Reduced-dimension constraints in ICONST

[apolet]

Hi folks,

Another question from me about setting up constraints in ICONST. How does one specify in ICONST the flags to bias a distance between two atoms but only along 1 or 2 axes, but not all three?
I am guessing that it will be some combination of "primitive" flags X/Y/Z or cX/cY/cZ with the "complex" flag C, but I do not think I understand how the X/Y/Z or cX/cY/cZ are used.

I am ideally looking for an example, the same way that the VASP wiki provides helpful examples on the other flags.

The context for this is trying to create a metadynamics-biased simulation of atoms sliding past one another in an anisotropic environment. It is possible to bias the three-dimensional distance, but that is quite inefficient.

Thank you!

[apolet]

Following up in absence of answers - is there any way to do this at all?

[andreas.singraber]

Hello!

I am not entirely sure what you are trying to achieve with this constraint. Please be aware that by constraining only one or two components of a distance vector you are breaking rotational invariance in your system, i.e. your Hamiltonian is not rotationally invariant.

Nevertheless, your constraint still can be applied with the ICONST file, for example like this:

Code: Select all

cZ 8 0
cZ 9 0
S 1 -1 0

In the first two lines this defines the z-components of the Cartesian coordinates of atoms 8 and 9 as primitive coordinates (not constrained) and then constrains the difference between them. Hence, the z-distance between atoms 8 and 9 will stay constant throughout the simulations while x- and y-distances may vary.

Best,
Andreas Singraber

[apolet]

Thank you! I was looking to apply metadynamics biasing along selective dimensions to increase the effectiveness of such biasing. If I know that I want to explore only a subset of displacements, then it makes a lot of sense to bias them along only the directions I care about.