Question about thermodynamic integration

LAMMPS LAMMPS Mailing List Mirror
1

Greetings,
I have been reading the documentation and the archives re: thermodynamic integration and have a question. I would like to perform a TI in which the LJ epsilon and sigma for one atom type are changed. My reading of compute ti is I must write the potential as f(lambda) U, i.e., a scaled potential energy. I don’t see how to do this for the LJ potential unless the potential energy is written as

U =U_0 + lambda (U_1 -U_0)

with U_1 and U_0 representing the endpoints of the transformation. I do not know if it is possible to evaluate U_1 and U_0 for use in a compute ti command. My question is then: does LAMMPS have the ability to perform an alchemical transformation in which both epsilon and sigma are changed for one LJ atom type?

Randy

2

I asked Aidan to look at this - he is more familiar with

compute ti than I am.

Steve

3

Yes, compute_ti requires that "scaling is done via a prefactor on the
energy." fix_adapt is a little more general, in that specific
parameters can be adjusted during the simulation. However, if the LJ
sigma is adjusted, then the quantity dUs/dlambda is not simply
proportional to Us, but is instead proportional to the sum of
dulj/dsigma for all pairs. So, I don't see an easy way to allow sigma
to be adjusted.