[gmx-developers] Re: Shift for MARTINI force field

[Michael Shirts]

Good discussion!

>>> Unfortunately this half step/full step issue will soon be reintroduced
>>> in Gromacs,
>>> because the velocity verlet integrator uses full time step velocities
>>> for coupling.
>> This might not be a good idea! Can't you leave leap frog?
> Velocity-verlet is an additional option, it will not replace leap-frog.
> This issue by itself is enough for me to prefer leap-frog in nearly all
> cases.

It's very possible to use an average kinetic energy temperature
control in velocity verlet -- but to do it in the full Trotter
formalism may require iterating over the constraints.  Leapfrog gets
away with this currently by using a linear approximation that isn't
measure preserving (but is reversible).  I'm trying to figure out if
it can be done w/o iterating over constraints right now; I'll probably
wait on this until the velocity verlet framework passes Berk's
approval -- I'm tired of always updating my local code to match
rapidly changing Gromacs :)  But certainly both leapfrog and velocity
verlet would both be options until there's conclusive evidence that
one is better than the other in all circumstances; right now, there's
different situations where one or the other may be useful.

>> I don't see any reason for having the option to output the full step
>> kinetic energy,
>> except for exact comparison against more incorrect results.

See, I would argue that by allowing a comparison vs more inaccurate
results in the same code base, it makes it easier to prove it's their
code that's wrong, not Gromacs.  So I'd argue for putting a switch in,
but turned off by default, and with a warning printed explaining the
issues.

Best,
Michael