My Community

Dear VASP-users,

Is it possible to restrict the movement of an atom to an arbitrary direction, using the "Selective dynamics"-tag, without defining that direction as an lattice vector ? For example, if my lattice vectors in the POSCAR read

bcc:
2.831000000000000
2.0000000000000000 0.0000000000000000 0.0000000000000000
0.0000000000000000 2.0000000000000000 0.0000000000000000
0.0000000000000000 0.0000000000000000 2.0000000000000000
16
Selective dynamics
Direct
...

Is it possible to restrict the movement of one atom to the (1,1,1)-direction ?

Best regards

Dan Fors

I dont think that is possible.
You might have to implement that yourself using Lagrangian Parameters to enforce that some atom can only move in a certain direction.

One possible way might be to set your (111) direction as one of the base vectors.

I am not sure if the link below fits your need.
http://cms.mpi.univie.ac.at/vasp-forum/ ... .php?4.955

In principle, you can specify the movement of each atom in the POSCAR by a adding a table of velocity vectors following the positions
http://cms.mpi.univie.ac.at/vasp/vasp/node63.html

For example this can be a POSCAR to simulate the collision of H+H:
-----------------
H+H
1.0
10.0 0.0 0.0
0.0 10.0 0.0
0.0 0.0 10.0
2
Selective dynamics
Cartesian
0.25 0.00 0.00 T T F
0.00 0.00 0.00 F F F
Cartesian
-0.01 0.00 0.00
0.00 0.00 0.00
-----------------

Thank you all for your replies,

The task I was considering to accomplish was to calculate diffusion barriers at 0 K. The main idea was to perform structural relaxations for the initial and final states and then choose the reaction coordinate as a linear interpolation between these two states. The particle would then be constrained to move in a plane orthogonal to the reaction coodinate.

The problem from my point of view arises if I specify the reaction coordinate as a base vector, I must find two vectors orthogonal to the reaction coordinate that can span a unit cell of reasonable size. In the (111)-direction case this may be possible to accomplish but if the reaction coordinate is more arbitrary due to the structural relaxations the probability for success decreases. I now wonder if there is an easy solution to this problem or if the best way is to use the implemented NEB-method instead.

Best regards

Dan Fors

With the NEB method the atoms are allowed to relax perpendicular to the path (with spring forces constraining relaxation along the path). This will converge towards the minimum energy path, which is probably what you're trying to calculate anyway.

Yes, NEB is your friend.

If you have a really well-defined reaction path then you can simulate such a thing with a series of optimisations with your migrating atoms held fixed at points along the reaction path, but you have to very carefuly fix other atoms to define your reference frame. (Otherwise the unit cell simply translates until you get your potenial minimum configuration back.)

But it really sounds like you want NEB.