY that
 | initial guess, for example in case of interpolation 
 | through intermediate values of a torsional angle...
 | That's the origin of my concern. Maybe it would not be a bad idea to
 | add this option...

at once I was also in the need of specifying EXACTLY a given initial path. 
The reason, in my case, was that I wanted to merge two paths from 
different calculations in a single one. Of course, I didn't want a 
re-interpolation to spoil the accuracy already reached.

Then, I build the path file "by hand" (but awk or the like is helpful). 
That's not that hard.

One possibility for you could be to edit the interpolated path file 
which you get with nstep=0, and replace the coordinate of some images with 
the ones you want. Then you can restart from this path file.

The real problem, in my opinion, is that a path build up this way will 
generally have images not at the right spacing, therefore 
feeling too much the springs in the direction parallel to the path. This 
could require a few additional optimization steps, and possibly compensate 
the advantages.

Another solution could be to generate a set of path files linking the 
various intermediate images, to merge them, and finally to start the 
optimization of this merged path.
I feel that could work better than my previous suggestion, because 
possibly no image would feel to strong springs, if the initial spacings of 
the various paths is comparable.

Hope this helps more than confuses
Guido

-- 
Guido Fratesi

Dipartimento di Scienza dei Materiali
Universita` degli studi di Milano-Bicocca
via Cozzi 53, 20125 Milano, Italy