Free Energy Perturbation

Currently, the free energy code in GROMACS is only set up for thermodynamic integration (TI). Various free energy perturbation methods (exponential averaging/the Zwanzig relation, or Bennett acceptance ratio, which in many contexts is substantially more efficient than TI or exponential averaging) require some additional information that is not currently output by GROMACS -- in particular, the potential energy of trajectory snapshots evaluated at other lambda values.

For free energy perturbation, and for various other methods (i.e. WHAM for free energy calculations) it would be extremely useful to put this into GROMACS.

All that needs to be done is:

• Add a mdp option to store how often to evaluate the energy at alternate lambda values (this needn't be done every step) and dump this to the energy file (or a separate energy file for just this)
• Add the ability to specify a set of lambda values in the mdp file: One to run at, others to evaluate the energy at.
• Add the necessary energy evaluations and dumps. Note that there will probably need to be additional calls to PME, etc., for these energy evaluations.

Erik Lindahl has apparently put some effort in this direction, but it is not finished.

A free energy perturbation tutorial is currently located here: http://www.dillgroup.ucsf.edu/group/wiki/index.php/Free_Energy:_Tutorial