[gmx-users] Coupling only a variable solvent subgroup to the thermostat

[Leandro Bortot]

Dear GROMACS users,

     In a NPT molecular dynamics simulation of a system composed by a water
box with ions and a protein I want to couple only water molecules which are
far from the protein to the Nosé-Hoover thermostat and leave the remaining
system uncoupled.

     I managed to couple a subset of the solvent molecules by specifying a
group in the mdp under tc-grps and by giving an index file to mdrun. The
problem is actually how to create the appropriate (variable) selection of
the water molecules which are to be coupled to the thermostat.

     While probably an inefficient way of doing this, I can only think of a
script which (1) stops the simulation just before the thermostat acts over
the system; (2) makes the index file with the appropriate selection with
g_select and then (3) continues the simulation, allowing the thermostat to
act over the appropriate subset of the solvent.


     I'm currently using 5 ps as the period of the contact between the
system and the thermal bath, meaning that I would have to stop the
simulation, say, each 4.9ps in order to make the new index file. Actually,
this is probably overcautious since I don't think the waters will move so
much in 5ps. Maybe I can make a new index file each, say, 20ps.


     Any suggestions and/or comments about how to do this and my proposed
script would be very helpful.


Thank you in advance,
Leandro Bortot