USER-COLVAR Package: handling of cartesian CVs with extended Lagrangian


A (hopefully) brief question on the USER-COLVAR package. If I use an extended Lagrangian on a Cartesian CV comprised of multiple atoms, e.g.:

Hello Spencer, ordinarily I would say that the two scripts should give the same result i.e. there are 9 distinct Langevin thermostats.

However, the Langevin thermostat code for the extended-Lagrangian variables does not support higher than one-dimensional variables, i.e. using cartesian for both a single atom (3 dimensions) and three atoms (9 dimensions) would raise an error.

To be honest, it’s a rather uncommon combination of features: cartesian is mostly there to support variables written by the user (a NAMD and VMD feature, soon to be added to LAMMPS).

There is no reason for that constraint, though, so I updated the Colvars library code here:

an added a version of the current LAMMPS source that includes this change:

I’ll make sure that it gets merged into the official LAMMPS at the next update.




Thanks very much for your response….

I agree, it is a bit of an edge case, however I think it will be useful to have the ability to use an n-dimentional langevin thermostat on extended Lagrangian DOFs to apply TAMD / DAFED-like algorithms to Cartesian collective variables. I find that Cartesian CVs can be useful if the choice of a suitable composite CV is not obvious and one does not wish to introduce an artificial bias by imposing a defined CV.

Thanks again!

Best regards, Spence

Spencer T. Stober, Ph.D.

Corporate Strategic Research

ExxonMobil Research and Engineering

1545 Route 22 East

Annandale, NJ 08801

(908) 335-2711