Working on NonbondedForce

[jchodera]

This is actually pretty complicated, so won't finish tonight. Will review my progress with you tomorrow.

[jchodera]

It looks like we will have a number of special cases for the NonbondedForce interactions involving the small molecule being alchemically transformed:

Sterics

• Pair interactions where epsilon and sigma do not change from A to B states should not be softened at intermediate lambda
• Pair interactions in which one atom is disappearing in one state (A or B) but not the other will have epsilon = 0 in that state, and should remained alchemically softcore in that state to avoid sampling issues, using a schedule like (1-lambda) or lambda depending on the state (A or B) in which the atom disappears.
• Pair interactions in which the parameters are just changing from one set to another set should be softened at intermediate lambda using a schedule like lambda*(1-lambda) so that physical interactions are recovered at lambda = 0 and lambda = 1.

Electrostatics

• Pair interactions where chargeProd does not change should not be softened
• Pair interactions in which chargeProd = 0 at one state (A or B) but not the other will remain alchemically softcore in that endstate
• Pair interactions in which parameters are changing will be softened at intermediate lambda using a schedule like lambda*(1-lambda) so that physical interactions are recovered at lambda = 0 and lambda = 1.

If we want the full generality of being able to tweak all of these choices individually, I can separate out some options. Also, we can separate the lambda values into different components if we want to play with different alchemical pathways.

[pgrinaway]

Should this discussion be moved to examol?

[jchodera]

Whoops, yes! Will do.