I would like to perform a constraint dynamics to calculate PMF on a distance variable. So I would like to calculate the mean force on the constraint. Is it possible to do it with fix shake? Or do you have some other suggestion.

I would like to perform a constraint dynamics to calculate PMF on a distance variable. So I would like to calculate the mean force on the constraint. Is it possible to do it with fix shake? Or do you have some other suggestion.

How about fix colvars? It has several options to compute free energies.

Thank you for your suggestion. While, for my case, I am interested in the calculation of PMF for Na Cl pair in water. And, the literature like JCP, 132, 114510 using constraint dynamics to calculate PMF (for FPMD case). And, due to my experience, for simple system like this, constraint dynamics seems to give a much faster convergence than method like umbrella sampling. I am not sure about whether the adaptive bias force or metadynamics can give a better performance. And, also constraint dynamics is much simpler.

Is it possible to add a constraint like shake in colvar. Or, is it possible that I use a large spring constant on the constraint and calculate that force. Would you please comment on this?

Thank you for your suggestion. While, for my case, I am interested in the
calculation of PMF for Na Cl pair in water. And, the literature like JCP,
132, 114510 using constraint dynamics to calculate PMF (for FPMD case). And,
due to my experience, for simple system like this, constraint dynamics seems
to give a much faster convergence than method like umbrella sampling. I am
not sure about whether the adaptive bias force or metadynamics can give a
better performance. And, also constraint dynamics is much simpler.

you are doing classical MD not FPMD, so sampling and convergence is
not something to worry about. not for such a simple system. i don't
know anything that is simpler than umbrella sampling.

Is it possible to add a constraint like shake in colvar. Or, is it possible
that I use a large spring constant on the constraint and calculate that
force. Would you please comment on this?

i don't know. if you want it that badly, dig into the code for fix
shake and implement it. that is why LAMMPS is open source. people can
implement what is important to them and in the end everybody benefits,
when this is integrated back into the code.

Hello Yi, the question of which one has the highest convergence rate between thermodynamic integration and umbrella sampling is a long debated issue which there is no point getting into for a simple case such as NaCl. Any method will work very well for it.

In any case, implementing a hard constraint and use it to perform thermodynamic integration should be done as Axel suggests (i.e. source code changes).

However, you can also use the same exact free energy estimator of thermodynamic integration in an unconstrained simulation. This can be done (among others) by the ABF method, the specific formalism used in the LAMMPS + colvars implementation is described here: