Dear all:
I want to use dimer method to identify a TS.
Now, I have the initial state and the guess TS.
How to use diminit.pl script to generate the input files?
The 'POSCAR' in diminit.ps command line is the TS's stucture
, right? I did it, but I got two meaningless structure files.
Thanks !
question about dimer run
Moderator: moderators
If you have a guess for the transition state then simply place that POSCAR in two different directories, named 01 and 02. Have the necessary KPOINTS, POTCAR and INCAR where the 01/ and 02/ lie, i.e. 01/, 02/, KPOINTS, POTCAR, and INCAR are all at the same specific location. Then format the INCAR file according to http://theory.cm.utexas.edu/vtsttools/dimer/
A dimer calculation needs two images to get started. The midpoint of these images can be your guess at the saddle point, and the vector between the images your guess at the negative mode (the reaction coordinate at the saddle).
If you are starting from an NEB calculation, there is a neb2dim.pl script which is probably the most convenient to use. This will find the location of the maximum energy along the NEB, and form a dimer at that point, oriented along the local tangent.
If you are starting from scratch, and you just have a guess of a saddle point, one thing you can do is put this POSCAR in both the 01 and 02 directories. In this way, you have provided no information about the negative curvature direction, so the algorithm will start with a random direction.
It sound like you have some idea of the reaction coordinate, so it would be best to get that information into your calculation. If you only have the initial state and TS guess, you could use the vector between these two configurations as a guess at the negative mode. To generate your two initial dimer images, you can use:
interpolate.pl POSCAR_IS POSCAR_TS 0.99
interpolate.pl POSCAR_IS POSCAR_TS 1.01
to form two images very close to the saddle point (POSCAR_TS) along a line to the initial state (POSCAR_IS).
We use the diminit.pl script primarily in conjunction with our adaptive KMC method. Here the idea is to generate randomly distributed initial dimer configuration around a known minimum. It is appropriate to use this script when you don't know about the reaction mechanism (expect perhaps which atoms are involved). To use the diminit.pl, you give it the POSCAR for your initial state, and a DISPLACECAR file, which describe which atoms should be displaced, and by how much (on average). Since you already know something about the saddle point, this is not the best script to use.
If you are starting from an NEB calculation, there is a neb2dim.pl script which is probably the most convenient to use. This will find the location of the maximum energy along the NEB, and form a dimer at that point, oriented along the local tangent.
If you are starting from scratch, and you just have a guess of a saddle point, one thing you can do is put this POSCAR in both the 01 and 02 directories. In this way, you have provided no information about the negative curvature direction, so the algorithm will start with a random direction.
It sound like you have some idea of the reaction coordinate, so it would be best to get that information into your calculation. If you only have the initial state and TS guess, you could use the vector between these two configurations as a guess at the negative mode. To generate your two initial dimer images, you can use:
interpolate.pl POSCAR_IS POSCAR_TS 0.99
interpolate.pl POSCAR_IS POSCAR_TS 1.01
to form two images very close to the saddle point (POSCAR_TS) along a line to the initial state (POSCAR_IS).
We use the diminit.pl script primarily in conjunction with our adaptive KMC method. Here the idea is to generate randomly distributed initial dimer configuration around a known minimum. It is appropriate to use this script when you don't know about the reaction mechanism (expect perhaps which atoms are involved). To use the diminit.pl, you give it the POSCAR for your initial state, and a DISPLACECAR file, which describe which atoms should be displaced, and by how much (on average). Since you already know something about the saddle point, this is not the best script to use.